Supplemental materials for

Nitrogen detected TROSY at high field yields high resolution and sensitivity for protein NMR

Koh Takeuchi†,§,‡, Haribabu Arthanari &,‡, Ichio Shimada*,†, ‖, and Gerhard Wagner*,&

†Molecular Profiling Research Center for Drug Discovery, National Institute for Advanced Industrial Science and Technology, Tokyo 135-0063, Japan, § JST, PRESTO, Tokyo 135-0063, Japan, &

Department of Biochemistry and Molecular Pharmacology, Harvard Medical School, Boston, MA 02115. ‖ Graduate Schools of Pharmaceutical Sciences, The University of Tokyo, Tokyo 113-0033,

Japan

*; equal contribution

Materials and methods

All chemicals were purchased from Sigma (St. Louis, MO) unless otherwise noted. All stable-isotope-labeled

materials were acquired from Cambridge Isotope laboratories (Cambridge, MA).

Expression and purification of the B domain of protein G (GB1)

The gene for His6-tagged GB1, consisting of 64 amino acid residues, was cloned into the pET9d vector (Novagen,

San Diego, CA) as previously described (Frueh et al., 2005). GB1 was expressed in commercially available BL21

(DE3) E.coli cells (Novagen) at 37 ˚C and protein expression was induced for 6 h at the same temperature. For

uniformly 2H15N13C labeled samples, the cells were cultured in 2H, 15N, 13C M9 media containing 8.5 g/L Na2HPO4, 3

g/L KH2PO4, 0.5 g/L NaCl, 2mM MgCl2, 0.1 mM CaCl2 in D2O, which was supplemented with 2g/L 2H13C glucose and

1 g/L of 15NH4Cl. For uniformly 15N13C labeled samples, the cells were cultured in 15N, 13C M9 media containing 8.5

g/L Na2HPO4, 3 g/L KH2PO4, 0.5 g/L NaCl, 2mM MgCl2, 0.1 mM CaCl2 in H2O, which was supplemented with 2g/L

13C glucose and 1 g/L of 15NH4Cl. The protein was purified with Ni-NTA affinity chromatography as previously

described (Frueh et al., 2005).

Expression and purification of MBP

A 42-kDa maltose binding protein (MBP), which has been extensively studied in E. coli for perdeuteration

and methyl-selective protonation proteins was used in this study(Gardner et al., 1998). MBP was expressed in

commercially available BL21 (DE3) E.coli cells (Novagen) at 37 ˚C and protein expression was

induced for overnight at the 30°C. For uniformly 2H15N13C labeled samples, the cells were

cultured in 2H, 15N, 13C M9 media containing 8.5 g/L Na2HPO4, 3 g/L KH2PO4, 0.5 g/L NaCl, 2mM

MgCl2, 0.1 mM CaCl2 in D2O, which was supplemented with 2g/L 2H13C glucose and 1 g/L of 15NH4Cl. MBP was purified in a buffer containing 25 mM Tris–HCl (pH 7.6), 300 mM NaCl, and 1

mM EDTA, and was then loaded onto amylose resin (3 ml, New England Biolabs, Beverly, MA,

USA), which was equilibrated with the buffer described above. After washing the resin, MBP

was eluted with the same buffer supplemented with 10 mM maltose.

NMR Experiments

NMR spectra were recorded on a Bruker (Billerica, MA) Avance 500 or Avance III 800 spectrometer equipped with

a triple-resonance cryogenic probe (TXO) equipped with the carbon/nitrogen inner coil and cryogenic preamps, which

are designed for heteronuclear-detection experiments. Spectra of the uniformly labeled GB1 sample were recorded at

286 K in buffer containing 5 mM sodium phosphate (pH 6.8), 50 mM NaCl. Spectra of the uniformly labeled MBP

sample were recorded at 290 K and 278 K in buffer containing 10 mM HEPES-NaOH (pH 7.4), 1 mM EDTA, 1 mM

β-cyclodextrin. The 2D 15N-detected HN HSQC experiments (15N 2D TROSY, 15N 2D IPAP and 15N 2D proton-

decoupled HSQC) were recorded with spectral widths of 40 ppm ( 15N, direct) and 6 ppm (1H, indirect). The two-

dimensional CON experiments were recorded with spectral widths of 40 ppm ( 15N, direct) and 24 ppm (13C, indirect).

