Dynamic metabolic labeling of DNA in vivowith arabinosyl nucleosidesAnne B. Neef and Nathan W. Luedtke1

Institute of Organic Chemistry, University of Zurich, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland

Edited by Carolyn R. Bertozzi, University of California, Berkeley, CA, and approved October 6, 2011 (received for review January 29, 2011)

Commonly used metabolic labels for DNA, including 5-ethynyl-2′-deoxyuridine (EdU) and BrdU, are toxic antimetabolites that causeDNA instability, necrosis, and cell-cycle arrest. In addition to per-turbing biological function, these properties can prevent metaboliclabeling studies where subsequent tissue survival is needed. Tobypass the metabolic pathways responsible for toxicity, whilemaintaining the ability to be metabolically incorporated into DNA,we synthesized and evaluated a small family of arabinofuranosyl-ethynyluracil derivatives. Among these, (2′S)-2′-deoxy-2′-fluoro-5-ethynyluridine (F-ara-EdU) exhibited selective DNA labeling, yethad a minimal impact on genome function in diverse tissue types.Metabolic incorporation of F-ara-EdU into DNA was readily detect-able using copper(I)-catalyzed azide–alkyne “click” reactions withfluorescent azides. F-ara-EdU is less toxic than both BrdU and EdU,and it can be detectedwith greater sensitivity in experiments wherelong-term cell survival and/or deep-tissue imaging are desired. Incontrast to previously reported 2′-arabino modified nucleosides andEdU, F-ara-EdU causes little or no cellular arrest or DNA synthesisinhibition. F-ara-EdU is therefore ideally suited for pulse-chase ex-periments aimed at “birth dating” DNA in vivo. As a demonstration,Zebrafish embryosweremicroinjectedwith F-ara-EdU at the one-cellstage and chased by BrdU at 10 h after fertilization. Following 3 dof development, complex patterns of quiescent/senescent cellscontaining only F-ara-EdU were observed in larvae along the dorsalside of the notochord and epithelia. Arabinosyl nucleoside deriva-tives therefore provide unique and effective means to introducebioorthogonal functional groups into DNA for diverse applicationsin basic research, biotechnology, and drug discovery.

chemical biology ∣ click chemistry ∣ fluorescent probe ∣ nucleosidemetabolism

The utilization of chemical techniques to address biologicalsystems is becoming increasingly important in basic research

and modern drug discovery (1). One underexplored area at thebiology–chemistry interface is the study of nucleic acids in theirnative environments. Traditional DNA and RNA imaging meth-odologies have utilized fluorescent fusion proteins, nonspecificstains for nucleic acids, immunostaining of BrdU-labeled DNA,or FISH (2). All of these approaches are limited in terms of theirlow throughput, large perturbations to native systems, and/orinability to be applied in unmodified cells and organisms.

Metabolic labeling of DNA has traditionally been performedusing [3H]thymidine or BrdU. These labels are limited in termsof their subsequent visualization, requiring either autoradiogra-phy, or DNA denaturation and antibody staining (3). BrdU im-munostaining is currently the most commonly used method, butit requires harsh chemical denaturation of cellular DNA and islimited by the poor tissue penetration of the BrdU antibody.In addition, BrdU itself is both toxic and mutagenic when appliedat high concentrations, and it can have a negative impact on DNAstability and the cell cycle (4).

The recent emergence of bioorthogonal chemical reporterstrategies has revolutionized the study of biological macromole-cules in their native environments (5–19). In this two-step ap-proach, a synthetic label containing a bioorthogonal functionalgroup is metabolically incorporated into the cellular target and

subsequently probed using a chemoselective reaction. Impor-tantly, only a small modification like an azide or a terminal alkyneis initially introduced, rendering the structure and function ofthe biomolecule virtually unchanged (5, 6). The most commonlyused chemoselective reactions for probe conjugation include theStaudinger ligation (7), copper(I)-catalyzed azide–alkyne cyclo-addition (CuAAC) (8), and strain-promoted azide–alkyne cyclo-addition reactions (9, 10). These reactions have been used toselectively label cellular proteins (11, 12), polynucleotides (13, 14),lipids (15, 16), and glycans (17–19) in living cells and animals.

A strategy for the bioorthogonal chemical labeling of DNAwasrecently reported using the deoxythymidine analog 5-ethynyl-2′-deoxyuridine (EdU) (13, 20, 21). After its metabolic incorpora-tion into DNA by living cells, the ethynyl groups of EdU couldbe visualized with high sensitivity following CuAAC stainingwith fluorescent azides (13). Unlike BrdU, this method does notrequire sample fixation or DNA denaturation prior to detection,and the sensitivity of EdU detection can exceed the limits ofclassical BrdU detection (22). Despite these advantages, EdU ismuch more toxic than BrdU, and its use over prolonged experi-mental times results in highly nonuniform CuAAC staining dueto cell-cycle arrest (22–27). In 2008, this EdU-based labelingstrategy was commercialized by Invitrogen and subsequently usedin cell proliferation and differentiation studies (28), for measur-ing nucleotide excision repair activity (29), for the analysis ofnuclear architecture (30), in tissue regeneration studies (31), forcell-cycle analysis (32, 33), and in at least 100 other publishedstudies to date.

