Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Deoxyfluorination with Sulfonyl Fluorides
Matthew K Nielsen
A DISSERTATION
PRESENTED TO THE FACULTY
OF PRINCETON UNIVERSITY
IN CANDIDACY FOR THE DEGREE
OF DOCTOR OF PHILOSOPHY
RECOMMENDED FOR ACCEPTANCE
BY THE DEPARTMENT OF CHEMISTRY
Advisor: Abigail G. Doyle
April 2018
© Copyright by Matthew K Nielsen, 2018. All rights reserved.
iii
Abstract
In drug design, the carbon-fluorine bond plays an important role in increasing metabolic
stability while modulating the reactivity, conformation, and solubility of pharmaceutical
candidates. The most straightforward method for selective aliphatic fluorination is the
deoxyfluorination of alcohols. Although reagents such as DAST are routinely used by medicinal
chemists for derivatization, the large scale deployment of deoxyfluorination is rare because
existing reagents suffer from a combination of high cost, poor selectivity, and thermal instability.
We have identified the sulfonyl fluoride motif as an inexpensive, thermally stable
deoxyfluorination reagent class that can be tuned to maximize selectivity for specific substrates.
For example, 2-pyridinesulfonyl fluoride (PyFluor) affords superior yields with unactivated
acyclic secondary alcohols by minimizing elimination side reactions. Further exploration of
existing and new reagents has revealed a complex reaction landscape in which all major classes
of alcohols may be fluorinated in moderate to high yield through the judicious selection of a
sulfonyl fluoride and base possessing complementary stereoelectronics.
Additionally, we demonstrate that machine learning algorithms can be used to identify
non-intuitive reactivity trends from high throughput screening data and predict the optimal
conditions for new, untested substrates. We also investigate the application of sulfonyl fluoride
deoxyfluorination to the radiosynthesis of 18F-labelled compounds, which are vital to medical
imaging with positron emission tomography.
iv
Acknowledgements
I wouldn’t be here without Michele. She did the research on how to prepare for and apply
for graduate schools, she gave me a timeline, helped me write essays, quizzed me with practice
GRE questions while we walked to the grocery store. She got me into every single school I
applied to. And for the past five years, she’s put up with the looming spectre of an organic
chemistry PhD, with me being gone for countless hours, frequently into the middle of the night.
She’s put up the fact that even when I was home, I spent half of it at my computer, and all of it
distracted, thanks to an infinite to-do list. It’s been miserable and painful, the never-ending
struggle between being with the people I love and putting in enough of an effort to secure our
future. And at the same time it’s been mysterious and wonderful and beautiful, we’ve grown in
ways that we never could have imagined, we’ve traveled around this corner of the world, seeing,
learning, feeling, tasting new things. We’ve raised an independent, spunky, self-confident girl
and have another fuzzy baby crawling around the house, exploring the piles of stuff we’ve never
quite managed to keep off the floor. Do I enjoy chemistry? Yes, some aspects of it, some of the
time, but looking back, everything I’ve done in the laboratory pales in comparison to what we
have accomplished and what I hope we will become.
Abby has been the best advisor that I could have hoped for. Somehow, despite my
profound ignorance and inability to maintain a balanced or predictable schedule, she stood back
and allowed me just the right amount of freedom, with just the right amount of prodding for
things to miraculously work out. She allowed me to escape Frick, to make the beautiful drive
down to West Point, to feel the excitement of working inside an actual living, breathing
pharmaceutical company. I don’t think anyone else would have put up with my meeting-induced
narcolepsy, my complete inability to use subtlety or filter my thoughts.
v
To my labmates, I wish I could have gotten to know more of you better. I wish I had gone
running more, played Frisbee more, gone to eat with you more. I wish I had helped keep the
laboratory running better. I tried to help when asked, but certainly did not go above and beyond
the call of duty. I realize it was the time and sacrifices of former and current graduate students
that made my research projects exist, that kept the instruments I needed functional. But with 60 –
80 hours per week and a family, I just couldn’t do more.
I don’t mean to play favorites, but Laura, you were definitely my favorite person to come
through the laboratory; thanks for talking to me and keeping things interesting. Thanks to the
old-timers like Jason, Dennis, and Erin who made the laboratory a friendly and accepting place.
Thanks again to Erin, Ben, Junyi, Derek, to the undergraduates Christian and Orestes, all with
whom I’ve had the privilege of collaborating. Thanks to Tom for mentoring me and passing on
some of his vast unconventional knowledge even when he was pretty sure I wouldn’t make it.
Thanks to my classmates Lucas who kept me well-fed and Julian who inexplicably didn’t
dismiss me as idiotic despite his reputation for doing so. And to everyone else, thanks for putting
up with me.
Thanks to Rob, Paul, and Dave for being on my committee. Thanks to István and Ken of
the NMR facility; my projects would never have succeeded without blatant abuse of the 19F
autosampler. Thanks to Meghan for talking me through admissions, to Clarice who dealt with all
of my receipts and suspicious packages, to Meredith who helped me wrap things up. Thanks to
the Graduate School and Princeton University, whose very generous support has kept us alive
these last five years. Thanks to my chemicals, who actually did what I asked of them most of the
time. And finally thanks to the stars, who kept a lonely vigil on my nocturnal efforts, who kept
me looking up when I finally walked home.
vi
To Michele
vii
Table of Contents
Abstract .......................................................................................................................................... iii
Acknowledgements........................................................................................................................ iv
Table of Contents.......................................................................................................................... vii
List of Figures ................................................................................................................................ ix
List of Schemes............................................................................................................................... x
List of Tables ................................................................................................................................. xi
List of Abbreviations .................................................................................................................... xii
Chapter 1. Discovery of 2-Pyridinesulfonyl Fluoride as a Deoxyfluorination Reagent .......1
1.1 Fluorine in Drug Development ..................................................................................2
1.2 Deoxyfluorination ......................................................................................................6
1.3 Development of 2-Pyridinesulfonyl Fluoride ..........................................................10
1.4 Conclusion ...............................................................................................................24
1.5 Experimental Section ...............................................................................................24
Chapter 2. A Systematic Investigation of Sulfonyl Fluoride Reactivity ..............................42
2.1 Limitations of PyFluor.............................................................................................43
2.2 Development of High-Throughput Screening Approach.........................................47
2.3 Multi-dimensional Screening of Alcohol Substrates ...............................................54
2.4 Modeling Sulfonyl Fluoride Reaction Space via Machine Learning.......................67
2.5 Conclusion ...............................................................................................................77
2.6 Experimental Section ...............................................................................................78
Chapter 3. Low-Temperature Radiofluorination Strategies ..............................................115
3.1 PET Radiochemistry ..............................................................................................116
3.2 Deoxyradiofluorination with Sulfonyl Fluorides...................................................120
viii
3.3 Radiofluorination of α-Diazocarbonyls .................................................................124
3.4 Experimental Section .............................................................................................132
Appendix A. Aryl Formylation via Photocatalytic Generation of Chlorine Radicals ......139
A.1 Aryl Formylation...................................................................................................140
A.2 Redox-Neutral Formylation of Aryl Chlorides with 1,3-Dioxolane.....................142
A.3 Experimental Section ............................................................................................149
ix
List of Figures
Figure 1.1 Examples of fluorinated pharmaceuticals......................................................................3
Figure 1.2 Sulfur(IV) reagents with improved stability..................................................................8
Figure 1.3 2-Pyridinesulfonate as a nucleophile-assisted leaving group. .....................................11
Figure 1.4 DSC trace for 2-pyridinesulfonyl fluoride...................................................................22
Figure 2.1 Sulfonyl fluorides selected for high-throughput screening. ........................................48
Figure 2.2 Kinetic study of sulfonyl fluorides. .............................................................................50
Figure 2.3 Hammett plot for deoxyfluorination with arylsulfonyl fluorides. ...............................51
Figure 2.4 Bases selected for high throughput screening. ............................................................52
Figure 2.5 Conjugate acid pKa trends. .........................................................................................52
Figure 2.6 Deoxyfluorination of primary unactivated alcohols. ...................................................54
Figure 2.7 Deoxyfluorination of amino alcohols via aziridinium intermediates. .........................56
Figure 2.8 Deoxyfluorination of secondary unactivated alcohols. ...............................................56
Figure 2.9 Deoxyfluorination of tertiary unactivated alcohols. ....................................................57
Figure 2.10 Effect of ring strain on SN2 transition state. ..............................................................58
Figure 2.11 Deoxyfluorination of 4-, 5-, and 7-membered cyclic alcohols..................................58
Figure 2.12 Deoxyfluorination of cyclohexanols. ........................................................................60
Figure 2.13 Cyclohexanol nucleophile approach trajectories. ......................................................60
Figure 2.14 Deoxyfluorination of benzylic alcohols. ...................................................................62
Figure 2.15 Deoxyfluorination of allylic alcohols. .......................................................................63
Figure 2.16 Deoxyfluorination of homobenzylic alcohols. ..........................................................64
Figure 2.17 Deoxyfluorination of homoallylic alcohols...............................................................65
Figure 2.18 Deoxyfluorination of α-hydroxycarbonyls. ...............................................................65
Figure 2.19 Deoxyfluorination of β-hydroxycarbonyls. ...............................................................66
Figure 2.20 Deoxyfluorination of hemiacetals. ............................................................................66
Figure 2.21 Simple vs. complex decision trees. ............................................................................71
Figure 2.22 Designated shared substrate atoms and descriptors...................................................73
Figure 2.23 Calibration plot of test set with random forest model. ..............................................74
Figure 2.24 External validation set and calibration plot. ..............................................................75
Figure 3.1 Positron emission tomography. ................................................................................116
x
List of Schemes
Scheme 1.1 Nucleophilic fluorination via halide metathesis. .........................................................5
Scheme 1.2 Nucleophilic fluorination of sulfonate esters with potassium fluoride........................6
Scheme 1.3 Deoxyfluorination with Yarovenko’s Reagent. ...........................................................6
Scheme 1.4 Deoxyfluorination with DAST. ...................................................................................7
Scheme 1.5 Deoxyfluorination with PhenoFluor. .........................................................................10
Scheme 1.6 Silver-catalyzed SN1 fluorination. .............................................................................11
Scheme 1.7 Synthesis of 2-pyridinesulfonate esters. ....................................................................12
Scheme 1.8 Proposed deoxyfluorination with 2-pyridinesulfonyl fluoride. .................................12
Scheme 1.9 Conjugate acid-assisted sulfonylation mechanism. ...................................................14
Scheme 1.10 Non-operative nucleophilic catalysis mechanism....................................................15
Scheme 1.11 Deoxyfluorination with tosyl fluoride and TBAF. ..................................................16
Scheme 1.12 Deoxyfluorination with PBSF. ................................................................................17
Scheme 1.13 Base nucleophilicity trends in deoxyfluorination of primary alcohols....................19
Scheme 1.14 Pfizer synthesis of 2-pyridinesulfonyl fluoride. ......................................................23
Scheme 1.15 Modified, two-step synthesis 2-pyridinesulfonyl fluoride.......................................23
Scheme 1.16 Examples of PyFluor application in the pharmaceutical industry. ..........................24
Scheme 3.1 Synthesis of [18F]FDG. ...........................................................................................118
Scheme 3.2 Deoxyradiofluorination with purified [18F]PBSF. ..................................................121
Scheme 3.3 Deoxyradiofluorination with [18F]PyFluor. ............................................................123
Scheme 3.4 Deoxyradiofluorination of unactivated alcohols with [18F]ArFSF. .........................123
Scheme 3.5 Copper-catalyzed fluorination of α-diazo carbonyls. .............................................125
Scheme 3.6 Radiofluorination of α-trifluoromethyl diazo compounds. .....................................126
Scheme 3.7 Prosthetic groups in radiosynthesis. ........................................................................131
Scheme A.1 Approaches to aryl formylation. ............................................................................140
Scheme A.2 Arylation of Csp3–H bonds with nickel metallaphotoredox. ..................................142
Scheme A.3 Proposed chlorine atom photoelimination mechanism. .........................................143
Scheme A.4 Redox-neutral formylation via chlorine atom photoelimination. ..........................144
Scheme A.5 Formylation with 1,3,5-trioxane. ...........................................................................146
xi
List of Tables
Table 1.1 Intrinsic halide reactivity.................................................................................................4
Table 1.2 Base screen for 2-pyridinesulfonyl fluoride deoxyfluorination. ...................................13
Table 1.3 Effect of sulfonyl fluoride structure on deoxyfluorination. ..........................................18
Table 1.4 Substrate scope for deoxyfluorination with 2-pyridinesulfonyl fluoride. .....................20
Table 2.1 Preliminary investigation of sulfonyl fluoride reactivity. .............................................44
Table 2.2 Comparison of PyFluor and ArFSF reactivity. ..............................................................45
Table 2.3 PyFluor vs. PBSF in the deoxyfluorination of cyclobutanols. ......................................46
Table 2.4 Initial sulfonyl fluoride vs. base screen for unactivated primary alcohols....................48
Table 2.5 Effect of temperature on deoxyfluorination of cyclic substrates. .................................61
Table 2.6 Isolation of deoxyfluorination products. .......................................................................68
Table 3.1 Development of copper catalyzed α-diazocarbonyl radiofluorination. ......................127
Table 3.2 Temperature screen for α-diazocarbonyl radiofluorination. ......................................128
Table 3.3 Preliminary scope of copper-catalyzed α-diazocarbonyl radiofluorination. ..............129
Table 3.4 Radiofluorination of α-diazoacetamides bearing free alcohols and amines. .............131
Table A.1 Optimization of formylation of aryl chlorides with 1,3-dioxolane. ..........................145
Table A.2 Substrate scope for aryl formylation. ........................................................................147
xii
List of Abbreviations
Å angstrom
Ac acetyl
AD Alzheimer’s disease
aq aqueous
Ar aryl
ArFSF pentafluorobenzenesulfonyl fluoride
ATP adenosine triphosphate
ATR attenuated total reflectance
β+ beta positive decay
B3LYP Becke, 3-parameter, Lee-Yang-Parr
BDE bond dissociation enthalpy
BDFE bond dissociation free energy
BHT butylated hydroxytoluene (2,6-di-tert-butyl-4-methylphenol)
Bn benzyl
Boc tert-butoxycarbonyl
BOX bisoxazoline
BTMG 2-tert-butyl-1,1,3,3-tetramethylguanidine
BTPP tert-butylimino-tri(pyrrolidino)phosphorane
Bu butyl
Bz benzoyl
°C degrees Celsius
Cbz carboxybenzyl
4-CF3PhSF 4-(trifluoromethyl)benzenesulfonyl fluoride
cis L.; on the same side
4-ClPhSF 4-chlorobenzenesulfonyl fluoride
CNPhSF cyanobenzenesulfonyl fluoride
CoA coenzyme A
cod 1,5-cyclooctadiene
Cy cyclohexyl
δ chemical shift in parts per million
d deuterium (in NMR solvents)
xiii
D dextrorotatory
DAST N,N-diethylaminosulfur trifluoride
DBN 1,5-diazabicyclo[4.3.0]non-5-ene
DBU 1,8-diazabicyclo[5.4.0]undec-7-ene
DCE 1,2-dichloroethane
DCM dichloromethane
DFI 2,2-difluoro-1,3-dimethylimidazolidine
DIBAL diisobutylaluminum hydride
DMAP 4-N,N-dimethylaminopyridine
DME 1,2-dimethoxyethane
DMF N,N-dimethylformamide
DMSO dimethylsulfoxide
dr diastereomeric ratio
DSC differential scanning calorimetry
dtbbpy 4,4′-di-tert-butyl-2,2′-dipyridyl
E Ger.; entgegen
E1 unimolecular elimination
E2 bimolecular elimination
ee enantiomeric excess
EOB at end-of-bombardment
EOS at end-of-synthesis
eq/equiv equivalent(s)
ESI-TOF electrospray ionization time-of-flight
Et ethyl
eV electron volt
FAO N5-fluoroacetylornithine
FDA (United States) Food and Drug Administration
FDG fludeoxyglucose
FPDA N-(2-(diethylamino)ethyl)-2-fluoropropanamide
FT-ATR Fourier-transform attenuated total reflectance
FTIR Fourier-transform infrared
GC gas chromatography
h hour(s)
xiv
HFIP 1,1,1,3,3,3-hexafluoroisopropanol
HPLC high-performance liquid chromatography
HRMS high-resolution mass spectrometry
Hz hertz
i iso
k rate constant
K equilibrium constant
K222 4,7,13,16,21,24-hexaoxa-1,10-diazabicyclo[8.8.8]hexacosane
KB kilobyte
L levorotatory
LAH lithium aluminum hydride
LC(/)MS liquid chromatography/mass spectrometry
LD50 lethal dose, 50%; (amount of substance sufficient to kill 50% of a test population)
LED light-emitting diode
M molar
mCi milliCurie
Me methyl
MeCN acetonitrile
MHz megahertz
min minute(s)
mol mole
MS mass spectrometry
MSE mean-squared error
MTBD 7-methyl-1,5,7-triazabicyclo[4.4.0]dec-5-ene
n normal (ie. straight-chain)
n.c.a. no carrier added
NfF nonaflyl (1,1,2,2,3,3,4,4,4-nonafluorobutane-1-sulfonyl) fluoride
NFP 4-nitrophenyl 2-fluoropropionate
NFSI N-fluorobenzenesulfonimide
NMR nuclear magnetic resonance
4-NsF 4-nitrobenzenesulfonyl fluoride
OTs para-toluenesulfonate
p para
xv
PBSF perfluorobutanesulfonyl (1,1,2,2,3,3,4,4,4-nonafluorobutane-1-sulfonyl) fluoride
PCC pyridinium chlorochromate
PET positron emission tomography
Ph phenyl
pKa negative base 10 logarithm of the acid dissociation constant
ppm parts per million
Pr propyl
R generic carbon group
R rectus
RCC radiochemical conversion
RCP radiochemical purity
RCY radiochemical yield
Rf retention factor
RMSE root-mean-squared error
rpm revolutions per minute
rr regioisomeric ratio
rt room temperature; 23–24 ºC
s second
S sinister
SCE saturated calomel electrode
SET single-electron transfer
SN1 nucleophilic substitution, unimolecular
SN2 nucleophilic substitution, bimolecular
SNAr nucleophilic aromatic substitution
t/tert tertiary
TBA n-tetrabutylammonium
TBAF n-tetrabutylammonium fluoride
TBAT n-tetrabutylammonium difluorotriphenylsilicate
TBD 1,5,7-triazabicyclo[4.4.0]dec-5-ene
temp temperature
TFA trifluoroacetic acid
THF tetrahydrofuran
TLC thin-layer chromatography
xvi
TMG 1,1,3,3-tetramethylguanidine
Torr Torricelli; (1/760th of one atmosphere)
tr retention time
trans L.; on the opposite side
Trt trityl (triphenylmethyl)
Ts para-toluenesulfonyl
UV ultra-violet
v/v volumetric solution composition
X generic halide
Z Ger.; zusammen
1
Chapter 1.
Discovery of 2-Pyridinesulfonyl Fluoride as a Deoxyfluorination Reagent1
1 Reproduced in part with permission from Nielsen, M. K.; Ugaz, C. R.; Li, W.; Doyle, A. G. J. Am. Chem. Soc. 2015, 137, 9571. © Copyright 2015 American Chemical Society.
2
1.1 Fluorine in Drug Development
Since its identification in 18132 and isolation in 1886,3 fluorine has made an ever-
increasing impact on society and the environment, featured in technologies such as the
refrigerant Freon,4 the chemoresistant fluoropolymer Teflon,5 and the monoisotopic uranium-235
hexafluoride species that enables uranium enrichment. 6 During the second half of the 20th
century, the emergence of selective C–F bond forming reactions for fine chemical synthesis led
to the proliferation of fluorinated pharmaceuticals7 and pesticides.8 Between 1955 and 2012, the
U.S. Food and Drug Administration approved 140 fluorinated drugs accounting for 11% of all
new molecular entities.9 The success of blockbuster drugs such as Lipitor (Figure 1.1) has made
fluorine one of the most easily recognizable pharmaceutical motifs and has ignited renewed
academic interest in fluorination methods, led by investigators such as Véronique Gouverneur,10
Melanie Sanford,11 Tobias Ritter,12 and Stephen Buchwald. 13 The carbon-fluorine bond is
2 Davy, H. Philos. Trans. R. Soc. London. 1813, 103, 263. Antoine Lavoisier had proposed that all acids (ie. H2SO4) contain oxygen (Lavoisier, A.-L. Mémoires de l'Académie des sciences 1778, 248). Davy failed to evolve oxygen from hydrofluoric acid, concluding that fluorine was a distinct element. 3 Moissan, H. C. R. Hebd. Seances Acad. Sci. 1886, 102, 1543. Moissan won the 1906 Nobel prize for preparing fluorine gas by electrolysis of anhydrous HF. 4 Midgley, T.; Henne, A. L. Ind. Eng. Chem. 1930, 22, 542. Midgely also discovered the infamous gasoline additive tetraethyl lead. 5 Plunkett, R. J. US Patent 2230654 A, Feb. 4, 1941. 6 Stahl, R. F. H.; Townend, R. V. US Patent 2953431 A, Sep. 20, 1960. Because fluorine is monoisotopic, UF6 isotopologues differ solely based on the mass of the uranium isotope. In contrast, mass separation of uranium oxides is complicated presence of 17O and 18O. 7 Ilardi, E. A.; Vitaku, E.; Njardarson, J. T. J. Med. Chem. 2014, 57, 2832. 8 Jeschke, P. ChemBioChem 2004, 5, 570. Fujiwara, T.; O’Hagan, D. J. Fluorine Chem. 2014, 167, 16. 9 Data retrieved from <http://www.accessdata.fda.gov/scripts/cder/daf/>. 10 Greedy, B.; Gouverneur, V. Chem. Commun. 2001, 3, 233. 11 Hull, K. L.; Anani, W. Q.; Sanford, M. S. J. Am. Chem. Soc. 2006, 128, 7134. 12 Furuya, T.; Kaiser, H. M.; Ritter, T. Angew. Chem., Int. Ed. 2008, 47, 5993 13 Watson, D. A.; Su, M.; Teverovskiy, G.; Zhang, Y.; Garcia-Fortanet, J.; Kinzel, T.; Buchwald, S. L. Science 2009, 325, 1661.
3
Figure 1.1 Examples of fluorinated pharmaceuticals.
particularly useful in drug design in that it can isosterically replace almost any carbon-hydrogen
bond while modulating metabolic resistance, reactivity of adjacent groups, conformation,14 and
solubility.15
Geologically, fluorine is the 15th most abundant element in earth’s crust with an average
concentration of 500 ppm. 16 Notwithstanding, fluorine is virtually absent in bioorganic
chemistry. In fact, only a dozen fluorinated natural products are known, none of which possess
particularly intriguing structures.17 Because organofluorine is so rare in nature, there has been
little selective pressure for the development of enzymes capable of recognizing and metabolizing
C–F bonds. Moreover, carbon and fluorine form the strongest organic single bond—typically
7 kcal/mol stronger than isosteric C–H bonds. As a result, fluorinated commodity chemicals such
as perfluorooctanesulfonyl fluoride (Scotchgard) are notorious for bioaccumulation. Medicinal 14 For example, fluorination of proline residues in collagen leads to enhanced stability arising from hyperconjugative conformational preferences: Shoulders, M. D.; Kramer, K. J.; Raines, R. T. Bioorg. Med. Chem. Lett. 2009, 19, 3859. 15 O’Hagan, D. Chem. Soc. Rev. 2008, 37, 308. Purser, S.; Moore, P. R.; Swallow, S.; Gouverneur, V. Chem. Soc. Rev. 2008, 37, 320. 16 Wedepohl, K. H. Geochim. Cosmochim. Acta 1995, 59, 1217. 17 O’Hagan, D.; Harper, D. B. J. Fluorine Chem. 1999, 100, 127. All are derived from 2-fluoroacetate except the antibiotic nucleocidin whose C–F bond is constructed by the unique fluorinase enzyme. (O’Hagan, D.; Schaffrath, C.; Cobb, S. L.; Hamilton, J. T. G.; Murphy, C. D. Nature 2002, 416, 279.) 2-Fluoroacetate itself is highly toxic with a human oral LD50 of 2–10 mg/kg. Upon entering the Krebs cycle via fluoroacetylCoA, fluoroacetate is converted to fluorocitrate, a suicide inhibitor of aconitase that arrests aerobic metabolism. (Egekeze, J. O.; Oehme, W. Vet. Hum. Toxicol. 1979, 21, 411.)
4
chemists, however, have harnessed this property to their advantage—the replacement of
oxidizable C–H bonds with inert C–F bonds can dramatically improve the metabolic stability and
lifetime of a pharmaceutical candidate in vivo.
Unfortunately, the economical construction of aliphatic C–F bonds is non-trivial. One
approach is to employ electrophilic fluorine sources including elemental fluorine gas and N–F
bond reagents (ie. NFSI, SelectFluor) that contain the reactive equivalent of a fluorine cation or
fluorine radical.18 As fluorine is the most electronegative element, these reagents are especially
reactive, but may be expensive or hazardous to deploy on large scale.
Alternatively, nucleophilic fluorination makes use of the fluoride anion through
traditional substitution pathways. Intrinsically, fluoride is the most nucleophilic halide because it
possesses the highest charge density; however, in practice, fluoride is often the least reactive due
to solvation effects (Table 1.1). In polar protic solvents, fluoride is surrounded by an almost
impenetrable solvent shell owing to its high enthalpy of solvation, thus rendering it inaccessible
for nucleophilic substitution. In aprotic solvents, most common fluoride salts (ie. KF) are highly
insoluble, resulting in similarly poor reactivity.19 Under both conditions, the order of halide
Table 1.1 Intrinsic halide reactivity.
18 Liang, T.; Neumann, C. N.; Ritter, T. Angew. Chem., Int. Ed. 2013, 52, 8214. 19 Labban, A. K. S.; Yizhak, M. J. Solution Chem. 1991, 20, 221.
5
reactivity is reversed, and the large polarizable iodide anion displays the highest reactivity due to
its low degree of solvation and high solubility. However, when a soluble, weakly-coordinating
counterion such as tetra-n-butylammonium (TBA) is employed under stringently anhydrous
conditions, the unsolvated “naked” fluoride anion is revealed to be far more reactive than iodide
(see example in Table 1.1).20 Crown ethers or cryptands can similarly activate KF; however,
these phase transfer reagents are prohibitively expensive on scale. 21 Thus, under most
economically viable conditions, simple nucleophilic substitution with fluoride anion requires
forcing conditions.
The earliest synthesis of aliphatic fluorides was reported by Nobel laureate Henri
Moissan in 1888 and involved halide metathesis between simple alkyl iodides and silver fluoride,
driven by the formation of highly insoluble silver iodide (Scheme 1.1).22 Although the reaction
proceeds at room temperature, the requirement for stoichiometric silver and unstable iodide
electrophiles is impractical for large scale preparation and purification.
Scheme 1.1 Nucleophilic fluorination via halide metathesis.
By the 1950s, nucleophilic fluorination had progressed slowly. The state-of-the-art
methodology developed by Edgell and Parts employed the newly developed sulfonate ester
electrophile, which could be conveniently synthesized from abundant alcohol precursors
(Scheme 1.2).23 The transformation used inexpensive potassium fluoride but required forcing
conditions—heating at 180 °C in diethylene glycol—as well as continuous removal of product
20 Winstein, S.; Savedoff, L. G.; Smith, S. Tetrahedron Lett. 1960, 1, 24. 21 Liotta, C. L.; Harris, H. P. J. Am. Chem. Soc. 1974, 96, 2250. 22 Moissan, H. Ann. Chim. Phys. 1890, 19, 266. 23 Edgell, W. F.; Parts, L. J. Am. Chem. Soc. 1955, 77, 4899.
6
Scheme 1.2 Nucleophilic fluorination of sulfonate esters with potassium fluoride.
under vacuum. As a result, this chemistry was limited to largely unfunctionalized substrates;
complex, biologically-relevant structures remained inaccessible.
1.2 Deoxyfluorination
In 1957, a group of Soviet chemists quietly revolutionized the field of aliphatic
fluorination with the discovery of Yarovenko’s reagent (Scheme 1.3).24 This α-difluoroamine
was the first example of a deoxyfluorination reagent, a compound capable of converting alcohols
directly to aliphatic fluorides. In solution, Yarovenko’s reagent readily elmininates fluoride to
form an iminium species. The alcohol attacks the iminium carbon to generate a highly reactive
leaving group that is promptly fluorinated by the displaced fluoride. Ultimately, cleavage of the
strong alcohol C–O bond is driven thermodynamically by formation of the amide carbonyl.
Scheme 1.3 Deoxyfluorination with Yarovenko’s Reagent.
In comparison to the SN2 procedures detailed in Schemes 1.1 and 1.2, deoxyfluorination
conveniently bypasses the synthetic steps required to convert alcohols into alkyl halide or
24 Yarovenko, N. N.; Raksha, M. A.; Shemanina, V. N.; Vasileva, A. S. J. Gen. Chem. USSR 1957, 27, 2246. Example from Wang, Z.-H.; Zheng, C.; Li, F.; Zhao, L.; Chen, F.-E.; He, Q.-Q. Synthesis 2012, 699.
7
sulfonate electrophiles. However, the major advantage of deoxyfluorination is that it enables
access to highly reactive leaving groups and/or nucleophiles in situ, allowing for fluorination
under mild conditions which leads to broad substrate tolerance. In simple SN2 reactions such as
that shown in Scheme 1.2, the electrophile and nucleophile must be relatively stable compounds
(ie. tosylates, KF) that can be synthesized, isolated, and carried on to the next reaction,
fundamentally limiting substrate reactivity. In contrast, the α-amino-α-fluoro ether leaving group
generated by Yarovenko’s reagent is much more reactive than tosylate and likely unisolable at
room temperature, but since electrophile formation and fluorination are performed in the same
pot, isolation is unnecessary. Likewise, Yarovenko’s reagent generates anhydrous HF that reacts
with the reagent to form a highly nucleophilic ammonium fluoride that would rapidly hydrate
and decompose under prolonged storage.
Deoxyfluorination entered the mainstream in the 1970s with the discovery of DAST
(N,N-diethylaminosulfur trifluoride) (Scheme 1.4).25 DAST reacts through a mechanism similar
to that of Yarovenko’s reagent, the main difference being that the intermediate leaving group is
Scheme 1.4 Deoxyfluorination with DAST.
25 Middleton, W. J. J. Org. Chem. 1975, 40, 574. Markovskij, L. N.; Pashinnik, V. E.; Kirsanov, A. V. Synthesis 1973, 12, 787. DAST was preceded by the discovery of gaseous SF4: Hasek, W. R.; Smith, W. C.; Engelhardt, V. A. J. Am. Chem. Soc. 1960, 82, 543. Example from Brandes, A.; Loegers, M.; Schmidt, G.; Angerbauer, R.; Schmeck, C.; Bremm, K.-D.; Bischoff, H.; Schmidt, D.; Schuhmacher, J. German Patent DE 19627430 A1, Jan. 15, 1998.
8
so reactive that most reactions are complete within minutes at −78 °C, vastly improving
functional group tolerance.26
The major shortcoming of DAST is that it is thermally unstable and will detonate
catastrophically at temperatures as low as 108 °C, rendering it unsuitable for industrial scale
use.27 Additionally, DAST reacts violently with water to form HF and will fume when exposed
to air. The reagent must be stored sealed in a freezer to prevent decomposition. With secondary
and tertiary alcohols, DAST often generates significant quantities of alkene elimination side
products28 that can be quite challenging to separate from the desired alkyl fluoride due to the
similarity in polarity and boiling point. On a positive note, DAST is fairly inexpensive ($374 per
mol from Oakwood).29
In an effort to improve upon the thermal stability of DAST, several sulfur(IV) derivatives
have since been reported, notable examples of which are shown in Figure 1.2. Deoxo-Fluor
($1,080 per mol from Acros) is actually less reactive than DAST and has a lower decomposition
Figure 1.2 Sulfur(IV) reagents with improved stability.
26 Above room temperature, DAST will even react with ketones and aldehydes to form geminal difluorides. Alcohols can be selectively fluorinated at cryogenic temperatures. 27Messina, P. A.; Mange, K. C.; Middleton, W. J. J. Fluorine Chem. 1989, 42, 137. L’Heureux, A.; Beaulieu, F.; Bennett, C.; Bill, D. R.; Clayton, S.; LaFlamme, F.; Mirmehrabi, M.; Tadayon, S.; Tovell, D.; Couturier, M. J. Org. Chem. 2010, 75, 3401. DSC data indicates a single sharp spike corresponding to an energy release of 63 kcal/mol. 28 Elimination can be minimized through rigorous exclusion of water and careful temperature control. Reported results for identical substrates vary widely among users, indicative of a steep learning curve. 29 DAST is synthesized from the reaction of SF4 gas with N,N-diethyltrimethylsilylamine.
9
onset temperature, but the exotherm is not as sudden.30 XtalFluor-E ($769 per mol from Alrich)
is a crystalline salt with improved stability during storage and transport; however, the reagent
simply dissociates into DAST and BF3 in solution and is subject to the same thermal onset
temperature.31 Fluolead ($3,710 per mol from Millipore-Sigma) displays higher reactivity than
DAST, albeit with a substantial increase in cost.32
Dozens of novel deoxyfluorination reagents have been reported since the 1970s, although
most have received little attention from the synthetic community.33 One notable exception is
PhenoFluor, a fluoroimidazolium reagent developed by the Ritter laboratory in 2011
(Scheme 1.5).34 Remarkably, PhenoFluor is capable of fluorinating phenols through a concerted
SNAr mechanism wherein the leaving group delivers fluoride through a four-membered transition
state. 35 Moreover, PhenoFluor can be used to fluorinate alcohols in densely-functionalized,
complex natural products with unprecedented selectivity and efficiency (Scheme 1.5).36 The
major limitation of PhenoFluor is its prohibitive cost, currently $325,000 per mol from
Millipore-Sigma. PhenoFluor also suffers from poor bench stability, although the Ritter
30 Lal, G. S.; Pez, G. P., Pesaresi, R. J.; Prozonic, F. M.; Cheng, H. J. Org. Chem. 1999, 64, 7048. 31 Beaulieu, F.; Beauregard, L.-P.; Courchesne, G.; Couturier, M.; LaFlamme, F.; L’Heureux, A. Org. Lett. 2009, 11, 5050. 32 Umemoto, T.; Singh, R. P.; Xu, Y.; Saito, N. J. Am. Chem. Soc. 2010, 132, 18199. For example, DAST reacts with carboxylic acids to form acid fluorides; Fluolead will convert carboxylic acids to trifluoromethyl groups. 33 Olah’s reagent: Olah, G. A.; Nojima, M.; Kerekes, I. Synthesis. 1973, 12, 786. SeF4: Olah, G. A.; Nojima, M.; Kerekes, I. J. Am. Chem. Soc. 1974, 96, 925. Ishikawa’s Reagent: Takaoka, A.; Iwakiri, H.; Ishikawa, N. Bull. Chem. Soc. Jpn. 1979, 52, 3377. Perfluorobutane ylides: Pasenok, S. V.; de Roos, M. E.; Appel, W. K. Tetrahedron 1996, 52, 2977. 34 Tang, P.; Wang, W.; Ritter, T. J. Am. Chem. Soc. 2011, 133, 11482. 35 Neumann, C. N.; Hooker, J. M.; Ritter, T. Nature 2016, 534, 369. The structurally similar reagent DFI was previously reported to fluorinate electron-deficient phenols: Hayashi, H.; Sonoda, H.; Fukumura, K.; Nagata, T. Chem. Commun. 2002, 1618. 36 Sladojevich, P.; Arlow, S. I.; Tang, P.; Ritter, T. J. Am. Chem. Soc. 2013, 135, 2470.
10
Scheme 1.5 Deoxyfluorination with PhenoFluor.
laboratory has addressed this by commercializing the reagent as a solution in toluene37 and
developing stable precursor salt mixtures that form the active reagent in situ.38 PhenoFluor may
prove valuable in medicinal chemistry for accessing small quantities of high-value targets that
would be otherwise inaccessible through conventional methodology but is unlikely to be
economically viable on preparatory scale.
Although the patent literature indicates that sulfur(IV) reagents are routinely employed
for derivatization by medicinal chemists, there are no FDA-approved drugs for which
deoxyfluorination is employed in final synthetic route. We suggest that this is because there are
no established reagents that are simultaneously selective, inexpensive, and thermally stable.
1.3 Development of 2-Pyridinesulfonyl Fluoride
Our laboratory’s interest in deoxyfluorination arose somewhat serendipitously. In the
spring of 2014, we identified an interesting transformation in which a silver Lewis acid could
catalyze the SN1 fluorination of a heterocyclic sulfonate ester (Scheme 1.6). We had selected the
pyridinesulfonate ester based on reports that it could behave as a nucleophile-assisted leaving
37 Fujimoto, T.; Becker, F.; Ritter, T. Org. Process. Res. Dev. 2014, 18, 1041. 38 PhenoFluor Mix: Fujimoto, T.; Ritter, T. Org. Lett. 2015, 17, 544. AlkylFluor: Goldberg, N. W.; Shen, X.; Li, J.; Ritter, T. Org. Lett. 2016, 18, 6102.
11
Scheme 1.6 Silver-catalyzed SN1 fluorination.
group as shown in Figure 1.3.39 In our proposed mechanism, the silver cation would coordinate
to the pyridine nitrogen, facilitating ionization under catalytically controlled conditions while
simultaneously delivering fluoride.40 Studies with enantioenriched starting material indicated that
a stereoablative SN1 mechanism was operative.41 Control experiments demonstrated that the
reaction could be carried with stoichiometric silver fluoride, supporting our theory that silver
might be responsible for fluoride delivery. 42 Our intention was to develop a catalytic
enantioselective SN1 fluorination, but unfortunately, we were never able to identify a ligand or
chiral counterion that would give measureable enantioselectivity.
Figure 1.3 2-Pyridinesulfonate as a nucleophile-assisted leaving group.
One frustrating aspect of this project was that the substrate sulfonate ester would
decompose rapidly even when stored in the freezer and had to be freshly prepared on a weekly
39 Hanessian, S.; Kagotani, M.; Komaglou, K. Heterocycles 1989, 28, 1115. Lepore, S. D.; Mondal, D.; Li, S. Y.; Bhunia, A. K. Angew. Chem., Int. Ed. 2008, 47, 7511. Ortega, N.; Feher-Voelger, A.; Brovetto, M.; Padron, J. I.; Martín, V. S.; Martín, T. ́ Adv. Synth. Catal. 2011, 353, 963. 40 The neodecanoate cation helps solubilize the silver salt, but may also initiate esterification of benzoyl fluoride. 41 Racemization of recovered sulfonate ester during the course of the reaction indicates that sulfonate ionization is reversible. 42 Under catalytic conditions, anhydrous HF is formed via esterification of benzoyl fluoride and hexafluoroisopropanol.
12
basis from the corresponding alcohol and sulfonyl chloride (Scheme 1.7). 43 2-Pyridinesulfonyl
chloride is even less stable and had to be prepared and used immediately or frozen in benzene.
Scheme 1.7 Synthesis of 2-pyridinesulfonate esters.
A senior graduate student, Thomas Graham, suggested that even if we succeeded in
developing the desired method, it was unlikely to be adopted due to the complexity of substrate
synthesis. He proposed that we could streamline the method by replacing the sulfonyl chloride
with a sulfonyl fluoride, which would generate both our substrate ester and fluoride in situ in a
formal deoxyfluorination (Scheme 1.8).
Scheme 1.8 Proposed deoxyfluorination with 2-pyridinesulfonyl fluoride.
Initially, we greeted this proposal with some skepticism. The proposed sulfonyl fluoride
structure is quite similar to tosyl chloride, the reagent used to generate tosylate esters. When
employing tosyl chloride, one does not usually expect chloride to come back around and displace
tosylate to give the deoxychlorinated product. Nevertheless, with the assistance of visiting
undergraduate Christian Ugaz, we began to investigate this proposal in the summer of 2014.
Following a reported synthesis for 2-pyridinesulfonyl fluoride,44 we selected alcohol 1.1 and
began to screen bases while assaying for formation of ester 1.2 or product 1.3 as shown in
Table 1.2.
43 Corey, E. J. Posner, G. H.; Atkinson, R. F.; Wingard, A. K.; Halloran, D. J.; Radzik, D. M.; Nash, J. J. J. Org. Chem. 1989, 54, 389. 44 Wright, S. W.; Hallstrom, K. N. J. Org. Chem. 2006, 71, 1080.
13
Table 1.2 Base screen for 2-pyridinesulfonyl fluoride deoxyfluorination.
OS
O O
N
Me
OH
Me base (2 equiv)toluene (0.4 M)room temp, 48 h
N
O O
SF
PhPh
F
MePh
1.1 1.2 1.3
basetosylate ester (1.2)
% yieldfluoride (1.3)
% yield% conversion
pyridine
triethylamine
0 0
16 0
13
33
DMAP 25 1 55
sodium hydride 0 0 0
DBU 97 77 100
(at 5 minutes) (at 48 hours)N
N
N
NMe2
In stark contrast to the unstable 2-pyridinesulfonyl chloride, 2-pyridinesulfonyl fluoride
appeared to be almost completely unreactive. With pyridine and triethylamine, bases typically
used for forming sulfonate esters from sulfonyl chlorides, low conversions were observed with
little ester formation. The nucleophilic catalyst DMAP provided our first glimpse of fluorinated
product in approximately 1% yield, but we were unable to further improve yield by modifying
conditions or employing other DMAP analogues. Most perplexingly, quantitatively preforming
the alkoxide with sodium hydride failed to provide any ester formation or conversion; in fact, we
observed full recovery of the sulfonyl fluoride. If the fully-deprotonated alkoxide was not
nucleophilic enough to react with the sulfonyl fluoride at room temperature, it seemed unlikely
that our transformation would be successful as proposed. Nevertheless we continued screening
until we stumbled across the amidine base DBU, which quite unexpectedly afforded almost
quantitative yield of the sulfonate ester within just a few minutes. Moreover, if the reaction was
allowed to run for 48 hours, we obtained 77% yield of the desired fluorinated product. In a single
experiment, we had gone from almost no reactivity to near optimal conditions.
14
A few months later, the Sharpless laboratory released a review on sulfonyl fluoride
reactivity that explained the unexpected success of DBU. 45 As we had suspected, sulfonyl
fluorides (S–F BDE: 91 kcal/mol) are much less reactive than sulfonyl chlorides (S–Cl BDE: 46
kcal/mol). In order for the sulfonyl fluoride to react, the departing fluoride must be stabilized by
a positively charged protic species as shown in Scheme 1.9. Referring to the bases in Table 1.2,
pyridine (pKa (conjugate acid, aq) = 5.3), DMAP (pKa (conjugate acid, aq) = 9.6), and triethylamine (pKa
(conjugate acid, aq) = 11.0) are simply not strong enough bases for there to exist any significant
quantity of protonated conjugate acid (secondary alcohol pKa (aq) = 16.5). Sodium hydride (pKa
(conjugate acid, aq) = 36), is certainly strong enough to fully deprotonate the alcohol; however, the
conjugate acid is hydrogen gas which bubbles out of solution and is not particularly useful for
stabilizing negative charge. On the other hand, amidine DBU is one of the strongest neutral
organic bases with an aqueous pKa of 13.5. Although equilibrium still favors the free base, there
is a high enough concentration of conjugate acid for the mechanism shown in Scheme 1.9 to
proceed at a reasonable rate. In addition to generating the desired sulfonate electrophile, the
protonated amidine forms a delocalized soluble counterion for fluoride, resulting in a strongly
nucleophilic species.
Scheme 1.9 Conjugate acid-assisted sulfonylation mechanism.
An alternative mechanistic possibility is that DBU serves as a nucleophilic catalyst and
displaces fluoride to form a sulfonyl transfer reagent in similar fashion to DMAP catalyzed
45 Dong, J.; Krasnova, L.; Finn, M. G.; Sharpless, B. K. Angew. Chem., Int. Ed. 2014, 53, 9430.
15
acylations (Scheme 1.10).46 To investigate this proposal, we exposed 2-pyridinesulfonyl fluoride
to DBU and found that the expected DBU adduct gradually grew in to high conversion over the
course of approximately two hours. However, when alcohol 1.1 was added, we observed no
conversion to the sulfonate ester 1.2. Since sulfonate ester formation is complete within minutes
under our reaction conditions, we believe that esterification proceeds almost exclusively through
the mechanism presented in Scheme 1.9.