In both spectra, 1H, 13C and 15N frequencies were centered at 8.5 ppm, 172 ppm, and 120 ppm, respectively.

Acquisition times were set to 0.63 sec (15N, direct dimension), 0.043 sec (1H, indirect dimension) and 0.037 sec (13C,

indirect dimension) for GB1. Whereas, 0.32 sec (15N, direct dimension) and 0.093 sec (1H, indirect dimension) were

used for 15N TRSOY experiment of MBP and 0.32 sec (15N, direct dimension) and 0.042 sec (1H, indirect dimension)

were used for IPAP and proton-decoupled HSQC experiments of MBP. For GB1, 2D 15N-detected HN HSQC

experiments (15N 2D TROSY, 15N 2D IPAP and 15N 2D proton-decoupled HSQC) experiments were recorded for 3.5 hr

in 800 MHz and 11 hr in 500 MHz. As for MBP, 15N TRSOY experiment was recorded for 8 hr in 800 MHz. IPAP and

proton-decoupled HSQC experiments of MBP were recorded for 13hr and 18 hr in 800 and 500 MHz, respectively.

Deuterium decoupling are achieved by using WALTZ16 (Shaka et al., 1983) (1 kHz) and carbon decoupling was

achieved by using a p5m4 supercycle (Fujiwara and Nagayama, 1988) with an adiabatic CHIRP pulse of 2.5 ms length

and 25% smoothing(Bohlen and Bodenhausen, 1993). Zero filling was applied for each FID but no apodization

function was used for the direct dimension before Fourier transformation. For the indirect dimensions, zero filling and

a cosine-apodization function were applied. The recycling delay was set to 1.0 s for proton excited experiment and 2.0

sec for 13C exited experiments. All spectra were processed with TOPSPIN (Bruker; Billerica, MA) and analyzed with

the program Sparky (Goddard and Kneller).

Calculation of transverse relaxation rates

The transverse relaxations of decoupled spin I (R2 I ¿were calculated based on the standard expression as follows,

R2 I=∑j

n

p I S j

2 ¿(1)

p I S j=−(

μ0

2√2)

γ I γ S jℏ

rI S j

3 (2)

δ I=−(μ0

3√2)γ I B0 ∆ σ I (3)

∆ σ I is the difference between the axial and the perpendicular principal components of the axially symmetric chemical shift tensors. Spectrum density were estimated using equation 4

J(ω) = 2τc / (5 x (1+ (τc ω)2) (4)

where, ω is the Larmor frequencies of the spins 1H and 15N, or sum or difference of those.

The TROSY (R2 I T¿ and anti-TROSY (R2 I AT

¿ transverse relaxation of spin I were calculated as follows,

R2 I T=R2 I−¿C p IS δI [ 4 J (0 ) ]∨¿ (5)

R2 I AT=R2 I+¿C p IS δ I [4 J (0 ) ]∨¿ (6)

C=(3cos2Θ−1) (7)

where Θ is the angle between the unique tensor axes of the dipole-dipole interactions and CSA interaction. For

calculation of transverse relaxation rates, the dipole-dipole interactions with directly bonded nuclei as well as nuclei

within defined distances and chemical shift anisotropies were taken into consideration to mimic the typical proton

density in a protein. For the estimation of the dipole-dipole interactions in a uniformly 15N-labeled protein, the typical

distances in α-helix region were used with distance cut-off criteria as follows: < 4.5 Å for 1H-1H pairs and < 2.1 Å for 1H-15N pairs. These are sequential HNi-HNi+1 distance (2.8 Å), HNi-HNi+2 distance (4.2 Å), intra residue HNi-Hαi

distance (2.2 Å), sequential HNi-Hαi+1 distance (3.5 Å), HNi-Hαi+2 distance (4.4 Å) , HNi-Hαi+3 distance (3.4 Å), HNi-Hαi+3

distance (4.2 Å), intra residue HNi-Hβi distance (2.5 Å), and sequential HNi-Hβi+1 distance (3.0 Å) for proton pairs.