Despite its rapidly growing popularity, EdU is a highly toxicantimetabolite that perturbs DNA function and stability (22–24, 34). The EdU inhibitory concentrations for cellular growth(IC50 ¼ 0.2–4 μM) and G2∕M cell-cycle arrest (IC50 ¼ 0.2 μM)(22, 25–27) are approximately 50-fold lower than those requiredfor efficient DNA labeling and detection (10 μM) (13, 28–33).In addition to perturbing biological function in complex ways,these properties complicate or even prevent metabolic labelingstudies where subsequent tissue survival is needed. We speculatedthat nucleoside derivatives having a D-arabinose (ara) configura-tion might bypass the pathways responsible for the antimetabolicactivities of EdU, BrdU, and other deoxyribosyl nucleoside deri-vatives. According to previous studies, however, the metabolicincorporation of arabinosyl nucleoside analogs such as cytosinearabinonucleoside, 1-(2'-deoxy-2'-fluoro-β-D-arabinofuranosyl)uracil (FAU), or 9-(β-D-arabinofuranosyl)-2-fluoroadenine (F-ara-A) into DNA resulted in subsequent inhibition of DNAsynthesis, eventually resulting in cell death by apoptosis (35–37).It was therefore not clear a priori if an alkyne-modified arabinosylnucleoside could exhibit efficient labeling of cellular DNA with-

Author contributions: A.B.N. and N.W.L. designed research; A.B.N. performed research;A.B.N. and N.W.L. analyzed data; and A.B.N. and N.W.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: luedtke@oci.uzh.ch.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1101126108/-/DCSupplemental.

20404–20409 ∣ PNAS ∣ December 20, 2011 ∣ vol. 108 ∣ no. 51 www.pnas.org/cgi/doi/10.1073/pnas.1101126108

Dow

nloa

ded

by g

uest

on

Aug

ust 2

8, 2

020

out inhibiting cellular proliferation. Given the overriding impor-tance of genomic integrity, these properties may have been dia-metrically opposed by nature.

Here we report the design, synthesis, and evaluation of a smallfamily of 2′-arabino-modified 5-ethynyluridine derivatives. Amongthese, (2′S)-2′-deoxy-2′-fluoro-5-ethynyluridine (F-ara-EdU) ex-hibited selective DNA labeling, yet has a minimal impact on gen-ome function in diverse tissue types—including human, monkey,and mouse. Surprisingly, little or no cellular arrest or DNA synth-esis inhibition was observed following the incorporation of F-ara-EdU into DNA. It is therefore compatible with both long-term andpreparative labeling experiments of 60 d (or more) with constantfeeding of the label. These results provide direct evidence thatalkyne-containing nucleosides can be incorporated into cellulargenomes that remain replication competent. In addition to en-abling strategies in macromolecular engineering and drug discov-ery, arabinosyl nucleoside analogs can provide improvedmetaboliclabels for applications involving bioorthogonal chemical reporterstrategies. For example, F-ara-EdU is less toxic than both BrdUand EdU, and it can be detected with greater sensitivity in experi-ments where long-term cell survival and/or deep-tissue imaging areneeded. F-ara-EdU is also fully compatible with BrdU in pulse-chase experiments for birth dating of DNA in vivo. As a demon-stration of this technique, Zebrafish embryos were microinjectedwith F-ara-EdU at the one-cell stage, followed by a BrdU “chase”at 5 or 10 h postfertilization (hpf). At 72 hpf, the embryos werefixed and stained. Excellent colocalization of F-ara-EdU andBrdU was observed when the BrdU chase was applied at end ofblastula stage (5 hpf), whereas much less colocalization was ob-served in embryos chased with BrdU at the end of the gastrulastage (10 hpf). In the latter case, complex patterns of quiescent/senescent cells containing only F-ara-EdU were observed alongthe dorsal side of the notochord and epithelia. These observationsare consistent with the known timing of tissue development anddifferentiation in Zebrafish embryos (38, 39), and furnish high-resolution imaging of metabolically dated DNA in whole animals.Arabinosyl nucleoside derivatives therefore provide unique andeffective means to introduce bioorthogonal functional groups intoDNA in vivo. Due to its minimal impact on cellular proliferationand its persistence in labeled chromosomes, F-ara-EdU will

expand the dynamic range of DNA labeling experiments in basicresearch, biotechnology, and drug discovery.

ResultsDesign of 5-Ethynyl-Arabino-Uridine Derivatives with Minimal Cyto-toxic Activities. Despite the rapidly growing utilization of EdUin cell-based assays (13, 28–33), the mechanisms responsible forits potent cytotoxicity are highly diverse and only partially recog-nized (22, 24). EdU is known to inhibit a variety of enzymesinvolved in nucleoside metabolism (24, 40, 41), whereas its abilityto cause cell-cycle arrest is presumably mediated by DNA damageresponse mechanisms that regulate the G2∕M checkpoint (22, 42).Depending upon the cell type, EdU can cause mutagenic and gen-otoxic effects including sister chromatid exchange (23), cell-cyclearrest resulting in necrosis (22), and the inhibition of virus replica-tion (25). Taken together, these results suggest that EdU is capableof interacting with a wide variety of endogenous receptors, includ-ing nucleoside metabolizing enzymes and DNA binding proteinsthat specifically recognize the deoxyribosemoiety. This type of pro-miscuity is not surprising, because the structure of EdU is nearlyidentical to its natural counterpart deoxyribosyl-thymine (dT). Wereasoned that nucleoside and nucleotide derivatives having aD-arabinosyl (2′S) configuration should be capable of bypassingmany of the metabolic pathways that recognize deoxyribosyl moi-eties. Previous studies have shown that substituents at the 2′(S)position can have a modulating effect on the antimetabolic activ-ities of certain nucleoside derivatives (25), but it was not known apriori if any alkynyl nucleoside (arabinofuranosyl or otherwise)could be incorporated into cellular DNA at detectable levels with-out having a negative impact on cellular proliferation. To test thispossibility, we synthesized a small family of 5-ethynyl-arabino-uridine derivatives containing Me, OH, or F at the 2′(S) position(Fig. 1A and SI Appendix, Schemes 1–3). Given the potentialantimetabolic activities of these compounds, we used a standardAlamar Blue assay to assess the combined effects of proliferationand metabolism on total cellular respiration. Using this assay, weevaluated the acute cytotoxic activities of (2′S)-2′-deoxy-2′-C-methyl-5-ethynyluridine (Me-ara-EdU), 1-β-D-arabino-pentofura-nosyl-5-ethynyluracil (ara-EU), F-ara-EdU, and EdU in humancervical cancer cells (HeLa), mouse embryonic fibroblasts (3T3),and African green monkey epithelial kidney cells (Vero). After

