Mycobacterium tuberculosis WhiB3 responds to O2and nitric oxide via its [4Fe-4S] cluster and isessential for nutrient starvation survivalAmit Singh*, Loni Guidry*, K. V. Narasimhulu†, Deborah Mai*, John Trombley*, Kevin E. Redding†, Gregory I. Giles‡§,Jack R. Lancaster, Jr.‡, and Adrie J. C. Steyn*¶

*Department of Microbiology, University of Alabama at Birmingham, Birmingham, AL 35294; †Department of Chemistry, University of Alabama,Tuscaloosa, AL 35487; and ‡Departments of Anesthesiology, Physiology, Biophysics, and Environmental Health Sciences, Center forFree Radical Biology, University of Alabama at Birmingham, Birmingham, AL 35294

Edited by Barry R. Bloom, Harvard School of Public Health, Boston, MA, and approved May 8, 2007 (received for review January 18, 2007)

A fundamental challenge in the redox biology of Mycobacteriumtuberculosis (Mtb) is to understand the mechanisms involved insensing redox signals such as oxygen (O2), nitric oxide (NO), andnutrient depletion, which are thought to play a crucial role inpersistence. Here we show that Mtb WhiB3 responds to thedormancy signals NO and O2 through its iron-sulfur (Fe-S) cluster.To functionally assemble the WhiB3 Fe-S cluster, we identified andcharacterized the Mtb cysteine desulfurase (IscS; Rv3025c) anddeveloped a native enzymatic reconstitution system for assem-bling Fe-S clusters in Mtb. EPR and UV-visible spectroscopy analysisof reduced WhiB3 is consistent with a one-electron reduction ofEPR silent [4Fe-4S]2� to EPR visible [4Fe-4S]�. Atmospheric O2

gradually degrades the WhiB3 [4Fe-4S]2� cluster to generate a[3Fe-4S]� intermediate. Furthermore, EPR analysis demonstratesthat NO forms a protein-bound dinitrosyl–iron–dithiol complexwith the Fe-S cluster, indicating that NO specifically targets theWhiB3 Fe-S cluster. Our data suggest that the mechanism of WhiB34Fe-4S cluster degradation is similar to that of fumarate nitrateregulator. Importantly, Mtb �whiB3 shows enhanced growth onacetate medium, but a growth defect on media containing glucose,pyruvate, succinate, or fumarate as the sole carbon source. Ourresults implicate WhiB3 in metabolic switching and in sensing thephysiologically relevant host signaling molecules NO and O2

through its [4Fe-4S] cluster. Taken together, our results suggestthat WhiB3 is an intracellular redox sensor that integrates envi-ronmental redox signals with core intermediary metabolism.

dormancy � redox � metabolism � iron–sulfur

The ability of Mycobacterium tuberculosis (Mtb) to maintain astate of latent tuberculosis infection is responsible for its

remarkable success as a pathogen. Persistent infection is the resultof complex interplay between the host immune system and bacterialsurvival mechanisms, and little is known about the specific bacterialand host components involved in this process. How Mtb can persistin human tissues for months to decades without replicating and thenabruptly resume growth and cause active disease is a fundamentalquestion in the tuberculosis field. Understanding the mechanisticbasis of persistence has become the focus of much research,primarily because persistent bacilli are in a state of ‘‘drug unre-sponsiveness’’ wherein they are insensitive to antimycobacterialdrugs. The role of oxygen tension in Mtb persistence has receivedwide attention because it has been demonstrated that although thetubercle bacilli require oxygen for replication, the bacilli can survivefor years without oxygen in vitro (1). Accordingly, the mechanismof physiological adaptation in response to limiting O2 availabilityhas been studied extensively (1). Studies have revealed that a rapidshift from aerobic to an O2-deficient environment results in bac-terial death. However, the bacilli survive upon gradual depletion ofO2, indicating that an ordered metabolic shutdown is necessary foradaptation to anoxia (1). Several groups have suggested the role ofhypoxia-induced regulon (Dos) in the downshift of Mtb from the

replicative to the persistent state. Of particular interest is the findingthat a significant degree of overlap exists between Mtb geneexpression patterns seen in response to hypoxia and those observedafter exposure to NO. These studies revealed intriguing insights intothe respiratory networks of Mtb that presumably lead to decreasedenergy production during in vivo adaptation (2). Long-term per-sistence followed by resumption of metabolic activity argues for thecoordinated regulation of a subset of Mtb metabolic pathwaysduring persistence (3). However, the molecular mechanism of howNO, hypoxia, and nutrient stress are sensed to induce metabolicadaptation is not known. Recent work suggested the involvement ofa heme-containing cytochrome c oxidase and/or the presence of afumarate nitrate regulator (FNR)-like sensor (2, 4, 5). As a result,it is postulated that Mtb possesses one or more sensors, which detectenvironmental redox signals such as O2, NO, and fluctuations in theintracellular redox state, to promote bacterial persistence.