Scheme 1.10 Non-operative nucleophilic catalysis mechanism.
Optimization studies with substrate 1.1 indicated that the base stoichiometry is important
as well. Inexplicably, the reaction performs best with 2 equivalents of DBU (79% yield) with
yields declining to the mid-60s with either 1.25 or 3 equivalents. On the other hand, only a slight
excess of sulfonyl fluoride (~1.1 equivalents) is necessary. A surprisingly wide range of solvents
is tolerated—toluene and ethereal solvents performed best, but 1.3 was still formed in 40 – 50%
yield in the polar solvents DMSO, acetonitrile, and DMF. Reaction concentrations above 0.4 M
are necessary to obtain full conversion within 48 hours.
It was at this time that we became aware of some precedents in sulfonyl fluoride
deoxyfluorination. The earliest report from Shimizu and Yoshioka in 1985 detailed the
fluorination of primary alcohols in refluxing THF with tosyl fluoride and TBAF
(Scheme 1.11).47 In this case, tosyl fluoride does not necessarily serve as the principal fluoride
source. TBAF-trihydrate is sufficiently nucleophilic to fluorinate primary tosylates in high yield
46 Taylor, J. E.; Bull, S. D.; Williams, J. M. J. Chem. Soc. Rev. 2012, 41, 2109. 47 Shimizu, M.; Nakahara, Y.; Yoshioka, H. Tetrahedron Lett. 1985, 26, 4207.
16
at room temperature, so the high reaction temperature is most likely necessary to force the
alcohol to react with tosyl fluoride. The same effect could be achieved by pregenerating the
tosylate from tosyl chloride in situ at room temperature prior to addition of TBAF. Interestingly,
secondary and tertiary alcohols are tolerated as peripheral functional groups.
Scheme 1.11 Deoxyfluorination with tosyl fluoride and TBAF.
The more relevant and disconcerting precedent was reported by Helmut Vorbrüggen in
1995. Under shockingly similar conditions—with DBU as base in toluene—Vorbrüggen showed
that perfluorobutanesulfonyl fluoride (PBSF, also referred to as nonaflyl fluoride or NfF) could
fluorinate secondary cyclic alcohols in a steroid scaffold with reasonable efficiency although
some racemization and elimination was observed (Scheme 1.12).48 Since this initial report, PBSF
has been examined by several other groups49 and employed on kilogram scale.50 Alternative
procedures have arisen that use triethylamine-HF51 and tetrabutylammonium difluorotriphenyl-
silicate (TBAT)52 as supplementary fluoride sources.
48 Bennua-Skalmowski, B.; Vorbrüggen, H. Tetrahedron Lett. 1995, 36, 2611. See also an earlier iteration employing DMAP: Bennua-Skalmowski, B.; Krolikiewicz, K.; Vorbrüggen, H. Bull. Soc. Chim. Belg. 1994, 103, 453. 49 Marson, C. M.; Decréau, R. A.; Smith, K. E. Synth. Commun. 2002, 32, 2125. Decréau, R. A.; Marson, C. M. Synth. Commun. 2004, 34, 4369. Takamatsu, S.; Katayama, S.; Hirose, N.; De Cock, E.; Schelkens, G.; Demillequand, M.; Brepoels, J.; Izawa, K. Nucleosides, Nucleotides, Nucleic Acids 2002, 21, 849. Izawa, K.; Takamatsu, S.; Katayama, S.; Hirose, N.; Kozai, S.; Maruyama, T. Nucleosides, Nucleotides, Nucleic Acids 2003, 22, 507. Egli, M.; Pallan, P. S.; Allerson, C. R.; Prakash, T. P.; Berdeja, A.; Yu, J.; Lee, S.; Watt, A.; Gaus, H.; Bhat, B.; Swayze, E. E.; Seth, P. P. J. Am. Chem. Soc. 2011, 133, 16642. See also reviews: Vorbrüggen, H. Synthesis 2008, 1165. Vorbrüggen, H. Helv. Chim. Acta 2011, 94, 947. 50 Daubié, C.; Mutti, S. Tetrahedron Lett. 1996, 37, 7743.
17
Scheme 1.12 Deoxyfluorination with PBSF.
With this staggering amount of precedent, we were forced to ask ourselves how
2-pyridinesulfonyl fluoride would represent any improvement over PBSF with regards to the
criteria of selectivity, cost, or stability. PBSF is inexpensive ($154.79 per mol from Oakwood,
synthesized by electrolytic fluorination of sulfolane.) Vorbrüggen has since reported that PBSF
can decompose in the presence of tertiary amines or heterocycles to form perfluorobutane gas,
which could potentially lead to reaction overpressure;53 however, we have not found this to be a
significant issue in our hands. It still remained to be seen how the selectivity of PBSF compared
to that of 2-pyridinesulfonyl fluoride. Hitherto, we had proceeded under the assumption that the
2-pyridinesulfonate significantly enhanced reaction rate by behaving as a nucleophile-assisted
leaving group (refer back to Figure 1.3). Thoroughly chastened, we proceeded to screen several
dozen sulfonyl fluorides as summarized in Table 1.3.
To our surprise, 2-pyridinesulfonyl fluoride remained the highest-yielding reagent.
Initially, we were suspicious that our optimal conditions might represent a local maximum, and
that we had screened ourselves into a corner by performing one-dimensional optimization.
Despite investigating the various conditions previously reported for PBSF and performing wide-
ranging multi-dimensional screens, we never identified another sulfonyl fluoride that would
51 Yin, J.; Zarkowsky, D. S.; Thomas, D. W.; Zhao, M. M.; Huffman, M. A. Org. Lett. 2004, 6, 1465. 52 Zhao, X.; Zhuang, W.; Fang, D.; Xue, X.; Zhou, J. Synlett 2009, 779. 53 Bennua-Skalmowski, B.; Klar, U.; Vorbrüggen, H. Synthesis 2008, 1175.
18
reliably deliver 1.3 in significantly higher yield. We did find, however, that migrating the
nitrogen from the 2- to the 3- or 4-pyridyl positions had almost no effect at all on yield (entries 1
– 3), invalidating our nucleophile-assisted leaving group hypothesis. Instead, reactivity appeared
to correlate with electron-withdrawing capability of the aryl ring. For example, electron-rich
tosyl fluoride reacted at an excruciatingly slow rate, affording only 21% yield after 72 hours
(entry 4), whereas electron-deficient nosyl fluoride afforded a respectable 72% yield (entry 5).
Table 1.3 Effect of sulfonyl fluoride structure on deoxyfluorination.
SO2F
O2N
F3C SO2F
F F
F F
F F
SO2F
Me
72% yield 12:1
57% yield 6:1
21% yield 17:1
yieldselectivity
reagent
Ph Me
OH
Ph Me
Fsulfonyl fluoride (1.1 equiv)DBU (2 equiv)
toluene (0.4 M), rt, 72 h
6
4
5
entry
1.1 1.3
(fluorination : elimination)
N
SO2F
N
SO2F
N
SO2F
79% yield >20:1
74% yield 13:1
70% yield 12:1
1
2
3
PBSF proved to be only moderately selective, affording 57% yield of 1.3 with 10% yield
of elimination side products (entry 6, only 4% elimination was observed with 2-pyridinesulfonyl
fluoride). This poor selectivity arises because the perfluorobutanesulfonate ester is simply too
reactive at room temperature (with reaction times measured in seconds) and will spontaneously
ionize leading to E1 elimination as well as stereoerosion. In comparison to primary alcohols, the
fluorination of secondary alcohols requires a soft touch. The congested backside approach lowers
the rate of SN2 substitution relative to primary substrates. At the same time, secondary
19
electrophiles typically have more β hydrogens and can form more substituted alkenes, increasing
the rate and thermodynamic favorability of elimination side reactions. Based on our results, as
leaving group ability decreases, the rate of elimination decreases more rapidly than the rate of
substitution, leading to higher selectivity. The 2-pyridinesulfonate leaving group coincidentally
provides the best balance of reactivity and selectivity at room temperature within the 72 hour
screening time period.
Unfortunately, while our optimized conditions worked well for other secondary alcohols,
we obtained only mediocre yields from primary alcohols. To our bemusement, the major side
product was identified by LCMS as an adduct arising from SN2 substitution by DBU base
(Scheme 1.12). Ironically, the Millipore-Sigma catalogue describes DBU as “a non-nucleophilic,
sterically hindered, tertiary amine base”. Screening a range of amidine and guanidine bases
revealed a clear trend between yield and steric encumbrance (Scheme 1.13). More compact bases
such as TMG, TBD, and DBN predominantly formed the base adduct, but fortunately the bulkier
MTBD enabled us to access primary fluorides such as 1.4 in higher than 80% yield.
Scheme 1.13 Base nucleophilicity trends in deoxyfluorination of primary alcohols.
We proceeded to demonstrate that 2-pyridinesulfonyl fluoride can fluorinate a broad
range of primary and secondary alcohols including sugars, steroids, and amino acids (1.5 – 1.8)
20
(Table 1.4). The transformation tolerates medicinally relevant functionality including
heterocycles (1.9), protected nitrogens (1.10 – 1.12), and even unprotected tertiary amines (1.13,
1.14). Primary and secondary alcohols can be fluorinated in the presence of less acidic,
unprotected tertiary alcohols (1.15). On the other hand, homobenzylic and secondary benzylic
alcohols, were susceptible to elimination, driven by the formation of the stabilized styrene (1.19
– 1.21). Primary benzylic alcohols were particularly low-yielding and predominantly formed the
base substitution adduct even when MTBD was employed (1.22). Finally, β-hydroxy alcohols
underwent exclusive elimination to the α,β-unsaturated structure (1.23). In general, primary
substrates performed much better with MTBD, while sterically congested, secondary alcohols
Table 1.4 Substrate scope for deoxyfluorination with 2-pyridinesulfonyl fluoride.
21
were favored by DBU. Cyclic substrates required heating, which in the case of steroid 1.7 led to
some elimination and erosion of stereochemistry. Overall, deoxyfluorination with
2-pyridinesulfonyl fluoride demonstrated breadth and functional group tolerance that compares
quite favorably with DAST, Yarovenko’s reagent, and other inexpensive deoxyfluorination
reagents.
Next, we examined the thermal stability of 2-pyridinesulfonyl fluoride. Historically, the
stability of DAST and other deoxyfluorination reagents has been assessed with differential
scanning calorimetry (DSC). When DAST is heated from room temperature to 300 °C at a rate of
5 °C per minute, a sharp, δ-function-like exotherm of 63 kcal/mol is observed at 147 °C,
inidicating that on large scale, a cooling failure could quickly lead to a catastrophic detonation.54
Deoxo-Fluor has a similar exotherm of 55 kcal/mol, but decomposition occurs over a much
broader temperature range from 135°C – 190 °C. Even PhenoFluor undergoes a 62 kcal/mol
thermal decomposition, but the onset temperature is much higher, approximately 210°C.55
When we subjected 2-pyridinesulfonyl fluoride to DSC, endotherms corresponding to
melting and evaporation were observed at 24 °C and 262 °C, respectively, but no exothermic
decomposition was observed up to 350 °C (Figure 1.4). This result was not unexpected; as noted
in Table 1.2, sulfonyl fluorides are almost completely unreactive until exposed to the strong
bases DBU or MTBD. 2-Pyridinesulfonyl fluoride is a low-melting crystalline solid that unlike
DAST can be stored on the benchtop for months without significant decomposition. Moreover,
the reagent does not hydrolyze during aqueous workup and can be purified by silica gel
chromatography, both manipulations that would completely destroy DAST.
54 Messina, P. A.; Mange, K. C.; Middleton, W. J. J. Fluorine Chem. 1989, 42, 137. L’Heureux, A.; Beaulieu, F.; Bennett, C.; Bill, D. R.; Clayton, S.; LaFlamme, F.; Mirmehrabi, M.; Tadayon, S.; Tovell, D.; Couturier, M. J. Org. Chem. 2010, 75, 3401. 55 Sladojevich, F.; Arlow, S. I.; Tang, P.; Ritter, T. J. Am. Chem. Soc. 2013, 135, 2470.
22
Figure 1.4 DSC trace for 2-pyridinesulfonyl fluoride.
Much to our consternation, less than a week after obtaining our DSC results, we were in
the process of distilling 2-pyridinesulfonyl fluoride when a sudden pressure spike blew off the
distillation head and sent a cloud of thick yellow smoke spilling out of the fume hood. Our
follow-up investigation suggested that the excursion had arisen from decomposition of a side
product—2-pyridyl disulfide 1.24—found in our crude reaction extract. Up to this point, we had
been following a synthesis reported by Pfizer wherein oxidation of 2-mercaptopyridine and
halide exchange of the resultant sulfonyl chloride were performed in the same pot
(Scheme 1.14), a procedure likely designed in response to the aforementioned instability of
23
2-pyridinesulfonyl chloride.56 Because the reaction is biphasic, tetrabutylammonium hydrogen-
sulfate is added as a phase-transfer reagent and the mixture must be stirred vigorously to form an
emulsion. On small scale, yields as high as 83% were obtained. Unfortunately, attempts to scale
up the reaction resulted in diminishing returns due to phase separation, even when mechanical
stirrers were employed. In low-yielding reactions, the major side product was found to be
disulfide 1.24, an intermediate in the oxidation of 2-mercaptopyridine.
Scheme 1.14 Pfizer synthesis of 2-pyridinesulfonyl fluoride.
To resolve this issue, we decided to perform a two-step synthesis (Scheme 1.15). First,
2-mercaptopyridine ($38.91 per mol from Chem Impex) was oxidized to the sulfonyl chloride in
a monophasic system followed by a single ethyl acetate extraction. The concentrated residue
(containing little disulfide) was then stirred in a mixture of saturated potassium bifluoride and
acetonitrile resulting in almost quantitative formation of the sulfonyl fluoride. Overall, this
procedure consumes less than $200 per mol of reagents on laboratory scale; and on production
scale would likely be competitive with DAST ($374 per mol from Oakwood) and PBSF
($154.79 per mol from Oakwood). In collaboration with Sigma-Aldrich (now Millipore-Sigma)
we were able to commercialize 2-pyridinesulfonyl fluoride under the name of PyFluor.
Scheme 1.15 Modified, two-step synthesis 2-pyridinesulfonyl fluoride.
56 Wright, S. W.; Hallstrom, K. N. J. Org. Chem. 2006, 71, 1080.
24
1.4 Conclusion
Our work describing the development of PyFluor was published in July of 2015,57 and
the reagent has since been utilized in both academia and industry (Scheme 1.16).58 PyFluor is
one of the only deoxyfluorination reagents that simultaneously provides decent selectivity and
high thermal stability at low cost. We hope that the availability of this and other sulfonyl
reagents will enable deoxyfluorination on preparatory and manufacturing scale and expand the
available chemical space for pharmaceutical development.
Scheme 1.16 Examples of PyFluor application in the pharmaceutical industry.
1.5 Experimental Section
General Methods and Instrumentation. (These apply to all chapters.) Reactions were
monitored by thin-layer chromatography on EMD Silica Gel 60 F254 plates, visualizing with UV-
light (254 nm). Organic solutions were concentrated under reduced pressure using a rotary
evaporator (23 °C, <50 torr). Automated column chromatography was performed using silica gel
cartridges on a Biotage Isolera 4 (repacked with 40-53 μm silica from Silicycle). Proton nuclear
57 Nielsen, M. K.; Ugaz, C. R.; Li, W.; Doyle, A. G. J. Am. Chem. Soc. 2015, 137, 9571. 58 Erickson, L. W.; Lucas, E. L.; Tollefson, E. J.; Jarvo, E. R. J. Am. Chem. Soc. 2016, 138, 14006. Galatsis, P.; Henderson, J. L.; Kormos, B. L.; Kurumbail, R. G.; Reese, M. R.; Stepan, A. F.; Verhoest, P. R.; Wager, T. T.; Pettersson, M. Y.; Garnsey, M. R. World Patent WO2017046675 A1, Mar. 23, 2017. Zheng, W.; Zhu, X.; Du, H.; Postema, M.; Jiang, Y.; Li, J.; Yu, R.; Choi, H.-W.; Lee, J.; Fang, F.; Custar, D. World Patent WO 2017066633 A1, Apr. 20, 2017.
25
magnetic resonance (1H NMR) spectra and carbon nuclear magnetic resonance (13C NMR)
spectra were recorded on a Bruker 500 AVANCE equipped with a cryoprobe (500 and 125 MHz,
respectively). Chemical shifts for protons are reported in parts per million downfield from
tetramethylsilane and are referenced to residual protium in the NMR solvent (CHCl3 = δ 7.26
ppm, CHDCl2 = δ 5.32 ppm, C6HD5 = δ 7.16 ppm, DMSO-d5 = δ 2.50 ppm). Chemical shifts for
carbon are reported in parts per million downfield from tetramethylsilane and are referenced to
the carbon resonances of the solvent peak (CDCl3 = δ 77.16 ppm, CD2Cl2 = δ 53.84 ppm, C6D6 =
δ 128.06 ppm, DMSO-d6 = δ 39.52 ppm). 19F fluorine spectra were recorded on either a Bruker
300 AVANCE (282 MHz) or a Bruker-adapted 400 MHz Oxford magnet (376 MHz); chemical
shifts are reported in parts per million and are referenced to CFCl3 (δ 0 ppm). NMR data are
represented as follows: chemical shift (δ ppm), multiplicity (s = singlet, d = doublet, t = triplet,
q = quartet, p = pentet, h = hextet, hept = heptet, m = multiplet), coupling constant in Hertz (Hz),
integration). All NMR spectra were taken at 25 °C. High-resolution mass spectra were obtained
on an Agilent 6220 LC/MS with an electrospray ionization time-of-flight (ESI-TOF) detector.
FTIR and FT-ATR spectra were recorded on a Perkin-Elmer Spectrum 100 and are reported in
terms of frequency of absorption (cm–1) and intensity (s = strong, m = moderate, w = weak,
br = broad). Gas chromatography (GC) was performed on an Agilent 7890A series instrument
equipped with a split-mode capillary injection system and a flame ionization detector. Liquid
chromatograpy-mass spectrometry (LCMS) data was obtained on an Agilent 1260 Infinity
instrument with a binary pump, a diode array detector, and an Agilent 6120 quadrupole detector.
Materials. Potassium bifluoride was obtained from Acros. 13% Sodium hypochlorite was
purchased from Fisher. 4-Phenyl-2-butanol was obtained from TCI. 1,8-Diazabicyclo[5.4.0]
undec-7-ene (DBU) was purchased from Oakwood. 7-Methyl-1,5,7-triazabicyclo[4.4.0]dec-5-
26
ene (MTBD) was purchased from Millipore-Sigma. Toluene and dichloromethane were
dispensed from a dry solvent system. Acetonitrile was purchased from EMD. Suppliers for all
other materials are noted in the individual procedures.
Synthesis of PyFluor (2-pyridinesulfonyl fluoride):
A 2-L multi-neck flask fitted with a mechanical stirrer, an addition funnel, and a thermometer
was charged with sulfuric acid (160 mL, 95 – 98%, EMD) and 2-mercaptopyridine (11.1 g, 100
mmol, Chem-Impex) and left open to atmosphere. The reaction mixture was cooled to 0 °C and
13% aqueous sodium hypochlorite (475 mL, 10 equiv) was added dropwise over 4 hours while
stirring vigorously and maintaining an internal reaction temperature below 10 °C. Warning! This
addition generates chlorine gas. The system must not be closed and should be adequately
ventilated. Upon complete addition, the reaction mixture was extracted twice with 250 mL ethyl
acetate. The combined organic extracts were concentrated to afford crude 2-pyridinesulfonyl
chloride. This was immediately added to a 1 L flask containing 1:3 (v/v) acetonitrile:water
(160 mL) and potassium bifluoride (39.1 g, 5 equiv). Warning! Potassium bifluoride solutions
are hazardous and will etch glassware. This solution was stirred vigorously for 20 minutes and
was then extracted twice with 150 mL ethyl acetate. The organic extracts were dried with sodium
sulfate and filtered through a 10 g silica plug, eluting with ethyl acetate. The concentrate was
purified by vacuum distillation at 55 °C and 60 mTorr to afford product as a colorless crystalline
solid (11.7 g, 73% yield).59 1H NMR (500 MHz, CDCl3): δ 8.84 (d, J = 4.8 Hz, 1H), 8.14 (d,
J = 7.9 Hz, 1H), 8.06 (tt, J = 7.8, 1.5 Hz, 1H), 7.72 (ddd, J = 7.7, 4.7, 1.1 Hz, 1H). 13C NMR
59 Wright, S. W.; Hallstrom, K. N. J. Org. Chem. 2006, 71, 1080.
27
(125 MHz, CDCl3): δ 151.39 (d, J = 30.4 Hz), 151.14 (d, J = 1.2 Hz), 138.84, 129.35, 124.24 (d,
J = 2.1 Hz). 19F NMR (376 MHz, CDCl3): δ −144.29 (s). (Note: The actual 19F NMR peak
should appear around +60 ppm, but wraps around due to limited instrument range.)
General procedure for deoxyfluorination with PyFluor: A 1-dram vial is charged sequentially
with substrate (1.0 mmol), toluene (1 mL, 1.0 M), PyFluor (177 mg, 1.1 equiv), and DBU (300
μL, 2 equiv). The vial is closed with a phenolic cap. The mixture is stirred at 600 rpm at room
temperature. No precautions are taken to exclude air or moisture. Products are isolated by
automated column chromatography.
(3-fluorobutyl)benzene (1.3): Synthesized according to the general procedure with 4-phenyl-2-
butanol (150 mg, 1 mmol), toluene (1 mL, 1.0 M), PyFluor (177 mg, 1.1 equiv), and DBU
(300 μL, 2 equiv). The reaction was diluted with 20 mL water and extracted three times with
20 mL hexanes. The organic extracts were dried with sodium sulfate and concentrated under
reduced pressure. The residue was purified by automated column chromatography (25 g silica, 0
→ 10% ethyl acetate in hexanes) to afford (3-fluorobutyl)benzene as a colorless oil (120 mg,
79% yield).60 A duplicate run provided 119 mg in 78% yield. 1H NMR (500 MHz, CDCl3):
δ 7.33 – 7.27 (m, 2H), 7.23 – 7.18 (m, 3H), 4.67 (dm, J = 48.8 Hz, 1H), 2.86 – 2.65 (m, 2H),
2.07 – 1.74 (m, 2H), 1.35 (dd, J = 23.9, 6.2 Hz, 3H). 13C NMR (125 MHz, CDCl3): δ 141.63,
128.58, 128.57, 126.09, 90.20 (d, J = 164.9 Hz), 38.81 (d, J = 20.8 Hz), 31.52 (d, J = 4.8 Hz),
21.16 (d, J = 22.7 Hz). 19F NMR (282 MHz, CDCl3): δ −174.25 (ddqd, J = 48.0, 30.4, 23.9,
15.6 Hz).
60 Yin, J.; Zarkowsky, D. S.; Thomas, D. W.; Zhao, M. M.; Huffman, M. A. Org. Lett. 2004, 6, 1465. Giudecelli, M. B.; Picq, D.; Veyron, B. Tetrahedron Lett. 1990, 31, 6527.
28
1-(3-fluoropropyl)-4-methoxybenzene (1.4): Synthesized according to the general procedure
with 3-(4-methoxyphenyl)-1-propanol (83 mg, 0.5 mmol, Millipore-Sigma), toluene (1 mL,
0.5 M), PyFluor (89 mg, 1.1 equiv), and MTBD (144 μL, 2 equiv). The reaction mixture was
taken up in minimal dichloromethane and purified directly by automated column
chromatography (50 g silica, 0 → 15% ethyl acetate in hexanes) yielding product as a colorless
oil (68 mg, 81% yield). A replicate run afforded 64 mg (76% yield.) 1H NMR (500 MHz,
CDCl3): δ 7.14 (d, J = 8.6 Hz, 2H), 6.87 (d, J = 8.6 Hz, 2H), 4.47 (dt, J = 47.3, 6.0 Hz, 2H), 3.81
(s, 3H), 2.71 (dd, J = 8.5, 6.9 Hz, 2H), 2.00 (dm, J = 25.4 Hz, 2H). 13C NMR (125 MHz,
CDCl3): δ 158.02, 133.22, 129.49, 113.95, 83.22 (d, J = 164.5 Hz), 55.33, 32.38 (d, J = 19.6
Hz), 30.48 (d, J = 5.4 Hz). 19F NMR (282 MHz, CDCl3): δ − 220.05 (tt, J = 47.2, 25.1 Hz). IR
(film, cm−1): 2957 (m), 2837 (w), 1613 (m), 1584 (w), 1511 (s), 1465 (w), 1443 (w), 1390 (w),
1300 (w), 1243 (s), 1177 (m), 1113 (w), 1031 (s), 917 (w), 903 (w), 832 (w), 810 (m), 788 (w),
764 (w), 746 (w), 700 (w). HRMS (ESI+): Calculated for [C10H13FO − OCH3]+, 137.0761;
found, 137.0758.
2,3,4,6-tetra-O-benzyl-D-glucopyranosyl fluoride (1.5): Synthesized according to the general
procedure with 2,3,4,6-tetra-O-benzyl-D-glucopyranose (270 mg, 0.5 mmol, Millipore-Sigma),
toluene (1.0 mL, 0.5 M), PyFluor (89 mg, 1.1 equiv), and MTBD (90 μL, 1.25 equiv). The
reaction mixture was taken up in minimal dichloromethane and purified by automated column
chromatography (50 g silica, 0 → 20% ethyl acetate in hexane) to afford product as a colorless
29
oil (243 mg, 90% yield, 23:77 α:β as determined by 19F NMR).61 A replicate run afforded
248 mg (91% yield, 23:77 α:β). 1H NMR (500 MHz, CDCl3): [23:77 mixture of α:β anomer]:
δ 7.42 – 7.14 (m, 20H α and β), 5.58 (dd, J = 53.2, 2.6 Hz, 1H α), 5.28 (dd, J = 52.8, 6.8 Hz, 1H
β), 4.98 (d, J = 10.9 Hz, 1H α), 4.92 (d, J = 11.0 Hz, 1H β), 4.88 (d, J = 11.1 Hz, 1H α and β),
4.86 (d, J = 10.7 Hz, 1H α), 4.83 (d, J = 10.3 Hz, 1H α and β), 4.81 (d, J = 10.6 Hz, 1H β), 4.73
(d, J = 11.2 Hz, 1H α and β), 4.65 (d, J = 12.3 Hz, 1H β), 4.62 (d, J = 13.2 Hz, 1H α), 4.57 (d,
J = 12.1 Hz, 1H β), 4.56 (d, J = 10.9 Hz, 1H β). 4.53 (d, J = 12.5 Hz, 1H α), 4.50 (d, J = 12.1 Hz,
1H α), 4.01 (t, J = 9.4 Hz, 1H α), 3.97 (dt, J = 10.2, 2.7 Hz, 1H α), 3.80 – 3.77 (m, 1H α), 3.77 –
3.72 (m, 3H β, 1H α), 3.72 – 3.66 (m, 1H β, 1H α), 3.65 – 3.55 (m, 2H β, 1H α). 13C NMR (125
MHz, CDCl3): [23:77 mixture of α:β anomer, α denoted by *]: δ 138.56*, 138.36, 138.09*,
137.95, 137.94, 137.80, 137.77*, 137.77*, 128.68*, 128.57, 128.56*, 128.54, 128.54, 128.54,
128.54*, 128.53*, 128.29, 128.19*, 128.15*, 128.07, 128.07*, 128.07*, 128.06, 128.02, 128.02,
127.98, 127.98*, 127.96, 127.93*, 127.92*, 127.86, 127.86*, 109.98 (d, J = 215.9 Hz), 105.68*
(d, J = 226.8 Hz), 83.56 (d, J = 11.3 Hz), 81.59 (d, J = 21.6 Hz), 81.56*, 79.38* (d, J = 24.8 Hz),
77.00, 76.72*, 75.96*, 75.60, 75.30*, 75.13, 74.91 (d, J = 5.0 Hz), 74.58 (d, J = 2.3 Hz), 73.70,
73.67*, 73.63*, 72.77* (d, J = 4.0 Hz), 68.44 , 67.88*. 19F NMR (376 MHz, CDCl3): [23:77
mixture of α:β anomer, α denoted by *]: δ −138.43 (dd, J = 52.8, 11.7 Hz), −149.92 (dd,
J = 53.2, 25.6 Hz)*.
61 Huang, K.-T.; Winssinger, N. Eur. J. Org. Chem. 2007, 1887. Caddick, S.; Gazzard, L.; Motherwell, W. B.; Wilkinson, J. A. Tetrahedron 1996, 52, 149.
30
5-deoxy-5-fluoro-2,3-O-isopropylidene-D-ribono-1,4-lactone (1.6): Synthesized according to
the general procedure with 2,3-O-isopropylidene-D-ribono-1,4-lactone (18.8 mg, 0.1 mmol,
Millipore-Sigma), toluene (250 μL, 0.4 M), PyFluor (17.7 mg, 1.1 equiv) and MTBD (29 μL,
2 equiv) in a 8 × 40 mm vial. The reaction mixture was taken up in minimal dichloromethane
and loaded on to a 5 g Florisil plug. This was rinsed with 50 mL hexanes (discarded) followed by
50 mL of 50% ethyl acetate in hexanes, which was concentrated to afford the title compound as
colorless needle-like crystals (17.3 mg, 91% yield). 62 A second experiment afforded 17.4 mg
(91% yield). 1H NMR (500 MHz, CDCl3): δ 4.83 – 4.58 (m, 5H), 1.49 (s, 3H), 1.40 (s, 3H). 13C
NMR (125 MHz, CDCl3): δ 173.50, 113.92, 82.58 (d, J = 172.1 Hz), 80.41 (d, J = 18.1 Hz),
77.37 (d, J = 4.9 Hz), 75.28 (d, J = 3.7 Hz), 26.90, 25.75. 19F NMR (282 MHz, CDCl3): δ
−235.53 (tdd, J = 46.5, 34.6, 2.9 Hz).
3α-fluoro-5α-androstan-17-one (1.7): Synthesized according to the general procedure from
3β-hydroxy-5α-androstan-17-one (290 mg, 1 mmol, Millipore-Sigma), toluene (1 mL, 1.0 M),
PyFluor (177 mg, 1.1 equiv), and DBU (300 μL, 2 equiv) with a reaction temperature of 58 °C.
The reaction mixture was taken up in dichloromethane and filtered through a short silica plug,
eluting with 25% ethyl acetate in hexanes. The filtrate was concentrated under reduced pressure
to remove toluene, and the residue was subjected to automated column chromatography (25 g
silica, 0 → 15% ethyl acetate in hexanes, 3.5 L gradient volume). The elimination side products
(57 mg) eluted first and were followed by 32 mg of a mixture of elimination side product and α
diastereomer (1.00:2.92 elimination:α diastereomer), 144 mg of fully resolved α diastereomer,
62 Nasomjai, P.; O’Hagan, D.; Slawin, A. M. Z. Beilstein J. Org. Chem. 2009, 5, 37.
31
29 mg of a mixture of α and β diastereomers (1.79:1.00 α:β), and 15 mg of fully resolved
β diastereomer, all as white crystalline solids (73% yield, 7.4:1 dr, 24% elimination). 63 A
duplicate reaction afforded a combined 275 mg products (74% yield, 8.1:1 dr, 22% elimination
as determined by 1H NMR). 1H NMR (500 MHz, CDCl3): δ 4.77 (dt, J = 48.8, 2.7 Hz, 1H),
2.40 (dd, J = 19.2, 8.8 Hz, 1H), 2.03 (dt, J = 18.8, 9.1 Hz, 1H), 1.95 – 1.81 (m, 2H), 1.80 – 1.73
(m, 2H), 1.69 – 1.34 (m, 8H), 1.33 – 1.13 (m, 6H), 0.99 (qd, J = 12.6, 4.4 Hz, 1H), 0.83 (s, 3H),
0.81 – 0.73 (m, 1H), 0.78 (s, 3H). 13C NMR (125 MHz, CDCl3): δ 221.30, 89.33 (d, J = 165.9
Hz), 54.26, 51.48, 47.83, 39.44, 35.93, 35.90, 35.06, 33.91 (d, J = 21.2 Hz), 32.43, 31.59, 30.82,
28.10, 27.09 (d, J = 21.8 Hz), 21.81, 20.11, 13.88, 11.22 (d, J = 1.4 Hz). 19F NMR (282 MHz,
CDCl3): δ −181.12 (m).
N-Boc-cis-4-fluoro-L-proline methyl ester (1.8): Synthesized according to the general
procedure from N-Boc-trans-4-hydroxy-L-proline methyl ester (245 mg, 1 mmol, Bachem),
toluene (1 mL, 1.0 M), PyFluor (177 mg, 1.1 equiv), and DBU (300 μL, 2 equiv) with a reaction
temperature of 50 °C. The reaction was taken up in minimal dichloromethane, loaded directly on
a 25 g automated silica column, and subjected to a gradient of 5 → 35% ethyl acetate in hexanes,
affording the title compound as a colorless solid (178 mg, 72% yield).63 A second run provided
183 mg (74% yield). 1H NMR (500 MHz, CDCl3): [mixture of two rotamers] δ 5.18 (d,
J = 52.9 Hz, 1H, major and minor), 4.52 (d, J = 9.7 Hz, 1H, minor), 4.41 (d, J = 9.5 Hz, 1H,
major), 3.89 – 3.54 (m, 2H, major and minor), 3.72 (s, 3H, major and minor), 2.53 – 2.23 (m,
2H, major and minor), 1.46 (s, 9H, minor), 1.41 (s, 9H, major). 13C NMR (125 MHz, CDCl3):
63 Sladojevich, P.; Arlow, S. I.; Tang, P.; Ritter, T. J. Am. Chem. Soc. 2013, 135, 2470.
32
[mixture of two rotamers, minor rotamer denoted by *] δ 172.39, 171.99*, 154.12*, 153.74,
92.32* (d, J = 177.2 Hz), 91.25 (d, J = 177.5 Hz), 80.53*, 80.49, 57.74, 57.35*, 53.29* (d, J =
24.4 Hz), 52.96 (d, J = 24.3 Hz), 52.49*, 52.35, 37.58 (d, J = 22.0 Hz), 36.71* (d, J = 21.9 Hz),
28.49*, 28.38. 19F NMR (282 MHz, CDCl3): δ −173.13 (m).
3-(3-fluoropropyl)pyridine (1.9): Synthesized according to the general procedure with
3-pyridinepropanol (137 mg, 1 mmol, Millipore-Sigma), toluene (1 mL, 1.0 M), PyFluor
(177 mg, 1.1 equiv), MTBD (290 μL, 2 equiv). The reaction mixture was taken up in minimal
dichloromethane and purified directly by automated column chromatography (50 g silica, 20 →
35% ethyl acetate in hexanes with 3% triethylamine) to afford product as a pale yellow oil
(93 mg, 67% yield). A replicate run afforded 95 mg (68% yield). 1H NMR (500 MHz, CDCl3):
δ 8.52 – 8.43 (m, 2H), 7.52 (d, J = 7.6 Hz, 1H), 7.22 (t, J = 6.4 Hz, 1H), 4.46 (dt, J = 47.1, 5.8
Hz, 2H), 2.76 (t, J = 7.7 Hz, 2H), 2.01 (dp, J = 26.5, 6.4 Hz, 2H). 13C NMR (125 MHz, CDCl3):
δ 150.08, 147.77, 136.46, 136.04, 123.51, 82.82 (d, J = 165.5 Hz), 31.83 (d, J = 19.9 Hz), 28.70
(d, J = 5.2 Hz). 19F NMR (282 MHz, CDCl3): δ −220.52 (tt, J = 47.2, 25.6 Hz). IR (film, cm−1):
3397 (br w), 2965 (m), 1576 (m), 1479 (m), 1452 (w), 1423 (m), 1391 (w), 1194 (w), 1128 (w),
1108 (w), 1062 (w), 1022 (s), 951 (w), 902 (m), 828 (w), 792 (m), 756 (w), 713 (s). HRMS
(ESI+): Calculated for [C8H10FN + H]+, 140.0870; found, 140.0875.
N-(3-fluoropropyl)phthalimide (1.10): Synthesized according to the general procedure with
N-(3-hydroxypropyl)phthalimide (205 mg, 1 mmol, Lancaster), toluene (1 mL, 1.0 M), PyFluor
(177 mg, 1.1 equiv), MTBD (290 μL, 2 equiv). The reaction mixture was taken up in minimal
33
dichloromethane and purified directly by automated column chromatography (50 g silica, 0 →
20% ethyl acetate in hexanes with 1% triethylamine) to afford the title compound as a fluffy
white solid (175 mg, 84% yield).64 A second experiment provided 171 mg (83% yield). 1H NMR
(500 MHz, CDCl3): δ 7.85 (dd, J = 5.4, 3.1 Hz, 2H), 7.72 (dd, J = 5.5, 3.1 Hz, 2H), 4.52 (dt,
J = 47.0, 5.7 Hz, 2H), 3.85 (t, J = 6.9 Hz, 2H), 2.09 (dp, J = 25.5, 6.4 Hz, 2H). 13C NMR
(125 MHz, CDCl3): δ 168.28, 134.01, 132.07, 123.29, 81.68 (d, J = 165.9 Hz), 34.63 (d, J = 5.3
Hz), 29.49 (d, J = 19.9 Hz). 19F NMR (282 MHz, CDCl3): δ −220.83 (tt, J = 47.0, 26.2 Hz).
N-Boc-3-fluoropropylamine (1.11): Synthesized according to the general procedure with
tert-butyl-N-(3-hydroxypropyl)carbamate (88 mg, 0.5 mmol, Millipore-Sigma), toluene (1 mL,
0.5 M), PyFluor (89 mg, 1.1 equiv), and MTBD (144 μL, 2 equiv). The reaction mixture was
taken up in minimal dichloromethane and purified by automated column chromatography (50 g
Florisil, 0 → 40% ethyl acetate in hexanes with 1% triethylamine) to afford the title compound
as a colorless oil (67 mg, 76% yield). 65 A second experiment afforded 62 mg (70% yield).
1H NMR (500 MHz, CDCl3): δ 4.80 (s, 1H), 4.48 (dt, J = 47.2, 5.7 Hz, 2H), 3.23 (m, 2H), 1.85
(dp, J = 27.5, 6.2 Hz, 2H), 1.40 (s, 9H). 13C NMR (125 MHz, CDCl3): δ 156.08, 82.26 (d, J =
164.1 Hz), 79.30, 37.29 (d, J = 3.7 Hz), 30.82 (d, J = 19.4 Hz), 28.45. 19F NMR (282 MHz,
CDCl3): δ −220.88 (tt, J = 47.4, 27.1 Hz).
64 Liu, Y.; Chen, C.; Li, H.; Huang, K.-W.; Tan, J.; Weng, Z. Organometallics 2013, 32, 6587. 65 Howbert, J. J.; Dietsch, G.; Hershberg, R.; Burgess, L. E.; Doherty, G. A.; Eary, C. T.; Groneberg, Z. J. World Patent WO2011022509 A2, Feb. 24, 2011.
34
tert-butyl (E)-(4-(4-(2,3-difluoropropoxy)styryl)phenyl)(methyl)carbamate (1.12):
Synthesized according to the general procedure from (E)-tert-butyl (4-(4-(3-fluoro-2-hydroxy-
propoxy)-styryl)phenyl)(methyl)carbamate (201 mg, 0.5 mmol), toluene (1 mL, 0.5 M), PyFluor
(89 mg, 1.1 equiv), and DBU (300 μL, 4 equiv) with a reaction temperature of 50 °C. The
reaction was taken up in minimal dichloromethane, loaded directly on a 50 g automated silica
column, and subjected to a gradient of 0 → 30% ethyl acetate in hexanes to afford the title
compound as a white solid (132 mg, 65% yield). A second run provided 128 mg (64% yield).
1H NMR (500 MHz, CDCl3): δ 7.47 – 7.42 (m, 4H), 7.24 – 7.19 (m, 2H), 7.02 (d, J = 16.3 Hz,
1H), 6.96 (d, J = 16.4 Hz, 1H), 6.93 – 6.90 (m, 2H), 5.02 (ddm, J = 47.3, 21.5 Hz, 1H), 4.75
(ddm, J = 47.5, 23.6 Hz, 2H), 4.25 (dd, J = 18.4, 5.1 Hz, 2H), 3.27 (s, 3H), 1.46 (s, 9H).
13C NMR (125 MHz, CDCl3): δ 157.80, 154.83, 143.05, 134.64, 131.19, 127.88, 127.81,
126.60, 126.51, 125.62, 114.91, 89.49 (dd, J = 176.7, 20.1 Hz), 81.90 (dd, J = 173.1, 23.3 Hz),
80.54, 65.85 (dd, J = 26.0, 8.0 Hz), 37.36, 28.50. 19F NMR (282 MHz, CDCl3): δ −196.96
(dddtd, J = 49.8, 23.9, 23.1, 18.3, 13.2 Hz), −234.32 (tdd, J = 47.2, 21.5, 13.1 Hz). IR (ATR,
cm−1): 2979 (w), 2933 (w), 1691 (s), 1606 (m), 1577 (s), 1515 (m), 1477 (w), 1457 (w), 1435
(w), 1393 (w), 1349 (s), 1309 (w), 1297 (w), 1247 (s), 1150 (s), 1104 (s), 1067 (w), 1037 (m),
968 (m), 924 (w), 856 (w), 835 (s), 806 (w), 791 (w), 769 (w), 744 (w), 713 (w), 675 (w), 658
(w). HRMS (ESI+): Calculated for [C23H27F2NO3 − C4H9 + 2H]+, 348.1406; found, 348.1405.
(1S,2R)-1-fluoro-N,N-dimethyl-1-phenylpropan-2-amine (1.13): Synthesized according to the
general procedure with (1S,2R)-N-methylephedrine (90 mg, 0.5 mmol, Millipore-Sigma), toluene
(1 mL, 0.5 M), PyFluor (89 mg, 1.1 equiv), and MTBD (144 μL, 2 equiv). The reaction mixture
was taken up with dichloromethane, concentrated under reduced pressure, and purified by
35
automated column chromatography (25 g silica, 0 → 5% methanol in dichloromethane with
1% triethylamine) to afford the free base of the title compound (with retention of
stereochemistry) as a pale yellow oil (58 mg, 64% yield). 66 A replicate experiment provided 65
mg (72% yield). 1H NMR (500 MHz, CD2Cl2): δ 7.40 – 7.34 (m, 2H), 7.34 – 7.28 (m, 3H), 5.58
(dd, J = 48.0, 4.3 Hz, 1H), 2.89 (dqd, J = 22.9, 6.8, 4.4 Hz, 1H), 2.31 (s, 6H), 1.02 (dd, J = 6.8,
1.8 Hz, 3H). 13C NMR (125 MHz, CD2Cl2): δ 140.20 (d, J = 20.2 Hz), 128.55, 128.14 (d, J =
1.5 Hz), 125.87 (d, J = 8.1 Hz), 95.30 (d, J = 176.0 Hz), 64.28 (d, J = 22.8 Hz), 41.52 (d, J = 1.8
Hz), 7.39 (d, J = 5.7 Hz). 19F NMR (282 MHz, CD2Cl2): δ −190.57 (dd, J = 48.2, 22.9 Hz).