Fig. 1. Arabinofuranosyl-5-ethynyluridine derivatives, acute toxicity, and metabolic labeling. (A) Structures of synthetic nucleosides studied in this report.(B) HeLa cell metabolism according to the Alamar Blue assay after 72 h in DMEMþ 10% FCS solutions containing variable concentrations of each nucleoside.See SI Appendix, Fig. S2 for cellular proliferation assays. (C) Metabolic labeling of genomic DNA in HeLa cells following a 24 h incubation with 10 μM of eachnucleoside. After washing and fixing the cells, ethynyl-modified DNAwas stained using AlexaFluor 488 azide and Cu(I), and total cellular DNAwas stained withDAPI. Negative controls received identical treatments, but were not exposed to a synthetic nucleoside. (Scale bar: 50 μm.)

Neef and Luedtke PNAS ∣ December 20, 2011 ∣ vol. 108 ∣ no. 51 ∣ 20405

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Aug

ust 2

8, 2

020

24 h, relatively little impact on total cellular respiration was ob-served (SI Appendix, Fig. S1). After 72 h of incubation, however,EdU was a potent inhibitor with IC50 values ranging from 1 (3T3)to 10 μM (Vero) (Fig. 1B and SI Appendix, Fig. S1). Its delayedimpact on respiration is consistent with cell-cycle arrest as the pri-mary inhibitory mechanism of EdU (22). Me-ara-EdU, ara-EU,and F-ara-EdU, in contrast, exhibited very little, if any, acute in-hibition of cellular respiration even at concentrations of 1 mM for72 h (Fig. 1B and SI Appendix, Fig. S1). These results confirmedour hypothesis that synthetic modification of the 2′-arabino posi-tion can dramatically reduce the antimetabolic activities of alkyne-containing nucleosides. To evaluate changes in cellular prolifera-tion, HeLa cells were grown in the presence of variable concentra-tions of F-ara-EdU or EdU for up to 60 d. During passages onevery third or fourth day, cells were counted, seeded at a constantdensity, and provided with fresh nucleosides and growth media.This process was repeated approximately 20 times and the num-bers of viable cells were counted during each passage. Whereas1 μM of EdU caused a complete loss of cellular replication after7 d, 1 μMof F-ara-EdU had comparatively mild effects on cell divi-sion, causing a maximal decrease of 50% in the rate of cell pro-liferation as compared to untreated cells (SI Appendix, Fig. S2).Interestingly, no changes in the rates of cell division were observedin tissues cultivated in 100 nMof F-ara-EdU (SI Appendix, Fig. S2),despite the presence of detectable DNA labeling at this concentra-tion (Fig. 2).

Metabolic Incorporation and DNA Labeling.To examine the potentialabilities of Me-ara-EdU, ara-EU, and F-ara-EdU to be metaboli-cally phosphorylated and incorporated into DNA, cells were incu-bated with variable concentrations of each nucleoside for 24 h andsubsequently stained with a fluorescent azide (AlexaFluor 488)using CuAAC reactions (Fig. 1C and SI Appendix, Figs. S3–S5).Cells initially treated with 1–100 μM of Me-ara-EdU exhibited nodetectable DNA labeling (Fig. 1C and SI Appendix, Fig. S3). Theapplication of 100 μM ara-EU resulted in modest staining of cel-lular nuclei (SI Appendix, Fig. S4), but cells treated with 1–10 μMof ara-EU exhibited very weak or no detectable staining (Fig. 1Cand SI Appendix, Fig. S4). In contrast, cells initially exposed to0.1–100 μM F-ara-EdU gave intense nuclear staining (Fig. 1C and

SI Appendix, Fig. S5). These results hinted that, despite its loweredtoxicity, F-ara-EdUmight possess metabolic labeling and detectionefficiencies similar to or even greater than EdU.