Fe-S cluster proteins are known to play essential roles in sensingexternal signals as well as the intracellular redox state of microbialcells (6). Fe-S cluster transcription factors such as SoxR and FNRhave been shown to be the primary sensors of NO and hypoxia andare responsible for the metabolic adaptation of Escherichia coli toenvironmental stresses (6). Mtb homologs of FNR and SoxR wouldbe considered excellent candidates to sense and relay redox signals.However, a clear FNR-like protein is absent in Mtb. Therefore, oneor more, yet-to-be-identified ‘‘redox sensors’’ likely play a crucialrole in the downshift of Mtb to the dormant state. Of interest to usis the role of Mtb proteins in sensing environmental host signals. Inan accompanying paper [Kumar et al. (28)], we demonstrated thatMtb DosS and DosT are heme proteins that directly sense O2, NO,and CO to modulate expression of the Mtb dormancy regulon. Inaddition, we have previously discovered a gene, whiB3, that plays arole in Mtb virulence (7). Therefore, of particular interest to us isthe role of the Mtb WhiB family in sensing environmental hostsignals. In this work we hypothesize that Mtb WhiB3 is an Fe-Scluster protein capable of sensing redox signals through its Fe-Scluster. To test this hypothesis, we developed an endogenous

Author contributions: A.S. and A.J.C.S. designed research; A.S., L.G., K.V.N., D.M., J.T., andG.I.G. performed research; A.S., L.G., K.V.N., K.E.R., G.I.G., J.R.L., and A.J.C.S. analyzed data;and A.S. and A.J.C.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Abbreviations: DNIC, dinitrosyl–iron–dithiol; DTH, sodium dithionite; EDFS, echo-detectedfield sweep; FNR, fumarate nitrate regulator; Mtb, Mycobacterium tuberculosis.

§Present address: School of Chemistry, University of Sydney, Sydney, NSW 2006, Australia.

¶To whom correspondence should be addressed at: Department of Microbiology, Univer-sity of Alabama at Birmingham, 845 19th Street South, 308 BBRB, Birmingham, AL 35294.E-mail: asteyn@uab.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0700490104/DC1.

© 2007 by The National Academy of Sciences of the USA

11562–11567 � PNAS � July 10, 2007 � vol. 104 � no. 28 www.pnas.org�cgi�doi�10.1073�pnas.0700490104

reconstitution system to functionally assemble Mtb Fe-S clusters.We used EPR, UV-visible spectroscopy, and a sensitive radioactiveassay to characterize how WhiB3 responds to O2 and NO, and wedraw several important parallels with FNR. Finally, we analyzed thegrowth phenotype of Mtb�whiB3 under nutritional stress. Weanticipate that understanding the molecular and biochemical mech-anism of this redox pathway will contribute toward understandinghow Mtb maintains a state of latency.

ResultsMtb IscS (Rv3025c) Is a Functional Cysteine Desulfurase. Bacteriahave developed (at least) three distinct, highly conserved systems(Nif, Isc, and Suf) for the biogenesis of Fe-S clusters in vivo (8).In vitro, Fe-S clusters can be assembled anaerobically by usingL-cysteine (sulfur donor), an L-cysteine desulfurase (e.g., NifS/IscS), an iron source, and a reducing agent (8). Most enzymaticFe-S cluster assembly systems make use of NifS, the cysteinedesulfurase of Azotobacter vinelandii. We were concerned that anon-native enzyme might not appropriately assemble Mtb Fe-Sclusters. Therefore, we sought to identify the Mtb homolog ofNifS and use it in an Mtb Fe-S assembly system. To date, amycobacterial homolog of NifS/IscS remained uncharacterized,and no study has yet demonstrated the assembly of any Mtb Fe-Scluster protein using NifS/IscS. Rv3025c, an Mtb homolog of A.vinelandii NifS, shows the highest homology (37% identity, 55%homology), whereas three other Mtb homologs (Rv1464,Rv3778c, and Rv3700c) show relatively weak homology. Mtb IscS(Rv3025c) contains a conserved pyridoxal 5�-phosphate-bindingsite at Lys-205 and an active Cys-329 residue required for theformation of an enzyme-bound cysteinyl persulfide (8), findingsconsistent with Rv3025c encoding for an L-cysteine desulfurase.Overexpressed and purified IscS from E. coli exhibited a yellowcolor, and UV-visible spectroscopy of IscS showed a majorabsorbance peak at 420 nm, which are characteristic features ofpyridoxal 5�-phosphate-dependent enzymes (9). The addition ofL-cysteine, but not L-alanine or L-histidine, reproducibly in-creased absorbance at 315 nm and showed a concomitantdecrease in absorbance at 420 nm [supporting information (SI)Fig. 6A]. These properties are highly suggestive of L-cysteinedesulfurase activity (8, 9).

In the first step of Fe-S cluster assembly, NifS/IscS acquires sulfurby converting cysteine to alanine, resulting in the formation ofcysteinyl persulfide. Under reducing conditions, persulfide-boundsulfur is released as free sulfide for the reconstitution of Fe-S clusterin the target protein (8). To evaluate the ability of Mtb IscS to useL-cysteine as a substrate to generate persulfide, we developed asensitive 35S transfer assay. The results demonstrate the acquisitionof 35S from 35S-labeled L-cysteine by IscS (see Experimental Proce-dures and SI Fig. 6B, lane 1). To assess the role of the cysteinylresidue in sulfur acquisition, IscS was alkylated by using N-ethylmaleimide and included in the 35S transfer assay. A drasticreduction in 35S incorporation from 35S-labeled cysteine in alkylatedIscS was observed, thereby establishing a role for the highly reactiveIscS cysteinyl thiolate in the formation of cysteinyl persulfide (SIFig. 6B, lane 2). Finally, in the presence of the reducing agent DTT(SI Fig. 6B, lane 3), we demonstrated the partial loss of 35S fromIscS, which is caused by the release of sulfide from persulfide-boundsulfur intermediate. Taken together, these data confirm that MtbIscS has active cysteine desulfurase activity.