Contribution from remote protons that are more than 4 residues away in a.a. sequences are not considered. As for the

estimation of the dipole-dipole interactions in a uniformly 2H15N-labeled protein, the same nuclei that were taken into

consideration in the case of a uniformly 1H15N-labeled protein were used, while all non-exchangeable protons are

replaced to deuterium. Chemical shift anisotropies were also included in the calculation (15NH: -160ppm, 1HN: 16

ppm). The angle between the unique tensor axes of the dipole-dipole interactions and CSA interaction (Θ ¿was set to

17° for 15NH.

Supplemental Discussion

Experimental verification of 15 N TROSY effect

In addition to the 15N-detected TROSY-HSQC experiment, an 15N-detected 2D 1H-15N IPAP-HSQC experiment (15N

IPAP-HSQC) that separately records TROSY and anti-TROSY 15NH resonances, an 15N-detected proton-decoupled 2D

1H-15N HSQC (15N 1H-decoupled HSQC) experiment, and an 15N-detected 2D CON experiment (Takeuchi et al., 2010)

were also recorded. Figure S4 shows 15N-detected experiments of 2H15N13C-labeled GB1 (8 kDa) recorded in a D2O

buffer at 286K. The rotational correlation time of GB1 under these experimental conditions was 7 ns and the calculated

line width for TROSY 15NH, proton-decoupled 15NH, anti-TROSY 15NH, and deuterium-decoupled 15ND resonances are

0.7 Hz, 3.9 Hz, 7.1 Hz, and 1.7 s-1 for 800 MHz, and 1.1 Hz, 3.2 Hz, 5.4 Hz, and 1.0 Hz for 500 MHz, respectively.

These experiments were conducted on the GB1 protein that was incubated in the D2O buffer for >12 hrs at 298K to

make all 15NH sites that are rapidly exchanging with solvent to 15ND (Figure S4A and B). Therefore, only the residual

15NH groups that resist to solvent exchange were observed in the proton-excited experiments, which minimizes the

broadening due to the solvent exchange (Thakur et al., 2013). The experiments were performed in 800 MHz and 500

MHz magnets. Both were equipped with cryogenic TXO probes with a cold 15N preamplifier. The measuring times for

the proton-excited HN and carbon-excited CON experiments were 3.5 hr and 12 hr, respectively. The long recording

time for proton-excited 2D experiments is due to the fact that small fraction of sites are proton attached after >12 hrs of

H/D exchange. The acquisition times for the 15N-detected experiments were set to 0.63 sec, which yielded a digital

resolution of 0.8 Hz after Fourier transform.

As expected, TROSY 15N resonances in the 15N-detected 2D HN and CON experiments were substantially narrower

than the 1H-decoupled 15NH, anti-TROSY 15NH, and 2H-decoupled 15ND resonances at 800 MHz. For instance, the line

width of Asp-46 15N was 1.2 Hz, 3.7 Hz, 8.3 Hz, and 2.5 Hz for TROSY 15NH, proton-decoupled 15NH, anti-TROSY

15NH, and 2H-decoupled 15ND resonances, respectively (Figure S4C). The experimentally measured line widths were in

close agreement with the calculated linewidths within the margin of error from the digital resolution of the spectra (0.8

Hz). The narrow line width is directly reflected in the gain of signal height. TROSY 15NH signal of Asp-46 15N was 3.6-

and 6.3-times more intense than 1H-decoupled 15NH and anti-TROSY 15NH resonances that were recorded with the same

pulse scheme, respectively. The 15N-detected TROSY-HSQC experiment yield 1.2-times greater signal height

compared to the 15N IPAP-HSQC experiment presumably due to favorable TROSY relaxation properties in the indirect

proton dimension. In contrast, the 15N TROSY effect in a 500 MHz magnet is less significant compared to that in an

800 MHz and the 2H-decoupled 15ND resonances have almost same linewidth as the 15NH TROSY resonances (Figure

S4D). This clearly demonstrates the benefit of selecting TROSY 15NH in higher magnetic field.