A detailed comparison of EdU and F-ara-EdU was thereforeconducted using variable nucleoside concentrations from 0.01 to100 μM and exposure times from 24 to 72 h (Fig. 2 and SIAppendix, Figs. S5–S7). Consistent with previous reports (13,22), the addition of 10 μM of EdU to live cells for 24 h followedby addition of AlexaFluor 488 azide and Cu(I) resulted in strongstaining of nearly all nuclei present. Similar results were obtainedfor F-ara-EdU and EdU after 24 h of labeling (SI Appendix,Figs. S5 and S6), but large differences became apparent with in-creased labeling times. After 72 h in the presence of 10–100 μMof EdU, mostly all of the cells had become bloated and detachedfrom the surface due to its potent toxicity (Fig. 2). Tissues treatedwith only 1 μM of EdU for 72 h were somewhat more viable andadherent, but a large fraction of these cells were either unlabeledor contained large, highly fluorescent nuclei with aberrantmorphologies (Fig. 2 and SI Appendix, Fig. S7). In contrast, nearlyall cells treated over the entire range of 0.01–10 μMof F-ara-EdUfor 72 h were uniformly labeled, viable, and did not display anysignificant respiratory or morphological changes (Fig. 2 and SIAppendix, Fig. S7). The relative staining intensities of F-ara-EdU-treated cells were roughly proportional to its applied concentra-tions, whereas EdU treatment resulted in strongly labeled orcompletely unlabeled cells after 72 h (Fig. 2 and SI Appendix,Fig. S7). Similar trends in metabolic incorporation were observedusing HeLa (Figs. 1 and 2) and Vero cell cultures (SI Appendix,Fig. S8). No F-ara-EdU labeling could be detected in cellswhere DNA synthesis was selectively inhibited by aphidicoline(SI Appendix, Fig. S9), thus excluding the possibility of RNAincorporation.

Effects of F-ara-EdU on cell-cycle progression. EdU has previouslybeen reported to arrest cells during the G2∕M checkpoint ofthe cell cycle at concentrations of 100–200 nM (22). We thereforeinvestigated the potential ability of F-ara-EdU to initiate cell-cycle arrest. After incubating HeLa cells with 1 μM of F-ara-EdUor EdU for 72 h, the cells were washed and treated with 10 μM ofBrdU for an additional 24 h. The cells were then fixed and stainedfor alkyne groups as well as BrdU. About 90% of cells labeledwith F-ara-EdU had subsequently incorporated BrdU, whereasless than 10% of EdU-treated cells had incorporated both EdUand BrdU (Fig. 3 and SI Appendix, Fig. S10). These data showthat, unlike EdU, F-ara-EdU has very little impact on cell-cycleprogression and is compatible with pulse-chase labeling strategieswhere continued cellular division after labeling is required.

To characterize the potential limits of F-ara-EdU in long-termlabeling experiments, HeLa cells were grown with constant feed-ing of 1 μM of F-ara-EdU for over 60 d. Periodic CuAAC stainingrevealed uniformly labeled tissues, suggesting that cells did notacquire resistance to F-ara-EdU labeling. BrdU colabeling ex-periments further confirmed that labeled cells were not arrestedin their growth cycles under these conditions (SI Appendix,Fig. S11). F-ara-EdU is therefore compatible with both analyticaland preparative labeling of DNA where long-term tissue prolif-eration is required during and after metabolic labeling.

Persistence of F-ara-EdU in Labeled Chromosomes. There are a num-ber of potential reasons for the dramatic differences betweenEdU and F-ara-EdU (43). For example, F-ara-EdU might not beidentified by DNA binding proteins that can detect DNA damageand cause cell arrest at the G2∕M checkpoint (42). As an alter-native explanation, we were concerned that cellular DNA repairmechanisms might efficiently remove F-ara-EdU from labeledgenomes, and therefore limit its utility in long-term pulse-chaseexperiments in vivo. To evaluate this possibility, we pulse-labeledHeLa cells with 1–10 μM of F-ara-EdU or EdU for 24 h (ca. one

Fig. 2. Comparison of EdU and F-ara-EdU for metabolic labeling of DNA.HeLa cells were treated with variable concentrations of EdU or F-ara-EdUfor 72 h. After washing and fixing the cells, ethynyl-modified DNA wasstained using AlexaFluor 488 azide and Cu(I), and total DNAwas stained withDAPI. (Scale bar: 50 μm.) See SI Appendix, Figs. S6 and S7 for quantitativeanalyses by FACS at 24 and 72 h.

20406 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1101126108 Neef and Luedtke

Dow

nloa

ded

by g

uest

on

Aug

ust 2

8, 2

020

cell division), washed the cells to remove any residual nucleo-sides, and incubated the cells in rich media lacking label for48–120 h (ca. 2–5 cell divisions). After CuAAC staining, weobserved highly irregular, yet intense fluorescence from indivi-dual F-ara-EdU labeled chromosomes against a low backgroundof unlabeled chromosomes (SI Appendix, Fig. S12). These resultssuggest relatively little DNA repair or sister chromatid exchangeof chromosomes containing F-ara-EdU over a time period of5 d or more. For comparison, a 1–10 μM pulse of EdU for 24 hfollowed by a 120 h chase resulted in cells with highly irregularstaining where most cells exhibited either 0% or 100% labeledchromosomes (SI Appendix, Fig. S12). Together with our data fromthe BrdU chase experiments (Fig. 3), these results further demon-strate that EdU can cause irreversible cell-cycle arrest even after arelatively short pulse of only 24 h, whereas F-ara-EdU does not.