(E)-N-ethyl-N-(2-fluoroethyl)-4-(4-nitrostyryl)aniline (1.14): Synthesized according to the
general procedure with the 2-(ethyl(4-(2-(4-nitrophenyl)ethenyl)-phenyl)amino)ethanol (78 mg,
0.25 mmol, Millipore-Sigma), toluene (625 μL, 0.4 M), PyFluor (44 mg, 1.1 equiv), and MTBD
(72 μL, 2 equiv). The reaction mixture was taken up in minimal dichloromethane and purified
directly by automated column chromatography (50 g silica, 0 → 40% ethyl acetate in hexanes
with 3% triethylamine) to afford the title compound as a dark orange crystalline solid (58 mg,
74% yield). A second experiment afforded 61 mg (78% yield). 1H NMR (500 MHz, CDCl3): δ
8.17 (d, J = 8.8 Hz, 2H), 7.55 (d, J = 8.8 Hz, 2H), 7.42 (d, J = 8.8 Hz, 2H), 7.19 (d, J = 16.3 Hz,
1H), 6.91 (d, J = 16.2 Hz, 1H), 6.69 (d, J = 8.9 Hz, 2H), 4.61 (dt, J = 47.1, 5.4 Hz, 2H), 3.68 (dt,
J = 23.1, 5.4 Hz, 2H), 3.49 (q, J = 7.1 Hz, 2H), 1.22 (t, J = 7.1 Hz, 3H). 13C NMR (125 MHz,
CDCl3): δ 148.16, 145.94, 145.04, 133.54, 128.77, 126.17, 124.39, 124.24, 121.68, 111.91,
81.69 (d, J = 170.5 Hz), 50.43 (d, J = 21.9 Hz), 45.76, 12.22. 19F NMR (282 MHz, CDCl3): δ
66 Hamman, S.; Beguin, C. G.; Charlon, C.; Luu-Duc, C. J. Fluorine Chem. 1987, 37, 343.
36
−221.52 (tt, J = 46.8, 23.0 Hz). IR (ATR, cm−1): 2924 (m), 1605 (m), 1581 (s), 1512 (s), 1452
(w), 1433 (w), 1398 (m), 1377 (w), 1360 (w), 1337 (s), 1273 (w), 1242 (w), 1184 (m), 1138 (w),
1109 (w), 1080 (w), 1037 (w), 1005 (w), 956 (m), 935 (w), 923 (w), 908 (w), 864 (m), 855 (w),
833 (s), 813 (s), 790 (w), 748 (m), 719 (w), 708 (w), 684 (m). HRMS (ESI+): Calculated for
[C18H19FN2O2 + H]+, 315.1503; found, 315.1509.
2-benzyl-4-fluoro-1-phenylpentan-2-ol (1.15): Synthesized according to the general procedure
with 2-benzyl-1-phenylpentane-2,4-diol (270 mg, 1 mmol), toluene (1 mL, 1.0 M), PyFluor
(177 mg, 1.1 equiv), and MTBD (290 μL, 2 equiv). The reaction was taken up with minimal
dichloromethane and directly purified by automated column chromatography (50 g silica, 0 →
10% ethyl acetate in hexanes) to afford product as a colorless oil (217 mg, 80% yield). A
duplicate experiment provided 217 mg (80% yield). 1H NMR (500 MHz, CDCl3): δ 7.36 – 7.20
(m, 10H), 5.09 (dm, J = 48.8 Hz, 1H), 2.93 – 2.82 (m, 4H), 1.90 – 1.81 (m, 1H), 1.83 (s, 1 H),
1.55 (ddd, J = 39.5, 15.3, 1.9 Hz, 1H), 1.30 (dd, J = 24.2, 6.2 Hz, 3H). 13C NMR (125 MHz,
CDCl3): δ 137.30, 137.19, 131.02, 130.97, 128.37, 128.30, 126.67, 126.62, 88.70 (d, J = 162.6
Hz), 73.57, 46.53 (d, J = 1.3 Hz), 46.07 (d, J = 1.9 Hz), 44.84 (d, J = 18.8 Hz), 22.39 (d, J = 23.1
Hz). 19F NMR (282 MHz, CDCl3): δ −169.75 (ddqd, J = 56.0, 39.4, 24.2, 16.7 Hz). IR (film,
cm−1): 3573 (br w), 3062 (w), 3029 (w), 2979 (w), 2929 (w), 1602 (w), 1494 (m), 1454 (w),
1380 (w), 1186 (w), 1133 (w), 1087 (m), 1031 (w), 1005 (w), 940 (w), 914 (w), 880 (w), 821
(w), 786 (w), 752 (m), 726 (m, 699 (s). HRMS (ESI+): Calculated for [C18H21FO + Na]+,
295.1469; found, 295.1481.
37
1,4-bis(2-fluoroethoxy)benzene (1.16): Prepared according to the general procedure from
hydroquinone bis(2-hydroxyethyl) ether (99 mg, 0.5 mmol, Millipore-Sigma), toluene (1.25 mL,
0.4 M), PyFluor (169 mg, 2.1 equiv) and MTBD (251 μL, 3.5 equiv). The reaction mixture was
taken up in minimal dichloromethane and purified by automated column chromatography (50 g
silica, 5 → 35% ethyl acetate in hexanes) to afford 73 mg of the title compound as a fluffy white
solid (72% yield).67 A replicate run afforded 72 mg product (71% yield). 1H NMR (500 MHz,
CDCl3): δ 6.87 (s, 4H), 4.72 (dm, J = 47.5 Hz, 4H), 4.15 (dm, J = 28.2 Hz, 4H). 13C NMR
(125 MHz, CDCl3): δ 153.09, 115.80, 82.15 (d, J = 170.3 Hz), 67.90 (d, J = 20.3 Hz). 19F NMR
(282 MHz, CDCl3): δ −223.80 (tt, J = 47.5, 28.1 Hz).
3-fluoropropylene carbonate (1.17): Synthesized according to the general procedure with
4-(hydroxymethyl)-1,3-dioxolan-2-one (59 mg, 0.5 mmol, Millipore-Sigma), toluene (1 mL,
0.5 M), PyFluor (89 mg, 1.1 equiv), and MTBD (144 μL, 2 equiv). The reaction mixture was
taken up in minimal dichloromethane and loaded on a 5 g Florisil plug. After rinsing with
100 mL hexanes, the plug was eluted with 100 mL 60% ethyl acetate in hexanes which was
concentrated to afford product as a colorless oil (42 mg, 70% yield). 68 A replicate run afforded
44 mg (73% yield). 1H NMR (500 MHz, CDCl3): δ 4.90 (dddt, J = 23.7, 8.9, 6.0, 2.8 Hz, 1H),
4.71 (ddd, J = 47.6, 11.1, 2.6 Hz, 1H), 4.61 – 4.41 (m, 3H). 13C NMR (125 MHz, CDCl3): δ
154.43, 81.16 (d, J = 177.3 Hz), 74.38 (d, J = 20.3 Hz), 64.93 (d, J = 6.9 Hz). 19F NMR
(282 MHz, CDCl3): δ −236.87 (tdd, J = 46.8, 23.5, 1.0 Hz).
67 Roush, D. M.; Shaw, D. A.; Jones, M. L.; Chang, J. H. U.S. Patent 4,960,884, Oct. 2, 1990. 68 Nakano, T.; Shiono, K. U.S. Patent 5,750,730, May 12, 1998.
38
benzyl 2-fluoroacetate (1.18): Synthesized following the general procedure with benzyl
glycolate (83 mg, 0.5 mmol, Millipore-Sigma), toluene (1.25 mL, 0.4 M), PyFluor (89 mg,
1.1 equiv) and MTBD (126 μL, 1.75 equiv). The reaction mixture was taken up in minimal
dichloromethane and subjected to automated column chromatography (50 g silica, 0 → 30%
ethyl acetate in hexanes) yielding 56 mg of the title compound as a colorless oil (67% yield). 69 A
replicate run afforded 50 mg product (60% yield). 1H NMR (500 MHz, CDCl3): δ 7.41 – 7.33
(m, 5H), 5.25 (s, 2H), 4.88 (d, J = 47.0 Hz, 2H). 13C NMR (125 MHz, CDCl3): δ 167.77 (d, J =
22.0 Hz), 134.92, 128.78, 128.77, 128.65, 77.65 (d, J = 182.5 Hz), 67.15. 19F NMR (376 MHz,
CDCl3): δ −230.33 (t, J = 47.0 Hz).
4-(2-fluoroethyl)-1,2-dimethoxybenzene (1.19): Synthesized according to the general
procedure with 2-(3,4-dimethoxyphenyl)ethanol (91 mg, 0.5 mmol, Millipore-Sigma), toluene
(1 mL, 0.5 M), PyFluor (89 mg, 1.1 equiv) and MTBD (144 μL, 2 equiv). The reaction mixture
was taken up in minimal dichloromethane and purified by automated column chromatography
(50 g silica, 0 → 20% ethyl acetate in hexanes). First, 7 mg of the elimination side product 1,2-
dimethoxy-4-vinylbenzene eluted followed by 77 mg of the fully resolved title compound both
as colorless oils (84% yield, 9% elimination).70 A second run afforded a combined 80 mg of
products (81% yield, 6% elimination as determined by 1H NMR). 1H NMR (500 MHz, CDCl3):
δ 6.82 (d, J = 8.0 Hz, 1H), 6.80 – 6.74 (m, 2H), 4.61 (dt, J = 47.2, 6.5 Hz, 2H), 3.88 (s, 3H), 3.86
(s, 3H), 2.96 (dt, J = 23.3, 6.6 Hz, 2H). 13C NMR (125 MHz, CDCl3): δ 149.01, 147.91, 129.75
69 Dang, H.; Mailig, M.; Lalic, G. Angew. Chem., Int. Ed. 2014, 53, 6473. 70 Khrimian, A. P.; DeMilo, A. B.; Waters, R. M.; Liquido, N. J.; Nicholson, J. M. J. Org. Chem. 1994, 59, 8034. Falk, A.; Göderz, A.-L.; Schmalz, H.-G. Angew. Chem., Int. Ed. 2013, 52, 1576.
39
(d, J = 6.4 Hz), 121.00, 112.28, 111.37, 84.42 (d, J = 168.9 Hz), 56.01, 55.94, 36.63 (d, J = 20.3
Hz). 19F NMR (376 MHz, CDCl3): δ −215.41 (tt, J = 46.9, 23.3 Hz).
5-(2-fluoroethyl)-4-methylthiazole (1.20): Prepared according to the general procedure with
4-methyl-5-thiazoleethanol (72 mg, 0.5 mmol, Alfa), toluene (1.66 mL, 0.3 M), PyFluor (89 mg,
1.1 equiv), and MTBD (144 μL, 2 equiv). The reaction mixture was taken up in minimal
dichloromethane and purified by automated column chromatography (50 g silica, 0 → 30% ethyl
acetate in hexanes) to afford 5 mg of the elimination side product 4-methyl-5-vinylthiazole
followed by 45 mg of the fully resolved title compound as a colorless oil (62% yield, 8%
elimination). 71 A replicate run afforded 7 mg elimination side product and 48 mg product (66%
yield, 11% elimination). 1H NMR (500 MHz, CDCl3): δ 8.59 (s, 1H), 4.56 (dt, J = 46.8, 6.2 Hz,
2H), 3.15 (dt, J = 23.5, 6.2 Hz, 2H), 2.40 (s, 3H). 13C NMR (125 MHz, CDCl3): δ 150.19,
150.08, 125.89 (d, J = 3.8 Hz), 83.16 (d, J = 170.8 Hz), 27.69 (d, J = 22.0 Hz), 15.00. 19F NMR
(376 MHz, CDCl3): δ −215.80 (tt, J = 47.0, 23.5 Hz).
(1-fluorodecyl)benzene (1.21): Synthesized according to the general procedure from (S)-1-
phenyl-1-decanol (234 mg, 1 mmol, Millipore-Sigma), toluene (5 mL, 0.2 M), PyFluor (177 mg,
1.1 equiv), and DBU (187 μL, 1.25 equiv). The reaction was taken up in minimal
dichloromethane, loaded directly on a 50 g automated silica column, and subjected to a gradient
of 0 → 10% ethyl acetate in hexanes to afford 168 mg of an unresolved mixture of product and
the elimination side product 1-phenyl-1-decene as a pale yellow oil (59% yield, 13% elimination
71 Lowe, G.; Potter, B. V. L. J. Chem. Soc., Perkin Trans. 1 1980, 2026. D’Auria, M.; Esposito, V.; Mauriello, G. Tetrahedron 1996, 52, 14253.
40
as determined by 1H NMR). A second run afforded 178 mg (63% yield, 13% elimination).
1H NMR (500 MHz, CDCl3): δ 7.41 – 7.18 (m, 5H), 5.43 (ddd, J = 47.9, 8.1, 4.9 Hz, 1H), 2.07
– 1.92 (m, 1H), 1.90 – 1.75 (m, 1H), 1.54 – 1.22 (m, 14H), 0.90 (t, J = 6.9 Hz, 3H). 13C NMR
(125 MHz, CDCl3): δ 140.78 (d, J = 19.8 Hz), 128.52, 128.26 (d, J = 1.9 Hz), 125.69 (d, J = 6.8
Hz), 94.87 (d, J = 170.1 Hz), 37.40 (d, J = 23.5 Hz), 32.03, 29.67, 29.64, 29.53, 29.45, 25.27 (d,
J = 4.3 Hz), 22.83, 14.27. 19F NMR (282 MHz, CDCl3): δ −174.59 (ddd, J = 46.8, 28.7, 16.8
Hz). IR (film, cm−1): 3031 (w), 2924 (m), 2854 (m), 1496 (w), 1455 (m), 1378 (w), 1309 (w),
1211 (w), 1029 (w), 964 (w), 912 (w), 755 (m), 697 (s). HRMS (ESI+): Calculated for [C16H25F
− F]+, 217.1951; found, 217.1962.
4-nitrobenzyl fluoride (1.22): Synthesized according to the general procedure with
4-nitrobenzylalcohol (77 mg, 0.5 mmol, Millipore-Sigma), toluene (625 μL, 0.8 M), PyFluor
(89 mg, 1.1 equiv), and MTBD (101 μL, 1.4 equiv). The reaction mixture was taken up in
minimal dichloromethane and purified by automated column chromatography (50 g silica, 0 →
20% ethyl acetate in hexane) to afford product as a white crystalline solid (35 mg, 45% yield). 72
A second experiment afforded 32 mg (41% yield). 1H NMR (500 MHz, CDCl3): δ 8.26 (d,
J = 8.3 Hz, 2H), 7.53 (d, J = 8.3 Hz, 2H), 5.51 (d, J = 46.8 Hz, 2H). 13C NMR (125 MHz,
CDCl3): δ 148.04, 143.52 (d, J = 17.7 Hz), 127.15 (d, J = 7.1 Hz), 123.97, 83.00 (d, J = 170.8
Hz). 19F NMR (376 MHz, CDCl3): δ −216.12 (t, J = 46.8 Hz).
72 Blessley, G.; Holden, P.; Walker, M.; Brown, J. M.; Gouverneur, V. Org. Lett. 2012, 14, 2754.
41
methyl cinnamate (1.23): Synthesized according to the general procedure from methyl
3-hydroxy-3-phenylpropanoate 73 (180 mg, 1 mmol). The reaction mixture was diluted with
10 mL water and extracted three times with 10 mL dichloromethane. The organic extracts were
concentrated under reduced pressure and the residue purified by automated column
chromatography (25 g silica, 0 → 10% ethyl acetate in hexanes to afford the elimination product
methyl cinnamate as a white crystalline solid (155 mg, 96% elimination). 74 No fluorinated
product was detected. 1H NMR (500 MHz, CDCl3): δ 7.70 (d, J = 16.1 Hz, 1H), 7.56 – 7.49 (m,
2H), 7.42 – 7.35 (m, 3H), 6.45 (d, J = 15.8 Hz, 1H), 3.81 (s, 3H). 13C NMR (125 MHz, CDCl3):
δ 167.58, 145.02, 134.50, 130.44, 129.03, 128.21, 117.92, 51.87.
73 Padhi, S. K.; Chadha, A. Synlett 2003, 639. 74 De Sarkar, S.; Grimme, S.; Studer, A. J. Am. Chem. Soc. 2010, 132, 1190.
42
Chapter 2.
A Systematic Investigation of Sulfonyl Fluoride Reactivity
43
2.1 Limitations of PyFluor
During development, it became increasingly apparent that PyFluor did not represent the
optimal structure for sulfonyl fluoride deoxyfluorination across all substrate classes. PyFluor was
specifically optimized to maximize yield of secondary fluoride 2.1 (see Table 2.1). Despite
extensive efforts, we never identified conditions under which another sulfonyl fluoride would
reliably deliver substantially higher yields of 2.1; however, there are many alcohol classes for
which other sulfonyl fluoride structures prove superior to PyFluor.
Our first systematic attempt to understand sulfonyl fluoride reactivity is presented in
Table 2.1 in which fifteen sulfonyl fluorides were screened against five substrate classes. In the
case of fluoride 2.1—the product on which PyFluor was optimized—PyFluor was outperformed
by the electron-deficient derivative 5-chloro-2-pyridinesulfonyl fluoride (entry 4), which
consistently delivered 2.1 in ~2% higher yield than PyFluor. However, we chose not to develop
this reagent because we felt the small improvement in yield did not justify the approximate 270-
fold increase in reagent cost.75 Another reagent, pentafluorobenzenesulfonyl fluoride (ArFSF,
entry 2) afforded 85% yield of 2.1, a 7% improvement; however, replicate experiments
indicated significant variability with an average of 80% ±2% yield, which lies within error of
PyFluor performance. More significantly, primary fluoride 2.2 was obtained in 75% yield with
ArFSF compared to only 48% yield with PyFluor. For a time, we considered replacing PyFluor
with ArFSF in our initial communication; however, we quickly discovered that ArFSF did not
perform well with functionalized substrates. As shown in Table 2.2, ArFSF failed to improve
upon the yield of six key substrates from the initial PyFluor scope (2.6 – 2.11). Timepoint
75 As of December 21, 2017, PyFluor precursor 2-mercaptopyridine is available from Chem-Impex for $38.91 per mol. 5-Chloro-2-mercaptopyridine costs $10,614.97 per mol from Millipore-Sigma.
44
Table 2.1 Preliminary investigation of sulfonyl fluoride reactivity.
45
Table 2.2 Comparison of PyFluor and ArFSF reactivity.
experiments measuring the formation of 2.1 indicated that ArFSF was approximately 30 times
more reactive than PyFluor. Based on a second-order kinetic fit,76 the reaction half-life (t1/2)77 of
PyFluor was measured to be 4.5 hours while ArFSF exhibited a t1/2 of only 8.6 minutes. Upon
addition of base, ArFSF reactions underwent a rapid color change with precipitation of a red
viscous material that would occasionally prevent stirbar rotation, potentially leading the higher
variation observed among replicate reactions. Moreover, there have been many reports
documenting the propensity of perfluoroaryl rings to undergo SNAr substitution at the para-
position78 likely contributing to the poor substrate tolerance shown in Table 2.2. Ultimately,
ArFSF was abandoned when we discovered that PyFluor could fluorinate primary alcohols in
~80% yield by employing the slightly bulkier base MTBD.
76 Reaction order was determined by comparing the least-squares fit of time point yields to both a first- and second-order product formation curve. For both reagents, the fluorination of 2.1 exhibited clear second-order kinetics, consistent with a rate-limiting reaction between the sulfonate electrophile and a fluoride nucleophile. 77 Measured as time required to attain 50% of the maximum observed yield. 78 Dudutiené, V.; Zubriené, A.; Smirnov, A.; Gylyté, J.; Timm, D.; Manakova, E.; Gražulis, S.; Matulis, D. Bioorg. Med. Chem. 2013, 21, 2093.
46
Continued analysis of the screen in Table 2.1 indicates that PyFluor is clearly not optimal
for fluorinating benzylic alcohols, tertiary alcohols, or phenols. In the case of benzylic fluoride
2.3, PyFluor was again outperformed by as much as 20% yield by ArFSF and four other reagents.
Tertiary fluoride 2.4 was only generated in trace amounts with PyFluor; however, PBSF
(perfluorobutane-1-sulfonyl fluoride, entry 6) and α-toluenesulfonyl fluoride (entry 10) both
afforded >10% yield. Finally, only PBSF and 4-nitrobenzenesulfonyl fluoride (entry 3) afforded
any detectable yield of fluorinated nitrophenol 2.5 at elevated temperature. In the case of
4-nitrobenzenesulfonyl fluoride, the trace product may derive from desulfonylative fluorination
of the reagent itself.79 The 12% yield observed with PBSF is likely formed via SNAr substitution
of the sulfonate ester intermediate—there is no evidence to suggest a more exotic mechanism.
As another example of PyFluor’s limitations, cyclobutanol 2.12-OH was fluorinated in
only 2% yield with PyFluor at room temperature (Table 2.3). Raising the reaction temperature to
50 °C led to an improved 35% yield. We hypothesized that the ring-strain inherent in the ideally
trigonal planar SN2 transition state resulted in a higher energy transition state, and that this
energy could be lowered with a more reactive leaving group. Indeed, switching to the highly
Table 2.3 PyFluor vs. PBSF in the deoxyfluorination of cyclobutanols.
79 Roberts, D. W. Chem. Res. Toxicol. 1995, 8, 545.
47
reactive PBSF enabled us to obtain product 2.12 in 72% yield at room temperature.
Based on these results, we concluded that reaction outcome among various alcohol
classes depended heavily on the leaving group ability of the sulfonate intermediate. Additionally,
from our experience with PyFluor, it was evident that the steric environment of the basic
nitrogen was critical as well. Unencumbered bases risked behaving as competing nucleophiles,
whereas bulkier bases were less efficient at delivering fluoride to alcohols with sterically
congested backside trajectories. We proposed a systematic screening approach in which alcohols
would be simultaneously screened against an array of sulfonyl fluorides and bases.
2.2 Development of High-Throughput Screening Approach
In the summer of 2016, visiting undergraduate Orestes Riera was the first to undertake
this effort. With an eclectic assortment of commercially available sulfonyl fluorides and bases,
Orestes performed two-dimensional screens and generated heat maps such as that shown in
Table 2.4 to visualize yield trends. Notably, the heat map for 2.13-OH revealed a complex
reaction landscape wherein yields varied considerably with sulfonate leaving group ability and
base steric effects, the optimal yield lying in between either set of extremes.
These initial 78-reaction screens provided valuable strategic insight. First, it was clear
that we would be unable to economically screen a diverse substrate scope in any reasonable
time-frame with this many reagents. Fortunately, small changes in sulfonyl fluoride electronics
did not appear to result in a dramatic effect on reactivity (eg. compare the reactivity of
2-CNPhSF, 3-CNPhSF, and 4-CNPhSF). As such, we could reduce the number of screened
sulfonyl fluorides as long they were representative of a broad range of leaving group ability.
After considering reactivity and reagent cost, we selected five sulfonyl fluorides for the next
screening phase (Figure 2.1). PyFluor ($1,991 per mol from Millipore-Sigma) was included to
48
Table 2.4 Initial sulfonyl fluoride vs. base screen for unactivated primary alcohols.
Figure 2.1 Sulfonyl fluorides selected for high-throughput screening.
serve as a benchmark from our previous studies. PBSF ($155 per mol from Oakwood) was
selected primarily due to its high electron-withdrawing character, but also to enable comparison
to the body of literature detailing its development. Additionally, we chose to include
4-chlorobenzenesulfonyl fluoride (4-ClPhSF), 4-(trifluoromethyl)benzenesulfonyl fluoride
(4-CF3PhSF), and 4-nitrobenzene-sulfonyl fluoride (4-NsF). These arylsulfonyl fluorides can be
49
economically prepared by simple halide exchange with the corresponding sulfonyl chloride.80
The Hammett σpara parameters for the chloro-, trifluoromethyl-, and nitro- substituents are 0.23,
0.54, and 0.78, respectively, indicating a relatively even distribution across the electronic
spectrum.81 DSC data indicate that all three reagents are thermally stable up to 300 °C.
In order to measure the relative leaving group ability of the selected sulfonyl fluorides,
we conducted kinetic studies for the fluorination of 2.13-OH. Relative rate constants and t1/2
values were determined by taking product formation timepoints and fitting the curves to second
order kinetics (Figure 2.2). This model assumes rapid formation of the sulfonate ester followed
by rate-limiting SN2 fluorination; however, in the case of 4-ClPhSF and benzenesulfonyl
fluoride, it was necessary to add an induction period accounting for the slower sulfonate
formation. As expected, the arylsulfonyl fluorides cover a reasonably broad range of reactivity;
however, PBSF proved to be a staggering 3,000 times more reactive than 4-NsF, the most
electron-deficient arylsulfonyl fluoride. Reactions that required hours or days to reach
completion with arylsulfonyl fluorides were complete within seconds using PBSF.
Unfortunately, spanning this gap in reactivity proved non-trivial. More electron-deficient
arylsulfonyl fluorides were found to decompose rapidly through SNAr pathways (ie. Table 2.1,
entry 15), whereas partially fluorinated alkylsulfonyl fluorides that might be less reactive than
PBSF are not commercially available and required hazardous electrophilic fluorine sources to
synthesize.
80 Dong, J.; Krasnova, L.; Finn, M. G.; Sharpless, B. K. Angew. Chem., Int. Ed. 2014, 53, 9430. As of Dec. 22, 2017, 4-chlorobenzenesulfonyl chloride costs $50.61 per mol (Acros), 4-(trifluoromethyl)benzenesulfonyl chloride costs $151.66 per mol (Oakwood), and 4-nitrobenzenesulfonyl fluoride costs $81.30 per mol (Chem-Impex). The sulfonyl fluoride is generated by stirring the sulfonyl chloride in a mixture of acetonitrile and saturated potassium bifluoride at room temperature for 1 – 3 hours (see experimental section). 81 Hansch, C.; Leo, A. Substituent Constants for Correlation Analysis in Chemistry and Biology Wiley-Interscience: NY, 1979.
50
Figure 2.2 Kinetic study of sulfonyl fluorides.
Based on the rate data for PhSF, 4-ClPhSF, 4-CF3PhSF, and 4-NsF, a linear Hammett
correlation was observed (Figure 2.3). The positive ρ value of +2.27 indicates that
deoxyfluorination is accelerated by increasing sulfonate electron-deficiency and that the rate is
somewhat more sensitive to substituent effects than the equilibria of substituted benzoic acid
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0 20 40 60 80 100
Reaction Time (h)
Yield
PBSF 4-NsF 4-CF3Ph PyFluor 4-ClPhS PhSF PBSF 4-NsF 4-CF3Ph PyFluor 4-ClPhS PhSF
51
dissociations. Interestingly, if PBSF and PyFluor were fitted to the plot based on the value of
log(k/kH), PBSF would have a σpara value of 2.3, higher than that of a diazonium group; where as
the σpara value of PyFluor would be only 0.35, falling between 4-ClPhSF and 4-CF3PhSF in
reactivity. This latter result was unexpected; we had previously rationalized the success of
PyFluor based on the assumption that the pyridine structure was highly electron-withdrawing.
The current study instead indicated that PyFluor was comparatively unreactive. Again, PyFluor
was optimized for yield of unactivated secondary fluoride 2.1 which is prone to competing
elimination. In hindsight, the best way to minimize elimination was to identify the least reactive
sulfonyl fluoride that would allow the reaction to proceed to completion within a reasonable
timeframe, and PyFluor just happened to fulfill this criterion at room temperature.
Figure 2.3 Hammett plot for deoxyfluorination with arylsulfonyl fluorides.
In addition to the specified sulfonyl fluorides, we selected four so-called “superbases”—
DBU ($15.34 per mol, Chem-Impex), MTBD ($306.44 per mol, Digital), BTMG (a.k.a. Barton’s
base, $2,197 per mol, FluoroChem), and BTPP ($6,823 per mol, Millipore-Sigma) (Figure 2.4),
arranged in order of increasing sterics from left to right. Unfortunately, BTMG and BTPP cost
52
Figure 2.4 Bases selected for high throughput screening.
many times more than the sulfonyl fluoride reagents, calling into question our claim that sulfonyl
fluoride deoxyfluorinations are inexpensive. As always, it should be acknowledged that quoted
prices may reflect low demand rather than intrinsic manufacturing cost; for example, the
synthesis of BTMG is straightforward from inexpensive reagents. 82 Notwithstanding, when
comparing the cost of the various deoxyfluorination approaches, the sulfonyl fluoride and base
should be considered in combination as co-reagents.
Additionally, the pKa of each base’s conjugate acid has a subtle effect on reactivity.83
Following sulfonate ester formation, the protonated base serves as the counterion to fluoride.
Bases with higher pKa will act as softer cations, resulting in a more soluble, reactive fluoride
source. For example, in the deoxyfluorination of 2.13-OH with 4-CF3PhSF, the relative reaction
rate with BTPP is 1.9 times higher than with BTMG owing to the 4.8 unit difference in pKa in
acetonitrile. Among the selected bases, pKa correlates roughly with increasing electron
delocalization as illustrated by the resonance hybrids in Figure 2.5.
Figure 2.5 Conjugate acid pKa trends.
82 Barton, D. H. R.; Eliott, J. D.; Géro, S. D. J. Chem. Soc. Perkin Trans. I 1982, 2085. 83 Literature pKa values: (a) Ishikawa, T. Superbases for Organic Synthesis; John Wiley & Sons, Ltd.: Chichester, 2009. (b) Bandar, J. S.; Lambert, T. H. J. Am. Chem. Soc. 2012, 134, 5552.
53
We devised a screening protocol in which each of five sulfonyl fluorides would be
screened against the four bases, resulting in a total of twenty reactions per substrate. In an effort
to balance reproducibility with limitations on substrate availability, we chose to screen on
0.1 mmol scale in 250 μL solvent (0.4 M concentration). Although toluene was found to
optimize formation of 2.1 with PyFluor, we selected the more polar THF to expand the scope of
soluble substrates. Reactions were carried out in 8 × 40 mm glass vials with a 5 mm Teflon
stirbars—the same dimensions typically employed in 96-well plates—enabling us to run several
hundred reactions simultaneously. The sulfonyl fluoride loading was left at 1.1 equivalents, and
the base loading was decreased to 1.5 equivalents, which we considered more favorable in light
of the cost of BTMG and BTPP. Reaction time was arbitrarily set to 48 hours on the justification
that longer reaction times would be impractical. Although our work with PyFluor indicated that
heating could accelerate relunctant deoxyfluorinations, we chose to screen at room temperature
reasoning that with the wide range of sulfonyl fluoride reactivity, it was likely that at least one
reagent would provide reasonable yield at room temperature for most substrates. Yields were
determined by 19F NMR (with 1-fluoronaphthalene as an internal standard) on an instrument
equipped with an autosampler, thus allowing higher throughput analysis. Although sulfonyl
fluorides are fairly stable at room temperature, the four arylsulfonyl fluorides were stored sealed
at 2 °C as a precautionary measure. PBSF and the four bases were stored sealed on the benchtop
at room temperature. 84 All manipulations were performed on the benchtop with no special
precautions to exclude air or moisture.
84 The bases show visible deterioration from prolonged air exposure and should be replaced with fresh material every 3 – 6 months. Decomposition is accompanied by an increase in viscosity and a color change from colorless to yellow. BTPP specifically will form orange precipitates and begin to smell of free pyrrolidine.
54
2.3 Multi-dimensional Screening of Alcohol Substrates
In order to identify general reactivity trends, we chose to systematically screen common
alcohol classes. First, we examined unactivated primary alcohols 2.13-OH – 2.16-OH. For
interpretability, all yield data is displayed in heat maps with sulfonyl fluoride reactivity
increasing from left to right, and base steric encumbrance increasing from top to bottom. As
shown in Figure 2.6, both 2.13 and 2.14 were formed in highest yield with the non-nucleophilic
bases BTMG and BTPP and the moderately electron-deficient 4-CF3PhSF. Employing less-
encumbered bases or more reactive sulfonyl fluorides led to diminished yield, most likely due to
Figure 2.6 Deoxyfluorination of primary unactivated alcohols.
55
competing nucleophilic substitution by the base. Adenosine derivative 2.15-OH exhibited a
distinct profile in which reactivity improved with both increasing base sterics and sulfonyl
fluoride reactivity. In this case, we suspect that 2.15-OH is similarly susceptible to base
substitution with DBU or MTBD. However, the 2′ and 3′ benzoyl groups partially block the
sulfonate backside attack trajectory, thus reducing the rate of fluorination by the bulkier BTMG-
and BTPP-HF adducts and necessitating a more reactive leaving group. By reducing the loading
of BTPP to 1.25 equivalents, 2.15 was obtained in an improved 74% yield with 4-NsF.
In the case of the α-amino alcohol 2.16-OH, we suspected that aniline nitrogen was engaging
in neighboring group participation. Displacement of the sulfonate ester by nitrogen would form
an aziridinium ion intermediate that could subsequently undergo fluoride ring-opening. 85
Formation of 2.16 exhibited only weak dependence on sulfonate identity, consistent with an
aziridinium ion serving as the actual fluorination electrophile. To test this hypothesis, we
constructed α-amino alcohol 2.17-OH for which nitrogen attack on the sulfonate ester would
generate the unsymmetrical aziridinium ion shown in Figure 2.7. Consistent with an aziridinium
intermediate, we observed formation of both 2.17 as well as the rearranged product 2.17′, a ring-
expanded azepane that would form from nucleophilic attack at the tertiary azirdinium carbon.86
Across all conditions, rearranged 2.17′ was favored by ~2:1 over 2.17 owing to the higher partial
positive charge on the more substituted aziridinium carbon.
Screening of unactivated secondary alcohols 2.1-OH and 2.18-OH revealed a more
complex reaction landscape (Figure 2.8). In general, moderately electron-deficient sulfonyl
fluorides performed best due to competing elimination, and the more compact DBU and MTBD
were favored, consistent with the more congested backside trajectory of a secondary alcohol.
85 Hamman, S.; Beguin, C. G.; Charlon, C.; Luu-Duc, C. J. Fluorine Chem. 1987, 37, 343 86 Déchamps, I.; Pardo, D. G.; Cossy, J. Eur. J. Org. Chem. 2007, 4224.
56
Figure 2.7 Deoxyfluorination of amino alcohols via aziridinium intermediates.
Figure 2.8 Deoxyfluorination of secondary unactivated alcohols.
Notwithstanding, these screens failed to identify superior conditions to those originally reported
with PyFluor (with 2 equiv DBU in toluene) with which 2.1 was obtained in 79% yield and 2.18
in 85% yield.
Timepoint studies similar to those shown in Figure 2.2 indicated second order kinetics
were operative in the deoxyfluorination of primary and secondary alcohols, consistent with a rate
57
limiting SN2 step between the sulfonate ester and fluoride species. Under the highest-yielding
screening conditions, second order t1/2 values ranging from 2 – 5 hours were measured for the
formation of primary and secondary fluorides 2.13, 2.14, 2.16, and 2.18.
Unactivated tertiary alcohol 2.19-OH demonstrated low overall reactivity (Figure 2.9).
The highly reactive PBSF performed by far the best, delivering between 9 – 22% yield. Among
the arylsulfonyl fluorides, significant product formation was only observed with DBU and
MTBD, again consistent with an increasingly congested nucleophile approach trajectory. In
contrast to primary and secondary unactivated alcohols, tertiary alcohol 2.19-OH displayed
distinct first-order kinetics indicative of an SN1 mechanism and rate-limiting autodissociation of
the leaving group. Although SN2 reactions with PBSF are complete within seconds, ionization of
perfluorobutanesulfonate is slow; the reaction of 2.19-OH had a first order t1/2 of 1.5 hours.
Figure 2.9 Deoxyfluorination of tertiary unactivated alcohols.
Cyclic alcohols typically have higher energy SN2 transition states because the presence of
ring strain leads to deviations from the ideal trigonal bipyramidal geometry (Figure 2.10).
Screening results for cyclic alcohols 2.10-OH, 2.12-OH, and 2.20-OH – 2.25-OH (Figures 2.11
and 2.12) indicated that PBSF universally provided the highest yields. As with tertiary alcohols,
the highly active perfluorobutanesulfonate leaving group is needed to lower ΔG‡ sufficiently for
the reaction to proceed to completion at room temperature.
58
Figure 2.10 Effect of ring strain on SN2 transition state.
Figure 2.11 Deoxyfluorination of 4-, 5-, and 7-membered cyclic alcohols.
59
For cyclobutanol 2.12-OH and 5-membered cyclic alcohol isosorbide 2.20-OH, PBSF
afforded nearly quantitative yields without much preference for base (Figure 2.11). Moderate
reactivity was observed with 4-NsF—the most electron-deficient arylsulfonyl fluoride—and
BTPP, which generates the most nucleophilic HF adduct, but otherwise only trace yields were
obtained. Hydroxyproline diastereomers 2.10-OH and 2.21-OH showed a broader sulfonyl
fluoride tolerance (indicative of lower ring strain in the transition state) as well as a marked
preference for the unencumbered MTBD due to steric encroachment of the ring on the
nucleophile approach trajectory. Allofuranose 2.22-OH displayed absolutely no reactivity except
with PBSF. The measured second-order t1/2 was 32 minutes, indicating a reaction rate more than
100 times slower than typical PBSF fluorinations, likely arising from rigidity of the fused
bicyclic structure. Bridged bicyclic nortropine 2.23-OH was obtained in a modest 52% yield
with PBSF and BTMG, but interestingly, 4-CF3PhSF and 4-NsF performed almost as well with
the less bulky DBU.
Cyclohexanols are notoriously low-yielding deoxyfluorination substrates because the
absence of ring-strain in the ground state further increases the ΔG‡ for nucleophilic substitution
(Figure 2.12). Cyclohexanol diastereomers 2.24-OH and 2.25-OH afforded at most 50% and
13% yields, respectively. 2.24-OH bears an equatorial leaving group, essentially requiring the
nucleophile to approach from inside the ring (Figure 2.13). On the other hand, the axial leaving
group in 2.25-OH is accompanied by two anti-periplanar hydrogens, which substantially lowers
the entropic cost of E2 elimination, allowing it to become the dominant reaction pathway.
Complete stereochemical inversion was observed for each of the cyclic alcohols described above
including for hydroxyproline diastereomers 2.10-OH and 2.21-OH and cyclohexanol
diastereomers 2.24-OH and 2.25-OH, indicating an SN2 mechanism.
60
Figure 2.12 Deoxyfluorination of cyclohexanols.
Figure 2.13 Cyclohexanol nucleophile approach trajectories.
Our previous studies indicated that temperature is a critical variable, especially in the
deoxyfluorination of cyclic alcohols. For example, the fused equatorially-constrained
cyclohexanol 2.26-OH afforded only 4% yield with PyFluor at room temperature; however,
yield was found to increase dramatically with temperature to as high as 74% yield at 58 °C
(Table 2.5). Likewise, the yield of 2.24 with 4-NsF and BTMG improved from 11% at room
temperature to 34% at 50 °C, although a comprehensive temperature screen was not conducted.
Our results suggest that there may exist an inverse relationship between optimal leaving group
ability and reaction temperature. For deactivated substrates including cyclic and tertiary alcohols,
arylsulfonates are simply poor leaving groups at room temperature, again due to the high ΔG‡
associated with ring strain. Under these circumstances, raising reaction temperature is likely to
improve yield, as was seen with 2.24-OH and 2.26-OH. Likewise, for unactivated or activated
substrates (ie. benzylic alcohols) the perfluorobutanesulfonate ester is too reactive, facilitating
61
Table 2.5 Effect of temperature on deoxyfluorination of cyclic substrates.
HO
OMe
Me
HH
H
F
OMe
Me
HH
H
PyFluor (1.1 eq)DBU (2 eq)
toluene (0.4 M), 48 h
temperature yield
room temperature 4% yield
40 °C 6% yield
50 °C 50% yield
58 °C 74% yield
60 °C 69% yield
70 °C 55% yield
2.26-OH 2.26
elimination and unwanted substitution reactions. In this scenario, cooling the PBSF reactions
may improve selectivity for fluorination, although this hypothesis has not been tested. There
certainly exists the possibility that superior yields to those observed at room temperature may be
obtained with some strategic combination of temperature and sulfonyl fluoride.
Activated benzylic alcohols 2.27-OH – 2.31-OH afforded the highest yields with
comparatively electron-rich arylsulfonyl fluorides such as 4-CF3PhSF and 4-ClPhSF
(Figure 2.14). In the case of primary benzylic alcohols 2.27-OH – 2.30-OH, the major side
product arose from competing nucleophilic substitution with the base, and there was a clear trend
of increasing yield with increasing base sterics. With BTPP, simple benzylic fluorides 2.27 and
2.28 were obtained in near quantitative yield. The more complex heteroaryl benzylic alcohols
2.29-OH and 2.30-OH were lower yielding. Secondary benzylic alcohol 2.31-OH was
particularly susceptible to rearrangement and elimination to the corresponding styrene. Fluoride
2.32 was formed with retention of configuration, indicative of an aziridinium ion intermediate.87
In this case, stabilization of positive charge at the benzylic position led to an overwhelming
preference for ring-opening at the benzylic site.
87 Hamman, S.; Beguin, C. G.; Charlon, C.; Luu-Duc, C. J. Fluorine Chem. 1987, 37, 343.
62
Figure 2.14 Deoxyfluorination of benzylic alcohols.
For many of the screening substrates. PyFluor delivered anomalously low yields despite
having an intermediate leaving group ability that lies between 4-ClPhSF and 4-CF3PhSF (eg. see
2.29 and 2.30). The obvious difference between the 2-pyridinesulfonate leaving group and the
63
other arylsulfonates is that the pyridine is ionizable. Indeed, protonation of the pyridine ring
should dramatically enhance its electron-withdrawing capability, yet in the presence of excess
superbase, significant protonation is unlikely. PyFluor has a tendency to perform better in
toluene than THF; for example, a 10% boost in yield is observed in toluene for fluoride 2.1. The
polarizability of the pyridine ring may enhance reactivity in weakly polar solvents.
Primary and secondary allylic alcohols 2.33-OH – 2.34-OH were generally lower
yielding and were susceptible allylic rearrangement (Figure 2.15). Primary fluoride 2.33 was
Figure 2.15 Deoxyfluorination of allylic alcohols.
64
obtained in as high as 52% yield with 2% yield of the tertiary rearrangement isomer (>20:1 rr).
Secondary fluoride 2.34 was formed less selectively in 36% yield with 16% of the linear isomer
(2.2:1 rr). Tertiary fluoride 2.35 was generated in no more than 10% yield and was outnumbered
1.5:1 by the linear isomer. For 2.33 and 2.34, a clear trend of decreasing regioselectivity was
observed with increasing leaving group ability, suggesting a competition between the SN2
mechanism and a unimolecular pathway involving an allyl cation intermediate. Interestingly,
timepoint studies indicated that first-order kinetics were operative for both secondary fluoride
2.34 and tertiary fluoride 2.35.
Homobenzylic alcohols 2.36-OH and 2.37-OH were both generally high yielding (Figure
2.16); however, 2.37 was clearly flavored by PBSF whereas 2.36 showed little sulfonyl fluoride
dependence, similar to that observed with α-amino alcohol 2.16-OH. We suggest that product
may form via an arenium ion intermediate. Homoallylic alcohol 2.38-OH was observed to
undergo a similar homoallylic rearrangement (Figure 2.17). 88 The α-hydroxycarbonyl
compounds 2.39-OH and 2.40-OH (Figure 2.18) were both fluorinated in reasonable yield. The
Figure 2.16 Deoxyfluorination of homobenzylic alcohols.
88 Néder, Á.; Uskert, A.; Nagy, É.; Méhesfalvi, Z.; Kuszmann, J. Acta Chim. Acad. Sci. Hung. 1980, 103, 231.
65
Figure 2.17 Deoxyfluorination of homoallylic alcohols.
Figure 2.18 Deoxyfluorination of α-hydroxycarbonyls.
sulfonate ester of 2.40-OH did not undergo complete conversion, perhaps due to the presence of
an adjacent unprotected tertiary alcohol. In both cases, abnormally low yields were observed
with 4-NsF, perhaps arising from nucleophilic aromatic substitution of the highly electron
deficient ring by the adjacent carbonyl.
In contrast, most β-hydroxycarbonyl compounds afforded exclusively the α,β-unsaturated
elimination side product owing to the comparatively low pKa of the α proton. One exception,
N-trityl serine 2.41-OH afforded as high as 28% yield with PBSF and BTPP (Figure 2.19). In
this case, the bulky trityl group likely blocks the α proton from deprotonation by the equally
66
Figure 2.19 Deoxyfluorination of β-hydroxycarbonyls.
bulky BTPP. Interestingly, no elimination was observed under these conditions; the aziridine
formed by displacement of the sulfonate by the tritylated nitrogen accounted for most of the
mass balance. The aziridine does not appear to be an intermediate, since there is no consumption
of aziridine when exposed to base-HF adduct for extended periods of time.