The line width of the TROSY resonances were almost identical in H2O compared to D2O while 0.03-0.04 ppm

isotope shifts were observed presumably due to the protons vs deuterium isotope effect from sequentially neighboring

amides (Figure S8A and C). The isotope effect caused by the splitting of resonances when mixed H 2O/D2O are used

(Figure S8B). Therefore, minimal amount of D2O should be used for the field-frequency lock when 15N-detected

experiments are recorded in H2O. One could also prepare a H2O sample with a few % of non-exchangeable deuterated

solvent such as DMSO for the purpose of locking.

Since no proton decoupling was applied in the TROSY pulse schemes, the narrow line width of TROSY 15NH is also

sensitive to multiple-bond scalar couplings from nearby protons, such as Hα ( 2JNiHi and 3JNiHi-1) and Hβ (3JNiHi) in the case

of protonated sample. As it is shown in Figure S9, the TROSY 15NH line width of protonated GB1 looks broader

compared to deuterated GB1, while the line width of 1H-decoupled 15NH resonances are almost identical in deuterated

and protonated samples. This can be a drawback of using the protonated sample, however, it would not be an issue for

a larger molecular weight systems since these couplings are rather small (< 2 Hz) when compared to the line width.

The benefit of 15N-direct detection of a large protein at higher field magnets were assessed with 650 μM 2H13C15N-

labeled maltose binding protein (MBP) in complex with β-cyclodextrin (βCD) in 800 MHz and 500 MHz magnets at

290 K (Figure S10). The rotational correlation time of the system, at the above conditions was 25 ns as deduced by

TRACT experiments (Lee et al., 2006). The expected line width of TROSY and anti-TROSY components are 2.2 Hz

and 24 Hz at 800 MHz, and 3.5 Hz and 17 Hz in 500 MHz spectrometer, respectively. The maximum acquisition time

for 15N dimension was set to 320 ms in order to make sure the resultant spectra have adequate resolution reflecting their

relaxation rates. The 15N-detected TROSY-HSQC experiment of βCD-bound MBP in 800 MHz magnetic field is

shown in Figure S10A and B, which clearly demonstrates the narrow TROSY 15NH line widths even for the large

protein with 25 ns correlation time. There is a dramatic difference between TROSY and anti TROSY resonances

(Figure S10C, left). As expected, the benefit from TROSY is less significant in the experiment recorded at 500 MHz

(Figure S10C, right).

Supplemental Figures

Figure S1 Expected transverse relaxation rates of low -nuclei as a function of the overall correlation time for a uniformly 15N13C-labeled protein. The transverse relaxation rates were calculated based on the equation (1) in the supplemental materials and methods. Magnetic field strength is set to 18.8 T, corresponding to proton frequency of 800 MHz. The dipole-dipole interactions with directly bonded nuclei as well as nuclei within defined distances were used to mimic the typical proton density in a protein. Typical distances in -helix region were used for the estimates. Chemical shift anisotropies were also included for the calculation (15NH: -160ppm, 13C: 2.2 ppm, 13C': 170 ppm). Note: the relaxation rates are not for TROSY selections. The values were taken from reference 1 in the main text.