The most commonly used metabolic label, BrdU, can also beproblematic in long-term pulse-chase experiments because it isknown to be an efficient substrate for glycosylase-mediated repair,and it causes sister chromatid exchange mutagenesis (4). To com-pare the relative persistence of F-ara-EdU versus BrdU in labeledchromosomes, we pulsed HeLa cells with 10 μM of BrdU, F-ara-EdU, or a 1∶1 mixture of each label for 24 h. The cells were thenwashed to remove residual nucleosides, and “cold” chased for 48–120 h in rich media lacking synthetic nucleosides. The cells werethen stained for both alkyne groups and BrdU (Fig. 4 andSI Appendix, Figs. S13 and S14). In cells receiving the 1∶1mixture,F-ara-EdU and BrdU exhibited good colocalization after a shortchase of 48 h—except in highly condensed metaphase chromo-somes where F-ara-EdU was clearly superior (SI Appendix,Fig. S13). After 120 h, however, the number and intensity of BrdU-labeled chromosomes were significantly lower than those for F-ara-EdU (Fig. 4 and SI Appendix, Fig. S14). The persistence and sen-sitivity of F-ara-EdU detection in labeled chromosomes thereforeexceeds BrdU. These results also suggest that F-ara-EdU providesa “pulse” that is fully compatible with a BrdU chase for birth datingof DNA in vivo. EdU, in contrast, can act as a powerful antime-tabolite by inhibiting subsequent incorporation of BrdU in cellcultures and in vivo (Figs. 3 and 5C).

Metabolic DNA Labeling in Developing Zebrafish Embryos. Zebrafishat the one-cell (zygote) stage were microinjected with F-ara-EdUand/or BrdU into the yolk sac and development was allowed toproceed for 3–5 d. We found that embryos tolerate similaramounts of F-ara-EdU and BrdU (up to ca. 3 pmol per embryo)

without showing negative effects on development or viability.Following fixation and permeabilization, the embryos injectedonly with BrdU exhibited strong immunostaining throughout thetail region, whereas in other body parts, such as the head, BrdUcould only be visualized in the outmost cell layers (SI Appendix,Figs. S15 and S16). When stained with fluorescent azides inpresence of copper(I), the embryos injected with only F-ara-EdUexhibited intense fluorescent staining throughout the entire bodythat colocalized with the noncovalent stain DAPI (SI Appendix,Figs. S17 and S18). F-ara-EdU labeling was detectable atamounts as low as 0.3 pmol per embryo (SI Appendix, Fig. S19),and uniform DNA labeling was observed in embryos as early as5 hpf (SI Appendix, Fig. S20). To investigate the compatibility ofF-ara-EdU and BrdU for pulse-chase experiments, Zebrafishzygotes were microinjectected with an F-ara-EdU pulse, followedby a BrdU chase at 5 hpf (end of blastula stage) or 10 hpf (endof gastrula stage). After 3 d of development, fixation and stainingrevealed excellent colocalization of F-ara-EdU and BrdU inlarvae chased with BrdU at 5 hpf (Fig. 5A), whereas much lesscolocalization was observed for larvae chased after 10 hpf(Fig. 5B). These results are consistent with the known timing oftissue development and differentiation in Zebrafish embryos,where the first terminal divisions occur at the end of the gastrulastage (38, 39). Interestingly, cells containing only F-ara-EdU areobserved along the dorsal side of the notochord and epithelia(arrows in Fig. 5B). These information-rich images reveal thepresence and location of quiescent/senescent cells that have notdivided from 10 h to 3 d after fertilization. Similar staining pat-terns, albeit with lower quality, are observed when BrdU is usedas the pulse, followed by F-ara-EdU as the chase (SI Appendix,Fig. S21). Taken together, these results demonstrate the compat-ibility of F-ara-EdU and BrdU for birth dating of DNA in vivo.EdU pulses, in contrast, caused subsequent inhibition of BrdUincorporation in Zebrafish and therefore little colocalization wasobserved in vivo (Fig. 5C).

DiscussionVisualization of DNA synthesis has traditionally been performedusing BrdU, but the limited tissue permeability of anti-BrdU anti-bodies restricts imaging to the outermost layers of cells in wholeanimals (SI Appendix, Figs. S15 and S16). Small organic dyes, incontrast, are compatible with deep-tissue imaging (SI Appendix,

Fig. 3. Pulse-chase labeling of DNA in HeLa cells with EdU or F-ara-EdU(1 μM) for 72 h, followed by a BrdU chase (10 μM) for 24 h. Following fixation,ethynyl-modified DNA was stained using AlexaFluor 488 azide and Cu(I),BrdU was stained with BrdU antibody-AlexaFluor 647 conjugate, and totalDNA was stained with DAPI. (Scale bar: 100 μm.) See SI Appendix, Fig. S10for quantitative analyses by FACS.

Fig. 4. Pulse labeling of DNA in HeLa cells with F-ara-EdU (10 μM), BrdU(10 μM), or a 1∶1 mixture (10 μM each) followed by a cold chase. HeLa cellswere incubated with F-ara-EdU and/or BrdU for 24 h, washed, and grown inrich media lacking synthetic nucleosides for 5 d and fixed. Ethynyl-modifiedDNAwas stained using AlexaFluor 488 azide and Cu(I), BrdU was stained withBrdU antibody-AlexaFluor 647 conjugate, and total DNA was stained withDAPI. (Scale bar: 10 μm.) See SI Appendix, Figs. S13 and S14 for variable-mag-nification images.