Hemiacetal-derived fluorides such as 2.42 likely form via decomposition of the sulfonate
to an oxocarbenium ion. Evidence for this intermediate includes the observation that
diastereoselectivity improves with increasing steric bulk of the base-HF adduct (Figure 2.20).
Thus, the preference for the inverted α-fluoride arises from the favorable nucleophile approach
on the less-hindered face of the oxocarbenium ion rather than through concerted inversion.
Figure 2.20 Deoxyfluorination of hemiacetals.
67
Fluorides 2.1, 2.10, 2.12 – 2.25 and 2.27 – 2.42 were scaled up under their respective
highest-yielding conditions on 0.5 – 1.0 mmol scale as shown in Table 2.6. With the exception of
2.30, allylic alcohols 2.33 – 2.35, and 2.42, which are susceptible to decomposition, all of the
products were isolated, although in some cases with characterized impurities. Many substrates
afforded similar or even improved reactivity on this scale, but primary benzylic alcohols
experienced a noticeable 10 – 20% decline in yield that we were unable to resolve with dilution
or reduction of base equivalents. Additionally, most reactions involving PBSF were highly
exothermic as a result of being complete within seconds (see Figure 2.2). In at least one instance,
addition of base actually caused the top layer of THF to boil briefly, indicating an internal
temprature of at least 66 °C. Surprisingly, yields remained largely unchanged from those
observed on the 0.1 mol screening scale; however, it is evident that most PBSF reactions could
not be carried out on preparatory scale without dilution, active cooling, or slow additions.
2.4 Modeling Sulfonyl Fluoride Reaction Space via Machine Learning
The results summarized in Table 2.6 denote a complex reaction landscape. Among
32 substrates, we identified nine distinct sets of optimized conditions (not including variations in
stoichiometry, concentration, and reaction time) that included all five sulfonyl fluorides and four
bases. Across our entire data set, under the individual optimal conditions for 2.1, 2.10, 2.12 –
2.25 and 2.27 – 2.42, we obtained an average of 64% yield of deoxyfluorinated product. The
highest yielding single set of conditions was the combination of PBSF with BTPP, which gave
an average of 53% yield. Had we selected this as our single, “optimal” set of conditions, the
average of our reported yields would have been 17 percentage points89 lower than that obtained
89 A 17 percentage point decrease in yield is not the same as a decrease of 17% yield. For example, a decline from 10% yield to 5% yield represents a 50 percentage point decrease in yield and a decrease of 5% yield.
68
Table 2.6 Isolation of deoxyfluorination products.
The sulfonyl fluoride and base employed for each substrate are listed below the yield. a Run in toluene (1.0 M). b Co-isolated with side product (see experimental section). c Inversion observed. d >20:1 diastereoselectivity. e 19F NMR yield. f Retention observed. g 4:1 dr.
with individualized conditions. In the most extreme case, PBSF and BTPP generated secondary
benzylic fluoride 2.31 in only 9% yield, an 84 percentage point decrease from the highest
yielding conditions (57% with 4-CF3PhSF and BTMG). Additionally, as noted in the preceding
69
section, PBSF reactions are highly exothermic, which may be a deterrent for process or industrial
chemists. These results highlight an ongoing conundrum in methods development—the trade-off
between identifying a general protocol that performs well across many substrates vs. developing
highly optimized conditions tailored to individual substrates. Organic chemists generally tend to
report a single set of “optimal” conditions with the objective of rendering the method easily
adoptable. On the one hand, a 17 percentage point hit to yield is not a terrible penalty for
defining a universal set conditions. However, when these universal conditions are low-yielding
or fail to deliver product for a new substrate, a user is unlikely to risk the time and resources
required to optimize reaction conditions unless no other options are available.
On the other hand, when individually optimized sets of conditions are reported, new users
may be dissuaded from adopting the method due to the complexity of interpreting
multidimensional chemical data. Several strategies that have been advanced to help address the
multitude of options, including the publication of reviews featuring “how-to” manuals,90 the
development of high-throughput informer libraries, 91 and the commercialization of catalyst
mixtures.92 As a user-friendly alternative, our laboratory has sought to employ machine learning
to analyze existing reaction data in its full complexity and then predict optimal conditions on a
90 For example, Grubbs laboratory published a review describing how to select from the many published olefin metathesis catalysts based on substrate class and desired product E- or Z- selectivity. Chatterjee, A. K.; Choi, T.-L.; Sanders, D. P.; Grubbs, R. H. J. Am. Chem. Soc. 2003, 125, 11360. 91 Merck has championed the idea of using preplated arrays of complex substrates to standardize the evaluation of new catalysts. Kutchukian, P. S.; Dropinski, J. F.; Dykstra, K. D.; Li, B.; DiRocco, D. A.; Streckfuss, E. C.; Campeau, L.-C.; Cernak, T.; Vachal, P.; Davies, I. W.; Krska, S. W.; Dreher, S. D. Chem. Sci. 2016, 7, 2604. 92 The Buchwald laboratory published dozens of biarylphosphine ligands tailored to different classes of palladium-catalyzed amination. When industry researchers clamored for a one-size-fits-all set of conditions, Fors discovered that a precatalyst mixture containing two of their best performing ligands led to broad-spectrum reactivity. Fors, B. P.; Buchwald, S. L. J. Am. Chem. Soc. 2010, 132, 15914.
70
per substrate basis. This approach would enable researchers to evaluate the feasibility of a
specific transformation and identify high-yielding conditions for a new substrate a priori.93
Recently, Derek Ahneman from our laboratory demonstrated that machine learning
algorithms can successfully model reaction space encompassed in large data sets and with high
predictive ability.94 Derek evaluated the Buchwald-Hartwig amination of p-toluidine with 15
distinct aryl halides, 4 palladium catalysts, 3 bases, and 23 isoxazole additives. Using ultra-high-
throughput experimentation equipment at Merck, Derek performed the reactions and obtained
yields for all 4,130 combinations of the above reagents. This data set was then split into a
training set comprising 70% of the reactions and a test set containing the remaining 30%, which
was saved for final model assessment. The training set was then used to evaluate a number of
machine learning algorithms with k-fold cross-validation 95 including linear regression with
principal component analysis, k-nearest neighbors, support vector machine, Bayes generalized
linear model, neural network, and random forest.96
In Derek’s study, the random forest model proved superior, predicting the test set yields
with a root-mean-squared error (RMSE) 97 of 7.8% yield, which approaches the limit of
experimental error. The random forest algorithm is based on the concept a decision tree, which
93 Previous attempts include a largely unsuccessful broad spectrum model based on substrate descriptors (Skoraczyński, G.; Dittwald, P.; Miasojedow, B.; Szymkuć, S.; Gajewska, E. P.; Grzybowski, B. A.; Gambin, A. Sci. Rep. 2017, 7, 3582) and a model based on substituent thermodynamic descriptors (Emami, F. S.; Vahid, A.; Wylie, E. K.; Szymkuć, S.; Dittwald, P.; Molga, K.; Grzybowski, B. A. Angew. Chem., Int. Ed. 2015, 54, 10797). 94 Ahneman, D. T.; Estrada, J. ; Lin, S.; Dreher, S. D.; Doyle, A. G. Manuscript under review. 95 For example, in 4-fold cross-validation, the training set would be divided into four equal subsets. The model is then trained using three of the subsets and validated with fourth. This is performed four times so that each subset serves as the validation set one time. This technique can be used to assess and address the extent of overfitting during model training. 96 Each of these models is described in the following review: Mitchell, J. B. Wiley Interdiscip. Rev.: Comput. Mol. Sci. 2014, 4, 468. 97 Assuming a normal error distribution, 68% of the predicted values will lie within one RMSE unit of the observed values.
71
resembles the way that chemists intuitively analyze reactivity trends. For example, upon
examining substrate scope in Table 2.6, one might generate a simple decision tree such as that
shown in Figure 2.21 that predicts the yields of alcohols based on structural categorization. This
type of model is useful for identifying general trends, but cannot make numerical predictions
with high accuracy. Unfortunately, iterative refinement of decision trees rapidly leads to
overfitting wherein every member of a data set is represented by an individual “leaf” in the tree.
At this extreme, the model predicts training set reactions with perfect precision, but has
absolutely no predictive ability. The random forest algorithm trains an ensemble (or “forest”) of
hundreds or thousands of simple decision trees trained on random-selected variable subsets.
When making a prediction, each individual tree is queried, and the collective output is averaged.
This allows the model to obtain a high degree of complexity and precision without succumbing
to overfitting. 98
Figure 2.21 Simple vs. complex decision trees.
One limitation of Derek’s study was the necessity of relying on additive screens to assess
functional group compatibility. Initially, the objective was to assess electrophiles and
nucleophiles containing the isoxazole ring, a valuable pharmacophore that is frequently
incompatible with cross-coupling due to oxidative addition of palladium into the N—O bond.
Unfortunately, in order to obtain accurate HPLC yields, each of the potentially hundreds of
98 Kleinberg, E. M. Ann. Statist. 1996, 24, 2319. Breiman, L. Mach. Learn. 2001, 45, 5.
72
products would need to have been isolated in high purity, adding thousands of man-hours to an
already complicated project. As a compromise, Derek selected only 15 distinct products but
subjected each to an additive screen of 23 isoxazoles. The additive screen has been advanced by
Glorius and co-workers as a way for rapidly assessing functional group compatibility for new
methods.99 In lieu of laboriously generated vast tables of non-standardized substrates, Glorius
envisioned that a representative reaction for any new method could be tested in the presence of a
standardized list of functional group containing additives (ie. alkenes, alcohols, amines, etc.).
Diminished yield or reaction failure in the presence of a specific additive would indicate
incompatibility, allowing one to “score” a new method based on functional group tolerance.
However, a major limitation is that additive tolerance does not necessarily indicate that a
functional group or motif will be tolerated when incorporated into the substrate scaffold.
One advantage of studying deoxyfluorination is that fluorination yields may be accurately
determined from crude reaction mixtures by 19F NMR, enabling one to rapidly assess structurally
and stereochemically diverse structures. In this regard, we believed that deoxyfluorination would
prove ideal for evaluating the capabilities of predictive modeling with machine learning.
In order to develop a random forest model, we first constructed a descriptor table
describing each of the 640 screening reactions shown in Figures 2.6 – 2.20. In our final model,100
99 Collins, K. D.; Glorius, F. Acc. Chem. Res. 2015, 48, 619. 100 In our initial attempt, we used a Python script developed by Derek Ahneman to automatically extract 43 atomic, molecular and vibrational descriptors (available at <https://github.com/ doylelab/rxnpredict/>). We found that the model trained with these automated descriptors was less predictive than with the smaller set of hand-selected descriptors. Although the initial model predicted the test set with an RMSE of 9.3% yield, the external validation set had an RMSE of 24.6% yield with weak correlation (R2 = 0.163). (For comparison, our final model had a test set RMSE of 7.4% yield and a validation set RMSE of 16% yield (R2 = 0.712).) The automated descriptor model predicted that PBSF would be the best sulfonyl fluoride for all validation substrates. As discussed above, PBSF does have the highest average yield across all substrate classes; but is outperformed much of the time by arylsulfonyl fluorides. We suspect that the
73
we hand selected atomic and molecular descriptors101 relevant to the SN2 mechanism for each of
the 32 alcohols, 4 bases, and 5 sulfonyl fluorides. Additionally, we included a number of
categorical descriptors describing the alcohol structure (ie. primary vs. secondary vs. tertiary,
cyclic, benzylic, etc.).102 The complete list of 23 descriptors is shown below: (The designated
atoms refer to those shared among substrate classes highlighted in Figure 2.22.)
Figure 2.22 Designated shared substrate atoms and descriptors.
- Alcohol - *C1 electrostatic charge - Alcohol - 4-membered ring - Alcohol - *C1 exposed area (Å2) - Alcohol - 5-membered ring - Alcohol - electronegativity - Alcohol - 6-membered ring - Base - *N1 exposed area (Å2) - Alcohol - 7-membered ring - Sulfonyl fluoride - *S1 electrostatic charge - Alcohol - benzylic - Sulfonyl fluoride - *F1 electrostatic charge - Alcohol - allylic - Sulfonyl fluoride - *O1 electrostatic charge - Alcohol - homobenzylic --------------------------------------------------------- - Alcohol - homoallylic - Alcohol - primary - Alcohol - alpha-carbonyl - Alcohol - secondary - Alcohol - beta-carbonyl - Alcohol - tertiary - Alcohol - hemiacetal - Alcohol - cyclic - Alcohol - amino alcohol
With these descriptor values, a table was assembled with 640 rows corresponding to each
screening reaction and 23 columns for each of the descriptors. For example, the row describing
the reaction of alcohol 2.1-OH with DBU and PyFluor would contain the 19 alcohol descriptors
for 2.1-OH, the single base descriptor for DBU, and the 3 sulfonyl fluoride descriptors for
overabundance of 43 descriptors relative to 32 substrates resulted in overfitting such that continuous variables (ie. electrostatic charge) were instead treated as discrete variables. As such, a new substrate bearing slightly different electrostatic charge values would be unclassifiable, and the algorithm would be forced to make generic predictions representing overall averages. 101 Calculations were run using Spartan ’14 v. 1.1.14 with the B3LYP, 6-31G* basis set. 102 Categorical descriptors are represented in binary. If an alcohol fulfills the category the value is ‘1’, otherwise the value is ‘0’.
74
PyFluor. Following data assembly, an R script103 documented in the experimental section was
executed. After loading and normalizing the descriptor table values and observed yields, the
R script randomly generated a training set with 70% of the screening data and a test set
containing the remaining 192 reactions. The script then trained a random forest model with the
training set, which was subsequently used to predict the yield of the reactions contained in the
test set. As shown in the calibration plot104 in Figure 2.23, this model provided high accuracy
with an RMSE of 7.4% yield. Given the complexity of our data set, we were quite pleased with
this result. In comparison to the Buchwald-Hartwig amination data set, ours contained much
fewer data points (640 vs. 4,130), represented a much broader range of structural complexity, and
included multiple mechanisms (eg. SN1, SN2, anchimeric assistance).
Figure 2.23 Calibration plot of test set with random forest model.
103 R is an open-source computing software package <https://cran.r-project.org/mirrors.html>. 104 Plots observed values against predicted values. In a perfectly accurate model, all points would fall on the line x = y.
75
The largest overprediction in yield was for the reaction of primary benzylic alcohol
2.29-OH with PyFluor and MTBD (5% observed vs. 27% predicted.) As discussed earlier,
PyFluor frequently delivers low yields that break with the trend of sulfonyl fluoride reactivity
(see Figure 2.14). The largest underprediction was for cyclobutanol 2.12-OH (83% observed vs.
63% predicted), the only example of a 4-membered cyclic alcohol in the training set.
Unlike linear regression models,105 for which variable importance is proportional to the
magnitude of its polynomial coefficient, the random forest algorithm does not generate an easily
interpretable function. In our study, the trained model is an 850 KB string, equivalent to
approximately 280 pages of plain text. To assess variable importance, we randomly shuffled the
values for a specific descriptor, retrained the model, and then measured the increase in mean
squared error (MSE). In our model, three most important variables (with ΔMSE values greater
than 50 [% yield]2) were the alcohol exposed area, the base exposed area, and the alcohol carbon
electrostatic charge. These results are in line with the predominant SN2 mechanism where the
sterics of the nucleophile, the sterics of the approach trajectory, and the electronic interaction
with electrophile and nucleophile are critical to reaction outcome.
Our ultimate goal in developing a predictive model was to enable users with unreported
alcohol substrates to evaluate if the sulfonyl fluoride deoxyfluorination method would be
feasible, and if so to identify reasonably high-yielding conditions a priori. To test this capability,
we screened an external validation set of five new alcohols 2.43-OH — 2.47-OH (listed in
Figure 2.24) that did not appear in the training set106 and used the trained random forest model to
105 Milo, A. ; Neel, A. J.; Toste, F. D.; Sigman, M. S. Science 2015, 347, 737. 106 Because the original screening data (comprising 20 reactions per alcohol) was split 70/30 into the training and test sets, each alcohol appeared on average 14 times in the training set and 6 times in the test set, albeit under different conditions. The distinction between the test set and the
76
predict the outcome for each of the twenty screening conditions. As shown in the calibration plot
in Figure 2.24, yields for primary fluorides 2.43 — 2.45 were predicted accurately with RMSE
values ranging from 8 — 13% yield. Secondary unactivated α-oxy fluoride 2.46 and α-amino
fluoride lumefantrine (2.47) were predicted with substantially lower RMSEs of 20% yield and
23% yield, respectively. However, both substrates feature new functionality that was not
Figure 2.24 External validation set and calibration plot.
F
O N
FF
RMSE: 20%
PBSF BTPP
( )-2.46Predicted: 62% yieldObserved: 86% yield
RMSE: 8.4%
4-CF3PhSF BTMG
2.43Predicted: 87% yieldObserved: 85% yield
N
N N
2.44Predicted: 82% yieldObserved: 93% yield
2.45Predicted: 75% yieldObserved: 62% yield
N
4-CF3PhSF BTPP
PBSF MTBD
F
RMSE: 13%
RMSE: 11%
PhOMe
F
Me
N
MeF
Cl
Cl
Cl
RMSE: 23%
PBSF BTPP
( )-2.47Predicted: 65% yieldObserved: 82% yield
external validation set is that the trained model has not “seen” 2.43-OH — 2.47-OH under any set of conditions.
77
represented in the training set, indicating that the model can make extrapolative predictions with
reasonable accuracy. Moreover, the error appears to be largely systematic, meaning that the
model successfully identified major underlying trends.
Although the model failed to predict the absolute best conditions for any substrate, the
observed yield corresponding to the best predicted conditions was at most only 4% yield below
the actual highest observed conditions. For example, while the yields of the 20 screening
reactions for fluoride 2.44 varied from 16 — 97% yield, the best predicted conditions
(4-CF3PhSF with BTPP) afforded 93% yield, only 4% less than the highest yielding conditions
(97% yield for PBSF with BTPP, see annotation in plot in Figure 2.24). Had we followed the
reported PyFluor conditions, we would have obtained 2.44 in only 18% yield; however, the
random forest model successfully identified conditions capable of delivering above 90% yield.
This level of accuracy is more than sufficient for enabling new adopters to evaluate reaction
feasibility and select initial reaction conditions.
2.5 Conclusion
We have demonstrated that the sulfonyl fluorides can fluorinate a broad range of alcohols
with selectivity and efficiency rivaling that of other classes of deoxyfluorination reagents. Unlike
sulfur(IV) and fluoroimidazolium structures, the sulfonyl fluoride motif is highly modular,
allowing one to fine-tune reagent reactivity to specific substrates. Moreover, we have shown that
a random forest model trained on a set of structurally and mechanistically diverse
deoxyfluorinations may be employed to assess substrate feasibility and accurately predict yields
for new substrates.
As it stands, our model is largely interpolative and is unlikely to provide accurate
predictions for substrate classes not included in the training set. Moreover, we have not yet
78
identified descriptors that can accurately account for the stereochemistry of cyclic substrates.107
Based on Table 2.6, a trained user likely could have predicted the optimal conditions for many of
the validation substrates. However, the model was able to make accurate predictions that defied
intuition; for example, secondary unactivated alcohol 2.46-OH would have been expected to
perform best with PyFluor and DBU, yet the model accurately predicted that PBSF would
substantially outperform other reagents.
We envision that our model could be developed into a user-friendly software tool that
could aid pharmaceutical researchers in identifying optimal deoxyfluorination conditions without
resorting to costly high throughput experimentation. As new results are obtained, our initial data
set could be continuously expanded with new substrates and additional variables including
stoichiometry, concentration, solvent, and temperature, leading to a more comprehensive
coverage of the sulfonyl fluoride deoxyfluorination reaction space. Finally, we anticipate that the
application of predictive algorithms to large experimental databases (eg. Reaxys, SciFinder) will
enable increasingly accurate predictions of reaction outcome for commonly employed reaction
classes.
2.6 Experimental Section
Reagents and Methods. See Section 1.5 for General Methods and Instrumentation.
Perfluorobutane-1-sulfonyl fluoride (PBSF) was purchased from Acros. 2-Pyridinesulfonyl
fluoride (PyFluor), 2-tert-butyl-1,1,3,3-tetramethylguanidine (BTMG), and tert-butylimino-
tri(pyrrolidino)phosphorane (BTPP) were purchased from Millipore-Sigma.
1,8-Diazabicyclo[5.4.0]undec-7-ene (DBU) was obtained from both Millipore-Sigma and Acros.
7-Methyl-1,5,7-triazabicyclo[4.4.0]dec-5-ene (MTBD) was purchased from both Millipore-
107 For example, the model predicts that the diastereomer of 2.22-OH should be relatively high yielding with PBSF, but experimentally, only trace yields are observed.
79
Sigma and TCI America. 4-Chlorobenzenesulfonyl fluoride (4-ClPhSF), 4-(trifluoromethyl)
benzene-sulfonyl fluoride (4-CF3PhSF), and 4-nitrobenzenesulfonyl fluoride (4-NsF) were
synthesized as described in Section IV. PBSF and recently purchased bottles of DBU, MTBD,
BTMG, and BTPP were stored sealed at room temperature. The remaining sulfonyl fluorides,
were stored sealed in a fridge at 2 °C. Tetrahydrofuran (THF), toluene, and other common
solvents were dispensed from a dry solvent system. Suppliers for all other materials are noted in
the individual procedures.
Reagent synthesis:
4-chlorobenzenesulfonyl fluoride (4-ClPhSF): A 1000-mL round bottom flask was charged
with a stirbar, 4-chlorobenzenesulfonyl chloride (5.00 g, 23.7 mmol, Acros), potassium
bifluoride (9.25 g, 5 equiv, Millipore-Sigma), and 50 mL of 3:1 water:acetonitrile. The resulting
suspension was stirred vigorously for five hours at room temperature. Note: This reaction
contains toxic hydrofluoric acid and will slowly etch glassware. The mixture was diluted with
50 mL brine and extracted once with 100 mL ethyl acetate. The organic extract was directly
filtered through 30 g of silica on a fritted filter, rinsing with an additional 50 mL ethyl acetate.
The filtrate was concentrated to afford 4.38 g 4-chlorobenzenesulfonyl fluoride as a fluffy white
solid (95% yield). Compound has been previously characterized. 108 1H NMR (500 MHz,
CDCl3): δ 7.96 (d, J = 8.7 Hz, 2H), 7.62 (d, J = 8.4 Hz, 2H). 13C NMR (125 MHz, CDCl3): δ
142.80, 131.53 (d, J = 25.7 Hz), 130.25, 130.00. 19F NMR (282 MHz, CDCl3): δ 66.48 (s).
108 Tang, L.; Yang, Y.; Wen, L.; Yang, X.; Wang, Z. Green Chem. 2016, 18, 1224.
80
4-(trifluoromethyl)benzenesulfonyl fluoride (4-CF3PhSF): A 1000-mL round bottom flask
was charged with a stirbar, 4-(trifluoromethyl)benzenesulfonyl chloride (9.60 g, 39.2 mmol,
Oakwood), potassium bifluoride (15.33 g, 5 equiv, Millipore-Sigma), and 80 mL of 3:1
water:acetonitrile. The resulting suspension was stirred vigorously for one hour at room
temperature. Note: This reaction contains toxic hydrofluoric acid and will slowly etch glassware.
At one hour, the mixture was diluted with 50 mL brine and extracted once with 100 mL ethyl
acetate. The organic extract was directly filtered through 30 g of silica on a fritted filter, rinsing
with an additional 100 mL ethyl acetate. The filtrate was concentrated to afford 8.47 g
4-(trifluoromethyl) benzenesulfonyl fluoride as a fluffy white solid (95% yield). Compound has
been previously characterized.108 1H NMR (500 MHz, CDCl3): δ 8.17 (d, J = 8.2 Hz, 2H), 7.92
(d, J = 8.2 Hz, 2H). 13C NMR (125 MHz, CDCl3): δ 137.33 (q, J = 33.5 Hz), 136.65 (d, J = 27.2
Hz), 129.26, 127.04 (q, J = 3.7 Hz), 122.87 (q, J = 273.5 Hz). 19F NMR (282 MHz, CDCl3): δ
65.88 (s, 1F), −63.53 (s, 3F).
4-nitrobenzenesulfonyl fluoride (4-NsF): A 1000-mL round bottom flask was charged with a
stirbar, 4-nitrobenzenesulfonyl chloride (5.00 g, 22.6 mmol, Oakwood), potassium bifluoride
(8.81 g, 5 equiv, Millipore-Sigma), and 50 mL of 3:1 water:acetonitrile. The resulting suspension
was stirred vigorously for one hour at room temperature. Note: This reaction contains toxic
hydrofluoric acid and will slowly etch glassware. At one hour, the mixture was diluted with 50
mL brine and extracted once with 100 mL ethyl acetate. The organic extract was directly filtered
through 30 g of silica on a fritted filter, rinsing with an additional 50 mL ethyl acetate. The
filtrate was concentrated to afford 4.20 g 4-nitrobenzenesulfonyl fluoride as a light brown solid
81
(91% yield). Compound has been previously characterized.109 1H NMR (500 MHz, CDCl3): δ
8.49 (d, J = 8.5 Hz, 2H), 8.25 (d, J = 8.9 Hz, 2H). 13C NMR (125 MHz, CDCl3): δ 151.92,
138.48 (d, J = 27.0 Hz), 130.15, 125.02. 19F NMR (282 MHz, CDCl3): δ 66.21 (s).
General procedure for deoxyfluorination with sulfonyl fluorides: A 2-dram vial with a stirbar
is sequentially charged with the alcohol substrate (1 mmol), the specified sulfonyl fluoride
(1.1 equiv), and dry THF (2 mL, 0.5 M). The contents are briefly stirred (~10 s) to dissolve or
suspend substrate, followed by addition of the specified base (1.1 – 1.5 equiv).* The reaction
vessel is sealed by wrapping the vial threads with Teflon tape (prior to substrate addition) and
affixing a phenolic cap. The reaction mixture is stirred at room temperature at 600 rpm for the
designated reaction time (30 min – 48 hours). Purification is typically performed by
concentrating the reaction mixture, loading onto a silica column with minimal dichloromethane,
and purifying by automated silica column chromatography. All manipulations are performed on
the benchtop. Aside from using dry solvent, no additional measures are taken to exclude air or
moisture. *Note on order of addition: The base is always added last. In cases where the substrate
alcohol is a liquid, the reaction solvent is added prior to the sulfonyl fluoride addition. PBSF and
PyFluor are added as a solution in the reaction solvent.
(±)-(3-fluorobutyl)benzene (2.1): A 1-dram vial with a stirbar was sequentially charged with
4-phenyl-2-butanol (150.2 mg, 1 mmol, Millipore-Sigma), a solution of PyFluor (177 mg,
1.1 equiv, Millipore-Sigma) in toluene (1 mL, 1.0 M), and DBU (300 μL, 2 equiv). The
reaction was stirred at 600 rpm at room temperature for 48 hours, concentrated, taken up in
minimal dichloromethane, and purified by automated column chromatography (50 g silica, 0 →
109 Tang, L.; Yang, Y.; Wen, L.; Yang, X.; Wang, Z. Green Chem. 2016, 18, 1224.
82
10% ethyl acetate in hexanes) to afford 120.7 mg product as a colorless oil (79% yield).110 1H
NMR (500 MHz, CDCl3): δ 7.32 (t, J = 7.5 Hz, 2H), 7.25 – 7.20 (m, 3H), 4.67 (dm,
J = 48.8 Hz, 1H), 2.83 (ddd, J = 14.8, 9.9, 5.3 Hz, 1H), 2.72 (ddd, J = 13.9, 9.6, 6.9 Hz, 1H),
2.07 – 1.95 (m, 1H), 1.85 (ddddd, J = 30.8, 13.9, 10.4, 6.9, 3.9 Hz, 1H), 1.37 (dd, J = 23.9,
6.2 Hz, 3H). 13C NMR (125 MHz, CDCl3): δ 141.63, 128.58, 128.57, 126.09, 90.20 (d, J =
164.9 Hz), 38.81 (d, J = 20.8 Hz), 31.52 (d, J = 4.8 Hz), 21.16 (d, J = 22.7 Hz). 19F NMR (282
MHz, CDCl3): δ −174.26 (ddqd, J = 48.0, 30.4, 23.9, 15.6 Hz).
CO2MeN
F
Boc
1-(tert-butyl) 2-methyl (2S,4S)-4-fluoropyrrolidine-1,2-dicarboxylate (2.10): Following the
general procedure, a 2-dram vial with a stirbar was charged sequentially with N-Boc-trans-4-
hydroxy-L-proline methyl ester (245.3 mg, 1 mmol, Alfa Aesar), a solution of PBSF (332 mg,
1.1 equiv) in THF (2 mL, 0.5 M), and MTBD (215 μL, 1.5 equiv). The reaction was stirred at
600 rpm at room temperature for 30 minutes, concentrated, taken up in minimal
dichloromethane, and purified by automated column chromatography (25 g silica, 0 → 15%
ethyl acetate in hexanes with 3% triethylamine) to afford 185.8 mg product as a colorless oil
(75% yield).111 1H NMR (500 MHz, CDCl3): [1.2:1 ratio of two rotamers] δ 5.19 (dq, J = 52.7,
3.7 Hz, 1H), 4.58 – 4.36 (m, 1H), 3.88 – 3.74 (m, 1H), 3.73 (s, 3H), 3.72 – 3.56 (m, 1H), 2.52 –
2.22 (m, 2H), 1.44 (d, J = 25.2 Hz, 9H). 13C NMR (125 MHz, CDCl3): [1.2:1 ratio of two
rotamers, minor rotamer denoted with *] δ 172.39, 171.99*, 154.12*, 153.74, 92.33* (d, J =
177.3 Hz), 91.26 (d, J = 177.4 Hz), 80.53*, 80.49, 57.74, 57.35*, 53.29* (d, J = 24.3 Hz), 52.97
110 Yin, J.; Zarkowsky, D. S.; Thomas, D. W.; Zhao, M. M.; Huffman, M. A. Org. Lett. 2004, 6, 1465. 111 Sladojevich, F.; Arlow, S. I.; Tang, P.; Ritter, T. J. Am. Chem. Soc. 2013, 135, 2470.
83
(d, J = 24.3 Hz), 52.49*, 52.35, 37.58 (d, J = 22.0 Hz), 36.71* (d, J = 21.8 Hz), 28.49*, 28.38.
19F NMR (282 MHz, CDCl3): [1.2:1 ratio of two rotamers] δ −172.34 – −173.70 (m).
((trans-3-fluorocyclobutoxy)methyl)benzene (2.12): Following the general procedure, a
2-dram vial with a stirbar was charged sequentially with cis-3-(benzyloxy)cyclobutan-1-ol
(178.2 mg, 1 mmol, 12:1 dr), a solution of PBSF (332 mg, 1.1 equiv) in THF (2 mL, 0.5 M), and
then BTMG (220 μL, 1.1 equiv). The reaction was stirred at 600 rpm at room temperature for 30
minutes, concentrated, taken up in minimal dichloromethane, and purified by automated column
chromatography (50 g silica, 0 → 10% ethyl acetate in hexanes) affording 151.8 mg product as a
fragrant colorless oil (84% yield, 12:1 dr). 112 1H NMR (500 MHz, CDCl3): δ 7.40 – 7.28 (m,
5H), 5.25 (dtt, J = 56.4, 6.7, 3.7 Hz, 1H), 4.43 (s, 2H), 4.35 (ddt, J = 11.4, 6.9, 4.5 Hz, 1H), 2.55
– 2.36 (m, 4H). 13C NMR (125 MHz, CDCl3): δ 137.98, 128.56, 127.91, 127.87, 87.07 (d,
J = 198.6 Hz), 70.77, 70.04 (d, J = 9.1 Hz), 38.23 (d, J = 21.6 Hz). 19F NMR (282 MHz,
CDCl3): δ −176.44 (dpd, J = 56.3, 20.7, 4.0 Hz).
(4-fluorobutyl)benzene (2.13): Following the general procedure, a 2-dram vial with a stirbar
was sequentially charged with 4-phenyl-1-butanol (150.2 mg, 1 mmol, Combi-Blocks), THF
(2 mL, 0.5 M), 4-CF3PhSF (251 mg, 1.1 equiv), and BTPP (460 μL, 1.5 equiv). The reaction
was stirred at 600 rpm at room temperature for 24 hours, concentrated, taken up in hexanes, and
purified by automated column chromatography (25 g silica, 0 → 5% ethyl acetate in hexanes) to
112 Franck, D.; Kniess, T.; Steinbach, J.; Zitzmann-Kolbe, S.; Friebe, M.; Dinkelborg, L. M.; Graham, K. Bioorg. Med. Chem. 2013, 21, 643.
84
afford 135.2 mg product as a colorless oil (89% yield).113 1H NMR (500 MHz, CDCl3): δ 7.35 –
7.28 (m, 2H), 7.25 – 7.18 (m, 3H), 4.57 – 4.50 (m, 1H), 4.44 (t, J = 5.8 Hz, 1H), 2.69 (t, J = 7.3
Hz, 2H), 1.83 – 1.69 (m, 4H). 13C NMR (125 MHz, CDCl3): δ 142.11, 128.53, 128.47, 125.97,
84.10 (d, J = 164.4 Hz), 35.54, 30.08 (d, J = 19.6 Hz), 27.12 (d, J = 5.1 Hz). 19F NMR (282
MHz, CDCl3): δ −218.31 (tt, J = 47.3, 25.2 Hz).
2-(3-fluoropropyl)-4,5-diphenyloxazole (2.14): Following the general procedure, a 2-dram vial
with a stirbar was sequentially charged with 3-(4,5-diphenyloxazol-2-yl)propan-1-ol 114
(279.3 mg, 1 mmol), 4-CF3PhSF (251 mg, 1.1 equiv), THF (2 mL, 0.5 M), and BTMG (300 μL,
1.5 equiv). The reaction was stirred at 600 rpm at room temperature for 24 hours, concentrated,
taken up in minimal dichloromethane, and purified by automated column chromatography (50 g
silica, 0 → 30% ethyl acetate in hexanes) to afford 269.2 mg product as a white crystalline solid
(96% yield). 1H NMR (500 MHz, CDCl3): δ 7.69 – 7.64 (m, 2H), 7.62 – 7.58 (m, 2H), 7.42 –
7.29 (m, 6H), 4.62 (dt, J = 47.1, 5.8 Hz, 2H), 3.02 (t, J = 7.6 Hz, 2H), 2.28 (dtt, J = 25.8, 7.5, 5.8
Hz, 2H). 13C NMR (125 MHz, CDCl3): δ 162.49, 145.42, 135.19, 132.57, 129.07, 128.74,
128.66, 128.52, 128.15, 127.99, 126.52, 82.91 (d, J = 165.7 Hz), 27.91 (d, J = 20.2 Hz), 24.18
(d, J = 5.6 Hz). 19F NMR (282 MHz, CDCl3): δ −220.58 (tt, J = 47.4, 25.7 Hz). IR (ATR,
cm−1): 3067 (w), 2969 (w), 2899 (w), 1605 (w), 1590 (s), 1503 (w), 1485 (w), 1440 (m), 1384
(w), 1322 (w), 1290 (w), 1253 (w), 1216 (m), 1203 (m), 1144 (w), 1074 (w), 1058 (m), 1027 (s),
993 (m), 962 (m), 921 (m), 901 (w), 889 (s), 767 (s), 709 (w), 696 (s), 673 (m), 661 (w). HRMS
(ESI+): Calculated for C18H17FNO+ [M + H]+ : 282.1289; found: 282.1287.
113 Prakash, G. K. S.; Chacko, S.; Vaghoo, H.; Shao, N.; Gurung, L.; Mathew, T.; Olah, G. A. Org. Lett. 2009, 11, 1127. 114 Derived from LAH reduction of oxaprozin methyl ester.
85
(2R,3R,4S,5S)-2-(6-(N-benzoylbenzamido)-9H-purin-9-yl)-5-(fluoromethyl)tetrahydro-
furan-3,4-diyl dibenzoate (2.15): Following the general procedure, a 1-dram vial with a stirbar
was sequentially charged with N6,N6,O2',O3'-tetrabenzoyladenosine 115 (341.8 mg, 0.5 mmol),
4-NsF (113 mg, 1.1 equiv), THF (1 mL, 0.5 M), and BTPP (190 μL, 1.25 equiv). The reaction
was stirred at 600 rpm at room temperature for 48 hours. The reaction mixture was concentrated,
taken up in minimal dichloromethane, and purified by automated column chromatography (25 g
silica, 10 → 50% ethyl acetate in hexanes), affording 255.3 mg product as a white solid
(75% yield). 1H NMR (500 MHz, CDCl3): δ 8.67 (s, 1H), 8.43 (s, 1H), 7.99 (d, J = 7.6 Hz, 2H),
7.93 (d, J = 7.6 Hz, 2H), 7.87 (d, J = 8.1 Hz, 4H), 7.57 (dt, J = 15.7, 7.5 Hz, 2H), 7.48 (t, J = 7.5
Hz, 2H), 7.41 (t, J = 7.9 Hz, 2H), 7.40 – 7.33 (m, 6H), 6.65 (d, J = 5.7 Hz, 1H), 6.13 (td, J = 5.7,
1.4 Hz, 1H), 6.07 (dd, J = 5.6, 3.7 Hz, 1H), 4.86 (dqd, J = 47.7, 10.7, 2.6 Hz, 2H), 4.69 (dq, J =
29.4, 2.9 Hz, 1H). 13C NMR (125 MHz, CDCl3): δ 172.36, 165.46, 165.05, 153.08, 152.60,
152.18, 146.37, 134.07, 133.99, 133.98, 133.16, 129.99, 129.92, 129.59, 128.89, 128.73, 128.67,
128.65, 128.35, 127.68, 86.52, 82.52 (d, J = 18.5 Hz), 82.47 (d, J = 173.1 Hz), 74.60 (d, J = 2.5
Hz), 71.46 (d, J = 4.5 Hz). 19F NMR (282 MHz, CDCl3): δ −230.87 (td, J = 47.1, 29.3 Hz). IR
(ATR, cm−1): 3067 (w), 1718 (s), 1599 (m), 1579 (m), 1493 (w), 1450 (m), 1315 (w), 1241 (s),
1178 (m), 1092 (s), 1069 (m), 1025 (m), 1002 (w), 932 (w), 902 (w), 870 (w), 799 (w), 772 (w),
701 (s). HRMS (ESI+): Calculated for C38H29FN5O7+ [M + H]+ : 686.2046; found: 686.2051.
115 Debarge, S.; Balzarini, J.; Maguire, A. R. J. Org. Chem. 2011, 76, 105.
86
(E)-4-((2-chloro-4-nitrophenyl)diazenyl)-N-ethyl-N-(2-fluoroethyl)aniline (2.16): Following
the general procedure, a 2-dram vial with a stirbar was sequentially charged with Disperse
Red 13 (348.8 mg, 1 mmol, Millipore-Sigma), 4-CF3PhSF (251 mg, 1.1 equiv), THF (2 mL, 0.5
M), and BTMG (300 μL, 1.5 equiv). The reaction was stirred at 600 rpm at room temperature
for 24 hours, concentrated, taken up in minimal dichloromethane, and purified by automated
column chromatography (25 g silica, 0 → 50% ethyl acetate in hexanes with 5% triethylamine)
to afford 288.8 mg product as a intensely dark red/purple solid (82% yield). 1H NMR (500
MHz, CDCl3): δ 8.39 (d, J = 2.5 Hz, 1H), 8.15 (dd, J = 8.9, 2.4 Hz, 1H), 7.94 (d, J = 9.3 Hz,
2H), 7.77 (d, J = 8.9 Hz, 1H), 6.77 (d, J = 9.2 Hz, 2H), 4.66 (dt, J = 47.0, 5.2 Hz, 2H), 3.77 (dt, J
= 23.7, 5.2 Hz, 2H), 3.58 (q, J = 7.1 Hz, 2H), 1.28 (t, J = 7.1 Hz, 3H). 13C NMR (125 MHz,
CDCl3): δ 153.30, 151.81, 147.46, 144.70, 134.26, 127.17, 126.26, 122.85, 118.28, 111.74,
80.93, 50.83 (d, J = 21.7 Hz), 46.45, 12.39. 19F NMR (282 MHz, CDCl3): δ −221.69 (tt, J =
47.2, 23.8 Hz). IR (ATR, cm−1): 3097 (w), 2972 (w), 2921 (w), 1597 (m), 1510 (s), 1403 (m),
1376 (w), 1329 (s), 1258 (w), 1234 (w), 1192 (w), 1137 (w), 1119 (s), 1041 (w), 999 (s), 888
(m), 821 (s), 796 (w), 746 (m), 726 (m). HRMS (ESI+): Calculated for C16H17ClFN4O2+ [M +
H]+ : 351.1019; found: 351.1020.
(±)-2-(fluoromethyl)-1-(3-phenylpropyl)piperidine (2.17): Following the general procedure, a
2-dram vial with a stirbar was sequentially charged with (1-(3-phenylpropyl)piperidin-2-
yl)methanol (233.3 mg, 1 mmol), a solution of PBSF (332 mg, 1.1 equiv) in THF (10 mL, 0.1
87
M), and BTPP (460 μL, 1.5 equiv). The reaction was stirred at 600 rpm at room temperature for
24 hours. The reaction mixture was concentrated, taken up in minimal dichloromethane, and
purified by automated column chromatography (25 g silica, 3% triethylamine in hexanes). The
rearranged isomer 3-fluoro-1-(3-phenylpropyl)azepane (2.17′) eluted first (127.9 mg, 54% yield)
followed by the partially resolved title compound 2.17 (76.7 mg, 32% yield), both isolated as
colorless oils. 1H NMR (500 MHz, CDCl3): δ 7.28 (t, J = 7.6 Hz, 2H), 7.19 (d, J = 7.7 Hz, 3H),
4.56 – 4.27 (m, 2H), 2.89 (dt, J = 11.8, 4.1 Hz, 1H), 2.78 (ddd, J = 13.2, 9.6, 6.2 Hz, 1H), 2.67 –
2.55 (m, 2H), 2.51 (ddd, J = 13.6, 9.4, 5.6 Hz, 2H), 2.21 (td, J = 11.2, 3.1 Hz, 1H), 1.81 (ddt, J =
15.7, 8.8, 6.0 Hz, 2H), 1.71 (dt, J = 12.8, 4.2 Hz, 1H), 1.62 (tt, J = 13.5, 3.8 Hz, 2H), 1.51 (tdd,
J = 14.3, 7.6, 3.8 Hz, 1H), 1.45 – 1.35 (m, 1H), 1.36 – 1.23 (m, 1H). 13C NMR (125 MHz,
CDCl3): δ 142.40, 128.47, 128.41, 125.84, 85.99 (d, J = 170.2 Hz), 60.38 (d, J = 17.8 Hz), 54.23
(d, J = 2.0 Hz), 52.20, 33.95, 28.45 (d, J = 6.8 Hz), 27.55, 25.72, 23.72. 19F NMR (282 MHz,
CDCl3): δ −221.12 (td, J = 47.7, 21.6 Hz). IR (film, cm−1): 3026 (w), 2933 (m), 2857 (w), 2797
(w), 1603 (w), 1496 (w), 1454 (m), 1340 (w), 1278 (w), 1119 (w), 1084 (w), 1059 (w), 1011 (m),
984 (w), 944 (w), 907 (w), 875 (w), 744 (m) 698 (s). HRMS (ESI+): Calculated for C15H23FN+
[M + H]+: 236.1809; found: 236.1806.