Mtb IscS Can Assemble an Fe-S Cluster in WhiB3. Aerobically purifiedWhiB3 is reddish-brown in color, a characteristic feature of Fe-Scluster proteins. In addition, the primary sequence of WhiB3contains a likely Fe-S cluster-binding motif consisting of fourconserved cysteine residues C-X29-C-X2-C-X5-C (where X is anyamino acid) (10). To demonstrate that Mtb IscS is capable ofmobilizing S from L-cysteine to assemble the putative WhiB3 Fe-Scluster, we included 35S-labeled cysteine in the reaction mixture and

monitored incorporation of 35S into WhiB3. We successfully dem-onstrated IscS-dependent transfer of 35S from L-cysteine to WhiB3over time (see Fig. 1A). As expected, we identified multiple WhiB3species on native polyacrylamide gels, which are attributable tohomooligomerization and reconstitution intermediates (Fig. 1).Importantly, in our control reaction mixture we excluded IscS orcysteine and did not generate any signal (data not shown).

Next, we examined the requirement of Fe2� for IscS-mediatedFe-S cluster assembly. Upon addition of increasing amounts ofFe2�, we observed a gradual increase in signal (Fig. 1B). In contrast,removal of Fe2� with the iron chelator 2,2-dipyridyl caused agradual decrease in signal (Fig. 1C). Lastly, we replaced all fourcysteines in WhiB3 with alanine (WhiB3�Cys-1–4). WhiB3�Cys-1–4was aerobically purified and used in the 35S transfer assay with IscS.The absence of 35S incorporation into WhiB3�Cys-1–4 suggests thatthe four cysteine residues are ligands of the WhiB3 Fe-S cluster(Fig. 1D). Taken together, the above results show that WhiB3 isindeed an Fe-S cluster protein. The results also demonstrate that wehave successfully developed an Mtb Fe-S cluster assembly system forWhiB3 and indicate a role for IscS in the biogenesis of Mtb Fe-Sclusters.

Mtb WhiB3 Contains a 4Fe-4S Cluster. Having established that MtbWhiB3 is an Fe-S cluster protein, we now sought to determinethe cluster type. We used UV-visible and EPR spectroscopy andthoroughly characterized the biophysical and biochemical prop-erties of WhiB3. We found the UV-visible spectrum of aerobi-cally purified WhiB3 (data not shown) to be indistinguishablefrom that of the bacterial type 2Fe-2S ferredoxins, exhibitingpeaks at 413 nm and two broad shoulders near 458 and 330 nm

Fig. 1. Mtb WhiB3 is an Fe-S cluster protein. (A) WhiB3 Fe-S cluster recon-stitution catalyzed by IscS over time. At various time intervals (10, 30, 60, 120,and 180 min), aliquots from the reconstitution reactions were removed andanalyzed by native PAGE followed by autoradiography. (B and C) Reconstitu-tion of the WhiB3 Fe-S cluster was carried out for 1 h in the presence of 50, 100,250, and 500 �M Fe2� (B) or 10, 25, 50, and 100 �M Fe2� chelator (2,2-dipyridyl)(C). (D) The four conserved cysteine residues in WhiB3 are essential for Fe-Scluster reconstitution. Fe-S cluster reconstitution was carried out with WhiB3and WhiB3�Cys-1–4 as described earlier and analyzed by native PAGE.

Singh et al. PNAS � July 10, 2007 � vol. 104 � no. 28 � 11563

BIO

CHEM

ISTR

Y

(11). However, because WhiB3 was initially aerobically purified,we believe the 2Fe-2S cluster was most likely a degradationproduct. Therefore, to reconstitute the native WhiB3 Fe-Scluster, WhiB3 was then incubated under anaerobic conditionswith freshly purified IscS, L-cysteine, DTT, and FeCl2 as de-scribed in Experimental Procedures. During anaerobic reconsti-tution, a greenish-brown color developed within 30 min (Fig. 2 A,Inset). Upon further incubation, we observed a time-dependentincrease in the intensity �413 nm with no other resolved features(Fig. 2 A). Both the absorbance �413 nm and the color devel-opment are characteristic of FNR-type 4Fe-4S cluster formation(12). As shown in Fig. 2 A, IscS completed assembly of the WhiB3Fe-S cluster within 7 h, as is evidenced by the single broadshoulder at 413 nm and an A413/A280 ratio of 0.31. This ratio isconsistent with the reconstitution of 4Fe-4S cluster in WhiB3.The UV-visible spectrum of reconstituted WhiB3 confers amolar absorption coefficient (�) at 413 nm of 12,500 M�1 cm�1,which is similar to that of 4Fe-4S clusters (13). As additionalverification, we performed a reconstitution reaction usingWhiB3�Cys-1–4 and observed no color development. As expected,we observed no peak at 413 nm (Fig. 2B), confirming the absenceof a 4Fe-4S cluster in WhiB3�Cys-1–4. To characterize further theredox behavior of the WhiB3 4Fe-4S cluster, anaerobicallyreconstituted WhiB3 was subjected to one-electron reductionwith sodium dithionite (DTH). Reduction caused partial bleach-ing of the greenish-brown color with the concomitant loss of thepeak at 413 nm (Fig. 2 A). The spectral properties of DTH-

reduced WhiB3 are characteristic of the reduction of a [4Fe-4S]2� to a [4Fe-4S]� state (10).