Neef and Luedtke PNAS ∣ December 20, 2011 ∣ vol. 108 ∣ no. 51 ∣ 20407

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Aug

ust 2

8, 2

020

Figs. S17 and S18). After metabolic incorporation into DNA,EdU and F-ara-EdU can be visualized in vivo with high sensitivityfollowing click reactions with fluorescent azides (13). This meth-od does not require sample fixation or DNA denaturation priorto detection, and the sensitivity of EdU and F-ara-EdU detectionexceed the limits of classical BrdU detection (22). Despite theseadvantages, EdU is much more toxic than BrdU, and its use overprolonged experimental times results in highly nonuniform stain-ing due to cell-cycle arrest (22–27). EdU is also a potent antime-tabolite that inhibits subsequent incorporation of BrdU anddT into DNA (Figs. 3 and 5C, and SI Appendix, Fig. S12). Theseeffects limit the utility of EdU in pulse-chase experiments and inother metabolic labeling studies where subsequent tissue survivalis needed. In contrast, F-ara-EdU causes little or no cell-cyclearrest and can effectively label DNA at concentrations well belowits toxicity. F-ara-EdU is therefore ideally suited for pulse-chaseexperiments aimed at birth dating DNA in vivo. As a demonstra-tion of this technique, F-ara-EdU was used in an assay for iden-tifying quiescent and senescent cells in vivo (arrows in Fig. 5).

In addition to enabling strategies in macromolecular engineer-ing and drug discovery, arabinosyl nucleoside analogs can provideimproved metabolic labels for applications involving bioorthogo-nal chemical reporter strategies. F-ara-EdU, for example, is less

toxic than both BrdU and EdU, yet it can be detected with greatersensitivity under conditions of long-term growth. A comparisonof each compound in terms of its toxicity versus the concentrationneeded for detection has revealed a much more favorable “meta-bolic labeling index” (analogous to therapeutic index) of F-ara-EdU as compared to BrdU and EdU.

The presence of a fluorine atom at the 2′-arabino position ofnucleotides is known to exert a wide variety of physicochemicaleffects (43). These include a strengthening of the glycosidic bond(44) that will limit F-ara-EdU’s potential for catabolism. In thecase of EdU, glycosidic bond hydrolysis by pyrimidine phosphor-ylases generates 5-ethynyluracil (45), itself a toxic antimetabolite(46, 47). Differences in glycosidic bond stability will also influencethe relative stability of F-ara-EdU after its incorporation intoDNA. Consistent with the reported behaviors of the 2′S-fluoroarabinosyl nucleoside analogs FAU and 1-(2'-deoxy-2'-fluoro-β-D-arabinofuranosyl)-5-iodouracil (48), F-ara-EdU containingchromosomes exhibit some resistance to DNA repair. This find-ing is consistent with previous reports that 2′S-F-substituted nu-cleotides are inefficient repair substrates and/or inhibitors of theenzymes responsible for DNA repair (43). In the case of F-ara-A,the fluorine group serves to inhibit the action of cellular deami-nases to dramatically increase the cellular half-life and anticanceractivities of ara-A (36). This same 2-fluorine group, however, canalso lead to the loss of specificity at the polymerase level as F-ara-Ais incorporated into both cellular RNA and DNA, whereas ara-A(lacking fluorine) is incorporated exclusively into DNA (49). It wastherefore not clear, a priori, if 2′-arabino-modified 5-ethynyluri-dine derivatives would be incorporated selectively into cellularDNA, RNA, or a mixture of both. In all cases, CuAAC stainingof ara-EU and F-ara-EdU-treated cells was colocalized with thenoncovalent DNA probe DAPI (Figs. 1–4) and with BrdU (Fig. 4and SI Appendix, Fig. S13). In addition, F-ara-EdU incorporationcould not be detected in cells where DNA synthesis was inhibitedby aphidicoline (SI Appendix, Fig. S9). The specificity of F-ara-EdU for DNA incorporation is consistent with the presence ofa stereoelectronic effect that influences sugar conformation andpolymerase selectivity in vivo (50).

Due to its minimal impact on cellular proliferation and itspersistence in labeled chromosomes, F-ara-EdU will find broadapplications in DNA labeling experiments. In particular, unre-solved questions in developmental biology regarding the replica-tion and flow of genetic materials over long time periods can nowbe addressed using this highly sensitive and minimally disruptivemetabolic label for DNA.

Materials and MethodsHeLa, Vero, and 3T3 cells were cultivated at 37 °C∕5% CO2 in DMEM contain-ing 4.5 g∕L glucose, 10% FCS, 50,000 units penicillin and 50 mg streptomycinper liter. For CuAAC staining, a freshly prepared staining mixture containing10 μM AlexaFluor 488 azide, 1 mM CuSO4, and 10 mM sodium ascorbatein PBS was used. See the SI Appendix for all other experimental details andprocedures.

ACKNOWLEDGMENTS. We thank Dr. Andreas Jurgeit, Prof. Urs Greber,Dr. Anna Paula de Oliveira, Prof. Cornel Fraefel, Kara Dannenhauer, Dr. EddaKastenhuber, and especially Prof. Stephan Neuhauss for helpful discussionsand technical assistance. We also thank the Swiss National Science Founda-tion (130074 to N.W.L.) and the University of Zürich Forschungskredit(57131902 to A.B.N.) for financial support.

1. Bucci M, Goodman C, Sheppard TL (2010) A decade of chemical biology. Nat Chem Biol6:847–854.

2. Zink D, Cremer T (1998) Cell nucleus: Chromosome dynamics in nuclei of living cells.Curr Biol 8:R321–R324.

3. Farkan S (1976) High-resolution autoradiography as a tool for the localization ofnucleic acid synthesis and distribution in the mammalian cell nucleus. J Microsc106:159–171.