(±)-1-(5-fluorohexyl)-3,7-dimethyl-3,7-dihydro-1H-purine-2,6-dione (2.18): A 1-dram vial
with a stirbar was charged sequentially with (±)-lisofylline116 (280.3 mg, 1 mmol), a solution of
PyFluor (177 mg, 1.1 equiv) in toluene (1 mL, 1.0 M), and DBU (300 μL, 2 equiv). The reaction
was stirred at 600 rpm at room temperature for 48 hours, concentrated, taken up in minimal
116 Kala, E. P.; Wojcik, T. Acta Pol. Pharm. 2007, 64, 109.
88
dichloromethane, and purified by automated column chromatography (50 g silica, 0 → 10%
pmethanol in dichloromethane) affording 257.4 mg of an off-white solid containing 228.0 mg of
the title compound117 (77% yield) and 29.4 mg of elimination side products (11% elimination,
7.2:1 selectivity). 1H NMR (500 MHz, CDCl3): δ 7.47 (s, 1H), 4.60 (ddqd, J = 49.6, 7.9, 6.2,
4.3 Hz, 1H), 3.96 (t, J = 7.6 Hz, 2H), 3.94 (s, 3H), 3.52 (s, 3H), 1.68 – 1.32 (m, 6H), 1.26 (dd, J
= 24.0, 6.2 Hz, 3H). 13C NMR (125 MHz, CDCl3): δ 155.29, 151.48, 148.76, 141.48, 107.67,
90.81 (d, J = 164.4 Hz), 41.17, 36.57 (d, J = 20.8 Hz), 33.62, 29.71, 27.84, 22.54 (d, J = 5.0 Hz),
21.04 (d, J = 22.8 Hz). 19F NMR (282 MHz, CDCl3): δ −172.56 (ddqd, J = 48.8, 27.4, 23.9,
17.3 Hz).
(3-fluoro-3-methylbutyl)benzene (2.19): A 2-dram vial with a stirbar was sequentially charged
with 2-methyl-4-phenyl-2-butanol (164.2 mg, 1 mmol, TCI), a solution of PBSF (332 mg,
1.1 equiv) in THF (2 mL, 0.5 M), and BTPP (460 μL, 1.5 equiv). The reaction was stirred at 600
rpm at room temperature for 12 hours, concentrated, taken up in minimal hexanes, and purified
by automated column chromatography (25 g silica, 0 → 9% ethyl acetate in hexanes) to afford
29.1 mg product as a colorless oil (18% yield).118 1H NMR (500 MHz, CDCl3): δ 7.29 (dd, J =
8.3, 6.9 Hz, 2H), 7.23 – 7.17 (m, 3H), 2.75 – 2.70 (m, 2H), 1.93 (dm, J = 19.5 Hz, 2H), 1.41 (d, J
= 21.4 Hz, 6H). 13C NMR (125 MHz, CDCl3): δ 142.18, 128.57, 128.43, 126.00, 95.47 (d, J =
165.6 Hz), 43.49 (d, J = 23.0 Hz), 30.41 (d, J = 5.3 Hz), 26.84 (d, J = 24.8 Hz). 19F NMR (282
MHz, CDCl3): δ −138.84 (hept/t, J = 21.4, 19.8, Hz).
117 Tung, R. D.; Liu, J. F.; Harbeson, S. L. World Patent WO2011028835 A1, Mar. 10, 2011. 118 Dryzhakov, M.; Moran, J. ACS Catal. 2016, 6, 3670.
89
(3S,3aR,6S,6aS)-6-fluorohexahydrofuro[3,2-b]furan-3-yl acetate (2.20): Following the
general procedure, a 20-mL scintillation vial with a stirbar was charged sequentially with
isosorbide-2-acetate (188.2 mg, 1 mmol, Combi-Blocks), a solution of PBSF (332 mg, 1.1 equiv)
in THF (10 mL, 0.1 M), and BTPP (460 μL, 1.5 equiv). The reaction was stirred at 600 rpm at
room temperature for 30 minutes, concentrated, taken up in minimal dichloromethane, and
purified by automated column chromatography (25 g silica, 0 → 60% ethyl acetate in hexanes)
to afford 178.6 mg of the title compound (94% yield) (coisolated with 10.2 mg n-hexane.) 1H
NMR (500 MHz, CDCl3): δ 5.15 (s, 1H), 5.04 (dd, J = 50.8, 2.9 Hz, 1H), 4.71 (dd, J = 8.1, 3.8
Hz, 1H), 4.66 (d, J = 3.8 Hz, 1H), 4.07 (dd, J = 24.7, 11.5 Hz, 1H), 3.90 – 3.76 (m, 3H), 2.05 (s,
3H). 13C NMR (125 MHz, CDCl3): δ 169.93, 95.05 (d, J = 180.7 Hz), 85.23, 84.91 (d, J = 30.6
Hz), 77.56, 72.64 (d, J = 22.2 Hz), 72.53, 20.91. 19F NMR (282 MHz, CDCl3): δ − 186.35
(dddd, J = 49.9, 40.2, 24.8, 8.0 Hz). IR (film, cm−1): 2965 (w), 2881 (w), 1742 (s), 1462 (w),
1434 (w), 1369 (m), 1228 (s), 1100 (w), 1067 (s), 1043 (m), 1016 (w), 989 (w), 967 (s), 922 (m),
860 (m), 842 (w), 780 (m). HRMS (ESI+): Calculated for C8H12FO4+ [M + H]+ : 191.0714;
found: 191.0711.
CO2MeN
F
Boc
1-(tert-butyl) 2-methyl (2S,4R)-4-fluoropyrrolidine-1,2-dicarboxylate (2.21): Following the
general procedure, a 20-mL scintillation vial with a stirbar was charged sequentially with N-Boc-
cis-4-hydroxy-L-proline methyl ester (245.3 mg, 1 mmol, Synthonix), a solution of PBSF
(332 mg, 1.1 equiv) in THF (10 mL, 0.1 M), and MTBD (215 μL, 1.5 equiv). The reaction was
90
stirred at 600 rpm at room temperature for 3 hours, concentrated, taken up in minimal
dichloromethane, and purified by automated column chromatography (25 g silica, 0 → 15%
ethyl acetate in hexanes with 3% triethylamine) to afford 162.1 mg product as a colorless oil
(66% yield).119 1H NMR (500 MHz, CDCl3): [1.6:1 ratio of two rotamers] δ 5.19 (dt, J = 52.4,
3.8 Hz, 1H), 4.48 – 4.32 (m, 1H), 3.94 – 3.76 (m, 1H), 3.73 (d, J = 7.0 Hz, 3H), 3.60 (ddd, J =
36.9, 12.9, 3.4 Hz, 1H), 2.63 – 2.50 (m, 1H), 2.09 (dddd, J = 38.8, 14.0, 9.5, 4.0 Hz, 1H), 1.42
(d, J = 22.9 Hz, 9H). 13C NMR (125 MHz, CDCl3): [1.6:1 ratio of two rotamers, minor rotamer
denoted with *] δ 173.26, 173.10*, 154.24*, 153.64, 92.01* (d, J = 178.7 Hz), 91.19 (d, J =
178.7 Hz), 80.70, 80.68*, 57.76, 57.40*, 53.38* (d, J = 22.8 Hz), 53.05 (d, J = 22.6 Hz), 52.50*,
52.28, 37.66 (d, J = 22.7 Hz), 36.78* (d, J = 22.6 Hz), 28.44*, 28.32. 19F NMR (282 MHz,
CDCl3): [1.6:1 ratio of two rotamers, minor rotamer denoted with *] δ −176.62 – −177.29* (m),
−177.54 (dtdd, J = 52.2, 38.2, 22.6, 18.6 Hz).
(3aR,5R,6S,6aS)-5-((R)-2,2-dimethyl-1,3-dioxolan-4-yl)-6-fluoro-2,2-dimethyltetrahydro-
furo[2,3-d][1,3]dioxole (2.22): Following the general procedure, a 2-dram vial with a stirbar
was charged sequentially with 1,2:5,6-di-O-isopropylidene-α-D-allofuranose (260.3 mg, 1 mmol,
Ark Pharm), a solution of PBSF (332 mg, 1.1 equiv) in THF (2 mL, 0.5 M), and then MTBD
(215 μL, 1.5 equiv). The reaction was stirred at 600 rpm at room temperature for 12 hours. Upon
completion, the mixture was concentrated, taken up in minimal dichloromethane, and purified by
automated column chromatography (15 g silica, 0 → 20% ethyl acetate in hexanes) to afford
119 Nielsen, M. K.; Ugaz, C. R.; Li, W.; Doyle, A. G. J. Am. Chem. Soc. 2015, 137, 9571.
91
229.8 mg product as a colorless oil (88% yield).120 1H NMR (500 MHz, CDCl3): δ 5.93 (d, J =
3.7 Hz, 1H), 4.99 (dd, J = 49.8, 2.3 Hz, 1H), 4.68 (dd, J = 10.7, 3.8 Hz, 1H), 4.27 (ddd, J = 8.3,
6.1, 4.8 Hz, 1H), 4.10 (dd, J = 8.7, 6.1 Hz, 1H), 4.09 (ddd, J = 29.1, 8.2, 2.2 Hz, 1H), 4.01 (dd, J
= 8.8, 4.8 Hz, 1H), 1.48 (s, 3H), 1.43 (s, 3H), 1.35 (s, 3H), 1.31 (s, 3H). 13C NMR (125 MHz,
CDCl3): δ 112.48, 109.62, 105.30, 93.93 (d, J = 183.8 Hz), 82.66 (d, J = 32.8 Hz), 80.77 (d, J =
19.0 Hz), 72.01 (d, J = 7.2 Hz), 67.29 (d, J = 1.1 Hz), 26.98, 26.82, 26.30, 25.28. 19F NMR (376
MHz, CDCl3): δ −207.60 (ddd, J = 49.9, 29.2, 10.9 Hz).
(±)-tert-butyl (1R,3s,5S)-3-fluoro-8-azabicyclo[3.2.1]octane-8-carboxylate (2.23): Following
the general procedure, a 2-dram vial with a stirbar was charged sequentially with N-Boc-
nortropine (227.3 mg, 1 mmol, Millipore-Sigma), a solution of PBSF (332 mg, 1.1 equiv) in
THF (2 mL, 0.5 M), and BTMG (300 μL, 1.5 equiv). The reaction was stirred at 600 rpm at
room temperature for 30 minutes, concentrated, taken up in minimal dichloromethane, and
purified by automated column chromatography (25 g triethylamine-pretreated silica, 0 → 25%
ethyl acetate in hexanes) to afford 122.4 mg product as a white solid (53% yield). 1H NMR (500
MHz, CD2Cl2): δ 4.95 (dtt, J = 48.8, 10.6, 6.3 Hz, 1H), 4.29 – 4.16 (m, 2H), 2.12 – 2.01 (m,
2H), 2.00 – 1.88 (m, 2H), 1.82 – 1.69 (m, 2H), 1.64 – 1.55 (m, 2H), 1.45 (s, 9H). 13C NMR (125
MHz, CD2Cl2): δ 153.43, 87.36 (d, J = 172.7 Hz), 79.64, 52.93 (d, J = 79.6 Hz), 37.84 (d, J =
98.2 Hz), 28.52, 28.43 (d, J = 97.0 Hz). 19F NMR (282 MHz, CD2Cl2): δ −179.80 (dtt, J = 48.8,
14.8, 4.7 Hz). IR (ATR, cm−1): 2981 (w), 2957 (w), 2935 (w), 2893 (w), 2867 (w), 1682 (s),
1474 (w), 1463 (w), 1382 (s), 1369 (w), 1345 (m), 1325 (w), 1310 (w), 1296 (m), 1274 (w),
120 Sladojevich, F.; Arlow, S. I.; Tang, P.; Ritter, T. J. Am. Chem. Soc. 2013, 135, 2470.
92
1257 (w), 1164 (s), 1103 (m), 1083 (s), 1036 (s), 995 (m), 967 (m), 918 (m), 878 (w), 869 (w),
844 (w), 826 (w), 803 (w), 777 (w), 757 (w), 739 (w). HRMS (ESI+): Calculated for
C8H13FNO2+ [M − C4H9 + 2H]+ : 174.0925; found: 174.0925.
2-(cis-4-fluorocyclohexyl)isoindoline-1,3-dione (2.24): Following the general procedure, a
20-mL scintillation vial with a stirbar was charged sequentially with 2-(trans-4-hydroxycyclo-
hexyl)isoindol-ine-1,3-dione121 (122.6 mg, 0.5 mmol), a solution of PBSF (166 mg, 1.1 equiv) in
THF (5 mL, 0.1 M), and BTMG (150 μL, 1.5 equiv). The reaction was stirred at 600 rpm at
room temperature for 30 minutes, concentrated, taken up in minimal dichloromethane, and
purified by automated column chromatography (25 g silica, 0 → 15% ethyl acetate in hexanes),
affording 56.3 mg of the title compound (46% yield) and 0.9 mg of an unidentified primary
fluoride rearrangement product. 1H NMR (500 MHz, CDCl3): δ 7.82 (dd, J = 5.4, 3.1 Hz, 2H),
7.70 (dd, J = 5.4, 3.0 Hz, 2H), 4.86 (d, J = 47.1 Hz, 1H), 4.17 (t, J = 13.0 Hz, 1H), 2.64 (qd, J =
13.4, 4.3 Hz, 2H), 2.20 (t, J = 12.5 Hz, 2H), 1.71 – 1.51 (m, 4H). 13C NMR (125 MHz, CDCl3):
δ 168.38, 133.99, 132.12, 123.25, 86.98 (d, J = 169.3 Hz), 49.79, 30.57 (d, J = 21.5 Hz), 24.02.
19F NMR (376 MHz, CDCl3): δ −185.90 (qt, J = 45.7, 10.1 Hz). IR (ATR, cm−1): 2940 (m),
2879 (w), 1771 (w), 1759 (w), 1701 (s), 1614 (w), 1469 (m), 1445 (w), 1432 (w), 1395 (w), 1380
(s), 1357 (m), 1336 (w), 1323 (w), 1271 (w), 1246 (m), 1157 (m), 1134 (w), 1079 (s), 1031 (m),
1017 (s), 1006 (m), 954 (w), 936 (m), 909 (m), 882 (s), 830 (m), 818 (w), 798 (m), 731 (w), 713
121 Glennon, R. A.; Hong, S.-S.; Bondarev, M.; Law, H.; Dukat, M.; Rakhit, S.; Power, P.; Fan, E.; Kinneau, D.; Kamboj, R.; Teitler, M.; Herrick-Davis, K.; Smith, C. J. Med. Chem. 1996, 39, 314.
93
(s), 684 (w). HRMS (ESI+): Calculated for C14H15FNO2+ [M + H]+ : 248.1081; found:
248.1079.
2-(trans-4-fluorocyclohexyl)isoindoline-1,3-dione (2.25): Following the general procedure, a
2-dram vial with a stirbar was charged sequentially with 2-(cis-4-hydroxycyclohexyl)
isoindoline-1,3-dione122 (245.3 mg, 1 mmol), a solution of PBSF (332 mg, 1.1 equiv) in THF
(2 mL, 0.5 M), and BTPP (460 μL, 1.5 equiv). The reaction was stirred at 600 rpm at room
temperature for 30 minutes, concentrated, taken up in minimal dichloromethane, and purified by
automated column chromatography (25 g silica, 0 → 10% ethyl acetate in hexanes), affording
43.4 mg of a white solid consisting of 36.9 mg of the title compound (15% yield) and 6.5 mg of
the elimination side product 2-(cyclohex-3-en-1-yl)isoindoline-1,3-dione. 1H NMR (500 MHz,
CDCl3): δ 7.82 (dd, J = 5.4, 3.0 Hz, 2H), 7.71 (dd, J = 5.4, 3.0 Hz, 2H), 4.63 (dtt, J = 48.6, 11.1,
4.6 Hz, 1H), 4.16 (tt, J = 12.3, 4.0 Hz, 1H), 2.34 (q, J = 13.5 Hz, 2H), 2.28 – 2.18 (m, 2H), 1.88
– 1.74 (m, 2H), 1.71 – 1.57 (m, 2H). 13C NMR (125 MHz, CDCl3): δ 168.38, 134.09, 132.04,
123.30, 91.02 (d, J = 172.9 Hz), 49.29 (d, J = 1.7 Hz), 32.01 (d, J = 19.8 Hz), 26.90 (d, J = 12.5
Hz). 19F NMR (282 MHz, CDCl3): δ −172.58 (dtt, J = 48.9, 9.1, 4.7 Hz). IR (ATR, cm−1): 3458
(w), 2936 (m), 2863 (w), 1765 (m), 1701 (s), 1611 (w), 1466 (w), 1455 (w), 1375 (s), 1332 (w),
1283 (w), 1255 (w), 1188 (w), 1167 (w), 1153 (w), 1112 (w), 1080 (m), 1032 (m), 984 (m), 939
(w), 898 (w), 856 (m), 795 (m), 712 (s), 652 (m). HRMS (ESI+): Calculated for C14H15FNO2+
[M + H]+ : 248.1081; found: 248.1079.
122 Hwang, S. H.; Tsai, H.-J.; Liu, J.-Y.; Morisseau, C.; Hammock, B. D. J. Med. Chem. 2007, 50, 3825.
94
4-(fluoromethyl)-1,1'-biphenyl (2.27): Following the general procedure, a 2-dram vial with a
stirbar was charged sequentially with biphenyl-4-ylmethanol (184.2 mg, 1 mmol, Oakwood),
4-CF3PhSF (251 mg, 1.1 equiv), THF (2 mL, 0.5 M), and then BTPP (460 μL, 1.5 equiv). The
reaction was stirred at 600 rpm at room temperature for 1 hour, concentrated, taken up in
minimal dichloromethane, and purified by automated column chromatography (25 g silica, 0 →
5% ethyl acetate in hexanes) to afford 168.2 mg product as a white solid (90% yield).123 1H
NMR (500 MHz, CDCl3): δ 7.68 – 7.60 (m, 4H), 7.52 – 7.45 (m, 4H), 7.39 (t, J = 7.4 Hz, 1H),
5.45 (d, J = 47.9 Hz, 2H). 13C NMR (125 MHz, CDCl3): δ 141.85 (d, J = 3.2 Hz), 140.70 (d, J =
1.1 Hz), 135.23 (d, J = 17.1 Hz), 128.96, 128.20 (d, J = 5.7 Hz), 127.66, 127.49 (d, J = 1.4 Hz),
127.28, 84.53 (d, J = 165.9 Hz). 19F NMR (282 MHz, CDCl3): δ −206.15 (t, J = 47.9 Hz).
4-(fluoromethyl)-N,N-dipropylbenzenesulfonamide (2.28): Following the general procedure, a
2-dram vial with a stirbar was charged sequentially with 4-(hydroxymethyl)-N,N-
dipropylbenzenesulfonamide124 (271.4 mg, 1 mmol), 4-CF3PhSF (251 mg, 1.1 equiv), THF
(2 mL, 0.5 M), and finally BTPP (460 μL, 1.5 equiv). The reaction was stirred at 600 rpm at
room temperature for 3 hours, concentrated, taken up in minimal dichloromethane, and purified
by automated column chromatography (25 g silica, 0 → 5% ethyl acetate in hexanes) to afford
232.4 mg of the title compound as a translucent crystalline solid (85% yield). 1H NMR (500
MHz, CDCl3): δ 7.82 (d, J = 8.0 Hz, 2H), 7.48 (d, J = 7.9 Hz, 2H), 5.44 (d, J = 47.1 Hz, 2H),
123 Xia, J.-B.; Zhu, C.; Chen, C. J. Am. Chem. Soc. 2013, 135, 17494. 124 Obtained via LAH reduction of probenecid methyl ester.
95
3.07 (t, J = 7.7 Hz, 4H), 1.54 (h, J = 7.4 Hz, 4H), 0.86 (t, J = 7.4 Hz, 6H). 13C NMR (125 MHz,
CDCl3): δ 140.75 (d, J = 17.5 Hz), 140.38 (d, J = 2.4 Hz), 127.43, 127.22 (d, J = 6.7 Hz), 83.45
(d, J = 169.2 Hz), 50.12, 22.12, 11.28. 19F NMR (282 MHz, CDCl3): δ −213.42 (t, J = 47.1 Hz).
IR (ATR, cm−1): 2965 (w), 2936 (w), 2876 (w), 1464 (m), 1410 (w), 1377 (w), 1335 (s), 1309
(w), 1212 (w), 1188 (w), 1155 (s), 1089 (m), 990 (s), 904 (w), 867 (w), 856 (w), 821 (m), 798
(w), 773 (m), 734 (s), 716 (w), 668 (s). HRMS (ESI+): Calculated for C13H21FNO2S+ [M + H]+ :
274.1272; found: 274.1266.
5-(4'-((2-butyl-4-chloro-5-(fluoromethyl)-1H-imidazol-1-yl)methyl)-[1,1'-biphenyl]-2-yl)-1-
trityl-1H-tetrazole (2.29): Following the general procedure, a 2-dram vial with a stirbar was
charged sequentially with N-trityl losartan125 (332.6 mg, 0.5 mmol), 4-NsF (113 mg, 1.1 equiv),
THF (1 mL, 0.5 M), and BTMG (150 μL, 1.5 equiv). The reaction was stirred at 600 rpm at
room temperature for 1 hour, concentrated, taken up in minimal dichloromethane, and purified
by automated column chromatography (25 g silica, 0 → 30% ethyl acetate in hexanes) to afford
136.9 mg of the title compound (41% yield) co-isolated with 0.9 mg ethyl acetate. 1H NMR (500
MHz, CDCl3): δ 7.97 (dd, J = 7.3, 1.8 Hz, 1H), 7.53 – 7.44 (m, 2H), 7.35 (t, J = 6.8 Hz, 4H),
7.26 (dd, J = 8.6, 7.1 Hz, 6H), 7.13 (d, J = 8.3 Hz, 2H), 6.92 (d, J = 7.3 Hz, 6H), 6.75 (d, J = 7.9
Hz, 2H), 5.02 (s, 2H), 4.98 (d, J = 50.3 Hz, 2H), 2.52 (td, J = 8.4, 7.9, 2.4 Hz, 2H), 1.66 (p, J =
7.7 Hz, 2H), 1.31 (dq, J = 14.8, 7.4 Hz, 2H), 0.87 (t, J = 7.5 Hz, 3H). 13C NMR (125 MHz,
CDCl3): δ 163.98, 149.87 (d, J = 4.5 Hz), 141.39, 141.36, 141.30, 133.99, 131.02 (d, J = 8.0
125 Levin, J. I. US Patent 5298517, Mar. 29, 1994.
96
Hz), 130.83, 130.35, 130.31, 130.11, 130.08, 128.44, 127.94, 127.76, 126.31, 125.33, 121.09 (d,
J = 19.2 Hz), 82.98, 72.18 (d, J = 164.8 Hz), 47.43, 29.75, 26.88, 22.50, 13.88. 19F NMR (282
MHz, CDCl3): δ −201.74 (t, J = 50.3 Hz). IR (ATR, cm−1): 3060 (w), 2958 (w), 2929 (w), 1733
(w), 1570 (w), 1493 (w), 1446 (m), 1424 (w), 1358 (w), 1255 (s), 1188 (w), 1158 (w), 1073 (w),
1029 (w), 1004 (w), 948 (m), 904 (w), 880 (w), 821 (w), 784 (w), 746 (s), 697 (s), 678 (w).
HRMS (ESI+): Calculated for C41H37ClFN6+ [M + H]+ : 667.2747; found: 667.2735.
1-(fluoromethyl)-3,5-dimethyl-1H-pyrazole (2.30): Following the general procedure, a 2-dram
vial with a stirbar was charged sequentially with 3,5-dimethylpyrazole-1-methanol (126.2 mg,
1 mmol, Millipore-Sigma), 4-ClPhSF (214 mg, 1.1 equiv), THF (2 mL, 0.5 M), and then BTPP
(460 μL, 1.5 equiv). The reaction was stirred at 600 rpm at room temperature for 1 hour.
Following addition of 1-fluoronaphthalene as an external standard, the yield was determined by
19F NMR in CDCl3 to be 57% yield. 1H NMR (500 MHz, CDCl3): δ 5.78 (d, J = 54.5 Hz, 2H),
5.77 (s, 1H), 2.18 (s, 3H), 2.08 (s, 3H). 13C NMR (125 MHz, CDCl3): δ 147.93, 140.86 (d, J =
2.5 Hz), 107.56, 84.75 (d, J = 199.6 Hz), 13.25, 10.96. 19F NMR (282 MHz, CDCl3): δ −162.39
(t, J = 54.3 Hz). IR (film, cm−1): (crude reaction mixture) 2971 (w), 2875 (w), 1566 (w), 1475
(w), 1412 (w), 1394 (w), 1202 (s), 1083 (s), 1032 (m), 1020 (m), 1007 (s), 826 (w), 783 (m), 751
(s), 697 (w). HRMS (ESI+): Calculated for C6H10FN2+ [M + H]+ : 129.0823; found: 129.0823.
(±)-methyl 4-(4-(1-fluoroethyl)-2-methoxy-5-nitrophenoxy)butanoate (2.31): Following the
general procedure, a 2-dram vial with a stirbar was charged sequentially with methyl 4-(4-(1-
97
hydroxyethyl)-2-methoxy-5-nitrophenoxy)butanoate 126 (313.3 mg, 1 mmol), 4-CF3PhSF
(251 mg, 1.1 equiv), THF (2 mL, 0.5 M), and BTMG (300 μL, 1.5 equiv). The reaction was
stirred at 600 rpm at room temperature for 24 hours, concentrated, taken up in minimal
dichloromethane, and purified by automated column chromatography (50 g silica, 0 → 20%
ethyl acetate in hexanes) to afford 187.9 mg product as a pale yellow solid (60% yield). 1H
NMR (500 MHz, CDCl3): δ 7.65 (s, 1H), 7.15 (s, 1H), 6.32 (dq, J = 48.4, 6.1 Hz, 1H), 4.13 (td,
J = 6.3, 2.7 Hz, 2H), 3.98 (s, 3H), 3.70 (s, 3H), 2.56 (t, J = 7.2 Hz, 2H), 2.19 (p, J = 6.7 Hz, 2H),
1.68 (dd, J = 24.3, 6.1 Hz, 3H). 13C NMR (125 MHz, CDCl3): δ 173.45, 154.55 (d, J = 1.8 Hz),
147.45 (d, J = 1.3 Hz), 138.45 (d, J = 3.9 Hz), 134.00 (d, J = 20.9 Hz), 109.00, 107.86 (d,
J = 16.6 Hz), 87.90 (d, J = 169.5 Hz), 68.35, 56.55, 51.89, 30.47, 24.36, 23.14 (d, J = 25.2 Hz).
19F NMR (282 MHz, CDCl3): δ −170.25 (dq, J = 48.6, 24.5 Hz). IR (ATR, cm−1): 3106 (w),
2956 (w), 2887 (w), 1726 (s), 1614 (w), 1577 (m), 1517 (sm), 1501 (s), 1469 (w), 1449 (m),
1433 (w), 1409 (w), 1373 (m), 1315 (m), 1279 (m), 1264 (s), 1221 (s), 1197 (m), 1176 (s), 1105
(m), 1075 (s), 1054 (w), 1035 (m), 1013 (s), 990 (w), 971 (m), 884 (s), 852 (m), 813 (s), 758 (m),
705 (w), 681 (w), 660 (w). HRMS (ESI+): Calculated for C14H18NO6+ [M − F]+ : 296.1129;
found: 296.1125.
1-((1R,2S)-1-fluoro-1-phenylpropan-2-yl)pyrrolidine (2.32): Following the general procedure,
a 2-dram vial with a stirbar was charged sequentially with (1R,2S)-1-phenyl-2-(1-pyrrolidinyl)-1-
propanol (205.3 mg, 1 mmol, Millipore-Sigma), a solution of PBSF (332 mg, 1.1 equiv) in THF
(2 mL, 0.5 M), and BTPP (460 μL, 1.5 equiv). The reaction was stirred at 600 rpm at room
temperature for 3 hours, concentrated, taken up in minimal dichloromethane, and purified by
126 Formed from methylation of 2-hydroxyethyl photolinker.
98
automated column chromatography (25 g silica, 0 → 25% ethyl acetate in hexanes with 3%
triethylamine) to afford 186.9 mg product (with stereoretention) as a colorless oil (90% yield).
1H NMR (500 MHz, CDCl3): δ 7.38 (t, J = 7.5 Hz, 2H), 7.35 – 7.27 (m, 3H), 5.81 (dd, J = 48.2,
2.3 Hz, 1H), 2.83 – 2.69 (m, 5H), 1.90 – 1.80 (m, 4H), 1.04 (dd, J = 6.7, 1.4 Hz, 3H). 13C NMR
(125 MHz, CDCl3): δ 139.30 (d, J = 20.4 Hz), 128.23 (d, J = 1.4 Hz), 127.61 (d, J = 0.5 Hz),
125.07 (d, J = 9.0 Hz), 94.39 (d, J = 179.9 Hz), 63.52 (d, J = 20.7 Hz), 51.18, 23.53, 10.13 (d, J
= 6.7 Hz). 19F NMR (282 MHz, CDCl3): δ −199.69 (dd, J = 48.2, 28.2 Hz). IR (film, cm−1):
3032 (w), 2967 (m), 2876 (w), 2791 (w), 1498 (w), 1451 (m), 1378 (w), 1351 (w), 1288 (w),
1200 (w), 1144 (w), 1091 (w), 1065 (w), 1031 (w), 1001 (m), 962 (m), 914 (w), 894 (w), 746 (s),
698 (s), 670 (m). HRMS (ESI+): Calculated for C13H19FN+ [M + H]+ : 208.1496; found:
208.1495.
(2E,6E)-1-fluoro-3,7,11-trimethyldodeca-2,6,10-triene (2.33): Following the general
procedure, a 2-dram vial with a stirbar was sequentially charged with trans,trans-farnesol
(222.4 mg, 1 mmol, Alfa Aesar), a solution of 4-CF3PhSF (251 mg, 1.1 equiv) in THF (2 mL,
0.5 M), and BTPP (460 μL, 1.5 equiv). The reaction was stirred at 600 rpm at room temperature
for 1 hour. Following addition of 1-fluoronaphthalene as an external standard, the yield was
determined by 19F NMR in CD2Cl2 to be 46% yield.127 Additionally, 4% yield of the branched
isomer (E)-3-fluoro-3,7,11-trimethyldodeca-1,6,10-triene was observed. 1H NMR (500 MHz,
CD2Cl2): δ 5.48 (q, J = 7.2 Hz, 1H), 5.15 – 5.06 (m, 2H), 4.88 (dd, J = 47.9, 7.2 Hz, 2H), 2.17 –
2.01 (m, 6H), 2.01 – 1.94 (m, 2H), 1.72 (dd, J = 4.7, 1.4 Hz, 3H), 1.67 (s, 3H), 1.60 (s, 6H).
13C NMR (125 MHz, CD2Cl2): δ 144.56 (d, J = 11.4 Hz), 135.92, 131.64, 124.66, 123.95,
127 Walkowiak, J.; Tomas-Szwaczyk, M.; Koroniak, H. J. Fluorine Chem. 2012, 143, 189.
99
119.36 (d, J = 17.0 Hz), 79.80 (d, J = 155.9 Hz), 40.09, 39.90 (d, J = 2.7 Hz), 27.11, 26.53 (d,
J = 3.6 Hz), 25.79, 17.76, 16.55 (d, J = 4.7 Hz), 16.08. 19F NMR (282 MHz, CD2Cl2): δ
−208.02 (tdq, J = 47.9, 9.6, 4.8 Hz).
3-fluorodec-1-ene (2.34): Following the general procedure, a 2-dram vial with a stirbar was
sequentially charged with 1-decen-3-ol (156.3 mg, 1 mmol, SAFC), THF (2 mL, 0.5 M),
4-CF3PhSF (251 mg, 1.1 equiv), and BTPP (460 μL, 1.5 equiv). The reaction was stirred at 600
rpm at room temperature for 24 hours. Following addition of 1-fluoronaphthalene as an external
standard, the yield was determined by 19F NMR in CDCl3 to be 39% yield.128 Additionally, 10%
yield of the linear isomer (E)-1-fluorodec-2-ene was observed. 1H NMR (500 MHz, CDCl3): δ
5.93 – 5.85 (m, 1H), 5.30 (ddd, J = 17.3, 3.4, 1.6 Hz, 1H), 5.20 (dd, J = 10.7, 1.4 Hz, 1H), 4.86
(ddd, J = 48.7, 12.6, 6.2 Hz, 1H), 1.78 – 1.57 (m, 2H), 1.47 – 1.17 (m, 10H), 0.94 – 0.79 (m,
3H). 13C NMR (125 MHz, CDCl3): δ 136.98 (d, J = 19.7 Hz), 116.82 (d, J = 11.9 Hz), 93.91 (d,
J = 166.6 Hz), 35.40 (d, J = 22.0 Hz), 31.95, 29.53, 29.36, 24.85 (d, J = 4.6 Hz), 22.84, 14.27.
19F NMR (282 MHz, CDCl3): δ −176.66 – −177.11 (m).
(±)-3-fluoro-3,7-dimethylocta-1,6-diene (2.35): Following the general procedure, a 2-dram vial
with a stirbar was sequentially charged with linalool (154.2 mg, 1 mmol, Alfa Aesar), a solution
of PBSF (332 mg, 1.1 equiv) in THF (2 mL, 0.5 M), and BTPP (460 μL, 1.5 equiv). The reaction
was stirred at 600 rpm at room temperature for 48 hours. Following addition of
1-fluoronaphthalene as an external standard, the yield was determined by 19F NMR in CDCl3 to
128 Chu, C. K.; Ziegler, D. T.; Carr, B.; Wickens, Z. K.; Grubbs, R. H. Angew. Chem., Int. Ed. 2016, 55, 8435.
100
be 9% yield. Additionally, 13% yield of the linear isomer (E)-1-fluoro-3,7-dimethylocta-2,6-
diene was observed along with 1% yield (Z)-1-fluoro-3,7-dimethylocta-2,6-diene.129 1H NMR
(500 MHz, CDCl3): δ 5.89 (td, J = 17.7, 11.0 Hz, 1H), 5.27 (d, J = 17.4 Hz, 1H), 5.15 – 5.07 (m,
2H), 2.14 – 1.97 (m, 2H), 1.74 – 1.64 (m, 5H), 1.61 (s, 3H), 1.42 (d, J = 21.5 Hz, 3H). 13C NMR
(125 MHz, CDCl3): δ 140.91 (d, J = 22.8 Hz), 132.08, 123.93, 113.35 (d, J = 11.4 Hz), 96.09 (d,
J = 169.8 Hz), 40.46 (d, J = 23.2 Hz), 25.81, 25.41 (d, J = 25.0 Hz), 22.45 (d, J = 4.9 Hz), 17.72.
19F NMR (282 MHz, CDCl3): δ −148.51 (dqd, J = 39.6, 21.7, 17.6 Hz).
(4-chlorophenyl)(3-(2-fluoroethyl)-5-methoxy-2-methyl-1H-indol-1-yl)methanone (2.36):
Following the general procedure, a 2-dram vial with a stirbar was charged sequentially with
(4-chlorophenyl)(3-(2-hydroxyethyl)-5-methoxy-2-methyl-1H-indol-1-yl)methanone 130
(343.8 mg, 1 mmol), 4-NsF (226 mg, 1.1 equiv), THF (2 mL, 0.5 M), and BTMG (300 μL,
1.5 equiv). The reaction was stirred at 600 rpm at room temperature for 24 hours, concentrated,
taken up in minimal dichloromethane, and purified by automated column chromatography (25 g
silica, 0 → 20% ethyl acetate in hexanes) to afford 295.8 mg product as a yellow solid
(86% yield). 1H NMR (500 MHz, CDCl3): δ 7.66 (d, J = 8.5 Hz, 2H), 7.47 (d, J = 8.5 Hz, 2H),
6.92 (d, J = 2.5 Hz, 1H), 6.88 (d, J = 9.0 Hz, 1H), 6.67 (dd, J = 9.0, 2.5 Hz, 1H), 4.61 (dt, J =
129 129 L’Heureux, A.; Beaulieu, F.; Bennett, C.; Bill, D. R.; Clayton, S.; LaFlamme, F.; Mirmehrabi, M.; Tadayon, S.; Tovell, D.; Couturier, M. J. Org. Chem. 2010, 75, 3401. Lee, E.; Yandulov, D. V.; J. Fluorine Chem. 2009, 130, 474. 130 Wey, S.-J.; Augustyniak, M. E.; Cochran, E. D.; Ellis, J. L.; Fang, X.; Garvey, D. S.; Janero, D. R.; Letts, L. G.; Martino, A. M.; Melim, T. L.; Murty, M. G.; Richardson, S. K.; Schroeder, J. D.; Selig, W. M.; Trocha, A. M.; Wexler, R. S.; Young, D. V.; Zemsteva, I. S.; Zifcak, B. M. J. Med. Chem. 2007, 50, 6367.
101
47.0, 6.8 Hz, 2H), 3.84 (s, 3H), 3.08 (dt, J = 21.3, 6.8 Hz, 2H), 2.36 (s, 3H). 13C NMR (125
MHz, CDCl3): δ 168.40, 156.11, 139.30, 135.79, 134.12, 131.26, 131.06, 130.98, 129.23,
115.16, 114.60 (d, J = 7.8 Hz), 111.45, 101.19, 82.83 (d, J = 170.2 Hz), 55.86, 25.66 (d, J = 21.8
Hz), 13.38. 19F NMR (282 MHz, CDCl3): δ −213.49 (tt, J = 47.0, 21.4 Hz). IR (ATR, cm−1):
3105 (w), 3071 (w), 3038 (w), 2996 (w), 2954 (w), 2892 (w), 2834 (w), 1892 (w), 1673 (s), 1619
(w), 1597 (m), 1478 (w), 1465 (s), 1433 (m), 1402 (w), 1353 (s), 1329 (m), 1289 (m), 1251 (w),
1230 (m), 1211 (s), 1156 (m), 1142 (w), 1085 (m), 1064 (m), 1035 (s), 1014 (m), 1007 (m), 978
(m), 953 (w), 938 (m), 862 (m), 848 (s), 827 (w), 802 (s), 755 (s), 734 (w), 715 (w), 692 (w), 672
(w). HRMS (ESI+): Calculated for C19H18ClFNO2+ [M + H]+ : 346.1005; found: 346.1001.
1-(2-fluoroethyl)-2-methyl-5-nitro-1H-imidazole (2.37): Following the general procedure, a
2-dram vial with a stirbar was charged sequentially with metronidazole (171.2 mg, 1 mmol,
Millipore-Sigma), a solution of PBSF (332 mg, 1.1 equiv) in THF (2 mL, 0.5 M), and BTMG
(220 μL, 1.1 equiv). The reaction was stirred at 600 rpm at room temperature for 30 minutes,
concentrated, taken up in minimal dichloromethane, and purified by automated column
chromatography (50 g silica, 0 → 10% methanol in dichloromethane) to afford 166.6 mg of a
colorless oil containing 152.9 mg of the title compound 131 (88% yield) and 13.7 mg of
elimination side product 2-methyl-5-nitro-1-vinyl-1H-imidazole. 1H NMR (500 MHz, CDCl3):
δ 7.97 (s, 1H), 4.75 (ddd, J = 47.0, 5.0, 4.0 Hz, 2H), 4.61 (dt, J = 26.0, 4.4 Hz, 2H), 2.50 (s, 3H).
13C NMR (125 MHz, CDCl3): δ 151.74, 138.30, 133.40, 82.24 (d, J = 171.6 Hz), 46.85 (d,
131 Goldberg, N. W.; Shen, X.; Li, J.; Ritter, T. Org. Lett. 2016, 18, 6102.
102
J = 20.0 Hz), 14.47 (d, J = 3.3 Hz). 19F NMR (282 MHz, CDCl3): δ −224.07 (tt, J = 47.3,
26.1 Hz).
3-((3S,8R,9S,10R,13S,14S)-3-fluoro-10,13-dimethyl-2,3,4,7,8,9,10,11,12,13,14,15-dodeca-
hydro-1H-cyclopenta[a]phenanthren-17-yl)pyridine (2.38): Following the general procedure,
a 1-dram vial with a stirbar was sequentially charged with abiraterone (174.8 mg, 0.5 mmol), a
solution of PBSF (166 mg, 1.1 equiv) in THF (1 mL, 0.5 M), and MTBD (108 μL, 1.5 equiv).
The reaction was stirred at 600 rpm at room temperature for 30 minutes. The reaction mixture
was concentrated, taken up in minimal dichloromethane, and purified by automated column
chromatography (25 g triethylamine-pretreated silica, 0 → 15% ethyl acetate in hexanes)
affording 49.2 mg of a white solid consisting of 22.5 mg of the title compound (13% yield, 3.9:1
dr favoring retention of configuration) and 25.9 mg of the homoallylic rearrangement isomer, 3-
((1aR,3aR,3bS,5aS,8aS,8bR,10R,10aR)-10-fluoro-3a,5a-dimethyl-1,1a,2,3,3a,3b,4,5,5a,8,8a,8b,9,
10-tetradecahydro-cyclopenta[a]cyclopropa [2,3]cyclopenta[1,2-f]naphthalen-6-yl) pyridine.
1H NMR (500 MHz, CDCl3): δ 8.62 (s, 1H), 8.46 (s, 1H), 7.64 (dq, J = 7.9, 1.8 Hz, 1H), 7.22
(dd, J = 8.0, 4.8 Hz, 1H), 5.99 (td, J = 3.3, 1.8 Hz, 1H), 5.43 (dd, J = 4.8, 2.8 Hz, 1H), 4.40 (dm,
J = 50.4 Hz, 1H), 2.49 – 2.42 (m, 1H), 2.26 (dddd, J = 15.8, 6.6, 3.4, 1.6 Hz, 1H), 2.15 – 1.96
(m, 4H), 1.82 – 1.72 (m, 2H), 1.72 – 1.52 (m, 6H), 1.48 (tdd, J = 12.1, 5.1, 3.0 Hz, 1H), 1.09 (s,
3H), 1.08 – 1.04 (m, 4H), 0.93 (ddd, J = 13.8, 11.8, 7.9 Hz, 1H). 13C NMR (125 MHz, CDCl3):
δ 151.77, 148.03, 148.00, 139.84 (d, J = 12.6 Hz), 133.81, 133.07, 129.36, 123.19, 122.82 (d,
J = 1.1 Hz), 92.89 (d, J = 174.0 Hz), 57.61, 50.33 (d, J = 1.7 Hz), 47.46, 39.54 (d, J = 19.3 Hz),
103
36.86 (d, J = 1.2 Hz), 35.33, 33.07, 31.94, 31.67, 30.53, 28.88 (d, J = 17.7 Hz), 21.03, 19.40,
16.72. 19F NMR (282 MHz, CDCl3): δ −167.96 (dm, J = 50.2 Hz). IR (ATR, cm−1): (1:1.2
mixture of product (3.9:1 dr):homoallylic rearrangement isomer): 3038 (w), 3010 (w), 2932 (m),
2856 (w), 1716 (w), 1599 (w), 1561 (w), 1474 (w), 1462 (w), 1442 (w), 1408 (w), 1399 (w),
1375 (m), 1319 (w), 1296 (w), 1280 (w), 1245 (w), 1192 (w), 1160 (w), 1128 (w), 1104 (w),
1080 (w), 1062 (w), 1050 (w), 1006 (s), 973 (m), 947 (m), 924 (w), 914 (w), 887 (w), 869 (m),
841 (w), 825 (w), 799 (s), 730 (w), 711 (s), 678 (w). HRMS (ESI+): Calculated for C24H31FN+
[M + H]+: 352.2435; found: 352.2440.
(±)-3-fluorodihydrofuran-2(3H)-one (2.39): According to the general procedure, a 2-dram vial
with a stirbar was charged sequentially with α-hydroxy-γ-butyrolactone (102.1 mg, 1 mmol,
Millipore-Sigma), a solution of PBSF (332 mg, 1.1 equiv) in THF (2 mL, 0.5 M), and MTBD
(215 μL, 1.5 equiv). The reaction was stirred at 600 rpm at room temperature for 30 minutes,
concentrated, taken up in minimal dichloromethane, and purified by automated column
chromatography (25 g silica, 0 → 40% ethyl acetate in hexanes) to afford 81.4 mg product of a
colorless oil (78% yield).132 1H NMR (500 MHz, CDCl3): δ 5.08 (dt, J = 51.2, 7.7 Hz, 1H), 4.38
(td, J = 8.9, 3.9 Hz, 1H), 4.20 (q, J = 8.2 Hz, 1H), 2.57 (ttd, J = 14.1, 7.3, 3.8 Hz, 1H), 2.40 (ddq,
J = 21.8, 13.4, 8.1 Hz, 1H). 13C NMR (125 MHz, CDCl3): δ 171.82 (d, J = 20.8 Hz), 85.34 (d,
J = 189.7 Hz), 64.97 (d, J = 5.9 Hz), 29.51 (d, J = 20.1 Hz). 19F NMR (282 MHz, CDCl3):
δ −195.68 (ddd, J = 51.0, 22.9, 13.8 Hz).