Finally, to identify conclusively the type of Fe-S cluster, weexploited echo-detected field sweep (EDFS) EPR and analyzedanaerobically reconstituted WhiB3. We observed no signal, sug-gestive of the presence of a nonparamagnetic species. However,upon reduction with DTH, WhiB3 exhibits a signal at g � 2.06 and1.94 (Fig. 2C). This signal is consistent with a one-electron reduc-tion of an EPR silent [4Fe-4S]2� species to a paramagnetic [4Fe-4S]� cluster with an electron spin S � 1/2 (10), and corroboratedour UV-visible spectroscopic data. As an additional verification ofthe IscS-dependent reconstitution of the WhiB3 Fe-S cluster, wealso tested the ability of the widely used A. vinelandii NifS toreassemble the WhiB3 Fe-S cluster. The A. vinelandii NifS-treatedWhiB3 samples showed UV-visible and EPR characteristics virtu-ally identical to those of the IscS-treated WhiB3 samples. However,the intensity of the absorbance at 413 nm and the EPR signals weremuch stronger compared with the IscS-treated WhiB3 sample (Fig.2D). Altogether, the data generated by several independent ap-proaches convincingly demonstrate that Mtb WhiB3 contains aredox-responsive 4Fe-4S cluster.

WhiB3 Responds to NO and O2. It is widely believed that NO and O2are key signaling molecules required for induction of the Mtbdormancy program. Therefore, determining the identity of Mtbproteins that react with these diatomic gases is important. In E. coli,the Fe-S cluster-containing transcription factors, FNR (4Fe-4S)and SoxR (2Fe-2S), have been shown to be direct sensors of NO andO2 (6). To establish whether WhiB3 is capable of reacting witheither gas, in vitro reconstituted WhiB3 was independently exposedto O2 or NO. To assess sensitivity to atmospheric O2, air wasbubbled through the sample for 2 min, and the oxygen-inducedchanges in the UV-visible spectrum were monitored over time. Atan early time point (5 min), we observed a nominal but highlyreproducible increase in absorbance at 413 nm followed by aprogressive decrease over 60 min accompanied by an increase inabsorbance in the 500- to 600-nm region (Fig. 3A). At subsequenttime points (90–180 min), there was a significant reduction in theabsorbance at 413 nm with the appearance of a new peak at 330 nmand a broad shoulder at 458 nm (Fig. 3A). These spectral featuresare similar to the UV-visible spectra of aerobically purified WhiB3and are consistent with an oxygen-dependent conversion of a4Fe-4S to a 2Fe-2S cluster. A plot of �A413 nm against time revealedthe loss of �75% of the 4Fe-4S cluster within 180 min (Fig. 3A,Inset).

More importantly, EDFS-EPR spectra of WhiB3 after air expo-sure (5 min) showed the appearance of a sharp signal centered atg � 2.01, which disappeared after prolonged exposure to air (60min) (Fig. 3B). These observations indicate that the small increasein the absorbance at 413 nm at 5 min after air exposure was causedby the conversion of [4Fe-4S]2� to the [3Fe-4S]� intermediate.However, prolonged incubation (60 min) of air-exposed WhiB3 ledto the conversion of the [3Fe-4S]� intermediate to the EPR silent[2Fe-2S]2� cluster followed by the loss of Fe-S cluster. Interestingly,a [3Fe-4S]� intermediate was also observed during air exposure ofthe FNR Fe-S cluster (13).

Furthermore, to assess the effect of O2 on the WhiB3 Fe-S clusterin more detail, the anaerobically reconstituted WhiB3 Fe-S clusterwas labeled with 35S, and the loss of incorporated 35S was measuredupon exposure of WhiB3 to O2. As anticipated, we observed agradual decline of radioactive signal with the complete loss of thesignal at 120 min after air exposure (Fig. 3C). These results provideclear evidence that the Mtb WhiB3 4Fe-4S cluster is degradedby O2.

To investigate whether NO is a ligand of the WhiB3 Fe-S clusterand to assess the NO responsiveness of WhiB3, aerobically purifiedWhiB3 was treated with the fast NO-releasing compound prolineNONOate and analyzed with continuous-wave EPR at liquid

Fig. 2. Spectroscopic characterization of WhiB3. UV-visible spectroscopy ofWhiB3 Fe-S cluster reconstitution is shown. (A) Reconstitution was carried outinside an anaerobic glove box as described in Experimental Procedures. Timepoints at which samples were scanned by a UV-visible spectrophotometer areindicated. Note the time-dependent increase in the characteristic 4Fe-4Sabsorption peak at 413 nm. After completion of the Fe-S cluster reassembly,the Fe-S cluster was reduced by adding 5 mM DTH. (Inset) Development of agreenish-brown color during the WhiB3 Fe-S cluster reconstitution before (leftcuvette) and after (right cuvette) the reconstitution reaction. (B) The fourconserved cysteine residues (C23-C53-C56-C62) in WhiB3 are required for Fe-Scluster assembly. Fe-S cluster reconstitution in WhiB3 and WhiB3�Cys-1–4 wascarried out as described in Experimental Procedures and analyzed by UV-visible spectroscopy. (C and D) EDFS EPR spectra of IscS- (C) and NifS- (D)reconstituted WhiB3 after reduction with DTH. EPR spectra were obtained at12 K, the EPR experimental conditions; �/2 and � pulses are 16 and 32 ns, � �180 ns. The spectra were acquired with 60 shots with a two-step phase cycle ata repetition rate of 1 kHz. Microwave frequency � 9.8046 GHz.