4. Taupin P (2007) BrdU immunohistochemistry for studying adult neurogenesis: Para-digms, pitfalls, limitations, and validation. Brain Res Rev 53:198–214.

5. Best MD (2009) Click chemistry and bioorthogonal reactions: Unprecedented selectiv-ity in the labeling of biological molecules. Biochemistry 48:6571–6584.

6. Sletten EM, Bertozzi CR (2009) Bioorthogonal chemistry: Fishing for selectivity in asea of functionality. Angew Chem Int Ed Engl 48:6974–6998.

7. Saxon E, Bertozzi CR (2000) Cell surface engineering by a modified Staudinger reac-tion. Science 287:2007–2010.

8. Wang Q, et al. (2003) Bioconjugation by copper(I)-catalyzed azide–alkyne [3þ 2]cycloaddition. J Am Chem Soc 125:3192–3193.

9. Baskin JM, et al. (2007) Copper-free click chemistry for dynamic in vivo imaging. ProcNatl Acad Sci USA 104:16793–16797.

10. Ning X, Guo J, Wolfert MA, Boons G-J (2008) Visualizing metabolically labeled glyco-conjugates of living cells by copper-free and fast Huisgen cycloadditions.Angew ChemInt Ed Engl 47:2253–2255.

Fig. 5. Pulse-chase labeling of Zebrafish embryos using F-ara-EdU andBrdU. (A) Embryos microinjected with F-ara-EdU at 0 hpf and with BrdU at5 hpf. (B) Embryos microinjected with F-ara-EdU at 0 hpf and with BrdU at10 hpf. (C) Embryos microinjected with EdU at 0 hpf and with BrdU at 10 hpf.Negative controls (D) remained untreated. Larvae were allowed to developuntil 3 d after fertilization, fixed, and stained using AlexaFluor 488 azide andCu(I), and BrdU antibody-AlexaFluor 647 conjugate. Pictures show lateralviews of tail regions. The white arrows indicate cells surrounding the noto-chord labeled with only F-ara-EdU; arrowheads indicate cells associated withepithelia labeled with only F-ara-EdU. Abbreviations: dor, dorsal; cra, cranial;ven, ventral; cau, caudal. (Scale bar: 50 μm.)

20408 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1101126108 Neef and Luedtke

Dow

nloa

ded

by g

uest

on

Aug

ust 2

8, 2

020

11. Kiick KL, Saxon E, Tirrell DA, Bertozzi CR (2001) Incorporation of azides into recombi-nant proteins for chemoselective modification by the Staudinger ligation. Proc NatlAcad Sci USA 99:19–24.

12. Ngo JT, et al. (2009) Cell-selective metabolic labeling of proteins. Nat Chem Biol5:715–717.

13. Salic A, Mitchison TJ (2008) A chemical method for fast and sensitive detection of DNAsynthesis in vivo. Proc Natl Acad Sci USA 105:2415–2420.

14. Jao CY, Salic A (2008) Exploring RNA transcription and turnover in vivo by using clickchemistry. Proc Natl Acad Sci USA 105:15779–15784.

15. Neef AB, Schultz C (2009) Selective fluorescence labeling of lipids in living cells.AngewChem Int Ed Engl 48:1498–1500.

16. Jao CY, RothM,Welti R, Salic A (2009)Metabolic labeling and direct imaging of cholinephospholipids in vivo. Proc Natl Acad Sci USA 106:15332–15337.

17. Prescher JA, Dube DH, Bertozzi CR (2004) Chemical remodelling of cell surfaces inliving animals. Nature 430:873–877.

18. Sawa M, et al. (2006) Glycoproteomic probes for fluorescent imaging of fucosylatedglycans in vivo. Proc Natl Acad Sci USA 103:12371–12376.

19. Laughlin ST, Baskin JM, Amacher SL, Bertozzi CR (2008) In vivo imaging of membrane-associated glycans in developing zebrafish. Science 320:664–667.

20. Gierlich J, Burley GA, Gramlich PME, Hammond DM, Carell T (2006) Click chemistry asa reliable method for the high-density postsynthetic functionalization of alkyne-modified DNA. Org Lett 8:3639–3642.

21. Gierlich J, et al. (2007) Synthesis of highly modified DNA by a combination of PCR withalkyne-bearing triphosphates and click chemistry. Chem Eur J 13:9486–9494.

22. Diermeier-Daucher S, et al. (2009) Cell type specific applicability of 5-ethynyl-2′-deox-yuridine (EdU) for dynamic proliferation assessment in flow cytometry. Cytometry A75A:535–546.

23. Cassiman JJ, De Clercq E, van den Berghe H (1983) Induction of sister-chromatidexchange by 5-substituted 2′-deoxyuridines. Mutat Res 117:317–327.

24. Danenberg PV, Bhatt RS, Kundu NG, Danenberg K, Heidelberger C (1981) Inter-action of 5-ethynyl-2′-deoxyuridylate with thymidylate synthetase. J Med Chem24:1537–1540.

25. De Clercq E, et al. (1983) Nucleic acid related compounds. 40. Synthesis and biologicalactivity of 5-alkynyluracil nucleosides. J Med Chem 26:661–666.