132 Sander, K.; Galante, E.; Gendron, T.; Yiannaki, E.; Patel, N.; Kalber, T. L.; Badar, A.; Robson, M.; Johnson, S. P.; Bauer, F.; Mairinger, S.; Stanek, J.; Wanek, T.; Kuntner, C.; Kottke, T.; Weizel, L.; Dickens, D.; Erlandsson, K.; Hutton, B. F.; Lythgoe, M. F.; Stark, H.; Langer, O.; Koepp, M.; Arstad, E. J. Med. Chem. 2015, 58, 6058.
104
(8S,9S,10R,13S,14S,17R)-17-(2-fluoroacetyl)-17-hydroxy-10,13-dimethyl-7,8,9,10,12,13,14,
15,16,17-decahydro-3H-cyclopenta[a]phenanthrene-3,11(6H)-dione (2.40): Following the
general procedure, a 2-dram vial with a stirbar was charged sequentially with prednisone (358.4
mg, 1 mmol, Millipore-Sigma), 4-CF3PhSF (251 mg, 1.1 equiv), THF (2 mL, 0.5 M), and BTPP
(380 μL, 1.25 equiv). The reaction was stirred at 600 rpm at room temperature for 3 hours,
concentrated, taken up in minimal dichloromethane, and purified by automated column
chromatography (25 g silica, 10 → 60% ethyl acetate in hexanes) to afford 229.7 mg of a white
solid consisting of 170.2 mg of the title compound (47% yield), and 59.5 mg of the unconverted
sulfonate ester intermediate 2-((8S, 9S,10R,13S,14S,17R)-17-hydroxy-10,13-dimethyl-3,11-
dioxo-6,7,8,9,10,11,12,13,14,15,16,17-dodecahydro-3H-cyclopenta[a]phenanthren-17-yl)-2-
oxoethyl 4-(trifluoromethyl)benzenesulfon-ate. 1H NMR (500 MHz, CDCl3): δ 7.59 (d, J = 10.2
Hz, 1H), 6.10 (dd, J = 10.2, 1.9 Hz, 1H), 6.01 (t, J = 1.8 Hz, 1H), 5.82 (s, 1H), 5.39 (dd, J = 47.6,
17.3 Hz, 1H), 5.10 (dd, J = 47.2, 17.3 Hz, 1H), 2.86 (d, J = 12.4 Hz, 1H), 2.58 – 2.45 (m, 2H),
2.39 – 2.27 (m, 2H), 2.18 (d, J = 11.2 Hz, 1H), 2.08 (s, 1H), 2.05 – 1.94 (m, 2H), 1.82 – 1.72 (m,
1H), 1.67 (ddd, J = 15.1, 9.6, 5.7 Hz, 1H), 1.43 – 1.36 (m, 1H), 1.35 (s, 3H), 1.25 – 1.18 (m, 1H),
0.53 (s, 3H). 13C NMR (125 MHz, CDCl3): δ 210.01, 206.30 (d, J = 12.4 Hz), 185.09, 167.18,
155.08, 127.04, 123.80, 87.41, 84.94 (d, J = 176.3 Hz), 58.81, 50.74, 49.53, 48.83, 41.95, 35.51,
33.84, 33.15, 31.55, 22.84, 18.76, 15.43. 19F NMR (282 MHz, CDCl3): δ −231.16 (t, J = 47.4
Hz). IR (ATR, cm−1): (4.5:1 mixture of product:sulfonate ester) 3383 (br, s), 2941 (m), 1730
(m), 1704 (m), 1656 (s), 1616 (m), 1598 (w), 1445 (w), 1373 (w), 1354 (w), 1324 (w), 1301 (w),
1243 (m), 1185 (m), 1136 (m), 1116 (w), 1088 (m), 1063 (w), 1043 (s), 943 (w), 899 (s), 821 (s),
105
774 (w), 750 (w), 727 (w), 709 (w), 691 (m). HRMS (ESI+): Calculated for C21H26FO4+ [M +
H]+ : 361.1810; found: 361.1811.
methyl (R)-3-fluoro-2-(tritylamino)propanoate (2.41): According to the general procedure, a
20-mL scintillation vial with a stirbar was charged sequentially with N-trityl-L-serine methyl
ester (361.4 mg, 1 mmol, TCI), a solution of PBSF (332 mg, 1.1 equiv) in THF (10 mL, 0.1 M),
and BTPP (460 μL, 1.5 equiv). The reaction was stirred at 600 rpm at room temperature for
3 hours, concentrated, taken up in minimal dichloromethane, and purified by automated column
chromatography (25 g silica, 0 → 15% ethyl acetate in hexanes), affording 294.3 mg of a white
foamy solid containing 87.9 mg product (24% yield) and 206.4 mg of the methyl (S)-1-
tritylaziridine-2-carboxylate. 1H NMR (500 MHz, CDCl3): δ 7.56 (d, J = 7.5 Hz, 6H), 7.32 (dt,
J = 7.8, 5.7 Hz, 6H), 7.29 – 7.22 (m, 3H), 4.68 (ddd, J = 47.0, 8.8, 4.7 Hz, 1H), 4.50 (ddd, J =
47.1, 8.8, 6.1 Hz, 1H), 3.67 (dddd, J = 19.4, 10.6, 6.1, 4.7 Hz, 1H), 3.29 (s, 3H), 2.89 (d, J = 10.3
Hz, 1H). 13C NMR (125 MHz, CDCl3): δ 172.79 (d, J = 3.5 Hz), 145.61, 128.79, 128.07,
126.72, 85.01 (d, J = 176.7 Hz), 71.02, 56.70 (d, J = 21.1 Hz), 52.19. 19F NMR (282 MHz,
CDCl3): δ −224.37 (td, J = 47.0, 19.4 Hz). IR (ATR, cm−1): (1:2.5 mixture of product:aziridine)
3057 (w), 3028 (w), 2951 (w), 1740 (s), 1596 (w), 1489 (m), 1446 (m), 1392 (w), 1285 (w),
1241 (w), 1198 (s), 1177 (s), 1080 (m), 1032 (m), 1014 (w), 972 (w), 903 (w), 839 (w), 776 (w),
745 (s), 705 (s), 696 (w). HRMS (ESI+): Calculated for C4H9FNO2+ [M − C(C6H5)3 + 2H]+ :
122.0612; found: 122.0611.
106
(3R,5aS,6R,9R,10S,12S,12aR)-10-fluoro-3,6,9-trimethyldecahydro-12H-3,12-epoxy[1,2]
dioxepino[4,3-i]isochromene (2.42): Following the general procedure, a 1-dram vial with a
stirbar was charged sequentially with dihydroartemisinin (284.3 mg, 1.0 mmol, TCI), 4-NsF
(226 mg, 1.1 equiv), THF (2 mL, 0.5 M), and BTMG (300 μL, 1.5 equiv). The reaction was
stirred at 600 rpm at room temperature for 30 minutes. Following addition of
1-fluoronaphthalene as an external standard, the yield was determined by 19F NMR to be
54% yield with a 3.8:1 dr.133 1H NMR (500 MHz, C6D6): δ 5.10 (s, 1H), 5.02 (dd, J = 53.9, 9.0
Hz, 1H), 2.72 (dtdd, J = 14.0, 11.6, 8.0, 5.9 Hz, 1H), 2.30 (ddd, J = 14.4, 13.4, 4.0 Hz, 1H), 1.69
(ddd, J = 14.3, 5.0, 3.0 Hz, 1H), 1.55 – 1.49 (m, 1H), 1.35 (s, 3H), 1.31 – 1.12 (m, 4H), 1.04 –
0.95 (m, 1H), 0.93 – 0.84 (m, 1H), 0.80 – 0.73 (m, 1H), 0.71 – 0.66 (m, 6H), 0.53 – 0.47 (m,
1H). 13C NMR (125 MHz, C6D6): δ 108.68 (d, J = 208.3 Hz), 104.50, 91.39 (d, J = 6.2 Hz),
79.70, 51.44 (d, J = 0.5 Hz), 45.21 (d, J = 9.4 Hz), 36.98, 36.53, 34.24, 33.10 (d, J = 19.2 Hz),
25.93, 25.04, 21.85, 20.19, 11.71. 19F NMR (282 MHz, C6D6): δ −141.58 (ddd, J = 53.9, 10.7,
4.7 Hz).
Random forest predictive model:
The procedure used to prepare data and train the model is described below:
1) The following programs and files were downloaded from the indicated source and installed:
- R (<https://cran.r-project.org/mirrors.html>)
- R Studio (<https://www.rstudio.com/products/rstudio/download/>)
133 Woo, S. H.; Parker, M. H.; Ploypradith, P.; Northrop, J.; Posner, G. H. Tetrahedron Lett. 1998, 39, 1533. Guar, R.; Cheema, H. S.; Kumar, Y.; Singh, S. P.; Yadav, D. K.; Darokar, M. P.; Khan, F.; Bhakuni, R. S. RSC Adv. 2015, 5, 47959.
107
- The “rxnpredict” root directory containing “rxnpredict.R” (code shown below) and two
directories named “R input” and “R output”.
2) The descriptors described in Figure 2.22 were saved in the file
“rxnpredict\R_input\descriptor_table.csv”. This file includes 23 columns corresponding to each
of the descriptors, a header row containing the descriptor titles, and 640 rows corresponding to
each reaction in the following order:
1 – alcohol 2.13-OH, base DBU, sulfonyl fluoride 4-ClPhSF 2 – alcohol 2.13-OH, base DBU, sulfonyl fluoride PyFluor 3 – alcohol 2.13-OH, base DBU, sulfonyl fluoride 4-CF3PhSF 4 – alcohol 2.13-OH, base DBU, sulfonyl fluoride 4-NsF 5 – alcohol 2.13-OH, base DBU, sulfonyl fluoride PBSF 6 – alcohol 2.13-OH, base MTBD, sulfonyl fluoride 4-ClPhSF 7 – alcohol 2.13-OH, base MTBD, sulfonyl fluoride PyFluor 8 – alcohol 2.13-OH, base MTBD, sulfonyl fluoride 4-CF3PhSF 9 – alcohol 2.13-OH, base MTBD, sulfonyl fluoride 4-NsF 10 – alcohol 2.13-OH, base MTBD, sulfonyl fluoride PBSF 11 – alcohol 2.13-OH, base BTMG, sulfonyl fluoride 4-ClPhSF 12 – alcohol 2.13-OH, base BTMG, sulfonyl fluoride PyFluor 13 – alcohol 2.13-OH, base BTMG, sulfonyl fluoride 4-CF3PhSF 14 – alcohol 2.13-OH, base BTMG, sulfonyl fluoride 4-NsF 15 – alcohol 2.13-OH, base BTMG, sulfonyl fluoride PBSF 16 – alcohol 2.13-OH, base BTPP, sulfonyl fluoride 4-ClPhSF 17 – alcohol 2.13-OH, base BTPP, sulfonyl fluoride PyFluor 18 – alcohol 2.13-OH, base BTPP, sulfonyl fluoride 4-CF3PhSF 19 – alcohol 2.13-OH, base BTPP, sulfonyl fluoride 4-NsF 20 – alcohol 2.13-OH, base BTPP, sulfonyl fluoride PBSF 21 – alcohol 2.13-OH, base DBU, sulfonyl fluoride 4-ClPhSF […] 640 – alcohol 2.42-OH, base BTPP, sulfonyl fluoride PBSF
Alcohols were listed in the following order: 2.13-OH, 2.1-OH, 2.19-OH, 2.14-OH, 2.18-OH, 2.15-OH, 2.16-OH, 2.17-OH, 2.12-OH, 2.21-OH, 2.10-OH, 2.22-OH, 2.20-OH, 2.24-OH, 2.25-OH, 2.23-OH, 2.27-OH, 2.28-OH, 2.29-OH, 2.30-OH, 2.31-OH, 2.32-OH, 2.33-OH, 2.34-OH, 2.35-OH, 2.36-OH, 2.37-OH, 2.38-OH, 2.39-OH, 2.40-OH, 2.41-OH, 2.42-OH. 3) Descriptors for an external test set corresponding to alcohols 2.43-OH — 2.47-OH were
generated as in Step 2 and saved in “rxnpredict\R_input\ descriptor_table_external_set.csv”.
4) The experimental reaction yields for alcohols 2.1-OH, 2.10-OH, 2.12-OH – 2.25-OH and
2.27-OH – 2.42-OH were saved in the file ““rxnpredict\R_input\observed_yields.csv” as a
single column corresponding to the rows in “rxnpredict\R_input\descriptor_table.csv” as shown
108
above. The experimental yields for alcohols 2.43-OH — 2.47-OH are included in
““rxnpredict\R_input\external_set_observed_yields.csv”.
5) The “rxnpredict\rxnpredict.R” R script was opened in R Studio.
6) The “rxnpredict\rxnpredict.R” script was executed in R studio. The script with documentation
is as follows:
# Install packages (if necessary) and load them. if (!require("pacman")) install.packages("pacman") pacman::p_load(ggplot2, caret, ModelMetrics, scales) # Set the working directory to the location of the rxnpredict folder. setwd("C:\\Users\\matt\\Desktop\\rxnpredict") # ============================================================================ # Load descriptor and yield data and prepare data for modeling. # ============================================================================ # Load user-created table containing reaction descriptors. descriptor.table <- read.csv("R_input\\descriptor_table.csv", header=TRUE) # Scale the descriptor data. Scale parameters are saved in descriptor.data. descriptor.data <- scale(descriptor.table) descriptor.scaled <- as.data.frame(descriptor.data) # Load user-created yield data. yield.data <- as.numeric(unlist(read.csv("R_input\\observed_yields.csv", header=FALSE, stringsAsFactors=FALSE))) # Append the yield data to the descriptor table. descriptor.scaled$yield <- yield.data # ============================================================================ # Split data and train random forest model. # ============================================================================ # Split into training and test set (70/30). set.seed(1751) size <- round(0.70*nrow(descriptor.scaled)) training <- sample(nrow(descriptor.scaled), size=size, replace=FALSE) training.scaled <- descriptor.scaled[training,] test.scaled <- descriptor.scaled[-training,] # 10-fold cross-validation. train_control <- trainControl(method="cv", number=10, savePredictions=TRUE) # Train the random forest model. rfFit <- train(yield ~ ., data=training.scaled, trControl=train_control, method="rf", importance=TRUE) # Save the trained random forest model. saveRDS(rfFit, "R_output\\rfFit.rds") # ============================================================================ # Calculate R^2 and RMSE using test set and generate calibration plot. # ============================================================================ # Predict yields for test set. rf.pred <- predict(rfFit, test.scaled)
109
# Generate *.csv showing predicted and observed yields for the test set (saves to test_set_predicted_yields.csv). predicted.yields <- as.data.frame(rf.pred) predicted.yields$rf.pred <- round(predicted.yields$rf.pred, digits=1) predicted.yields$yield <- test.scaled$yield predicted.yields["Error"] <- predicted.yields$yield-predicted.yields$rf.pred names(predicted.yields)[names(predicted.yields) == 'rf.pred'] <- 'Predicted Yield' names(predicted.yields)[names(predicted.yields) == 'yield'] <- 'Observed Yield' write.csv(predicted.yields, 'R_output\\test_set_predicted_yields.csv') # Calculate R^2 and RMS error for test set. rf.r2 <- cor(rf.pred, test.scaled$yield)^2 rf.rmse <- rmse(rf.pred, test.scaled$yield) # Calculate R^2 and RMS error for training set (included to compare accuracy of training vs. test sets). rftrain.pred <- predict(rfFit, training.scaled) rftrain.r2 <- cor(rftrain.pred, training.scaled$yield)^2 rftrain.rmse <- rmse(rftrain.pred, training.scaled$yield) # Generate calibration plot of test set (saves to test_set-calibration_plot.png). df <- data.frame(x = rf.pred, y = test.scaled$yield) rsq <- paste(round(rf.r2, digits = 3)) rms <- paste(round(rf.rmse, digits = 1)) p1 <- ggplot(df, aes(x = x, y = y)) + geom_point(alpha = 0.4) + scale_x_continuous(breaks = seq(0,100,25), lim=c(0, 100)) + labs(x='Predicted Yield', y='Observed Yield', caption = bquote(R^2 ~ " = " ~ .(rsq) * "; RMS error = " ~ .(rms)
* "%")) + theme(plot.caption = element_text(hjust = 0.5, size = 8)) + geom_segment(aes(x=0,xend=100,y=0,yend=100), linetype="dashed") ggsave(file="R_output\\test_set-calibration_plot.png", width=5, height=4) # ============================================================================ # Create Variable importance plot. # ============================================================================ # Read in variable importance from trained rf model. rf_imp <- importance(rfFit$finalModel) rf.imp.df <- cbind(as.data.frame(rf_imp), names(rf_imp[, 1])) colnames(rf.imp.df)[1] <- "IncMSE" colnames(rf.imp.df)[3] <- "descriptor" # For descriptor names, replace "_" with " " and "." with "*". rf.imp.df$descriptor <- gsub("_", " ", rf.imp.df$descriptor) rf.imp.df$descriptor <- gsub("[.]", "*", rf.imp.df$descriptor) # Capitalize descriptor names. simpleCap <- function(x) { s <- strsplit(x, " ")[[1]] paste(toupper(substring(s, 1, 1)), substring(s, 2), sep="", collapse=" ") } rf.imp.df$descriptor <- sapply(rf.imp.df$descriptor, simpleCap) # Plot variable importance (saves to variable_importance_plot.png). # USER: change '10' on next line to modify minimum percentage cutoff for IncMSE. p2 <- ggplot(rf.imp.df[rf.imp.df$IncMSE>10, ], aes(x=reorder(descriptor, IncMSE), y=IncMSE)) + geom_bar(stat="identity") + scale_y_continuous(labels = comma) + labs(x="", y="Increase in Mean Squared Error (%)") + coord_flip() # USER: change 'width' and 'height' parameter on next line to control plot dimensions. ggsave(file="R_output\\variable_importance_plot.png", width=8, height=4) # ============================================================================ # Load descriptors and predict yields for external test set. # ============================================================================ # Load external test set.
110
externalset.table <- read.csv("R_input\\descriptor_table_external_set.csv", header=TRUE) # Scale the external set data using the same scaling as for the training and test sets. externalset.data <-
scale(externalset.table,attr(descriptor.data,"scaled:center"),attr(descriptor.data,"scaled:scale"))
externalset.scaled <- as.data.frame(externalset.data) # Predict yields for external test set. rf.externalset <- predict(rfFit, externalset.scaled) # Create table with predicted yields for external test set (saves to external_set_predicted_yields.csv). externalset.predictedyields <- as.data.frame(rf.externalset) externalset.predictedyields $rf.externalset <- round(externalset.predictedyields $rf.externalset, digits=1) names(externalset.predictedyields )[names(externalset.predictedyields ) == 'rf.externalset'] <- 'Predicted Yield' write.csv(externalset.predictedyields , 'R_output\\external_set_predicted_yields.csv') # ============================================================================ # Calibration plots for external test substrates. # ============================================================================ # Load external set observed yields. external_obs <- as.numeric(unlist(read.csv("R_input\\external_set_observed_yields.csv", header=FALSE,
stringsAsFactors=FALSE))) # Store predicted yields for external substrates. external_pred <- c(rf.externalset) # Generate calibration plot for external substrates (saves to external-calibration_plot.png). ex.r2 <- cor(external_pred, external_obs)^2 ex.rmse <- rmse(external_pred, external_obs) alcohol <- rep(c("1ag", "1ah", "1ai", "1aj","1ak"), times = c(20,20,20,20,20)) ex.df <- data.frame(x = external_pred, y = external_obs, substrate = alcohol) exrsq <- paste(round(ex.r2, digits = 3)) exrms <- paste(round(ex.rmse, digits = 1)) ex.p1 <- ggplot(ex.df, aes(x = x, y = y, color = substrate)) + scale_color_manual(values=c("red", "darkorange1", "blue", "darkgreen", "purple3", "black")) + geom_point(alpha = 0.6) + scale_x_continuous(breaks = seq(0,100,25), lim=c(0, 100)) + labs(x='Predicted Yield', y='Observed Yield', caption = bquote(R^2 ~ " = " ~ .(exrsq) * "; RMS error = " ~
.(exrms) * "%")) + theme(plot.caption = element_text(hjust = 0.5, size = 8), legend.position="none") + geom_segment(aes(x=0,xend=100,y=0,yend=100,color="black"), linetype="dashed") ggsave(file="R_output\\external-calibration_plot.png", width=5, height=4) # Calculate R^2 and RMSE for external test substrates. ex1.r2 <- cor(external_pred[1:20], external_obs[1:20])^2 ex1.rmse <- rmse(external_pred[1:20], external_obs[1:20]) ex2.r2 <- cor(external_pred[21:40], external_obs[21:40])^2 ex2.rmse <- rmse(external_pred[21:40], external_obs[21:40]) ex3.r2 <- cor(external_pred[41:60], external_obs[41:60])^2 ex3.rmse <- rmse(external_pred[41:60], external_obs[41:60]) ex4.r2 <- cor(external_pred[61:80], external_obs[61:80])^2 ex4.rmse <- rmse(external_pred[61:80], external_obs[61:80]) ex5.r2 <- cor(external_pred[81:100], external_obs[81:100])^2 ex5.rmse <- rmse(external_pred[81:100], external_obs[81:100]) # Generate table containing R^2 and RMSE for external test substrates (saves to external_set_stats.csv). externalsetstats <-
matrix(c(1,ex1.r2,ex1.rmse,2,ex2.r2,ex2.rmse,3,ex3.r2,ex3.rmse,4,ex4.r2,ex4.rmse,5,ex5.r2,ex5.rmse),ncol=3,byrow=TRUE)
colnames(externalsetstats) <- c("Ext. Substrate","R^2","RMS Error") externalsetstats <- as.table(externalsetstats) write.csv(externalsetstats , 'R_output\\external_set_stats.csv') # ============================================================================ # Combined calibration plot (used to generate Figure 3 in manuscript).
111
# ============================================================================ # Generate calibration plot of test set and external substrates (saves to combined-calibration_plot.png). combined_obs <- c(external_obs,test.scaled$yield) combined_pred <- c(external_pred,rf.pred) combined_alcohol <- rep(c("1ag", "1ah", "1ai", "1aj","1ak","test set"), times = c(20,20,20,20,20,192)) combined.df <- data.frame(x = combined_pred, y = combined_obs, substrate = combined_alcohol) combined.p1 <- ggplot(combined.df, aes(x = x, y = y, color = substrate, shape = substrate, fill = substrate)) + scale_color_manual(values=c("#ff3d3d", "#fc9f28", "#415ce2", "#1c9102", "#8841a8", "black", "black")) + scale_shape_manual(values=c(22, 24, 21, 25, 23, 20)) + scale_fill_manual(values=c("#ff3d3d", "#fc9f28", "#415ce2", "#1c9102", "#8841a8", "black")) + geom_point(alpha = 0.7) + theme_bw() + scale_x_continuous(breaks = seq(0,100,25), lim=c(0, 100)) + labs(x='Predicted Yield', y='Observed Yield', caption = bquote("Test set:" ~ R^2 ~ " = " ~ .(rsq) * "; RMSE = "
~ .(rms) * "%")) + theme(plot.caption = element_text(hjust = 0.5, size = 8), legend.position="none") + geom_segment(aes(x=0,xend=100,y=0,yend=100,color="black"), linetype="dashed") ggsave(file="R_output\\combined-calibration_plot.png", width=4, height=3.2, dpi=600)
External validation substrates:
6-(3-fluoropropyl)-[1,2,4]triazolo[1,5-a]pyrimidine (2.43): 3-[1,2,4]Triazolo[1,5-a]pyrimidin-
6-ylpropan-1-ol (Millipore-Sigma) was evaluated according to the general procedure. A sample
for characterization was obtained by subjecting the crude reaction mixtures to automated column
chromatography (25 g silica, 30 → 70% ethyl acetate in hexanes), which afforded the title
compound as a white solid. 1H NMR (500 MHz, CDCl3): δ 8.74 (s, 1H), 8.70 (s, 1H), 8.48 (s,
1H), 4.54 (dt, J = 47.1, 5.6 Hz, 2H), 2.94 (t, J = 7.7 Hz, 2H), 2.10 (dm, J = 27.6 Hz, 2H). 13C
NMR (125 MHz, CDCl3): δ 156.62, 156.28, 154.53, 134.05, 123.92, 82.26 (d, J = 166.7 Hz),
31.33 (d, J = 20.1 Hz), 25.85 (d, J = 4.7 Hz). 19F NMR (282 MHz, CDCl3): δ −221.05 (tt, J =
47.3, 26.4 Hz). IR (ATR, cm−1): 3104 (w), 3063 (w), 2966 (w), 2924 (w), 2904 (w), 1891 (w),
1626 (m), 1569 (w), 1537 (m), 1511 (s), 1449 (w), 1439 (m), 1425 (w), 1391 (m), 1361 (m),
1319 (m), 1288 (w), 1268 (s), 1250 (s), 1236 (w), 1221 (w), 1176 (s), 1124 (m), 1094 (w), 1067
(w), 1021 (s), 950 (m), 923 (m), 889 (s), 843 (m), 785 (s), 757 (w), 740 (w), 666 (s). HRMS
(ESI+): Calculated for C8H10FN4+ [M + H]+: 181.0884; found: 181.0884.
112
3-(fluoromethyl)-5-(4-fluorophenyl)isoxazole (2.44): 5-(4-Fluorophenyl)isoxazole-3-methanol
(Millipore-Sigma) was evaluated according to the general procedure. A sample for
characterization was obtained by subjecting the crude reaction mixtures to automated column
chromatography (25 g silica, 0 → 15% ethyl acetate in hexanes), which afforded the title
compound as a white solid. 1H NMR (500 MHz, CDCl3): δ 7.81 – 7.72 (m, 2H), 7.20 – 7.12 (m,
2H), 6.59 (s, 1H), 5.49 (d, J = 46.9 Hz, 2H). 13C NMR (125 MHz, CDCl3): δ 169.98, 164.00 (d,
J = 251.6 Hz), 160.54 (d, J = 23.1 Hz), 128.08 (d, J = 8.6 Hz), 123.50 (d, J = 3.4 Hz), 116.41 (d,
J = 22.2 Hz), 98.36, 76.19 (d, J = 166.8 Hz). 19F NMR (282 MHz, CDCl3): δ −108.96 (tt, J =
8.5, 5.2 Hz, 1F), −221.36 (t, J = 46.8 Hz, 1F). IR (ATR, cm−1): 3120 (w), 2973 (w), 1915 (w),
1619 (m), 1601 (m), 1513 (m), 1462 (m), 1439 (m), 1371 (w), 1306 (w), 1236 (m), 1186 (w),
1161 (m), 1095 (w), 1047 (w), 1035 (s), 992 (s), 949 (w), 909 (m), 844 (s), 817 (s), 754 (s), 724
(w), 679 (m). HRMS (ESI+): Calculated for C10H8F2NO+ [M + H]+: 196.0568; found: 196.0565.
2-fluoro-2,3-dihydro-1H-indene (2.45): 2-Indanol (Millipore-Sigma) was evaluated according
to the general procedure. A sample for characterization was obtained by subjecting the crude
reaction mixtures to automated column chromatography (25 g silica, 0 → 25% ethyl acetate in
hexanes), which afforded the title compound as a volatile colorless oil.134 1H NMR (500 MHz,
CDCl3): δ 7.31 – 7.26 (m, 2H), 7.24 – 7.17 (m, 2H), 5.58 – 5.41 (m, 1H), 3.32 – 3.23 (m, 2H),
3.22 – 3.18 (m, 2H). 13C NMR (125 MHz, CDCl3): δ 140.11, 127.00, 124.97, 94.85 (d, J =
176.7 Hz), 40.70 (d, J = 23.1 Hz). 19F NMR (282 MHz, CDCl3): δ −173.84 (dtt, J = 53.3, 31.6,
28.7 Hz).
134 Ventre, S.; Petronijevic, F. R.; MacMillan, D. W. C. J. Am. Chem. Soc. 2015, 137, 5654.
113
(±)-(2-fluoropropoxy)benzene (2.46): 1-Phenoxy-2-propanol (TCI) was evaluated according to
the general procedure. A sample for characterization was obtained by subjecting the crude
reaction mixtures to automated column chromatography (25 g silica, 0 → 10% ethyl acetate in
hexanes), which afforded the title compound as a colorless oil.135 1H NMR (500 MHz, CDCl3):
δ 7.34 – 7.27 (m, 2H), 6.98 (t, J = 7.4 Hz, 1H), 6.94 (d, J = 7.8 Hz, 2H), 5.07 (dpd, J = 48.6, 6.3,
3.5 Hz, 1H), 4.15 – 3.96 (m, 2H), 1.47 (dd, J = 23.6, 6.4 Hz, 3H). 13C NMR (125 MHz, CDCl3):
δ 158.61, 129.65, 121.32, 114.73, 88.60 (d, J = 169.3 Hz), 70.81 (d, J = 23.5 Hz), 17.60 (d, J =
22.3 Hz). 19F NMR (282 MHz, CDCl3): δ −180.43 (dtq, J = 48.5, 23.5, 19.7 Hz).
(±)-(Z)-N-butyl-N-(2-(2,7-dichloro-9-(4-chlorobenzylidene)-9H-fluoren-4-yl)-2-fluoroethyl)
butan-1-amine (2.47): Lumefantrine (Acros) was evaluated according to the general procedure.
A sample for characterization was obtained by subjecting the crude reaction mixtures to
automated column chromatography (25 g silica, 0 → 20% ethyl acetate in hexanes), which
afforded the title compound as a yellow solid.136 1H NMR (500 MHz, CDCl3): δ 7.73 (d, J = 2.0
Hz, 1H), 7.69 (d, J = 8.4 Hz, 1H), 7.60 (s, 1H), 7.49 (d, J = 1.9 Hz, 1H), 7.46 (s, 4H), 7.44 (d, J =
2.0 Hz, 1H), 7.34 (dd, J = 8.3, 1.9 Hz, 1H), 6.14 (ddd, J = 48.3, 7.6, 2.3 Hz, 1H), 3.01 – 2.91 (m,
1H), 2.91 – 2.77 (m, 1H), 2.70 – 2.61 (m, 2H), 2.60 – 2.50 (m, 2H), 1.48 – 1.38 (m, 4H), 1.38 –
1.30 (m, 4H), 0.92 (t, J = 7.3 Hz, 6H). 13C NMR (125 MHz, CDCl3): δ 141.48, 138.43, 136.04
135 Doyle, M. P.; Whitefleet, J. L.; Bosch, R. J. J. Org. Chem. 1979, 44, 2923. 136 Goldberg, N. W.; Shen, X.; Li, J.; Ritter, T. Org. Lett. 2016, 18, 6102.
114
(d, J = 20.8 Hz), 135.74, 134.93, 134.84, 134.64 (d, J = 5.1 Hz), 134.12, 133.42, 132.87, 130.66,
129.23, 128.72, 128.28, 125.90 (d, J = 13.1 Hz), 124.55 (d, J = 2.7 Hz), 123.74, 120.75, 92.21
(d, J = 175.9 Hz), 59.37 (d, J = 22.2 Hz), 54.40 (d, J = 1.4 Hz), 29.09, 20.76, 14.25. 19F NMR
(282 MHz, CDCl3): δ −180.14 (ddd, J = 48.2, 33.2, 21.6 Hz).
115
Chapter 3.
Low-Temperature Radiofluorination Strategies137
137 Reproduced in part with permission from Nielsen, M. K.; Ugaz, C. R.; Li, W.; Doyle, A. G. J. Am. Chem. Soc. 2015, 137, 9571 and Gray, E. E.; Nielsen, M. K.; Choquette, K. A.; Kalow, J. A.; Graham, T. J. A.; Doyle, A. G. J. Am. Chem. Soc. 2016, 138, 10802. © Copyright 2015–2016 American Chemical Society.
116
3.1 PET Radiochemistry
Positron emission tomography or PET is a non-invasive imaging technique that maps the
three-dimensional distribution of specific molecules within a biological system.138 Central to
PET is the synthesis of radiotracers—biologically active compounds such as metabolites or drug-
like structures that contain a positron-emitting isotope. When a proton-rich nucleus such as
fluorine-18 undergoes β+ decay, a proton decays to a neutron and ejects a positron, the anti-
matter equivalent of an electron (Figure 3.1).139 The positron rapidly collides with an electron,
resulting in an annihilation event that produces two 511 keV gamma rays traveling away from
each other at an almost perfect 180° angle. A ring of scintillation detectors, such as that shown in
Figure 3.1, can detect coincident gamma rays arising from β+ decay, which through a
computationally intensive algorithm can be used to reconstruct the three-dimensional distribution
of the tracer.
Figure 3.1 Positron emission tomography.
PET is routinely employed in oncology for identifying and following the progression of
tumors. The most common radiotracer is [18F]FDG or [18F]2-deoxy-2-(18F)fluoro-D-glucose
(Scheme 3.1). [18F]FDG is taken up with glucose and phosphorylated in the first step of
glycolysis. Once phosphorylated, [18F]FDG-phosphate cannot leave the cell, but neither can it
138 Ametamey, S. M.; Honer, M.; Schubiger, P. A. Chem. Rev. 2008, 108, 1501. 139 Zanzonico, P. Semin. Nucl. Med. 2004, 34, 87.
117
serve as a substrate for glucose-6-phosphate isomerase due to the absence of a 2-hydroxy group.
As such, the [18F]FDG tends to accumulate in cells with high glucose demand. Most tumors
rapidly outgrow the blood supply and are unable to maintain aerobic metabolism through the
Krebs cycle. Under oxygen-deficient conditions, the tumor must rely almost solely on glycolysis,
which requires 19 times as much glucose to produce ATP. Combined with the demand of
sustained growth, tumors may consume more than 200 times as much glucose as surrounding
tissues. As a result, PET scans with [18F]FDG can reveal the location and size of tumors to
within millimeter precision. Notably, this process is non-invasive, requiring no surgery or
biopsy.
PET is also serves as a valuable tool in pharmaceutical development. For example, PET
tracers can be used to quantify the brain receptor occupancy of candidate neurological drugs,
aiding in early go/no-go decisions.140 If a neurological drug candidate is underperforming in
preclinical trials, it may be that the dose, and thus receptor occupancy, is too low. On the other
hand, the drug may be successfully engaging the target receptor but simply fail to produce the
desired therapeutic effect. Without PET, researchers would have little recourse but to continue
increasing the dose in hopes that they are facing the first scenario, potentially wasting millions of
dollars of resources. As another example, clinical trials for Alzheimer’s disease (AD) drug
candidates are notoriously challenging to conduct due to the incapacity to discern between AD
and other forms of dementia in living patients. The development of a PET tracer for β amyloid
has enabled researchers to screen for and accurately select patient groups for trials.141
140 Van Laere, K. J.; Sanabria-Bohórquez, S. M.; Mozley, D. P.; Burns, D. H.; Hamill, T. G.; Van Hecken, A.; De Lepeleire, I.; Koole, M.; Bormans, G.; de Hoon, J.; Depré, M.; Cherchio, K.; Plalcza, J.; Han, L.; Renger, J.; Hargreaves, R. J.; Iannone, R. J. Nucl. Med. 2014, 55, 65 141 Klunk, W. E.; Engler, H.; Nordberg, A.; Wang, Y.; Blomqvist, G.; Holt, D. P.; Bergström, M.; Savitcheva, I.; Huang, G. F.; Estrada, S.; Ausén, B.; Debnath, M. L.; Barletta, J.; Price, J. C.;
118
The vast majority of PET tracers are synthesized via the nucleophilic displacement of
tosylates, mesylates, or triflates with [18F]KF and the phase transfer reagent Kryptofix K222 as
shown in Scheme 3.1.142 Kryptofix K222 is a cryptand, an amine bridged bicyclic derivative of
18-crown-6 that completely envelops potassium to form a highly soluble fluoride source. The
combination of KF/K222 is one of the most reactive neutral fluoride sources identified to date;
unfortunately, the sheer cost ($39,700 per mol from Millipore-Sigma) precludes its use on large
scale. This is not a significant problem for PET radiosynthesis, which usually only employs
milligram quantities of substrate. A typical PET reaction might be conducted with
1000 milliCuries (mCi) of 18F (approximately 10 ng of 18F), meaning that fluoride is by far the
limiting reagent and pseudo-first-order kinetics are operative with respect to fluoride.
Scheme 3.1 Synthesis of [18F]FDG.
In practice, it is impossible to obtain isotopically pure fluorine-18 because fluorine-19 is
ubiquitous in the environment in ppm quantities. In typical [18F]KF, fluorine-19 outnumbers
fluorine-18 by 1000:1 (resulting in a total fluoride mass of ~10 μg). Specific activity is used as a
measure of isotopic purity; for example, pure fluorine-18 would have a specific activity of
1,710,000 mCi/μmol, but with 1000-fold dilution by fluorine-19, the specific activity is only
1,710 mCi/μmol. With current imaging technology, specific activities of ≥1,000 mCi/μmol are
necessary to maintain an acceptable signal-to-noise ratio. In some reported methods, fluorine-19
Sandell, J.; Lopresti, B. J.; Wall, A.; Koivisto, P.; Antoni, G.; Mathis, C. A.; Långström, B. Ann. Neurol. 2004, 55, 306. 142 Hamacher, K.; Coenen, H. H.; Stöcklin, G. J. Nucl. Med. 1986, 27, 235. Gangadharmath, U.; Walsh, J.; Kolb, H. US Patent 20130005956 A1; Jan. 3, 2013.
119
is intentionally added as a “carrier” to enhance fluorine-18 uptake; however, these approaches
are unsuitable for most imaging applications. Ideally, new technologies should be “no-carrier-
added” protocols in which the reagents contain no labile exogenous fluorine-19.
Because fluorine is the limiting reagent, reaction yields are measured in terms of activity
and are reported as radiochemical yield (RCY), which is the percentage of initial activity that is
incorporated into the product, usually corrected to account for radionuclide decay during
synthesis. In contrast to synthetic reactions, a RCY of ≥5% is acceptable for most PET
applications. In order to obtain optimal signal-to-noise, PET scans typically require that patients
be injected with approximately 10 mCi of activity, and a radiosynthesis can easily begin with
1,000 – 5,000 mCi. Far more important is reproducibility; the radiosynthetic methodology should
be as robust as possible because a reaction failure can waste an entire day’s worth of time for
both researcher and patient.
Another complicating factor of PET is that radiosyntheses must be quick, typically 10 –
20 minutes. This is due in part to the short half-life of fluorine-18 (109.8 minutes), but also a
result of the extensive purification and quality-control experiments that must be conducted prior
to injection in a patient. One of the major shortcomings of current PET methodology is that high
temperatures—typically 100 – 200 °C—are necessary for the reaction to occur within the desired
timeframe. As a result, many candidate PET substrates simply fail due to decomposition under
the harsh reaction conditions.
Interestingly, we have observed that 100 mg-scale KF/K222 reactions (with non-
radioactive anhydrous KF) are complete within seconds at room temperature. We suspect that
high temperatures are required for PET synthesis because the [18F]KF is partially hydrated.
Fluorine-18 is typically synthesized in aqueous solution, exchanged in a quaternary ammonium
120
resin with potassium carbonate to afford [18F]KF, and then dried azeotropically with acetonitrile.
Given fluoride’s high enthalpy of hydration, it is likely that some water remains within the
fluoride hydration sphere, thus attenuating its nucleophilicity.
3.2 Deoxyradiofluorination with Sulfonyl Fluorides
We hypothesized that low-temperature radiofluorination could be achieved by
deoxyfluorination with sulfonyl fluorides. In our proposal, standard partially hydrated [18F]KF
would react with excess sulfonyl chloride to form both the [18F]sulfonyl fluoride and sulfonic
acid, thereby removing water. Subsequent addition of alcohol and base the same pot would result
in deoxyradiofluorination under anhydrous conditions.
Although there have been a few isolated reports of deoxyradiofluorination, the
methodology has never found practical use. One issue is that most deoxyfluorination reagents
contain multiple equivalents of reactive fluoride (see Chapter 1, Scheme 1.3 – 1.5 and
Figure 1.2), which reduces the maximum theoretical radiochemical yield by 1/n (where n is the
number of reactive fluorines per molecule).143 More importantly, all previous attempts at deoxy-
radiofluorination have required carrier addition, resulting in low specific activity products.
Straatman and Welch successfully deoxyradiofluorinated ethyl alcohol with [18F]DAST, but
synthesized the reagent from the electrophilic fluorine source [18F]F2, which cannot be made
with high specific activity.144
More relevantly, Jelinski et al. synthesized and purified the sulfonyl fluoride [18F]PBSF
as shown in Scheme 3.2 but found that they could only fluorinate hydroxyproline to form 3.1
143 For example, consider DAST (Et2NSF3) which containes three reactive fluorines. Since fluorine-19 typically outnumbers fluorine-18 by 1000:1, the chances that radiosynthetic DAST would contain more than one fluorine-18 are 1,000,000:1. Each time mono-[18F]DAST reacts, there is a 2 in 3 chance that the fluorine-18 will not be incorporated, but will end up as a reaction byproduct. 144 Straatmann, M. G.; Welch, M. J. J. Nucl. Med. 1977, 18, 151.
121
when exogenous non-radioactive PBSF was added. 145 The reason that the no-carrier-added
reaction fails comes down to stoichiometry. In a typical radiofluorination with 1000 mCi of
activity and a specific activity of 1,500 mCi/μmol, there are approximately 500 nanomols of total
fluoride, meaning that at most 500 nanomols of PBSF reagent are generated. This will react with
the alcohol to form nominally 500 nanomols each of sulfonate ester and base-HF adduct, which
in 250 μL toluene corresponds to a concentration of 0.002 M. Typical synthetic PBSF
deoxyfluorinations are performed in 1 M toluene, thus the second order deoxyradiofluorination
will proceed at a rate approximately 250,000 times lower than that observed under synthetic
conditions. Furthermore, because the substrate is in excess, the free alcohol will compete with
fluoride for sulfonate displacement. By adding an equivalent of exogenous PBSF, the
deoxyradiofluorination can generate a full equivalent of sulfonate ester and fluoride, and can thus
proceed at the same rate as under synthetic conditions.
Scheme 3.2 Deoxyradiofluorination with purified [18F]PBSF.
145 Jelinski, M.; Hamacher, K.; Coenen, H. H. J. Labelled Compd. Radiopharm. 2001, 44, S151.
122
We decided to synthesize [18F]PyFluor from the sulfonyl chloride 146 instead of the
sulfonimide, which we achieved in almost quantitative radiochemical conversion (RCC) 147
(Scheme 3.3). The radiochemical purity of our [18F]PyFluor was found to be 96%, indicating that
in reactions affording higher than 4% RCC, the fluoride must be derived in part from the reagent
(as opposed to residual [18F]KF/K222). We chose not to isolate the sulfonyl fluoride prior to the
deoxyfluorination on the rationale that upon addition of alcohol and base, the unreacted sulfonyl
chloride would enable formation of a full equivalent of sulfonate ester. Under these conditions,
fluorination would exhibit pseudo-first-order kinetics in analogy to the standard SN2
radiofluorination.