11564 � www.pnas.org�cgi�doi�10.1073�pnas.0700490104 Singh et al.

nitrogen temperatures. A weak EPR signal centered at g � 2.03 wasobserved, which is strongly indicative of a monomeric dinitrosyl–iron–dithiol (DNIC) complex (Fig. 3D, Inset) wherein the sulfideligands are displaced by the formation of Fe-NO bonds and the ironatoms remain bound to the protein via cysteine ligands (14). In starkcontrast to the aerobically purified WhiB3 sample, exposure ofanaerobically reconstituted WhiB3 to proline NONOate showed astrong EPR signal at g � 2.036 (Fig. 3D). Furthermore, sequentialaddition of increasing amounts of proline NONOate significantlyenhanced the EPR signal of monomeric DNIC in a concentration-dependent manner (Fig. 3D). In sum, these data conclusively showthat the WhiB3 4Fe-4S cluster can react directly with NO and O2and suggest that the WhiB3 Fe-S cluster functions as a built-insensor of the proposed dormancy signals.

Role of Mtb WhiB3 in Metabolic Stress. The results indicate that NOis a ligand of WhiB3 and that O2 reacts with WhiB3, both via its4Fe-4S cluster. Nonetheless, Mtb�whiB3 does not show a lethalphenotype upon exposure to bacteriostatic concentration of NO or

in the Wayne model of gradual oxygen depletion (data not shown),suggesting that bacterial survival is not critically affected by theseparticular in vitro conditions. Importantly, recent studies haveshown that nutrient depletion is a signal for mycobacterial persis-tence (15) and influences the expression of respiratory/metabolicenzymes responsible for modulating the intracellular redox state ofMtb (15). We tested the hypothesis that WhiB3 plays a role in thebacilli’s response to nutritional stress. Growth of wild-type (WT)Mtb and Mtb�whiB3 was compared by normalizing the cell densityand spotting an equal volume (i.e., number) of cells onto solidDubos medium (basal) or onto Dubos plates containing glucose,succinate, fumarate, pyruvate, citrate, or acetate as sole carbonsources. Dubos basal medium contains asparagine and caseinhydrolysate as sole source of carbon and nitrogen. The spot colonyphenotype indicates that Mtb�whiB3 grew poorly on this mediumas judged by the noticeably decreased colony diameter and fewersurviving colonies per spot, compared with WT Mtb (Fig. 4). Moreimportantly, supplementation of Dubos basal medium with glucose,succinate, or fumarate did not rescue the Mtb�whiB3 growth defect(Fig. 4). Also, Mtb�whiB3 showed a slight but reproducible growthdefect on pyruvate but not on citrate media (data not shown). Instark contrast, Mtb�whiB3 grows significantly better than WT Mtbon basal medium supplemented with the short-chain fatty acid (C2)acetate. Furthermore, Mtb�whiB3 showed a striking difference inspot colony rugosity compared with WT Mtb. For example, onglucose plates, WT Mtb exhibited a cord-like structure with anorganized inner ring, well defined edges, and an overall rugoseappearance, whereas the Mtb�whiB3 colony displayed less cordingwith irregular edges and structure (Fig. 4). Stable expression ofwhiB3 in Mtb�whiB3 resulted in complementation of the growthdefect on Dubos basal medium (SI Fig. 7).

Furthermore, liquid growth assays confirmed that Mtb�whiB3was compromised for growth in 7H9 basal media supplementedwith either glucose or succinate as sole carbon sources (SI Fig. 8).Finally, growth was restored by complementing Mtb�whiB3 with

Fig. 3. The Mtb WhiB3 Fe-S cluster responds to O2 and NO. (A) Anaerobicallyreconstituted WhiB3 was purified by size-exclusion column and exposed to O2

by bubbling air through the sample for 2 min. UV-visible spectra were acquiredbefore and after air exposure at various time intervals as indicated. (Inset) Rateof the 4Fe-4S cluster loss was determined by calculating the percent loss ofabsorbance at 413 nm upon exposure to air at various time intervals. (B) TheO2-treated sample of reconstituted WhiB3 was withdrawn immediately (5min) after exposure to O2 (red line) and 60 min after exposure (blue line) andanalyzed by EPR. EDFS EPR spectra were measured at 12 K; the �/2 and � pulsesused were 16 and 32 ns with, � � 180 ns. The spectrum was acquired with 60shots with a two-step phase cycling at repetition rate of 1 kHz. Microwavefrequency � 9.806 GHz. The appearance of a sharp signal at g � 2.01 indicatesa [3Fe-4S]� cluster. (C) Loss of the 4Fe-4S cluster in WhiB3 was also confirmedby exposing 35S-labeled WhiB3 to air as described before. At various timeintervals (0, 30, 60, 90, and 120 min), samples were removed and analyzed bynative PAGE followed by autoradiography (Upper) and SDS/PAGE followed byCoomassie blue staining to rule out degradation of WhiB3 upon air exposure(Lower). (D) Effect of NO on WhiB3. Reconstituted WhiB3 was exposed tovarious concentrations of proline NONOate (WhiB3:NO) as indicated andanalyzed by EPR. (Inset) Aerobically purified WhiB3 was deoxygenated andexposed to a 10-fold molar excess of proline NONOate. EPR data were ac-quired by using a continuous-wave EPR spectrometer at a microwave fre-quency of 9.669 GHz operating at 100-kHz field modulation and 150 K. Amodulation amplitude of 0.5 mT and microwave power of 2 mW were used.The peak at 2.036 g is consistent with the formation of monomeric DNIC (14).

Fig. 4. Growth phenotype of Mtb�whiB3. Identical volumes (30 �l) andequal number of cells (3 � 105, 3 � 104) of WT Mtb and Mtb�whiB3 werespotted on Dubos–agar medium either in the absence [basal (A)] or in thepresence of 50 mM glucose (B), 50 mM succinate (C) or 25 mM acetate (D) assole carbon substrate. Growth was analyzed every week after inoculation, andphotographs were taken at 4–6 weeks after inoculation at �7 magnificationby using a Zeiss stereo microscope. (Left) WT Mtb. (Right) Mtb�whiB3.