26. Meneni S, et al. (2007) 5-Alkynyl-2′-deoxyuridines: Chromatography-free synthesisand cytotoxicity evaluation against human breast cancer cells. Bioorg Med Chem15:3082–3088.

27. Cappella P, Gasparri F, Pulici M, Moll J (2008) A novel method based on click chemistry,which overcomes limitations of cell cycle analysis by classical determination of BrdUincorporation, allowing multiplex antibody staining. Cytometry A 73A:626–636.

28. Cinquin O, Crittenden SL, Morgan DE, Kimble J (2010) Progression from a stem cell-likestate to early differentiation in the C. elegans germ line. Proc Natl Acad Sci USA107:2048–2053.

29. Limsirichaikul S, et al. (2008) A rapid non-radioactive technique for measurement ofrepair synthesis in primary human fibroblasts by incorporation of ethynyl deoxyuri-dine (EdU). Nucleic Acids Res 37:e31.

30. Cseresnyes Z, Schwarz U, Green CM (2009) Analysis of replication factories in humancells by super-resolution light microscopy. BMC Cell Biol 10:88.

31. Leatherman JL, DiNardo S (2010) Germline self-renewal requires cyst stem cells andstat regulates niche adhesion in Drosophila testes. Nat Cell Biol 12:806–811.

32. Sugiyama M, et al. (2009) Illuminating cell-cycle progression in the developingzebrafish embryo. Proc Natl Acad Sci USA 106:20812–20817.

33. Lu JJ, Li F, Murphy CS, Davidson MW, Gilbert DM (2010) G2 phase chromatin lacksdeterminants of replication timing. J Cell Biol 189:967–980.

34. Ross H, et al. (2011) Ethynyldeoxyuridine (EdU) suppresses in vitro populationexpansion and in vivo tumor progression of human glioblastoma cells. J Neurooncol105:1–14.

35. Iwasaki H, Huang P, Keating MJ, Plunkett W (1997) Differential incorporation ofara-C, gemcitabine, and fludarabine into replicating and repairing DNA in proliferat-ing human leukemia cells. Blood 90:270–278.

36. Plunkett W, Chubb S, Alexander L, Montgomery JA (1980) Comparison of the toxicityandmetabolism of 9-beta-D-arabinofuranosyl-2-fluoroadenine and 9-beta-D-arabino-furanosyladenine in human lymphoblastoid cells. Cancer Res 40:2349–55.

37. Collins JM, Klecker RW, Katki AG (1999) Suicide prodrugs activated by thymidylatesynthase: Rationale for treatment and noninvasive imaging of tumors with deoxyur-idine analogues. Clin Cancer Res 5:1976–1981.

38. Kimmel CB, Ballard WW, Kimmel SR, Ullmann B, Schilling TF (1995) Stages of embryo-nic development of the zebrafish. Dev Dyn 203:253–310.

39. Mendelson B (1986) Development of reticulospinal neurons of the zebrafish. I. Time oforigin. J Comp Neurol 251:260–271.

40. Garvey EP, Lowen GT, Almond MR (1998) Nucleotide and nucleoside analogues asinhibitors of cytosolic 5′-nucleotidase I from heart. Biochemistry 37:9043–9051.

41. Balzarini J, De Clercq E, Mertes MP, Shugar D, Torrence PF (1982) 5-Substituted 2′-deoxyuridines: Correlation between inhibition of tumor cell growth and inhibitionof thymidine kinase and thymidylate synthetase. Biochem Pharmacol 31:3673–3682.

42. Sancar A, Lindsey-Boltz LA, Unsal-Kacmaz K, Linn S (2004) Molecular mechanismsof mammalian DNA repair and the DNA damage checkpoints. Annu Rev Biochem73:39–85.

43. Watts JK, Damha MJ (2008) 2′F-arabinonucleic acids (2′F-ANA)—history, properties,and new frontiers. Can J Chem 86:641–656.

44. Lee S, Bowman BR, Ueno Y, Wang S, Verdine GL (2008) Synthesis and structure ofduplex DNA containing the genotoxic nucleobase lesion N7-methylguanine. J AmChem Soc 130:11570–11571.

45. Niedzwicki JG, el Kouni MH, Chu SH, Cha S (1983) Structure-activity relationships ofligands of the pyrimidine nucleoside phosphorylases. Biochem Pharmacol 32:399–415.

46. Porter DJT (1994) Reaction of 5-ethynyluracil with rat liver xanthine oxidase. J BiolChem 269:27932–27940.

47. Thornburg LD, Stubbe J (1993) Mechanism-based inactivation of thymine hydroxylase,an alpha-ketoglutarate-dependent deoxygenase, by 5-ethynyluracil. Biochemistry32:14034–14042.

48. Klecker RW, Katki AG, Collins JM (1994) Toxicity, metabolism, DNA incorporation withlack of repair, and lactate production for 1-(2′-fluoro-2′-deoxy-beta-D-arabinofurano-syl)-5-iodouracil in U-937 and MOLT-4 cells. Mol Pharmacol 46:1204–1209.

49. Huang P, Plunkett W (1991) Action of 9-beta-D-arabinofuranosyl-2-fluoroadenine onRNA metabolism. Mol Pharmacol 39:449–455.

50. Williams AA, et al. (2009) Impact of sugar pucker on base pair and mispair stability.Biochemistry 48:11994–12004.

Neef and Luedtke PNAS ∣ December 20, 2011 ∣ vol. 108 ∣ no. 51 ∣ 20409

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Aug

ust 2

8, 2

020