Unfortunately, we found that unactivated alcohols afforded no significant yield after
20 minutes with this protocol. This was not wholly unexpected, as synthetic PyFluor
deoxyfluorinations may take up to 48 hours. In order to demonstrate the proof-of-concept, we
selected the hemiacetal tetra-O-benzylglucose, which was by far the most reactive substrate from
our initial PyFluor report. We were pleased to find that at 80 °C we obtained the desired product
3.2 in 15% RCC. Notably, this was the first reported example of a deoxyradiofluorination with
146 The synthesis of [18F]sulfonyl fluorides has been described: Matesic, L.; Wyatt, N. A.; Fraser, B. H.; Roberts, M. P.; Pham, T. Q.; Greguric, I. J. Org. Chem. 2013, 78, 11262. Inkster, J. A. H.; Liu, K.; Ait-Mohand, S.; Schaffer, P.; Guerin, B.; Ruth, T. J.; Storr, T. ́ Chem. - Eur. J. 2012, 18, 11079. Ironically, both groups were employing the sulfonyl fluoride motif for its stability. In one embodiment, the authors made [18F]benzenesulfonyl fluoride 4-carbaldehyde as a prosthetic group and then reacted it with complex amines to form the Schiff’s base. The [18F]sulfonyl fluoride was shown to be stable in vivo. 147 In radiochemistry, radiochemical yield (RCY) implies that the radiotracer has been isolated. Radiochemical conversion (RCC) generally refers to the percentage of activity corresponding to the product in a radio TLC plate. However, this will not account for insoluble activity (for example, fluoride baked into the glass of the reaction vessel), meaning that RCCs are generally higher than RCYs. In our methodology, we transferred the activity and filtered it into a secondary vial, allowing us to measure the percentage of soluble activity, which multiplying by the TLC yields affords a good approximation of the actual RCY. Nonetheless, our yields are designated as RCCs.
123
no carrier addition. It is noteworthy that the substrate tosylate ester is highly unstable and
decomposes within hours,148 meaning that the traditional one-step SN2 radiofluorination would
have been unsuitable for generating fluoride 3.2.
Scheme 3.3 Deoxyradiofluorination with [18F]PyFluor.
In an effort to expand reactivity to unactivated substrates, we chose to investigate a
number of more reactive sulfonyl fluorides. We found that the highly electron-deficient
[18F]ArFSF was capable of generating hydroxyproline derivative 3.1 in 22% RCC after
10 minutes at 40 °C, representing a significant improvement in comparison to the typical
>100 °C temperatures required for SN2 radiofluorination (Scheme 3.4).
Scheme 3.4 Deoxyradiofluorination of unactivated alcohols with [18F]ArFSF.
As it stands, our fledgling sulfonyl fluoride deoxyradiofluorination faces numerous
challenges and does not provide any clear advantage over existing methodology except in cases
where the sulfonate ester of the alcohol substrate is unstable. The one-pot two-step
transformation is more operationally complex, requiring a second reagent addition after
148 Eby, R.; Schuerch, C. Carbohydr. Res. 1974, 34, 79.
124
formation of the sulfonyl fluoride. Although radiochemical yields in the 10 – 20% range are
acceptable, these could certainly be improved. Synthetically, we have observed that the sulfonyl
chloride tends to react directly with DBU and MTBD, and may not be particularly efficient at
forming the desired sulfonate electrophile. Moreover, some of the displaced chloride likely reacts
to form alkyl chloride, which may be challenging to separate from the desired fluorinated
product. The reagent [18F]ArFSF may also be susceptible to SNAr substitution, which would
release free fluorine-19 and erode specific activity.
We anticipate that the method could be improved with a more thorough examination of
sulfonyl fluoride identity and reaction conditions. For example, we have observed that the
Verkade superbase 149 (2,8,9-Trimethyl-2,5,8,9-tetraaza-1-phosphabicyclo[3.3.3]undecane; see
Chapter 2, Table 2.4) forms HF adducts that react orders of magnitude faster than DBU-HF or
MTBD-HF owing to high charge delocalization and solubility. Although the Verkade superbase
is much too expensive for synthetic scale ($32,300 per mol from Aldrich), the cost would be
acceptable on radiosynthetic scale, and the additional boost in fluoride nucleophilicity might
enable deoxyradiofluorination to occur at room temperature.
3.3 Radiofluorination of α-Diazocarbonyls
As an alternative approach to low-temperature radiofluorination, we investigated the
copper-catalyzed insertion of α-diazocarbonyl compounds into H–18F. In the presence of
transition metal catalysts, α-diazocarbonyl compounds can lose nitrogen with concurrent
formation of a reactive metal carbenoid species.150 Numerous reports have emerged describing
the catalytic insertion of copper and rhodium carbenoids into N–H, O–H, and even C–H bonds,
149 Tang, J.; Dopke, J.; Verkade, J. G. J. Am. Chem. Soc. 1993, 115, 5015. 150 Ford, A.; Miel, H.; Ring, A.; Slattery, C. N.; Maguire, A. R.; McKervey, M. A. Chem. Rev. 2015, 115, 9981.
125
in many cases with high enantioselectivity.151 Although Olah and others have demonstrated that
diazo compounds can be fluorinated directly under acidic conditions,152 a selective, catalytic
diazo fluorination had not been achieved.153 In our laboratory, between 2013 and 2015, Julia
Kalow, Kim Choquette, and Erin Gray developed the first example of a catalytic H–F insertion
(Scheme 3.5). 154 The reaction employs a copper(I) catalyst, and under its highest-yielding
conditions, a bisoxazoline (BOX) ligand. Kim and Erin additionally identified a phosphoramidite
ligand capable of delivering product in as high as 86% ee, although the procedure was never
generalized. The fluoride source is KF, which is typically completely insoluble in dichloroethane
(DCE); however, it was found that addition of hexafluoroisopropanol (HFIP) as a co-solvent led
to dramatically improved solubility.155 Most reactions were complete within 30 minutes at or just
Scheme 3.5 Copper-catalyzed fluorination of α-diazo carbonyls.
151 Gillingham, D.; Fei, N. Chem. Soc. Rev. 2013, 42, 4918. Zhao, X.; Zhang, Y.; Wang, J. Chem. Commun. 2012, 48, 10162. Zhu, S.-F.; Zhou, Q.-L. Acc. Chem. Res. 2012, 45, 1365. 152 Curtius, T. J. Prakt. Chem. 1888, 38, 396. Olah, G.; Kuhn, S. Chem. Ber. 1956, 89, 864. Ynunyants, I. L.; Kisel, Y. M.; Bykhovskaya, E. G. Bull. Acad. Sci. USSR, Div. Chem. Sci. 1956, 5, 363. Olah, G. A.; Welch, J. Synthesis 1974, 896. Setti, E. L.; Mascaretti, O. A. J. Chem. Soc., Perkin Trans. 1 1988, 2059. Ohno, M.; Itoh, M.; Ohashi, T.; Eguchi, S. Synthesis 1993, 793. Pasceri, R.; Bartrum, H. E.; Hayes, C. J.; Moody, C. J. Chem. Commun. 2012, 48, 12077. 153 Huw Davies demonstrated a vinylogous fluorination of vinyl diazoacetates in 2013, but fluorination at the diazo position was not observed. Qin, C.; Davies, H. M. L. Org. Lett. 2013, 15, 6152. 154 Gray, E. E.; Nielsen, M. K.; Choquette, K. A.; Kalow, J. A.; Graham, T. J. A.; Doyle, A. G. J. Am. Chem. Soc. 2016, 138, 10802. 155 The hexaflouroisopropoxy ether arising from O–H insertion was typically observed in yields ranging from approximately 1 – 10%. This side product proved almost impossible to separate from the desired fluoride with column chromatography.
126
above room temperature, suggesting that the reaction would proceed to completion within the
timeframe of a typical radiosynthesis. Notably, this approach would bypass the high
temperatures and potential elimination and decomposition side products observed in SN2
radiofluorinations.
The Gouverneur laboratory has demonstrated that α-aryl-α-trifluoromethyl diazo
compounds can be radiofluorinated at room temperature (Scheme 3.6). 156 However, the acid-
induced transformation is racemic and requires the addition of oxogenous HF∙pyridine (Olah’s
reagent). A 1 μL volume of Olah’s reagent contains 731 μg of fluorine, meaning that in a typical
1,000 mCi radiosynthesis, the method would deliver product with a specific activity no higher
than 25 mCi/μmol, well below the acceptable limit of ~1,000 mCi/μmol.
Scheme 3.6 Radiofluorination of α-trifluoromethyl diazo compounds.
In order to translate our synthetic α-diazocarbonyl fluorination to a radiosynthetic
technique, we began our investigations with methyl 2-diazo-2-phenylacetate 3.3 (Table 3.1). Our
hope was that since the synthetic reaction employed KF, we would be able to use the standard
[18F]KF/K222 fluoride source without modification. Initially, we employed copper(I) triflate
toluene complex as the catalyst, which had been shown to give higher enantioselectivity in the
synthetic method. We were pleased to find that with the same ratios of copper, ligand, and HFIP
relative to substrate, we were able to obtain 74% RCC at 50 °C in our first attempt (Table 3.1,
entry 1). However, we observed that the copper triflate complex would undergo visible
156 Emer, E.; Twilton, J.; Tredwell, M.; Calderwood, S.; Collier, T. L.; Liégault, B.; Taillefer, M.; Gouverneur, V. Org. Lett. 2014, 16, 6004.
127
degradation during weighing and ligand precoordination with widely varying results. Instead,
tetrakis(acetonitrile)copper(I) hexafluorophosphate proved to be more resistant to oxidation and
hydrolysis and provided more reproducible yields (entry 2). In the absence of copper catalyst, a
trace amount of fluorination was observed, probably arising from acid-induced diazo
composition by HFIP (pKa = 9.3) (entry 3). Without HFIP, no radiofluorinated product was
detected whatsoever, highlighting the importance of HFIP in solvating fluoride (entry 4).
Although rhodium catalysts (such as [Rh(OAc)2]2) provided ~50% yield under synthetic
conditions with a total reaction time of just a few seconds, no radiochemical conversion was
observed under analogous radiofluorination conditions.
Table 3.1 Development of copper catalyzed α-diazocarbonyl radiofluorination.
When we measured the specific activity of 3.4 under our standard conditions (entry 2),
we observed a low specific activity of only 24 mCi/μmol, indicating that one of the fluorinated
reagents (either hexafluorophosphate or HFIP) was acting as an exogenous source of fluorine-19.
HFIP, which contains strong C–F bonds, has been previously used in radiosyntheses with no
detrimental effect on specific activity.157 As such, we suspected that the hexafluorophosphate
anion was non-innocent, perhaps dissociating into PF5 and fluoride. Indeed, under standard
157 Revunov, E.; Zhuravlev, F. J. Fluorine Chem. 2013, 156, 130.
128
synthetic conditions, 3.4 was obtained in 1.0% yield in the absence of KF, confirming that
hexafluorophosphate is non-innocent. However, when we employed tetrakis(acetonitrile)
copper(I) perchlorate,158 no fluorinated product was detected without KF. Under radiosynthetic
conditions, the perchlorate catalyst delivered 74% RCC with an acceptable specific activity of
1300 mCi/μmol. The highly oxidizing perchlorate anion is generally avoided on large scale;
however, under radiosynthetic conditions with ~1 mg catalyst present, we believe this is an
acceptable solution for maintaining high specific activity.
Interestingly, at 40 °C, only trace amounts of 3.1 are formed and the solution remains
yellow-orange, the color of the diazo substrate (Table 3.2). However, at 50 °C, almost
quantitative radiochemical yield is obtained with full consumption of the diazo substrate (evident
as the solution changes from yellow to colorless). At higher temperatures, the color change is
more rapid and is accompanied by visible evolution of nitrogen gas. These data suggest an
unusually high dependence of reaction rate on temperature.
Table 3.2 Temperature screen for α-diazocarbonyl radiofluorination.
With this radiofluorination protocol in hand, we set out on a preliminary investigation of
scope (Table 3.3). The more complex and electron-rich 2-arylacetate 3.5 was obtained in
158 Liang, H.-C.; Karlin, K. D.; Dyson, R.; Kaderli, S.; Jung, B.; Zuberbühler, A. D. Inorg. Chem. 2000, 39, 5884.
129
similarly high yield, suggesting that the transformation is fairly general for this substrate class.
However, we questioned the overall utility of the 2-aryl-2-fluoroacetate structure for
radiosynthesis. This motif does not show up commonly in drugs or drug-like molecules and does
not represent an obvious disconnection. Separation of the diazo starting materials from the
methylene precursor can be extraordinarily challenging. Finally, even if the transformation were
highly enantioselective, radiochemists tend to steer away from non-stereospecific reactions
because on tracer scale, even a small impurity of the other enantiomer could complicate
interpretation of a PET scan due to unexpected on- and off-target receptor interactions.
Table 3.3 Preliminary scope of copper-catalyzed α-diazocarbonyl radiofluorination.
R
O
N2
R
O
X X18F
3.564 ±6% RCC (n = 3)
3.853 ±25% RCC (n = 3)
3.936 ±9% RCC (n = 3)
3.738 ±11% RCC (n = 4)
OBzOBzO
HN
OBz
OOBz
18F
18F
Me
O
OBn
Me
O
HN
PhO
NH
18F
O
NH
3.480 ±7% RCC (n = 3)
18F
O
NH
3.632 ±9% RCC (n = 4)
O Ot-Bu
NHBoc
NEt2
O
N
O
Oi-Pr
18F
O
OMePh
18F
O
OMe
[Cu(MeCN)4]PF6 (10 mol%)
(S,S)-t-BuBOX (15 mol%)
[18F]KF-K222, HFIP (10 equiv)
DCE (300 L), 50 °C, 10 min28 mol
We were quite pleased to find that 2-diazoacetamides 3.6 – 3.9 also proved competent in
this reaction. Notably, 3.6 is a one-carbon homologue of reported tracer [18F]FPDA,159 and 3.7 is
identical to the structure of [18F]FAO. 160 In both cases, our transformation represents a
significant improvement, which will be discussed hereafter. From a synthetic perspective,
159 Liu, H.; Liu, S.; Miao, Z.; Jiang, H.; Deng, Z.; Hong, X.; Cheng, Z. Mol. Pharmaceutics 2013, 10, 3384. 160 Turkman, N.; Gelovani, J. G.; Alauddin, M. M. J. Labelled Compd. Radiopharm. 2011, 54, 33.
130
synthesizing 2-fluoroacetamides from 2-diazoacetamides is highly impractical, given that one
could simply conduct a peptide coupling between 2-fluoroacetate and the amine. However, in
radiosyntheses, the radionuclide should ideally be incorporated in the final step. Because of the
fluorine-18’s short half-life and operational time constraints, multi-step radiosyntheses are quite
involved and are generally avoided.
Radiochemists have previously taken two approaches to synthesizing α-fluoroamides.
The first is to perform the standard SN2 radiofluorination, which is inefficient due to the high
nucleophilicity of the α-carbon. For example, [18F]FAO (3.7) was previously synthesized from
the α-bromoamide in only 8% RCY (vs. 38% RCC with our method). The second approach is to
make a “prosthetic group” a simple fluorine-18 containing synthon that can then be appended to
the desired substrate in a subsequent step. As an example, the prosthetic group [18F]NFP was
used to generate the αvβ3 integrin receptor ligand [18F]Galacto-RGD (Scheme 3.7). 161 This
approach has proven quite versatile for the synthesis of macromolecules, but in this instance
requires four radiochemical manipulations. [18F]FPDA (the homologue of 3.6) was likewise
synthesized from [18F]NFP with an RCY of only 3% over four steps (vs. 32% RCC with our
method).
The advantage of our copper-catalyzed radiofluorination of α-diazoacetamides is that
fluorine is introduced in the last step under mild conditions, leading to substantial improvements
in RCY and operational simplicity. Moreover, we have demonstrated synthetically that HF
outcompetes alcohols and even amines for insertion, with only trace amounts of competing
insertion products observed (Table 3.4). Diazoacetamides can be readily synthesized from
161 Haubner, R.; Kuhnast, B.; Mang, C.; Weber, W. A.; Kessler, H.; Wester, H. J.; Schwaiger, M. Bioconjug. Chem. 2004, 15, 61.
131
Scheme 3.7 Prosthetic groups in radiosynthesis.
Table 3.4 Radiofluorination of α-diazoacetamides bearing free alcohols and amines.
complex amine substrates using diazoacetate transfer reagents,162 which have even been used to
diazoacetylate native proteins.163 One challenge in employing such substrates will be identifying
compatible solvent systems in which the macromolecule would be soluble, although it should be
noted that the combination of DCE with HFIP is surprisingly polar. Additionally, even if O–H
and N–H insertion from unprotected amines and alcohols is negligible on synthetic scale, it may
be more significant when fluoride is limiting as under PET conditions. Although this
162 Ouihia, A.; René, L.; Guilhem, J.; Pascard, C.; Badet, B. J. Org. Chem. 1993, 58, 1641. 163 Josa-Cullere, L.; Wainman, Y. A.; Brindle, K. M.; Leeper, F. J. ́ RSC Adv. 2014, 4, 52241.
132
methodology will require further development, we envision that this methodology will allow
near room-temperature radiofluorination of complex biomolecules under mild conditions.
3.4 Experimental Section
Reagents and Methods. See Section 1.5 for General Methods and Instrumentation. Acetonitrile
was drawn from an Millipore-Sigma Sure/Seal bottle. Potassium oxalate, potassium carbonate,
Kryptofix 222, sodium acetate trihydrate, and glacial acetic acid were purchased from Acros.
Radiochemistry experiments were performed in a shielded lead hot-cell using a Gilson
automated liquid handler. RadioTLCs were visualized by sectioning aluminum-backed TLC
plates and analyzing each segment on a PerkinElmer 1480 Wizard 3 automated gamma counter.
RadioHPLC was performed on an Agilent 1100 system. Shielded QMA resins charged with
[18F]fluoride were delivered by Siemens Molecular Imaging, Inc., North Wales, PA 19454.
Deoxyradiofluorination:
[18F]PyFluor ([18F]2-pyridinesulfonyl fluoride): A QMA resin containing 482 mCi of
[18F]fluoride was delivered @ 9:53 (Siemens Molecular Imaging, Inc., North Wales,
Pennsylvania 19454). Solution A was prepared with 200 mg potassium oxalate and 3 mg
potassium carbonate in 5 mL water. An eluent was prepared from 1.2 mL acetonitrile, 250 μL
water, and 50 μL of solution A. The resin was eluted with 1.0 mL of this eluent, affording 228
mCi @ 11:42 (94% elution efficiency). A Wheaton 1 mL conical glass vial with a septum cap
was clamped in a microwave cavity. Using a Gilson automated liquid handler, 100 μL of the
eluted activity and 200 μL of a Kryptofix 222 solution (36 mg/mL in acetonitrile) were
transferred to the vial. The solvent was removed under a stream of argon with microwave heating
(60 °C, 50 W, 240 s). Residual water was removed azeotropically by adding 300 μL of
133
acetonitrile and evaporating under argon (60 °C, 50 W, 120 s); this step was performed three
times. Solution B was prepared with 100 μL of a benzene solution containing 50 mg/mL of
2-pyridinesulfonyl chloride, and 900 μL acetonitrile. The reaction vial containing dry activity
was charged with 200 μL of solution B and 200 μL acetonitrile (1 mg 2-pyridinesulfonyl
chloride (6 μmol) in total volume of 400 μL 1:1 benzene:acetonitrile). The vial was then heated
by microwave (80 °C, 70 W, 300 s). The end-of-synthesis (EOS) activity was 1.205 mCi @
17:05. For the purposes of analysis, the reaction mixture was taken up in ~1 mL of acetonitrile
and was filtered through a short cotton plug into a secondary container. The amount of soluble
activity was 1.090 mCi @ 17:10. The product identity was confirmed by spiking the reaction
mixture with unlabeled PyFluor and subjecting to radioHPLC (XBridge Phenyl 3.5 μm (4.6 ×
150 mm), 1.5 mL/min, 25% acetonitrile: 10 mM pH 4 sodium acetate buffer. Product tr = 4.7
min). The soluble mixture was spotted on a 10 cm TLC plate and developed with 40% ethyl
acetate in hexanes (product Rf = 0.7). RadioTLC analysis indicated that 94% of the soluble
material was product, corresponding to an overall radiochemical conversion (RCC) of 88%.
Run 2: 89% RCC (7.85 mCi EOS). Run 3: 86% RCC (20.2 mCi EOS).
[18F]2,3,4,6-tetra-O-benzyl-D-glucopyranosyl fluoride (3.2): The crude reaction mixture in the
synthesis of [18F]PyFluor was filtered through a short (~1 g) activated silica plug. By radioTLC,
this sample had a radiochemical purity (RCP) of ≥96%. (Important! This solution contains as
much as 5 mg/mL of unreacted 2-pyridinesulfonyl chloride in addition to [18F]PyFluor. The
sulfonyl chloride should not be separated from the [18F]PyFluor.) Solution C was prepared with
2,3,4,6-tetra-O-benzyl-D-glucopyranose (11.1 mg, 21 μmol, Millipore-Sigma) and MTBD (6.5
134
mg, 42 μmol) in 1 mL acetonitrile. A Wheaton 1 mL conical glass vial with a septum cap was
clamped in a microwave cavity. Using a Gilson automated liquid handler, the vial was charged
with 100 μL of the filtered [18F]PyFluor solution (containing ≤0.5 mg 2-pyridinesulfonyl
chloride), and 200 μL of solution C (2.2 mg substrate (1.5 equiv) and 1.3 mg MTBD (3 equiv)
(with respect to 2-pyridinesulfonyl chloride) in a total of 300 μL acetonitrile (with some residual
benzene)). The reaction mixture was heated by microwave (80 °C, 70 W, 1200 s). The EOS
activity was 1.490 mCi @ 14:27. The reaction mixture was taken up in ~1 mL of 15% ethyl
acetate in hexanes and was filtered through a short cotton plug into a secondary container. The
amount of soluble activity was 378 μCi @ 14:27. The product identity was confirmed by spiking
the reaction mixture with unlabeled product and subjecting to radioHPLC (XBridge Phenyl 3.5
μm (4.6 × 150 mm), 1.5 mL/min, 65% acetonitrile: 10 mM pH 4 sodium acetate buffer. Product
tr = 11.2 min). The soluble mixture was spotted on a 10 cm TLC plate and developed with 15%
ethyl acetate in hexanes (product Rf = 0.6). RadioTLC analysis indicated that 52% of the soluble
material was product, corresponding to an overall radiochemical conversion (RCC) of 13%. Run
2: 21% RCC (785 μCi EOS). Run 3: 11% RCC (622 μCi EOS).
Radiofluorination of α-diazocarbonyls:
[18F]methyl 2-fluoro-2-phenylacetate (3.4): A QMA resin containing ~300 mCi of
[18F]fluoride was delivered @ 10:40 (Siemens Molecular Imaging, Inc., North Wales,
Pennsylvania 19454). The resin was eluted with a 1.0 mL solution containing 2 mg potassium
oxalate monohydrate and 30 μg potassium carbonate in 4:1 acetonitrile:water (v/v), affording
148.9 mCi @ 12:50. A Wheaton 1 mL conical glass vial with a septum cap was clamped in a
microwave cavity and fitted with an 18g outlet needle. Using a Gilson automated liquid handler,
135
50 μL of the eluted activity and 40 μL of a Kryptofix 222 solution (36 mg/mL in acetonitrile)
were transferred to the vial (1.1 μmol potassium, 3.8 μmol cryptand (approximately 4 equiv with
respect to potassium). The solvent was removed under a stream of argon with microwave heating
(65 °C, 50 W, 240 s). Residual water was removed azeotropically by adding 300 μL of
acetonitrile and evaporating under argon (65 °C, 50 W, 120 s); this step was performed three
times. The 18g vent needle was removed prior to addition of reagents. A catalyst solution was
prepared with 10.6 mg tetrakis(acetonitrile)copper(I) hexafluorophosphate (stored in a
dessicator) and 12.5 mg tBuBOX in 1000 uL DCE. The solution was sealed in a 1-dram vial with
a stirbar and allowed to stir for 30 minutes prior to use. The catalyst solution should be
colorless—a blue solution indicates oxidation to copper(II). A substrate solution was prepared
with 50 mg methyl 2-diazo-2-phenylacetate, 700 μL DCE, and 300 μL HFIP. The reaction vial
containing azeotropically dried activity was charged sequentially with 100 μL substrate solution
and 100 μL catalyst solution (5 mg substrate (28 μmol), 10 mol% copper (1.06 mg), 15 mol%
ligand (1.25 mg), 10 equiv HFIP (30 μL) in a total volume of 200 μL in DCE). Using more than
10 mol% catalyst relative to the substrate may result in diminished radiochemical yield. The vial
was heated in a microwave cavity (50 °C, 40 W, 600 s). During the course of the reaction, the
solution turned from yellow to colorless as the diazo was consumed. The EOS activity was 866
μCi @ 15:42. The soluble activity was transferred to a secondary vial rinsing with 40% ethyl
acetate in hexanes and had an activity of 787 μCi @ 15:44 (92% of activity is soluble). The
product identity was confirmed by subjecting both the unlabeled authentic product and crude
reaction mixture to radioHPLC (XBridge Phenyl 3.5 μm (4.6 × 150 mm), 1.5 mL/min, 20%
acetonitrile: 10 mM Na2HPO4. Product tr = 17.9 min). The soluble content was spotted on a 10
cm TLC plate and developed with 10% ethyl acetate in hexanes (product Rf = 0.8). RadioTLC
136
analysis indicated that 87% of the soluble material was product, corresponding to an overall
RCC of 80% (693 μCi EOS). Run 2: 87% RCC (466 μCi EOS). Run 3: 73% RCC (395 μCi
EOS). With tetrakis(acetonitrile)copper(I) hexafluorophosphate, the observed specific activity
was 24 mCi/μmol EOB. With tetrakis(acetonitrile)copper(I) perchlorate,164 the observed specific
activity was 1300 mCi/μmol EOB.
(±)-[18F]Methyl 2-fluoro-2-(4-((3-isopropyl-2-oxooxazolidin-5-yl)methoxy)phenyl)acetate
(3.5): This compound was labeled following the same procedure described in the synthesis of
(3.4) using 9.5 mg (±)-methyl 2-diazo-2-(4-((3-isopropyl-2-oxooxazolidin-5-yl)methoxy)phenyl)
acetate (28 μmol), 10 mol% [Cu(MeCN)4]PF6, 15 mol% tBuBOX and 10 equiv HFIP in a total
volume of 300 μL in DCE, microwaving at 50 °C (40 W, 600 s). Run 1: 68% RCC (1.07 mCi
EOS). Run 2: 68% RCC (775 μCi EOS). Run 3: 57% RCC (542 μCi EOS). RadioHPLC:
XBridge C18 5 μm (4.6 × 150 mm), 1.5 mL/min, 5% → 55% acetonitrile:0.1% TFA; Product tr =
7.4 min. Radio TLC: 70% ethyl acetate:hexanes; product Rf = 0.4.
[18F]N-(2-(diethylamino)ethyl)-2-fluoroacetamide (3.6):
This compound was labeled following the same procedure described in the synthesis of (3.4)
using 5.3 mg N-(2-(diethylamino)ethyl)-2-diazoacetamide (28 μmol), 10 mol%
[Cu(MeCN)4]PF6, 15 mol% tBuBOX and 10 equiv HFIP in a total volume of 300 μL in DCE,
164 Liang, H.-C.; Karlin, K. D.; Dyson, R.; Kaderli, S.; Jung, B.; Zuberbühler, A. D. Inorg. Chem. 2000, 39, 5884.
137
microwaving at 50 °C (40 W, 600 s). Run 1: 43% RCC (742 μCi EOS). Run 2: 28% RCC (412
μCi EOS). Run 3: 35% RCC (424 μCi EOS). Run 4: 23% RCC (219 μCi EOS). RadioHPLC:
Phenomenex Gemini C18 3 μm (4.6 × 150 mm), 1.5 mL/min, 5% → 90% acetonitrile:10 mM
Na2HPO; Product tr = 3.6 min. Radio TLC: 20% methanol: dichloromethane with 3%
triethylamine; product Rf = 0.6.
[18F]N2-Boc-N5-(2-fluoroacetyl)ornithine-tert-butyl ester (3.7): This compound was labeled
following the same procedure described in the synthesis of (3.4) using 10.2 mg N2-Boc-N5-(2-
diazoacetyl)ornithine-tert-butyl ester (28 μmol), 10 mol% [Cu(MeCN)4]PF6, 15 mol% tBuBOX
and 10 equiv HFIP in a total volume of 300 μL in DCE, microwaving at 50 °C (40 W, 600 s).
Run 1: 53% RCC (1.81 mCi EOS). Run 2: 42% RCC (386 μCi EOS). Run 3: 33% RCC (317 μCi
EOS). Run 4: 26% RCC (580 μCi EOS). RadioHPLC: XBridge Phenomenex Gemini C18 3 μm
(4.6 × 150 mm), 1.5 mL/min, 5% → 90% acetonitrile:10 mM Na2HPO4; Product tr = 6.4 min.
Radio TLC: 10% methanol: dichloromethane with 1.5% triethylamine; product Rf = 0.6.
[18F]benzyl (2-fluoroacetyl)-L-phenylalanyl-L-leucinate (3.8): This compound was labeled
following the same procedure described in the synthesis of (3.4) using 12.4 mg benzyl
(2-diazoacetyl)-L-phenylalanyl-L-leucinate (28 μmol), 10 mol% [Cu(MeCN)4]PF6, 15 mol%
tBuBOX and 10 equiv HFIP in a total volume of 200 μL in DCE, microwaving at 50 °C (40 W,
600 s). Run 1: 57% RCC (1.95 mCi EOS). Run 2: 76% RCC (1.56 mCi EOS). Run 3: 26% RCC
(289 μCi EOS). RadioHPLC: XBridge C18 5 μm (4.6 × 150 mm), 1.5 mL/min, 30% →
138
65% acetonitrile:10 mM NH4OAc; Product tr = 10.1 min. Radio TLC: 60% ethyl acetate:
hexanes; product Rf = 0.8.
[18F]N-(2-fluoroacetyl)-1,3,4,6-tetra-O-benzoyl-α-D-glucosamine (3.9): This compound was
labeled following the same procedure described in the synthesis of (3.4) using 18.9 mg
N-(2-diazoacetyl)-1,3,4,6-tetra-O-benzoyl-α-D-glucosamine (28 μmol), 10 mol%
[Cu(MeCN)4]PF6, 15 mol% tBuBOX and 10 equiv HFIP in a total volume of 300 μL in DCE,
microwaving at 50 °C (40 W, 600 s). Run 1: 46% RCC (1.79 mCi EOS). Run 2: 31% RCC
(694 μCi EOS). Run 3: 31% RCC (566 μCi EOS). RadioHPLC: XBridge C18 5 μm (4.6 × 150
mm), 1.5 mL/min, 5% → 90% acetonitrile:0.1% TFA; Product tr = 8.0 min. Radio TLC: 60%
ethyl acetate:hexanes; product Rf = 0.7.
139
Appendix A.
Aryl Formylation via Photocatalytic Generation of Chlorine Radicals165
165 Reproduced in part with permission from Nielsen, M. K.; Shields, B. J.; Liu, J.; Williams, M. J.; Zacuto, M. J.; Doyle, A. G. Angew. Chem., Int. Ed. 2017, 56, 7191. © Copyright 2017 Wiley-VCH.
140
A.1 Aryl Formylation
Aromatic aldehydes are versatile intermediates in the synthesis of pharmaceuticals,
fragrances, fine chemicals, and natural products.166 Aldehydes can be rapidly elaborated through
an ever-growing host of C–C and C–X bond-forming reactions including reductive amination,
olefination, aldol-type condensations, and Grignard addition. Despite the ubiquitous application
of aryl aldehydes, synthetic methods for their preparation are limited and generally require redox
manipulations. Straightforward redox approaches include DIBAL reduction of carboxylic acids
and selective oxidation of primary benzylic alcohols (Scheme 4.1A). Many of the earliest
syntheses involved electrophilic aromatic substitution with various in situ generated formyl
Scheme A.1 Approaches to aryl formylation.
166 In the synthesis of atorvastatin: Baumann, K. L.; Butler, D. E.; Deering, C. F.; Mennen, K. E.; Millar, A.; Nanninga, T. N.; Palmer, C. W.; Roth, B. D. Tetrahedron Lett. 1992, 33, 2283. In the synthesis of montelukast: McNamara, J. M.; Leazer, J. L.; Bhupathy, M.; Amato, J. S.; Reamer, R. A.; Reider, P. J.; Grabowski, E. J. J. J. Org. Chem. 1989, 54, 3718. In fragrances: Fráter, G.; Bajgrowicz, J. A.; Kraft, P. Tetrahedron 1998, 54, 7633. In total syntheses: Woodward, R. B.; Cava, M. P.; Ollis, W. D.; Hunger, A.; Daeniker, H. U.; Schenker, K. J. Am. Chem. Soc. 1954, 76, 4749. Magnus, P.; Sane, N.; Fauber, B. P.; Lynch, V. J. Am. Chem. Soc. 2009, 131, 16045.
141
electrophiles (Scheme 4.1B).167 These SEAr reactions have proven to be quite versatile, but are
fundamentally limited by the selectivity of directing groups. Addition of organolithium or
Grignard reagents to DMF can enable access to electronically disfavored regioisomers
(Scheme 4.1C); however, the use of organometallic nucleophiles requires cryogenic temperatures
and severely limits functionl group tolerance.168
To date, the most general method for the synthesis of aryl aldehydes is palladium
catalyzed reductive carbonylation of aryl iodides and bromides, which was first reported by Heck
in 1974 under 100 atm of syngas (1:1 H2:CO) at 150 °C (Scheme 4.1D). 169 Reductive
carbonylation proceeds through a two-electron coupling mechanism in which oxidative addition
of the aryl halide is followed by migratory insertion of the aryl group into a coordinated carbon
monoxide. In formylation mechanisms, the resulting palladium acyl species is simply reduced off
with hydrogen. Although this procedure has been performed on multi-ton scale,170 the use of
pressurized gases, particularly carbon monoxide, at high temperature is not ideal for benchtop
synthesis due to the accompanying health and safety hazards. Stille and Pri-Bar introduced the
use of tin hydrides and silanes as solid-phase reductants to replace hydrogen gas.171 Likewise,
various reports have emerged in which carbon monoxide is replaced with formate, N-formyl
saccharin, carbon dioxide, isocyanate, paraformaldehyde, or metal carbonyls, although these
167 Reimer, K.; Tiemann, F. Ber. Dtsch. Chem. Ges. 1876, 9, 1268. Gattermann, L.; Koch, J. A. Chem. Ber. 1897, 30, 1622. Gattermann, L.; Berchelmann, W. Ber. Dtsch. Chem. Ges. 1898, 31, 1765. Vilsmeier, A.; Haack, A. Chem. Ber. 1927, 60, 119. Duff, J. C.; Bills, E. J. J. Chem. Soc. 1932, 1987. 168 Bouveault, L. Bull. Soc. Chim. Fr. 1904, 31, 1306. 169 Schoenberg, A.; Heck, R. F. J. Am. Chem. Soc. 1974, 96, 7761. 170 Klaus, S.; Neumann, H.; Zapf, A.; Strübing, D.; Hübner, S.; Almena, J.; Riermeier, T.; Groß, P.; Sarich, M.; Krahnert, W.-R.; Rossen, K.; Beller, M. Angew. Chem., Int. Ed. 2006, 45, 154. 171 Tin hydrides: Baillargeon, V. P.; Stille, J. K. J. Am. Chem. Soc. 1983, 105, 7175. Silanes: Pri-Bar, I.; Buchman, O. J. Org. Chem. 1984, 49, 4009.
142
reagents generate low concentrations of carbon monoxide in situ. 172 Despite the forcing
conditions, this method has relatively broad scope, but will not tolerate electrophilic functionality
susceptible to oxidative addition, strongly nucleophilic groups, or motifs that may be
hydrogenated in the presence of palladium and hydrogen gas. Additionally, reductive
carbonylation typically does not work for aryl chlorides, which are generally more accessible and
less expensive than the corresponding bromides and iodides.
A.2 Redox-Neutral Formylation of Aryl Chlorides with 1,3-Dioxolane
Our interest in aryl formylation arose from a serendipitous discovery made by graduate
student Ben Shields in our laboratory. While investigating nickel metalla-photoredox,173 Ben
observed that under certain conditions, aryl chlorides would couple with the THF solvent to form
adducts such as that shown in Scheme A.2.174 This reactivity was wholly unexpected, as THF,
with a reduction potential of +1.75 V (vs. SCE in acetonitrile) is incapable of being oxidized by
the fluorinated heteroleptic iridium photocatalyst (+1.21 V vs. SCE in acetonitrile).
Scheme A.2 Arylation of Csp3–H bonds with nickel metallaphotoredox.
172 Formate anion: Okano, T.; Harada, N.; Kiji, J. Bull. Chem. Soc. Jpn. 1994, 67, 2329. N-formyl saccharin: Ueda, T.; Konishi, H.; Manabe, K. Angew. Chem., Int. Ed. 2013, 52, 8611. Carbon dioxide: Yu, B.; Zhao, Y.; Zhang, H.; Xu, J.; Hao, L.; Gao, X.; Liu, Z. Chem. Commun. 2014, 50, 2330. t-Butyl isocyanate: Jiang, X.; Wang, J.-M.; Zhang, Y.; Chen, Z.; Zhu, Y.-M.; Ji, S.-J. Org. Lett. 2014, 16, 3492. Paraformaldehyde: Natte, K.; Dumrath, A.; Neumann, H.; Beller, M. Angew. Chem., Int. Ed. 2014, 53, 10090. Iron pentacarbonyl: Iranpoor, N.; Firouzabadi, H.; Etemadi-Davan, E.; Rostami, A.; Moghadam, K. R. Appl. Organometal. Chem. 2015, 29, 719. 173 Zuo, Z.; Ahneman, D. T.; Chu, L.; Terrett, J. A.; Doyle, A. G.; MacMillan, D. W. C. Science 2014, 345, 437. Tellis, J. C.; Primer, D. N.; Molander, G. A. Science 2014, 345, 433. 174 Shields, B. J.; Doyle, A. G. J. Am. Chem. Soc. 2016, 138, 12719. A similar discovery was made concurrently by the Molander group: Heitz, D. R.; Tellis, J. C.; Molander, G. A. J. Am. Chem. Soc. 2016, 138, 12715
143
Although similar C–H activation had been conducted with the addition of H-atom
transfer (HAT) catalysts such as quinuclidine,175 this remarkable reactivity in the absence of a
HAT reagent defied explanation by any existing metallaphotoredox mechanism. Through
extensive mechanistic investigation, Ben developed the proposal shown in Scheme A.3. First, the
aryl chloride undergoes oxidative addition into Ni(0) species A.1 to form Ni(II) complex A.2.
Meanwhile, excitation of ground state Ir(III) photocatalyst A.3 generates excited state Ir(III)
species A.4 (half potential: +1.21 V vs. SCE in acetonitrile) which can oxidize Ni(II) complex
A.2 (peak potential: +0.85 V vs. SCE in acetonitrile) to form Ni(III) complex A.5. At this point,
Ben proposed a truly remarkable event, the absorption of a second photon by the nickel complex
that results in homolytic bond cleavage of the Ni–Cl bond, ejecting a chlorine radical. This
radical can abstract a C–H bond from the α-oxy position of THF, and the resultant carbon radical
will recombine with nickel to form A.7. Reductive elimination affords the observed product, and
finally the reduced Ir(II) catalyst A.9 reduces the Ni(I) A.8 species back to Ni(0) A.1.
Scheme A.3 Proposed chlorine atom photoelimination mechanism.
Ar ClLnNi
0
LnNiII
LnNiI A.8
Cl
HCl
O
H
SETIrIII A.3
*IrIII A.4
Ar IrII A.9
h
SET
LnNiIII
O
Ar LnNiIII
Cl
Ar
Cl
Ar
LnNiII ArLnNi
II Ar
h
aryl chloride
oxidant
reductant
A.1A.2
A.5
A.6
A.7
O
O
175 Shaw, M. H.; Shurtleff, V. W.; Terrett, J. A.; Cuthbertson, J. D.; MacMillan, D. W. C. Science 2016, 352, 1304.
144
Photoelimination of halides from Ni(III) had previously been observed in the Nocera
laboratory in the context of harvesting solar energy as chlorine gas,176 but had never been
invoked as an elementary step in a synthetic method. Ben was able to demonstrate that oxidation
of the Ni(II)aryl chloride complex A.2 to Ni(III) species A.5 with a stoichiometric oxidant led to
no product formation. However, when the oxidized mixture was irradiated, solvent
functionalization of THF was observed in 28% yield in the absence of a photocatalyst.
We proposed that this method could provide a redox-neutral alternative to reductive
formylation through the functionalization of formaldehyde acetal solvents. As shown in
Scheme A.4, functionalization of 1,3-dioxolane would lead generate an aryl acetal, which upon a
mild, acidic workup would afford the desired aldehyde.
Scheme A.4 Redox-neutral formylation via chlorine atom photoelimination.
Ar ClLnNi
0
LnNiII
LnNiI
Cl
HCl
O
OH
SETIrIII
*IrIII
ArIrII
h
SET
LnNiIII
OO
Ar LnNiIII
Cl
Ar
Cl
Ar
LnNiII ArLnNi
II Ar
h
aryl acetal
aryl chloride
oxidant
reductant
O O
aryl aldehyde
Haq
Ar
O
Workup
H
H
kcal mol-1
ab
86.888.2
BDFE
a b
O
O
We began our investigation with 4-chlorobenzonitrile A.9. Linear acetals such as
dimethoxymethane proved unreactive, largely due to poor catalyst solubility; however, with
1,3-dioxolane under slightly modified conditions (Ni(II)Cl2∙glyme vs. Ni(0)(cod)2 precatalyst,
176 Hwang, S. J.; Powers, D. C.; Maher, A. G.; Anderson, B. L.; Hadt, R. G.; Zheng, S. L.; Chen, S.; Nocera, D. G. J. Am. Chem. Soc. 2015, 137, 6472. Hwang, S. J.; Anderson, B. L.; Powers, D. C.; Maher, A. G.; Hadt, R. G.; Nocera, D. G. Organometallics 2015, 34, 4766.
145
1 mol% vs. 2 mol% photocatalyst, 0.05 M vs. 0.04 M) we found that the desired acetal A.10
could be obtained in 82% yield (Table A.1, entry 1). Under screening conditions with 25 W blue
LED strips, similar yield was obtained (entry 2). In the absence of light, photocatalyst, or nickel
catalyst, no yield was observed. Without a base or other proton sink, catalyst turnover was
limited and protodehalogenation product A.12 was observed (entry 6). Although 1,3-dioxolane is
inexpensive ($39 per liter from Acros), it may be employed in subsolvent quantities with
benzene (entry 7). Lowering catalyst loading below 10 mol% Ni and 1 mol% photocatalyst
resulted in a significant decrease in yield (entry 8). Although Ni(cod)2 was optimal for Ben’s
original arylation, with 1,3-dioxolane, the yield was diminished and biaryl A.13 was observed,
perhaps a byproduct of catalyst activation.
Table A.1 Optimization of formylation of aryl chlorides with 1,3-dioxolane.
Under our standard conditions (entry 1), we also observed 7% yield of the
4-functionalization product A.11, which we had suspected might occur due to the calculated
1.4 kcal/mol difference in the 2- and 4-position C–H bond dissociation free energies (see
Scheme A.4). Indeed, across all substrates investigated, the desired 2-functionalization product
was obtained with approximately 10:1 selectivity, roughly what would be expected based on
thermodynamics and the relative stoichiometry of the distinct C–H bonds. Although aryl acetal
146
A.10 hydrolyzes readily in acid at room temperature, formyl acetal A.11 is resistant to hydrolysis
even when heated to 50 °C. For polar substrates, the 4-functionalized side product proved
surprisingly difficult to separate from the aldehyde with standard column chromatography. In an
effort to avoid selectivity issues, we attempted to use 1,3,5-trioxane (50 equiv in benzene) which
would afford only a single regioisomer. The desired acetal A.14 was obtained in a modest 57%
yield (Scheme A.5), but almost no hydrolysis was observed under our standard workup
procedures, indicating that more forcing deprotection conditions would be necessary.
Scheme A.5 Formylation with 1,3,5-trioxane.
Despite our efforts, we were unable to identify improved conditions differing
substantially from those originally reported by Ben. Among common photocatalysts, the
fluorinated heteroleptic iridium catalyst Ir[dF(CF3)ppy]2(dtbbpy)PF6177 was almost uniquely
effective. Likewise, 4,4-di-tert-butyl-2,2-dipyridyl (dtbbpy) was by far the best ligand for nickel,
although bathophenanthroline performed reasonably well. Na2CO3 performed almost as well as
K3PO4, while the bases KOH, KF, Cs2CO3, NaHCO3 and Na3PO4 provided moderate yield.