Singh et al. PNAS � July 10, 2007 � vol. 104 � no. 28 � 11565

BIO

CHEM

ISTR

Y

WT WhiB3 but not WhiB3�Cys-1–4 (SI Fig. 8). These resultsdemonstrate that WhiB3 plays an important role in regulating thecore intermediary metabolism of Mtb. Because the intracellularredox state is dramatically influenced by nutrients (16), these resultsserve to implicate further the WhiB3 Fe-S cluster as a criticalsensory component of the intracellular redox state of Mtb.

DiscussionLittle is known regarding the biochemical and genetic mecha-nisms of redox sensing in Mtb. Here we show that Mtb WhiB3 isa redox-responsive protein that through its Fe-S cluster respondsto physiologically relevant host signals thought to be involved inpersistence. We provide biochemical evidence that Mtb WhiB3contains a redox-responsive 4Fe-4S cluster that specificallyreacts with O2 and NO. We also provide evidence for a linkamong environmental redox signals, nutritional stress, andWhiB3. These findings will provide mechanistic insight into howMtb senses and responds to protective host molecules andnutritional stress, to favor bacilli persistence.

Our first challenge was to accurately assemble the WhiB3 Fe-Scluster. Rather than exploiting the widely used A. vinelandii NifS forFe-S cluster assembly, we chose to identify and use the endogenousMtb cysteine desulfurase. Through a series of spectroscopic andsensitive labeling experiments, we demonstrated that IscS is apyridoxal 5�-phosphate-containing enzyme capable of assemblingthe WhiB3 Fe-S cluster. Intriguingly, Mtb does not contain other Isccomponents involved in Fe-S cluster assembly (e.g., IscA and IscU)but does possesses the complete repertoire of Suf machinery. Thecharacterization of the Mtb Fe-S cluster assembly components isparticularly interesting because in other systems NifS/IscS has beenimplicated in the repair of oxygen and NO-damaged Fe-S clusterproteins (8). We used the 35S transfer assay and UV-visible andEPR spectroscopy to characterize the WhiB3 Fe-S cluster thor-oughly. Upon treatment with DTH, the EPR silent [4Fe-4S]2�

cluster undergoes a one-electron reduction to generate a paramag-netic state, which results in an EPR signal distinctive of a [4Fe-4S]�cluster (10). The complete absence of 4Fe-4S cluster inWhiB3�Cys-1–4 confirms the role of four conserved residues (C23-C53-C56-C62) in coordinating an Fe-S cluster. Because the E. coliFe-S cluster-containing transcriptional regulators, FNR and SoxR,are sensors of O2 and NO, respectively (6), we studied the effect ofexposure to O2 and NO on the WhiB3 Fe-S cluster and were ableto draw several parallels between the mechanism of cluster degra-dation of WhiB3 and FNR. For example, in response to O2, thetransformation of the WhiB3 [4Fe-4S]2� cluster via [3Fe-4S]� and[2Fe-2S] cluster intermediates is mechanistically similar to theO2-induced modification of the FNR [4Fe-4S]2� cluster, whichinvolves two steps (13). First, oxidation of the WhiB3 cubane[4Fe-4S]2� cluster by O2 rapidly generates the [3Fe-4S]� interme-diate and releases one electron and Fe2� for the two-electronreduction of O2 to produce H2O2. Second, in a nonredox step, H2O2converts the [3Fe-4S]� intermediate to a planar [2Fe-2S]2� cluster(Fig. 5). Finally, H2O2 and/or superoxide destroys the 2Fe-2Scluster (17). These spectroscopic data are further supported by thesensitive Fe-S cluster 35S-labeling assay demonstrating that Fe2�

and S are essential for WhiB3 Fe-S cluster configuration. Impor-tantly, when anaerobically reconstituted WhiB3 was exposed to O2,we demonstrated an associated loss of the WhiB3 Fe-S cluster,suggesting that O2 gradually decays the Fe-S cluster. These datastrongly suggest that similar to FNR, WhiB3 activity is regulated viaits Fe-S cluster by O2 and thus functions as a redox sensor. Notably,the O2 exposure properties of the FNR and WhiB3 Fe-S clustersmay reflect important biological differences between E. coli that iscapable of a rapid transition from aerobic to anaerobic environmentand Mtb that requires a gradual transition. Crucial differencesbetween WhiB3 and FNR are also apparent. For example, whenexposed to O2, loss of the FNR [4Fe-4S]2� cluster is much morerapid. In the case of FNR, 10-min air exposure caused a �50%

decrease in the 4Fe-4S cluster absorbance (12), whereas 120 minwas required to cause a similar decrease for WhiB3. Finally, as is thecase for FNR and SoxR Fe-S clusters, we provide clear evidencethat NO also targets the WhiB3 Fe-S cluster to generate a char-acteristic DNIC complex. The interaction of NO with the WhiB34Fe-4S cluster could modulate its activity in vivo by direct nitrosy-lation of the Fe-S cluster as described for SoxR (6) and FNR (14).