Figure S2: Relaxation of TROSY 15NH, 15ND and TROSY 1HN of 7 ns, 60 ns, and 120 ns systems at various magnetic strength. (A) The calculated R2 for TROSY 15NH (gray) and 15ND (white) at indicated magnetic strength in proton frequency. (B) R2 ratio of 15ND over TROSY 15NH. (C) The calculated R2 for TROSY 1HN. The transverse relaxation rates were calculated based on the equation (1) and (5) in the supplemental materials and methods.

Figure S3: Effect of deuteration on R2 of TROSY component of 15NH and 1HN. The transverse relaxation rates of 20 ns system in 800 MHz magnet with indicated labeling condition are shown. The dipole-dipole interactions with directly bonded nuclei as well as nuclei within defined distances were used to mimic the typical proton density in a protein. Typical distances in α-helix region were used for the estimates. The transverse relaxation rates were calculated based on the equation (5) in the supplemental materials and methods.

Figure S4: 15N-detected 2D experiments on GB1 in 800 MHz and 500 MHz magnets at 286K. (A and B) Expanded (A) 15N 2D TROSY-HSQC spectrum (black), TROSY (red) and anti-TROSY (blue) spectrum deduced from 15N 2D IPAP-HSQC experiment, and 15N 2D 1H-decoupled HSQC spectrum (purple) of 2H15N13C-labeled GB1 in D2O buffer in 800 MHz at 286 K. The samples were equilibrated in D2O buffer >12 hr at 298K and only the sites that are resistant to H/D exchanges are observed. All spectra were recorded for 3.5 hr. The acquisition time for 15N and 1H dimensions were 0.63 sec and 0.043 sec, respectively. T1 relaxation delay was set to 1 sec. (B) Expanded 15N-detected 2D CON spectra that was recorded with the same sample as (A). The spectrum was recorded for 12 hr. (C) Expanded region of panel (A) and (B) showing Asp-46 resonances that are recorded in 800 MHz magnet with indicated experimental time. Color code of the spectra were same as (A) and (B). For proton-excited spectra, slice of each resonances are shown in the same intensity scale. (D) Same as (C) except the spectrum was recorded in 500 MHz magnet. Note that the 15N 2D TROSY-HSQC with ST2PT transfer selects the TROSY component in both 1H and 15N dimensions, while IPAP TROSY and anti-TROSY as well as decoupled peaks do not have TROSY selection in 1H dimension. Thus the signals in 15N TROSY 2D TROSY-HSQC spectrum are displaced by ~45 Hz in the proton dimension.

Figure S5: Comparison of linear-sampled and NUS-sampled 15N-detected TROSY-HSQC. (A) Linear sampling (B) 40% NUS sampling and (C) 25% NUS sampling spectra were shown. For (B) and (C), 40% and 25% of indirect points were extracted in the Poisson-gap sampling manner(Hyberts et al., 2010) to generate the NUS data from the linearly sampled data. The NUS data were reconstructed by hmsIST procedures(Hyberts et al., 2012). The 40% sampled spectrum reproduce all signals in linearly sampled data. Most of the signals were successfully reproduced even with 25% sampled data. Note that the NUS data were not time-equivalent acquisitions but represent 40% and 25% shorter overall measurement times.

Figure S6: Transverse relaxation rates of TROSY 15N and TROSY 1H resonances as a function of the rotational correlation time at 800 MHz. The transverse relaxation rates were calculated based on equation (5) in the supplemental materials and methods using the same parameters that were used to produce Figure 1.

Figure S7: Transverse relaxation rates of 15N resonances as a function of the rotational correlation time at 800 MHz. The transverse relaxation rates were calculated based on equation (1) and (5) in the supplemental materials and methods using the same parameters that were used to produce Figure 1.

Figure S8: The isotope shifts induced by sequential protons to TROSY 15NH

resonances. The TROSY 15NH resonances of Asp-46 and Val-6 recorded in the various H2O:D2O ratios are shown. H2O:D2O ratios were (A) 95:5, (B) 2:1, and (C) 0:100. The chemical shift difference between the resonances with 1H in both i-1 and i+1 sites compared to that with both 2H were 0.04 ppm for Asp-46 and 0.03 ppm for Val-6. The spectra were recorded at a field strength of 800 MHz. The magnitude of the isotope shift vary and depends on the site, presumably reflecting the relative orientation between the detected nuclei and the sequentially neighing amides protons.