Whereas the bond dissociation energy (BDE) of HCl is 103 kcal/mol, the BDE for HBR
is only 88 kcal/mol, approximately the same as the C–H bonds in dioxolane. When the aryl
bromide corresponding to A.9 was subjected to reaction conditions, A.10 was obtained in only
55% yield, albeit with higher selectivity against 4-functionalization. The aryl iodide afforded no
yield because the BDE of HI is only 73 kcal/mol, too weak for C–H abstraction. Preliminary data
suggests that acyl chlorides and alkyl chlorides may undergo formylation to glyoxals and
177 (4,4'-di-t-butyl-2,2'-bipyridine)bis[3,5-difluoro-2-[5-trifluoromethyl-2-pyridinyl-κN)phenyl-κC]iridium(III) hexafluorophosphate [870987-63-6]
147
homologated aliphatic aldehydes, respectively; however this methodology has not been
developed.
To investigate the scope of our redox-neutral formylation strategy, we subjected a
number of stereoelectronically distinct aryl chlorides to our optimal reaction conditions for A.9
followed by a mild acidic workup to obtain the aldehyde (Table A.2). The reaction tolerates both
electron-deficient and electron-rich aryl halides (A.15 – A.17) although the latter are obtained in
somewhat diminished yield due to the longer reaction time. Protic functionality including
Table A.2 Substrate scope for aryl formylation.
148
alcohols (A.18) and primary amides (A.19) are tolerated, which under palladium catalysis
conditions would likely compete for C–O and C–N bond formation. Various heterocycles may be
formylated as well (A.20 – A.21), although the highly electron-deficient pyridyl acetal A.21 did
not fully hydrolyze even when heated above 50 °C in concentrated acid. As a demonstration of
the breadth and mildness of these conditions, we formylated a number of pharmaceutical and
biological compounds (A.22 – A.33). Notably, A.22 was obtained in 89% yield with 9% yield of
the acetal isomer, a nearly quantitative yield of C–H functionalization products. Diformylation of
a dichloride was achieved in the case of A.24. Most remarkably, the chlorophenylalanine residue
in tripeptide A.29 containing an unprotected serine was formylated in 56% yield. This is an
unprecedented direct approach to biomacromolecules containing aryl aldehydes, which have
been invoked as a handle in bioconjugation.178
In conclusion, we have harnessed the photoelimination of chlorine radicals from
nickel(III) to develop a redox-neutral formylation of aryl chlorides. In comparison to reductive
carbonylation and other formylation approaches, this method proceeds under exceedingly mild
conditions with unprecedented substrate scope, which we attribute to the fact chlorine radical is
generated catalytically and only in proximity to the nickel catalyst, preventing the indiscriminate
functionalization that would occur during a stoichiometric C–H halogenation. We hope that the
capability to introduce aldehydes into functionally complex intermediates will enable
pharmaceutical researchers to employ reactions such as reductive amination and olefination with
advanced intermediates.
178 Carrico, I. S.; Carlson, B. L.; Bertozzi, C. R. Nat. Chem. Biol. 2007, 3, 321.
149
A.3 Experimental Section
Reagents and Methods. See Section 1.5 for General Methods and Instrumentation. 4,4'-Di-tert-
butyl-2,2'-bipyridine (dtbbpy) was purchased from Aldrich, Astatech, and Alfa Aesar. Nickel(II)
chloride, dimethoxyethane adduct (NiCl2∙glyme) and (4,4'-Di-t-butyl-2,2'-bipyridine)bis[3,5-
difluoro-2-[5-trifluoromethyl-2-pyridinyl-κN)phenyl-κC]iridium(III) hexafluorophosphate
(Ir[dF(CF3)ppy]2(dtbbpy)PF6) were purchased from Strem. Anhydrous potassium phosphate
tribasic (K3PO4) was purchased from Aldrich. All reagents were stored in an N2-filled glovebox.
1,3-Dioxolane (Alfa Aesar, 99.5% with up to 75 ppm BHT stabilizer) was degassed with argon,
brought into the glovebox, and stored on activated 4Å molecular sieves. Liquid aryl chlorides
were degassed and solid aryl chlorides were purged under vacuum and brought into the glovebox
during reaction setup.
General procedure for formylation: In the glovebox, a 1-dram vial with a teflon stirbar is
charged with NiCl2∙glyme (5.5 mg, 10 mol%), dtbbpy (10.1 mg, 15 mol%), and 1,3-dioxolane
(2 mL). This nickel catalyst solution is stirred for a minimum of 10 minutes prior to addition to
the reaction mixture and should form a light green homogeneous solution. Meanwhile, a
threaded 16 × 125 mm borosilicate reaction tube with a teflon stirbar is charged with substrate
(0.25 mmol), K3PO4 (106 mg, 2 equiv), Ir[dF(CF3)ppy]2(dtbbpy)PF6 photocatalyst (2.8 mg,
1 mol%), and 1,3-dioxolane (3 mL). Finally, the nickel catalyst solution is added (total reaction
volume: 5 mL, 0.05 M.) The reaction tube is capped with a septum cap, sealed with electrical
tape, and removed from the glovebox. The reaction is stirred at 500 rpm for 72 hours while
illuminating with a 34 W blue LED lamp (Kessil KSH150B, λmax = 450 nm flanked by a second
peak at λ = 422 nm) placed horizontally at 2 cm distance from the reaction tube. Owing to the
significant heat output of the lamp, a fan is used to cool the reaction tube to nominally room
150
temperature; however, the actual reaction temperature is typically around 30 °C. Upon
completion, the reaction mixture is concentrated on a rotary evaporator and transferred to a
20 mL scintillation vial in 5 mL acetone followed by addition of 5 mL aqueous 1 M HCl. The
mixture is stirred for 1 hour to hydrolyze the acetal to the desired aldehyde. The solution is then
diluted with 25 mL saturated sodium bicarbonate and extracted with 3 × 25 mL ethyl acetate.
The organic extracts are dried with anhydrous sodium sulfate, concentrated, and purified by
automated column chromatography.
4-(1,3-dioxolan-4-yl)benzonitrile (A.11): Isolated as a side product in the formation of
aldehyde A.15. 1H NMR (500 MHz, CDCl3): δ 7.66 (d, J = 8.3 Hz, 2H), 7.46 (d, J = 8.1 Hz,
2H), 5.28 (s, 1H), 5.09 (s, 1H), 5.06 (t, J = 6.5 Hz, 1H), 4.28 (dd, J = 8.6, 7.0 Hz, 1H), 3.68 (dd,
J = 8.2, 6.3 Hz, 1H). 13C NMR (125 MHz, CDCl3): δ 145.36, 132.61, 126.73, 118.76, 112.08,
96.45, 76.76, 71.90. IR (ATR, cm−1): 2923 (m), 2854 (m), 2229 (m), 1611 (w), 1505 (w), 1466
(w), 1414 (w), 1365 (w), 1305 (w), 1284 (w), 1211 (w), 1153 (m), 1084 (s), 1017 (m), 948 (s),
924 (s), 836 (s), 780 (w), 730 (w). HRMS (ESI+): Calculated for C10H10NO2+ [M + H]+:
176.0706; found: 176.0706
4-(1,3,5-trioxan-2-yl)benzonitrile (A.14): This reaction was performed general procedure from
4-chlorobenzonitrile (34.4 mg, 0.25 mmol, Aldrich) with a reaction time of 120 hours and with
the following modifications: (1) Benzene was used in place of 1,3-dioxolane as solvent, including
for the Ni/ligand prestir. (2) The reaction was also charged with 1,3,5-trioxane (1.13 g, 50 equiv,
151
Aldrich). Following acid hydrolysis, the residue was analyzed by NMR and found to contain 2%
yield of the aldehyde 4-formylbenzonitrile and 55% yield of the title compound, acetal adduct
4-(1,3,5-trioxan-2-yl)benzonitrile. This represents a combined effective yield of 57% and
indicates that more forcing hydrolysis conditions may be necessary to quantitatively obtain the
aldehyde. The acetal adduct can be purified by automated chromatography (25 g silica, 0 → 30%
ethyl acetate in hexanes) and isolated as a white crystalline solid. 1H NMR (500 MHz, CDCl3):
δ 7.70 (d, J = 8.5 Hz, 2H), 7.62 (d, J = 8.3 Hz, 2H), 5.90 (s, 1H), 5.38 (d, J = 6.6 Hz, 2H), 5.32
(d, J = 6.7 Hz, 2H). 13C NMR (125 MHz, CDCl3): δ 141.39, 132.37, 127.00, 118.58, 113.38,
99.99, 93.67. IR (ATR, cm−1): 2878 (w), 2231 (m), 1398 (m), 1190 (m), 1165 (s), 1084 (m),
1067 (s), 1025 (m), 987 (m), 973 (w), 960 (s), 945 (s), 891 (m), 826 (s), 775 (w), 734 (w), 708
(w). HRMS (ESI+): Calculated for C10H10NO3+ [M + H]+: 192.0655; found: 192.0653.
4-formylbenzonitrile (A.15): Synthesized following the general procedure from
4-chlorobenzonitrile (34.4 mg, 0.25 mmol, Aldrich). Purified by automated column
chromatography (25 g silica, 0 → 18% ethyl acetate in hexanes) to afford 27.6 mg product (84%
yield) as a white crystalline solid.179 A second run provided 27.0 mg (82% yield). 1H NMR (500
MHz, CDCl3): δ 10.10 (s, 1H), 8.00 (d, J = 8.2 Hz, 2H), 7.85 (d, J = 8.2 Hz, 2H). 13C NMR
(125 MHz, CDCl3): δ 190.73, 138.85, 133.04, 130.03, 117.84, 117.75.
179 C. Cheng, M. Brookhart Angew. Chem., Int. Ed. 2012, 51, 9422.
152
4-(tert-butyl)benzaldehyde (A.16): Synthesized from 1-(tert-butyl)-4-chlorobenzene (42.2 mg,
0.25 mmol, Accela ChemBio) following the general procedure. Purified by automated column
chromatography (50 g silica, 0 → 15% ethyl acetate in hexanes) to afford 30.0 mg product
(74% yield) as a colorless oil. 180 A second run provided a mixture of 30.2 mg product
(74% yield) and 4.8 mg of the acetal isomer side product 4-(4-(tert-butyl)phenyl)-1,3-dioxolane
(9% yield, 8.0:1 ratio). 1H NMR (500 MHz, CDCl3): δ 9.98 (s, 1H), 7.82 (d, J = 8.5 Hz, 2H),
7.55 (d, J = 8.4 Hz, 2H), 1.35 (s, 9H). 13C NMR (125 MHz, CDCl3): δ 192.19, 158.57, 134.20,
129.83, 126.12, 35.49, 31.21.
p-anisaldehyde (A.17): Synthesized following the general procedure from 4-chloroanisole
(35.6 mg, 0.25 mmol, Oakwood). Purified by automated column chromatography (25 g silica,
0 → 12% ethyl acetate in hexanes) to afford 19.8 mg product (58% yield) as a colorless oil.181 A
repeat experiment generated a mixture of 19.8 mg product (58% yield) and 3.6 mg of the acetal
isomer side product 4-(4-methoxyphenyl)-1,3-dioxolane (8% yield, 7.2:1 ratio). 1H NMR
(500 MHz, CDCl3): δ 9.89 (s, 1H), 7.84 (d, J = 8.4 Hz, 2H), 7.01 (d, J = 8.4 Hz, 2H), 3.89 (s,
3H). 13C NMR (125 MHz, CDCl3): δ 191.00, 164.72, 132.14, 130.06, 114.44, 55.74.
4-(3-hydroxypropyl)benzaldehyde (A.18): Synthesized according to the general procedure
from 3-(4-chlorophenyl)propan-1-ol (42.7 mg, 0.25 mmol, Matrix). Purified by automated
column chromatography (25 g silica, 0 → 45% ethyl acetate in hexanes) to afford 29.5 mg of a
180 C. B. Kelly, M. A. Mercadante, R. J. Wiles, N. E. Leadbeater, Org. Lett. 2013, 15, 2222. 181 R. Kawahara, K. Fujita, R. Yamaguchi Angew. Chem., Int. Ed. 2012, 51, 12790.
153
colorless oil. By NMR, this was found to be a partially resolved mixture containing 28.0 mg
product (69% yield) and 1.5 mg of the acetal isomer 3-(4-(1,3-dioxolan-4-yl)phenyl)propan-1-ol
(2% yield, 33:1 ratio).182 A second run afforded a combined 24.2 mg product (59% yield) and 3.2
mg acetal isomer (6% yield, 9.8:1 ratio). 1H NMR (500 MHz, CDCl3): δ 9.97 (s, 1H), 7.80 (d, J
= 8.1 Hz, 2H), 7.36 (d, J = 7.9 Hz, 2H), 3.69 (t, J = 6.4 Hz, 2H), 2.80 (t, J = 7.8 Hz, 2H), 1.92
(dt, J = 14.0, 6.4 Hz, 2H), 1.52 (s, 1H). 13C NMR (125 MHz, CDCl3): δ 192.17, 149.56, 134.67,
130.14, 129.26, 62.04, 33.87, 32.44.
2-(3-formylphenyl)acetamide (A.19): Synthesized according to the general procedure from
2-(3-chlorophenyl)acetamide (42.4 mg, 0.25 mmol, Aldrich). Purified by automated column
chromatography (50 g silica, 0 → 5% methanol in dichloromethane) to afford 32.4 mg of a white
solid. NMR indicated a mixture of 28.8 mg product (71% yield) and 3.6 mg of the acetal isomer
2-(3-(1,3-dioxolan-4-yl)phenyl)acetamide (6% yield, 11.1:1 ratio). A second run provided 28.5
mg product (70% yield) and 5.1 mg acetal isomer side product (10% yield, 7.1:1 ratio). 1H NMR
(500 MHz, acetone-d6): δ 10.03 (s, 1H), 7.86 (s, 1H), 7.79 (d, J = 7.6 Hz, 1H), 7.65 (d, J = 7.6
Hz, 1H), 7.53 (t, J = 7.6 Hz, 1H), 7.01 (s, 1H), 6.41 (s, 1H), 3.63 (s, 2H). 13C NMR (125 MHz,
acetone-d6): δ 192.99, 172.60, 138.56, 137.73, 136.15, 130.86, 129.82, 128.68, 42.75. IR (ATR,
cm−1): 3362 (br, m), 3170 (br, m), 2929 (w), 2826 (w), 2742 (w), 1692 (m), 1642 (s), 1604 (m),
1483 (w), 1415 (m), 1391 (m), 1303 (m), 1235 (s), 1191 (w), 1140 (m), 1089 (w), 1002 (w), 915
(m), 885 (w), 813 (w), 777 (s). HRMS (ESI+): Calculated for C9H10NO2+ [M + H]+ : 164.0706;
found: 164.0705.
182 B. H. Lipshutz, M. Hageman, J. C. Fennewald, R. Linstadt, E. Slack, K. Voightritter Chem. Commun. 2014, 50, 11378.
154
2-methylbenzo[d]thiazole-5-carbaldehyde (A.20): Synthesized following the general
procedure from 5-chloro-2-methylbenzothiazole (45.9 mg, 0.25 mmol, TCI). Purified by
automated column chromatography (50 g silica, 0 → 19% ethyl acetate in hexanes to afford 31.5
mg product (71% yield) as a white crystalline solid. 183 A repeat experiment afforded an
unresolved mixture of 30.0 mg product (68% yield) and 4.0 mg of the acetal isomer 5-(1,3-
dioxolan-4-yl)-2-methylbenzo[d]thiazole (7% yield, 9.3:1 ratio). 1H NMR (500 MHz, CDCl3): δ
10.11 (s, 1H), 8.38 (d, J = 1.5 Hz, 1H), 7.95 (d, J = 8.2 Hz, 1H), 7.88 (dd, J = 8.3, 1.5 Hz, 1H),
2.87 (s, 3H). 13C NMR (125 MHz, CDCl3): δ 191.84, 169.10, 153.60, 142.35, 134.94, 125.20,
124.26, 122.21, 20.47.
3-(1,3-dioxolan-2-yl)isonicotinonitrile (A.21): Synthesized according to the general procedure
from 3-chloro-4-cyanopyridine (34.6 mg, 0.25 mmol, Matrix). Aldehyde formation was not
observed under the standard hydrolysis conditions. The residue was by automated column
chromatography (25 g silica, 0 → 40% ethyl acetate in hexanes with 5% triethylamine) to afford
33.6 mg of a white crystalline solid. NMR analysis indicated that this consisted of 30.4 mg
desired acetal product (69% yield) and 3.2 mg of the acetal isomer 3-(1,3-dioxolan-4-
yl)isonicotinonitrile (7% yield, 9.5:1 ratio). A replicate experiment provided 32.0 mg product
(73% yield) and 3.8 mg of the acetal isomer (9% yield, 8.4:1 ratio). 1H NMR (500 MHz,
CDCl3): [9.5:1 mixture of product:acetal isomer] δ 8.88 (s, 1H), 8.80 (d, J = 5.0 Hz, 1H), 7.58
183 U. Klar, B. Buchmann, W. Schwede, W. Skuballa, J. Hoffmann, R. B. Lichtner Angew. Chem., Int. Ed. 2006, 45, 7942.
155
(dd, J = 5.0, 0.8 Hz, 1H), 5.98 (s, 1H), 4.34 – 4.23 (m, 2H), 4.18 – 4.05 (m, 2H). 13C NMR
(125 Hz, CDCl3): [9.5:1 mixture of product:acetal isomer]: δ 151.53, 149.87, 134.71, 126.53,
119.67, 115.07, 100.80, 66.35. IR (ATR, cm−1): 2970 (w), 2899 (m), 2234 (w), 1593 (m), 1559
(w), 1482 (w), 1386 (s), 1291 (w), 1241 (m), 1091 (s), 1047 (w), 1020 (w), 966 (m), 941 (m),
912 (s), 831 (s), 743 (w), 725 (w), 689 (w). HRMS (ESI+): Calculated for C9H9N2O2+ [M + H]+:
177.0659; found: 177.0657.
4-(3-methyl-1,1-dioxido-4-oxo-1,3-thiazinan-2-yl)benzaldehyde (A.22): Synthesized
according to the general procedure from chlormezanone (68.4 mg, 0.25 mmol, Enamine).
Purified by automated column chromatography (50 g silica, 0 → 4% methanol in
dichloromethane) affording 66.1 mg of a pale orange solid. By NMR, this consisted of 59.7 mg
desired product (89% yield) and 6.4 mg of the acetal isomer 2-(4-(1,3-dioxolan-4-yl)phenyl)-3-
methyl-1,3-thiazinan-4-one 1,1-dioxide (8% yield, 10.8:1 ratio). A second run provided a
mixture of 59.3 mg product (89% yield) and 7.6 mg acetal isomer side product (10% yield, 9.1:1
ratio). 1H NMR (500 MHz, CDCl3): δ 10.07 (s, 1H), 8.00 (d, J = 8.3 Hz, 2H), 7.58 (d, J = 8.1
Hz, 2H), 5.40 (d, J = 2.1 Hz, 1H), 3.35 – 3.10 (m, 4H), 2.96 (s, 3H). 13C NMR (125 MHz,
CDCl3): δ 191.23, 166.12, 137.79, 136.36, 130.42, 129.02, 80.57, 44.07, 36.41, 30.63. IR (ATR,
cm−1): 2930 (w), 2854 (w), 2746 (w), 1698 (m), 1646 (s), 1607 (m), 1580 (w), 1453 (w), 1420
(w), 1385 (m), 1319 (s), 1294 (w), 1238 (w), 1208 (m), 1170 (w), 1127 (s), 1034 (w), 1015 (w),
996 (w), 959 (w), 918 (w), 881 (m), 861 (w), 811 (w), 796 (w), 727 (m). HRMS (ESI+):
Calculated for C12H14NO4S+ [M + H]+ : 268.0638; found: 268.0643.
156
isopropyl 2-(4-(4-formylbenzoyl)phenoxy)-2-methylpropanoate (A.23): Synthesized
following the general procedure from isopropyl 2-(4-(4-chlorobenzoyl)phenoxy)-2-
methylpropanoate (fenofibrate, 90.2 mg, 0.25 mmol, Aldrich). Purified by automated column
chromatography (25 g silica, 0 → 17% ethyl acetate in hexanes) to afford 72.3 mg product (82%
yield) as a colorless oil. A second run provided 70.5 mg (80% yield). 1H NMR (500 MHz,
CDCl3): δ 10.11 (s, 1H), 7.97 (d, J = 8.3 Hz, 2H), 7.85 (d, J = 8.1 Hz, 2H), 7.75 (d, J = 8.8 Hz,
2H), 6.86 (d, J = 8.8 Hz, 2H), 5.07 (hept, J = 6.3 Hz, 1H), 1.65 (s, 6H), 1.19 (d, J = 6.3 Hz, 6H).
13C NMR (125 MHz, CDCl3): δ 194.59, 191.77, 173.10, 160.23, 143.39, 138.26, 132.24,
130.08, 129.80, 129.57, 117.36, 79.58, 69.49, 25.47, 21.63. IR (film, cm−1): 2984 (w), 2939 (w),
1728 (m), 1705 (m), 1655 (m), 1596 (s), 1500 (w), 1467 (w), 1419 (w), 1385 (w), 1278 (m),
1249 (s), 1204 (w), 1178 (m), 1142 (s), 1100 (s), 1011 (w), 973 (w), 929 (s), 852 (m), 830 (m),
790 (w), 763 (m), 692 (w). HRMS (ESI+): Calculated for C21H23O5+ [M + H]+ : 355.1540;
found: 355.1535.
5-(1,5-dimethyl-2,4-dioxo-3-azabicyclo[3.1.0]hexan-3-yl)isophthalaldehyde (A.24):
Synthesized according to the general procedure from 3-(3,5-dichlorophenyl)-1,5-dimethyl-3-
azabicyclo[3.1.0]hexane-2,4-dione (procymidone, 35.5 mg, 0.125 mmol, Aldrich). Purified by
automated column chromatography (50 g silica, 10 → 40% ethyl acetate in hexanes) affording
18.3 mg product as a white solid (54% yield). A second run provided 15.8 mg (47% yield). 1H
157
NMR (500 MHz, CDCl3): δ 10.09 (s, 2H), 8.34 (s, 1H), 8.09 (d, J = 1.5 Hz, 2H), 1.83 (d, J =
4.8 Hz, 1H), 1.54 (s, 6H), 1.27 (d, J = 4.7 Hz, 1H). 13C NMR (125 MHz, CDCl3): δ 190.07,
176.08, 137.84, 134.06, 132.01, 129.43, 33.07, 30.47, 10.09. IR (ATR, cm−1): 3078 (w), 2981
(w), 2936 (w), 2853 (w), 2752 (w), 1776 (w), 1691 (s), 1597 (m), 1464 (m), 1392 (m), 1371 (s),
1344 (w), 1257 (w), 1153 (w), 1139 (m), 1118 (s), 1087 (m), 1048 (w), 1011 (w), 967 (m), 920
(w), 880 (w), 809 (m), 764 (w), 733 (s), 677 (s). HRMS (ESI+): Calculated for C15H14NO4+
[M + H]+ : 272.0917; found: 272.0919.
(±)-4-(3-cyclopropyl-2-hydroxy-1-(1H-1,2,4-triazol-1-yl)butan-2-yl)benzaldehyde (A.25):
Synthesized following the general procedure from cyproconazole (72.9 mg, 0.25 mmol, 2.3:1 dr,
Alfa Aesar). Purified by automated column chromatography (50 g silica, 0 → 3% methanol in
dichloromethane, slow gradient) to afford 48.9 mg of a white solid. By NMR, this was found to
be a mixture of 44.0 mg product (62% yield, 2.5:1 dr) and 4.9 mg of the acetal isomer (±)-2-(4-
(1,3-dioxolan-4-yl)phenyl)-3-cyclopropyl-1-(1H-1,2,4-triazol-1-yl)butan-2-ol (6% yield, 10.2:1
ratio, dr not determined). A second run provided a mixture of 49.1 mg product (69% yield, 2.0:1
dr) and 5.6 mg of the acetal isomer (7% yield, 10.2:1 ratio). 1H NMR (500 MHz, CDCl3): δ
9.94 (s, 1H), 7.84 (s, 1H), 7.77 (s, 1H), 7.74 (d, J = 7.5 Hz, 2H), 7.55 (d, J = 8.3 Hz, 2H), 4.95
(d, J = 14.0 Hz, 1H), 4.73 (s, 1H), 4.54 (d, J = 14.1 Hz, 1H), 1.31 (dq, J = 9.8, 6.8 Hz, 1H), 1.07
(d, J = 6.8 Hz, 3H), 0.66 – 0.55 (m, 1H), 0.46 – 0.39 (m, 1H), 0.39 – 0.31 (m, 1H), 0.02 (dq, J =
9.6, 5.0 Hz, 1H), −0.09 (dq, J = 9.8, 5.0 Hz, 1H). 13C NMR (125 MHz, CDCl3): δ 191.93,
151.88, 148.99, 144.26, 135.35, 129.32, 126.97, 79.87, 57.04, 47.43, 14.68, 13.45, 6.52, 3.08. IR
(film, cm−1): 3389 (br, m), 3126 (w), 3078 (w), 2997 (w), 2969 (w), 2879 (w), 2739 (w), 1698
158
(s), 1607 (s), 1574 (w), 1509 (m), 1389 (w), 1309 (w), 1276 (m), 1213 (s), 1173 (w), 1138 (m),
1064 (w), 1017 (m), 967 (w), 924 (w), 888 (w), 829 (s), 733 (s), 679 (s), 658 (m). HRMS
(ESI+): Calculated for C16H20N3O2+ [M + H]+ : 286.1550; found: 286.1548.
N-(4-formylphenyl)-2,2-dimethylpentanamide (A.26): Synthesized according to the general
procedure from monalide (59.9 mg, 0.25 mmol, Aldrich). Purified by automated column
chromatography (50 g silica, 0 → 20% ethyl acetate in hexanes) to afford 44.4 mg of a colorless
oil. NMR analysis indicated a mixture of 38.6 mg product (66% yield) and 5.8 mg of the acetal
isomer N-(4-(1,3-dioxolan-4-yl)phenyl)-2,2-dimethylpentanamide (8% yield, 7.7:1 ratio). A
second run provided 37.8 mg product (65% yield) and 6.2 mg acetal isomer side product (9%
yield, 7.2:1 ratio). 1H NMR (500 MHz, CDCl3): 9.92 (s, 1H), 7.85 (d, J = 8.4 Hz, 2H), 7.72 (d,
J = 8.6 Hz, 2H), 7.50 (s, 1H), 1.60 (t, J = 4.3 Hz, 2H), 1.37 – 1.31 (m, 2H), 1.30 (s, 6H), 0.92 (t,
J = 7.0 Hz, 3H). 13C NMR (125 MHz, CDCl3): δ 191.12, 176.56, 143.65, 132.37, 131.27,
119.52, 43.91, 43.56, 25.56, 18.29, 14.71. IR (film, cm−1): 3355 (br, m), 2960 (w), 2931 (w),
2872 (w), 2737 (w), 1682 (s), 1587 (s), 1513 (s), 1474 (w), 1411 (w), 1390 (w), 1306 (m), 1243
(m), 1216 (w), 1163 (s), 1146 (w), 1112 (w), 1011 (w), 927 (w), 830 (m), 790 (m), 769 (w), 732
(m). HRMS (ESI+): Calculated for C14H20NO2+ [M + H]+: 234.1489; found: 234.1487.
(±)-(4R,5R)-N-cyclohexyl-5-(4-formylphenyl)-4-methyl-2-oxothiazolidine-3-carboxamide
(A27): Synthesized following the general procedure from hexythiazox (88.2 mg, 0.25 mmol,
159
Aldrich). Purified by automated column chromatography (25 g silica, 0 → 30% ethyl acetate in
hexanes) to afford 58.1 mg of a colorless oil. NMR analysis indicated that this was a partially
resolved mixture of 55.8 mg product (64% yield) and 2.3 mg of the acetal isomer (±)-(4R,5R)-5-
(4-(1,3-dioxolan-4-yl)phenyl)-N-cyclohexyl-4-methyl-2-oxothiazolidine-3-carboxamide
(2% yield, 27:1 ratio). A second run provided 60.3 mg product (70% yield) and 6.9 mg acetal
isomer side product (7% yield, 9.8:1 ratio). 1H NMR (500 MHz, CDCl3): δ 10.01 (s, 1H), 8.05
(d, J = 7.8 Hz, 1H), 7.88 (d, J = 8.2 Hz, 2H), 7.50 (d, J = 8.3 Hz, 2H), 4.90 (qd, J = 6.3, 1.1 Hz,
1H), 4.27 (d, J = 1.0 Hz, 1H), 3.75 – 3.63 (m, 1H), 1.98 – 1.88 (m, 2H), 1.75 – 1.67 (m, 2H),
1.63 (d, J = 6.3 Hz, 3H), 1.62 – 1.55 (m, 1H), 1.42 – 1.31 (m, 2H), 1.30 – 1.18 (m, 3H). 13C
NMR (125 MHz, CDCl3): δ 191.49, 172.46, 150.16, 147.42, 136.38, 130.70, 127.42, 61.62,
50.62, 49.28, 33.05, 33.00, 25.60, 24.77, 24.77, 20.55. IR (film, cm−1): 3324 (w), 2930 (m),
2854 (w), 1696 (s), 1660 (m), 1607 (m), 1525 (s), 1451 (w), 1364 (m), 1310 (w), 1209 (m), 1167
(s), 1103 (m), 1064 (w), 1006 (w), 913 (w), 891 (w), 828 (m), 754 (w), 729 (s), 672 (m). HRMS
(ESI+): Calculated for C18H23N2O3S+ [M + H]+ : 347.1424; found: 347.1418.
N-Cbz-DL-4-formylphenylalanine benzyl ester (A.28): Synthesized according to the general
procedure from N-Cbz-DL-4-chlorophenylalanine benzyl ester (106.0 mg, 0.25 mmol). The
acetal was only stirred in HCl for 10 minutes as significant benzyl ester hydrolysis was observed
at 1 hour. Purified by automated column chromatography (25 g silica, 0 → 40% ethyl acetate in
hexanes) to afford 75.1 mg of a white solid. By NMR, this was found to be a mixture of 66.5 mg
product (64% yield) and 8.6 mg of the acetal isomer N-Cbz-DL-4-(1,3-dioxolan-4-
yl)phenylalanine benzyl ester (7% yield, 8.5:1 ratio). A second run yielded 67.6 mg product
160
(65% yield) and 9.0 mg of the of the 1,3-dioxolan-4-yl acetal isomer (8% yield, 8.3:1 ratio). 1H
NMR (500 MHz, CDCl3): δ 9.94 (s, 1H), 7.68 (d, J = 7.7 Hz, 2H), 7.39 – 7.27 (m, 10H), 7.14
(d, J = 7.7 Hz, 2H), 5.31 (d, J = 8.1 Hz, 1H), 5.22 – 5.02 (m, 4H), 4.74 (q, J = 6.6 Hz, 1H), 3.18
(ddd, J = 44.1, 13.9, 6.1 Hz, 2H). 13C NMR (125 MHz, CDCl3): δ 191.92, 170.98, 155.60,
142.95, 136.20, 135.36, 134.91, 130.17, 130.00, 128.87, 128.87, 128.81, 128.69, 128.43, 128.27,
67.62, 67.22, 54.67, 38.56. IR (film, cm−1): 3339 (m br), 3034 (w), 2953 (w), 1696 (s), 1607
(m), 1578 (w), 1515 (m), 1499 (m), 1455 (w), 1387 (w), 1343 (w), 1307 (w), 1254 (w), 1211 (s),
1170 (s), 1107 (w), 1054 (m), 1027 (w), 911 (w), 843 (w), 823 (w), 777 (w), 739 (m), 697 (s).
HRMS (ESI+): Calculated for C25H24NO5+ [M + H]+ : 418.1649; found: 418.1644.
N-Cbz-L-alanyl-L-seryl-L-3-formylphenylalanine benzyl ester (A.29): Synthesized according
to the general procedure from N-Cbz-L-alanyl-L-seryl-L-3-chlorophenylalanine benzyl ester
(145.5 mg, 0.25 mmol). The acetal was stirred in HCl for only 10 minutes to limit ester
hydrolysis. Purified by automated column chromatography (50 g silica, 0 → 4% methanol in
dichloromethane) and isolated 99.3 mg of a pale yellow solid. NMR analysis indicated that this
consisted of a mixture of 79.3 mg product (55% yield), 10.1 mg of unreacted starting material
(7% of mass balance, some material not isolated), and 9.9 mg of the acetal isomer N-Cbz-L-
alanyl-L-seryl-L-3-(1,3-dioxolan-4-yl)phenylalanine benzyl ester (6% yield, 8.6:1 ratio). A
second run afforded 79.9 mg product (56% yield), 31.7 mg unreacted substrate (22% of mass
balance), and 11.5 mg acetal isomer side product (7% yield, 7.5:1 ratio). 1H NMR (500 MHz,
DMSO-d6): δ 9.95 (s, 1H), 8.33 (d, J = 7.6 Hz, 1H), 7.86 (d, J = 7.9 Hz, 1H), 7.78 – 7.71 (m,
161
2H), 7.53 (d, J = 7.7 Hz, 1H), 7.47 (t, J = 8.1 Hz, 2H), 7.38 – 7.22 (m, 10H), 5.07 (s, 2H), 5.02
(d, J = 4.9 Hz, 2H), 4.88 (t, J = 5.5 Hz, 1H), 4.61 (q, J = 7.4 Hz, 1H), 4.32 (q, J = 6.4 Hz, 1H),
4.10 (t, J = 7.2 Hz, 1H), 3.54 (t, J = 5.8 Hz, 2H), 3.16 (dd, J = 13.8, 6.1 Hz, 1H), 3.08 (dd, J =
13.8, 8.4 Hz, 1H), 1.17 (d, J = 7.3 Hz, 3H). 13C NMR (125 MHz, DMSO-d6): δ 193.13, 172.39,
170.82, 170.04, 155.74, 138.12, 136.98, 136.25, 135.63, 135.45, 130.71, 129.10, 128.39, 128.37,
128.06, 127.89, 127.81, 127.74, 127.49, 66.17, 65.43, 61.64, 54.93, 53.44, 50.06, 36.25, 18.21.
IR (ATR, cm−1): 3286 (br, s), 3065 (w), 3034 (w), 2935 (w), 1720 (w), 1689 (m), 1639 (s), 1530
(s), 1452 (m), 1388 (w), 1344 (w), 1241 (s), 1125 (w), 1051 (m), 1027 (m), 952 (w), 907 (w),
843 (w), 746 (m), 694 (s). HRMS (ESI+): Calculated for C31H34N3O8+ [M + H]+ : 576.2340;
found: 576.2341.
ethyl 4-(8-formyl-5,6-dihydro-11H-benzo[5,6]cyclohepta[1,2-b]pyridin-11-ylidene) piper-
idine-1-carboxylate (A.30): Synthesized following the general procedure from loratadine
(95.7 mg, 0.25 mmol, TCI). Purified by automated column chromatography (25 g silica, 20 →
60% ethyl acetate in hexanes with 5% triethylamine additive) to afford 78.0 mg of a colorless oil.
NMR analysis indicated this to be a mixture of 69.8 mg product (74% yield) and 8.2 mg of the
acetal isomer ethyl 4-(8-(1,3-dioxolan-4-yl)-5,6-dihydro-11H-benzo[5,6]cyclohepta[1,2-b]-
pyridin-11-ylidene)piperidine-1-carboxylate (8% yield, 9.2:1 ratio). A second run yielded
73.2 mg product (78% yield) and 7.3 mg of the of the 1,3-dioxolan-4-yl acetal isomer (7% yield,
11.4:1 ratio). 1H NMR (500 MHz, CDCl3): δ 9.94 (s, 1H), 8.41 (d, J = 4.4 Hz, 1H), 7.71 (s, 1H),
7.67 (d, J = 7.8 Hz, 1H), 7.45 (d, J = 7.6 Hz, 1H), 7.36 (d, J = 7.7 Hz, 1H), 7.11 (dd, J = 7.7, 4.8
162
Hz, 1H), 4.12 (q, J = 7.1 Hz, 2H), 3.89 – 3.74 (m, 2H), 3.48 (ddd, J = 16.3, 10.6, 4.7 Hz, 2H),
3.15 (ddt, J = 13.4, 8.7, 4.1 Hz, 2H), 2.97 – 2.81 (m, 2H), 2.50 (ddd, J = 14.2, 9.4, 4.6 Hz, 1H),
2.33 (dtt, J = 23.7, 13.8, 4.5 Hz, 3H), 1.23 (t, J = 7.1 Hz, 3H). 13C NMR (125 MHz, CDCl3): δ
192.01, 156.17, 155.56, 146.85, 146.09, 138.96, 138.45, 137.98, 135.70, 134.49, 133.54, 130.00,
129.91, 128.24, 122.61, 61.48, 44.86, 44.86, 31.70, 31.68, 30.97, 30.67, 14.79. IR (film, cm−1):
2979 (w), 2913 (w), 2858 (w), 2730 (w), 1689 (s), 1601 (w), 1566 (w), 1428 (m), 1384 (w), 1300
(w), 1277 (w), 1221 (s), 1171 (w), 1113 (m), 1061 (w), 1026 (w), 996 (m), 898 (m), 859 (w), 831
(w), 800 (w), 766 (w), 728 (m), 675 (w). HRMS (ESI+): Calculated for C23H25N2O3+ [M + H]+ :
377.1860; found: 377.1854.
1-methyl-2-oxo-5-phenyl-2,3-dihydro-1H-benzo[e][1,4]diazepine-7-carbaldehyde (A.31):
Synthesized according to the general procedure from diazepam (71.2 mg, 0.25 mmol). Purified
by automated column chromatography (50 g silica, 0 → 60% ethyl acetate in hexanes) to afford
50.2 mg product (72% yield) as a pale yellow solid. A second run provided a mixture of 51.4 mg
product (74% yield) and 6.3 mg of the acetal isomer side product 7-(1,3-dioxolan-4-yl)-1-
methyl-5-phenyl-1,3-dihydro-2H-benzo[e][1,4]diazepin-2-one (8% yield, 9.5:1 ratio). 1H NMR
(500 MHz, CDCl3): δ 9.93 (s, 1H), 8.08 (dd, J = 8.6, 2.0 Hz, 1H), 7.83 (d, J = 1.9 Hz, 1H), 7.58
(dd, J = 8.3, 1.4 Hz, 2H), 7.49 (dd, J = 10.5, 8.0 Hz, 2H), 7.42 (dd, J = 8.2, 6.8 Hz, 2H), 4.88 (d,
J = 10.9 Hz, 1H), 3.78 (d, J = 10.9 Hz, 1H), 3.46 (s, 3H). 13C NMR (125 MHz, CDCl3):
δ 190.19, 169.88, 169.75, 148.65, 138.35, 133.36, 131.72, 131.34, 131.00, 129.62, 129.17,
128.63, 121.80, 57.10, 35.08. IR (ATR, cm−1): 3057 (w), 2922 (w), 2852 (w), 1733 (w), 1676
(s), 1608 (s), 1596 (w), 1570 (w), 1489 (w), 1474 (w), 1446 (m), 1420 (m), 1374 (m), 1330 (s),
163
1275 (m), 1243 (w), 1208 (m), 1195 (w), 1123 (s), 1071 (s), 1025 (w), 985 (m), 913 (w), 871
(w), 831 (m), 783 (m), 757 (w), 746 (s), 721 (m), 698 (s), 672 (w). HRMS (ESI+): Calculated
for C17H15N2O2+ [M + H]+: 279.1128; found: 279.1128.
benzyl 2-(5-(4-formylbenzoyl)-1,4-dimethyl-1H-pyrrol-2-yl)acetate (A.32): Synthesized
according to the general procedure from benzyl 2-(5-(4-chlorobenzoyl)-1,4-dimethyl-1H-pyrrol-
2-yl)acetate (95.5 mg, 0.25 mmol). Purified by automated column chromatography (50 g silica, 0
→ 25% ethyl acetate in hexanes) to afford 69.9 mg product (74% yield) as a yellow solid. A
second run afforded a mixture of 68.4 mg product (73% yield) and 9.5 mg of the acetal isomer
side product benzyl 2-(5-(4-(1,3-dioxolan-4-yl)benzoyl)-1,4-dimethyl-1H-pyrrol-2-yl)acetate
(9% yield, 8.0:1 ratio). 1H NMR (500 MHz, CDCl3): δ 10.10 (s, 1H), 7.96 (d, J = 8.1 Hz, 2H),
7.80 (d, J = 7.9 Hz, 2H), 7.41 – 7.31 (m, 5H), 5.95 (s, 1H), 5.19 (s, 2H), 3.76 (s, 3H), 3.71 (s,
2H), 1.68 (s, 3H). 13C NMR (125 MHz, CDCl3): δ 191.93, 186.52, 169.27, 146.42, 138.03,
135.48, 133.84, 130.21, 129.85, 129.65, 129.39, 128.79, 128.67, 128.49, 113.10, 67.34, 33.45,
32.90, 14.69. IR (ATR, cm−1): 3032 (w), 2921 (w), 2851 (w), 1721 (s), 1700 (s), 1615 (s), 1567
(w), 1501 (w), 1485 (w), 1455 (m), 1425 (w), 1385 (s), 1376 (s), 1323 (m), 1301 (w), 1271 (m),
1220 (w), 1183 (s), 953 (m), 859 (w), 840 (m), 804 (m), 759 (m), 744 (s), 697 (s). HRMS
(ESI+): Calculated for C21H22O5+ [M + H]+ : 376.1543; found: 376.1547.
164
N-benzyl-6-formyl-3-indolyl-β-D-galactopyranoside tetraacetate (A.33): Synthesized
according to the general procedure from N-benzyl-6-chloro-3-indolyl-β-D-galactopyranoside
tetraacetate (147.0 mg, 0.25 mmol). Purified by automated column chromatography (50 g silica,
10 → 45% ethyl acetate in hexanes) and obtained 59.5 mg of a pale orange solid. NMR analysis
indicated that this consisted of 53.9 mg desired product (37% yield) and 5.6 mg of the acetal
isomer N-benzyl-6-(1,3-dioxolan-4-yl)-3-indolyl-β-D-galactopyranoside tetraacetate (4% yield,
10.3:1 ratio). A second run afforded 55.7 mg product (38% yield) and 7.2 mg acetal isomer side
product (5% yield, 8.1:1 ratio). 1H NMR (500 MHz, CDCl3): δ 10.00 (s, 1H), 7.82 (s, 1H), 7.68
(d, J = 8.2 Hz, 1H), 7.62 (d, J = 8.3 Hz, 1H), 7.36 – 7.28 (m, 3H), 7.13 (s, 1H), 7.10 (d, J = 6.3
Hz, 2H), 5.53 (dd, J = 10.5, 8.0 Hz, 1H), 5.45 (d, J = 3.1 Hz, 1H), 5.34 (s, 2H), 5.10 (dd, J =
10.5, 3.4 Hz, 1H), 4.90 (d, J = 8.0 Hz, 1H), 4.19 (qd, J = 11.3, 6.6 Hz, 2H), 3.98 (t, J = 6.6 Hz,
1H), 2.18 (s, 3H), 2.15 (s, 3H), 2.02 (s, 3H), 1.97 (s, 3H). 13C NMR (125 MHz, CDCl3): δ
192.32, 170.31, 170.22, 170.14, 169.41, 136.90, 136.50, 133.41, 131.58, 129.01, 128.12, 126.65,
124.88, 120.42, 119.95, 118.28, 112.70, 102.60, 71.17, 70.78, 68.74, 66.92, 61.45, 50.23, 20.88,
20.69, 20.61, 20.61. IR (ATR, cm−1): 2927 (w), 2855 (w), 1743 (s), 1686 (m), 1612 (w), 1545
(m), 1469 (w), 1454 (w), 1367 (m), 1214 (s), 1162 (m), 1048 (s), 953 (w), 903 (w), 812 (w), 772
(w), 741 (w), 704 (m). HRMS (ESI+): Calculated for C30H32NO11+ [M + H]+ : 582.1970; found:
582.1976.