The fact that WhiB3 contains a redox-sensitive 4Fe-4S clusterand is cytoplasmically located and constitutively expressed (18)suggests a role for WhiB3 in sensing physiological changes associ-ated with cellular metabolism. Because the intracellular redoxenvironment is directly coupled to central metabolism, we assessedthe role of WhiB3 in carbon source utilization. The severe growthimpairment of Mtb�whiB3 observed on various carbohydratesimplicates WhiB3 in sensing and responding to the changes in thecellular redox state (energy levels), thereby enabling Mtb to alle-viate the potential harmful effect of redox imbalance. Interestingly,Mtb�whiB3 grew better on acetate as the sole carbon source,pointing toward WhiB3 as a potential regulator of fatty acidmetabolism in Mtb. Several lines of evidence have demonstratedthat Mtb prefers fatty acids rather than carbohydrates as carbonsubstrates during infection (19, 20). These studies suggest that Mtbsenses through a yet-to-be identified mechanism, low concentra-tions of glycolytic substrate and/or oxygen, and initiates a metabolicswitchover to the preferred in vivo carbon source, fatty acids (19,21). The redox-sensing properties of WhiB3, together with thegrowth properties of Mtb�whiB3 on carbon-limiting media, inparticular acetate, implicate WhiB3 in the regulation of this met-abolic switchover event. Lastly, the dramatic effect of whiB3 oncolony rugosity (Fig. 4) suggests alterations in cell wall lipidcomposition and may well be in support of our earlier hypothesisthat whiB3 regulates the production of polyketides to modulate thehost immune system (7).

Nutritional stress is known to influence the electron transportchain, proton-motive force, nucleotide pools, and redox potential

Fig. 5. Hypothetical model depicting WhiB3 as an intracellular redox sensor.In an oxygen-rich environment, the WhiB3 [4Fe-4S]� cluster is oxidized to[4Fe-4S]2� and undergoes sequential conversion, yielding [3Fe-4S]� with theconcomitant production of H2O2 and [2Fe-2S]2� intermediates and eventuallycomplete loss of the cluster. Mtb IscS can activate apo-WhiB3 in vivo togenerate WhiB3 [4Fe-4S]2�. The [4Fe-4S]2� cluster can undergo reduction bythioredoxin and mycothiols to generate [4Fe-4S]�. In vivo growth of Mtb isaccompanied by carbon source and O2 depletion thereby causing fluctuationsin the intracellular redox state. WhiB3 senses these changes in the redox statethrough its [4Fe-4S] cluster to regulate catabolic [glycolysis, tricarboxylic acid(TCA) cycle, glyoxylate cycle] and anabolic (polyketides biosynthesis) metab-olism for maintaining redox balance and persistence. The O2- and NO-sensingproperties of WhiB3 and the growth defect exhibited by Mtb�whiB3 onvarious carbon sources suggest a role as redox sensor responsible for integrat-ing environmental signals to core intermediary metabolism.

11566 � www.pnas.org�cgi�doi�10.1073�pnas.0700490104 Singh et al.

(16). Previous studies have demonstrated a link between nutrientdepletion and Mtb persistence (15) and between aerobic carbonstarvation and anaerobiosis that tightly regulate the intracellularredox balance of bacteria (22). Interestingly, induction of the MtbDos dormancy pathway by hypoxia, NO (2, 23), and palmitate (areduced carbon source) (5) suggests an overlap between the Mtbadaptive response to extracellular redox signals and carbon sourceutilization, which cause changes in the intracellular redox state.

The observation that Mtb�whiB3 showed a growth defect ondifferent carbon sources, but not in the Wayne model (in whichthere is a sufficient carbon source, but oxygen is gradually de-pleted), and none upon exposure to bacteriostatic concentrations ofNO (which inhibits aerobic respiration), strongly supports ourmodel (Fig. 5) and current literature that aerobic respiration andcarbon source utilization are intricately linked to fluctuations in theintracellular redox environment. The latter is supported by elegantstudies showing that in response to carbon source starvation, E. coliFNR and ArcA (24, 25) regulate the synthesis of key enzymesinvolved in glycolysis and tricarboxylic acid cycle. Finally, the severegrowth impairment of Mtb�whiB3 on various carbon sources as wellas the induction of WhiB2 (15), WhiB1 (26), and WhiB3 (27) uponnutrient limitation, by exogenous cAMP, and during infection,respectively, strongly suggests that some Mtb WhiB family membersplay an important role in sensing fluctuations in the intracellularredox state.

In summary, we have identified and characterized a nativeenzymatic reconstitution system enabling the in vitro assembly ofFe-S clusters in Mtb. We have shown that Mtb WhiB3 contains aredox-responsive [4Fe-4S]2� cluster that upon oxygen exposuregenerates a [3Fe-4S]� species and is eventually destroyed, suggest-ing that WhiB3 is a sensor of O2. We have also shown that NO isa ligand of WhiB3. Finally, we have shown that WhiB3 is essentialfor optimal growth on a number of carbon sources, includingtricarboxylic acid cycle intermediates. Based on this cumulativeevidence, we propose that Mtb WhiB3 is a redox-dependent tran-scription factor that senses intracellular redox changes through itsFe-S cluster and integrates environmental signals with core inter-mediary metabolism essential for survival. The ability of WhiB3 toreact with diverse redox signals has broad implications for under-standing the mechanism of Mtb persistence.

Experimental ProceduresCulture Conditions. Mtb H37Rv and Mtb�whiB3 mutant were cul-tured in Middlebrook 7H9/7H10 medium as described previously(7). E. coli cultures were grown in LB medium. Whenever required,antibiotics were added to the culture medium (for E. coli, 100 �g/mlcarbenicillin and 150 �g/ml hygromycin; for Mtb, 50 �g/mlhygromycin).

Overexpression and Purification of WhiB3, WhiB3�Cys-1–4, and Mtb IscS.WhiB3, WhiB3�Cys-1–4, and Mtb IscS were purified as described (seeSI Experimental Procedures).