Figure S9: Effect of 2J and 3J multiple bond 15N-1H coupling on TROSY 15NH

resonances. The resonances of Val-6 recorded with 2H15N13C-labeled (top) and 15N13C-labeled (bottom) GB1. TROSY 15NH resonance (red) of protonated GB1 was a few Hz broader compared to deuterated GB1. This is not due to the additional dipole interaction, which would be expected to equally affect both the TROSY 15NH and proton-decoupled 15NH

resonance but the line width of proton-decoupled 15NH resonance is almost invariant (purple). This would reflect the multiple splitting induced by 2J or 3J scalar couplings to Hα and Hβ that are not decoupled in TROSY experiment but are effectively decoupled in proton-decoupled 15NH HSQC experiments.

Figure S10: 15N-detected TROSY-HSQC experiment of βCD-bound MBP at 800 MHz and 500 MHz magnets. (A) 15N-detected TROSY-HSQC spectrum of 650 μM 2H15N-labeled MBP in complex with 2 mM βCD in 95% H2O, 5% D2O

buffer in 800 MHz at 290 K. The experimental time was 8hr. (B) Expanded spectra of squared region of 15N-detected

TROSY-HSQC spectrum in (A). (C) Slice of IPAP experiment for the dotted resonance in 15N-detected TROSY-HSQC

spectrum. TROSY (upper) and anti-TROSY (lower) resonances are shown. The spectra were recorded for 13 hr for 800

MHz and 18 hr for 500 MHz. The noise levels were indicated by the transparent grey bar.

Supplemental References

Bohlen, J.M., and Bodenhausen, G. (1993). Experimental Aspects of Chirp NMR Spectroscopy. J Mag Res A 102, 293-301.

Frueh, D.P., Arthanari, H., and Wagner, G. (2005). Unambiguous assignment of NMR protein backbone signals with a time-shared triple-resonance experiment. J Biomol NMR 33, 187-196.

Fujiwara, T., and Nagayama, K. (1988). Composite inversion pulses with frequency switching and their application to broadband decoupling. J Mag Res (1969) 77, 53-63.

Gardner, K.H., Zhang, X., Gehring, K., and Kay, L.E. (1998). Solution NMR Studies of a 42 KDa Escherichia Coli Maltose Binding Protein/β-Cyclodextrin Complex:  Chemical Shift Assignments and Analysis. J Am Chem Soc 120, 11738-11748.

Goddard, T.D., and Kneller, D.G. SPARKY 3. University of California, San Francisco.Hyberts, S.G., Milbradt, A.G., Wagner, A.B., Arthanari, H., and Wagner, G. (2012). Application of

iterative soft thresholding for fast reconstruction of NMR data non-uniformly sampled with multidimensional Poisson Gap scheduling. J Biomol NMR 52, 315-327.

Hyberts, S.G., Takeuchi, K., and Wagner, G. (2010). Poisson-gap sampling and forward maximum entropy reconstruction for enhancing the resolution and sensitivity of protein NMR Data. J Am Chem Soc 132, 2145-2147.

Lee, D., Hilty, C., Wider, G., and Wüthrich, K. (2006). Effective rotational correlation times of proteins from NMR relaxation interference. J Magn Reson 178, 72-76.

Takeuchi, K., Heffron, G., Sun, Z.Y., Frueh, D.P., and Wagner, G. (2010). Nitrogen-detected CAN and CON experiments as alternative experiments for main chain NMR resonance assignments. J Biomol NMR 47, 271-282.

Thakur, A., Chandra, K., Dubey, A., D'Silva, P., and Atreya, H.S. (2013). Rapid characterization of hydrogen exchange in proteins. Angew Chem Int Ed 52, 2440-2443.