Iron-Sulfur Cluster Assembly. The Fe-S cluster reconstitution processwas performed under anoxic conditions in an anaerobic glove box.Aerobically purified WhiB3 and WhiB3�Cys-1–4 (30 �M each) werecompletely degassed by using compressed argon gas and anaero-bically equilibrated with buffer containing 50 mM sodium phos-phate, pH 7.5, and 150 mM NaCl. This was followed by the additionof a 10-fold molar excess Fe(II) chloride, 5 mM DTT, and 2.5 �gof IscS. The reaction was initiated by adding 2 mM L-cysteine andsubsequently monitored over time by UV-visible spectroscopy forFe-S cluster assembly. The reaction mixture was terminated bysize-exclusion chromatography to remove low molecular polymers.For reduction, a sample of fully reconstituted WhiB3, after passingthrough a size-exclusion column, was treated with 5 mM DTH andanalyzed by UV-visible spectroscopy. For EPR spectroscopy, re-constituted WhiB3 before (oxidized) and after (reduced) DTHtreatment was transferred to EPR tubes and immediately frozen inliquid nitrogen. To study the effect of O2 on Fe-S cluster stability,a sample of fully reconstituted WhiB3 was purified by size-exclusionchromatography, transferred to an anaerobic cuvette, and airbubbled through the mixture for 2 min. The cuvette was then sealedand monitored by UV-visible spectroscopy (DU800; BeckmanCoulter, Fullerton, CA). For EPR analysis at liquid helium tem-peratures, reconstituted WhiB3 was transferred to an EPR tube 3and 60 min after air exposure and immediately frozen in liquidnitrogen. For NO exposure, purified WhiB3 and anaerobicallyreconstituted WhiB3 were treated inside the anaerobic glove boxwith aliquots of freshly prepared proline NONOate (CaymanChemicals, Ann Arbor, MI). Aliquots of WhiB3 were placed in ananaerobic cuvette and titrated by injection with aliquots of a 2.5 mMstock solution of proline NONOate. Samples were then transferredto EPR tubes and immediately frozen in liquid nitrogen.

EPR Spectroscopy. EPR measurements were performed as de-scribed (see SI Experimental Procedures).

We thank Anton Poliakov for electrospray ionization MS analysis andmembers of the A.J.C.S. laboratory and Mary Hondalus for critical reviewof this manuscript. We also thank Denis Dean (Virginia Tech, Blacksburg,VA) for providing the A. vinelandii nifS plasmid. This work was supportedby National Institute of Allergy and Infectious Diseases, National Institutesof Health Grant R01AI058131 (to A.J.C.S.). A.S. is the recipient of apostdoctoral fellowship from the Heiser Foundation. L.G is the recipient ofNational Institutes of Health Cell and Molecular Biology Training GrantNIH T32 GM008111-19, and J.T is the recipient of Agent of InternationalHealth and Biodefense Concern Training Grant 1T32AI55438.

1. Wayne LG, Sohaskey CD (2001) Annu Rev Microbiol 55:139–163.2. Voskuil MI, Schnappinger D, Visconti KC, Harrell MI, Dolganov GM, Sherman

DR, Schoolnik GK (2003) J Exp Med 198:705–713.3. Loebel RO, Shorr E, Richardson HB (1933) J Bacteriol 26:167–200.4. Ohno H, Zhu G, Mohan VP, Chu D, Kohno S, Jacobs WR, Jr, Chan J (2003) Cell

Microbiol 5:637–648.5. Boshoff HI, Myers TG, Copp BR, McNeil MR, Wilson MA, Barry CE, III (2004)

J Biol Chem 279:40174–40184.6. Green J, Paget MS (2004) Nat Rev Microbiol 2:954–966.7. Steyn AJ, Collins DM, Hondalus MK, Jacobs WR, Jr, Kawakami RP, Bloom BR

(2002) Proc Natl Acad Sci USA 99:3147–3152.8. Johnson DC, Dean DR, Smith AD, Johnson MK (2005) Annu Rev Biochem

74:247–281.9. Heidenreich T, Wollers S, Mendel RR, Bittner F (2005) J Biol Chem 280:4213–4218.

10. Jakimowicz P, Cheesman MR, Bishai WR, Chater KF, Thomson AJ, Buttner MJ(2005) J Biol Chem 280:8309–8315.

11. Ta DT, Vickery LE (1992) J Biol Chem 267:11120–11125.12. Khoroshilova N, Popescu C, Munck E, Beinert H, Kiley PJ (1997) Proc Natl Acad

Sci USA 94:6087–6092.13. Crack J, Green J, Thomson AJ (2004) J Biol Chem 279:9278–9286.

14. Cruz-Ramos H, Crack J, Wu G, Hughes MN, Scott C, Thomson AJ, Green J, PooleRK (2002) EMBO J 21:3235–3244.

15. Betts JC, Lukey PT, Robb LC, McAdam RA, Duncan K (2002) Mol Microbiol43:717–731.

16. Taylor BL, Zhulin IB (1999) Microbiol Mol Biol Rev 63:479–506.17. Sutton VR, Stubna A, Patschkowski T, Munck E, Beinert H, Kiley PJ (2004)

Biochemistry 43:791–798.18. Hutter B, Dick T (1999) Res Microbiol 150:295–301.19. Munoz-Elias EJ, McKinney JD (2005) Nat Med 11:638–644.20. Graham JE, Clark-Curtiss JE (1999) Proc Natl Acad Sci USA 96:11554–11559.21. Bloch H, Segal W (1956) J Bacteriol 72:132–141.22. Nystrom T (1994) Mol Microbiol 12:833–843.23. Park HD, Guinn KM, Harrell MI, Liao R, Voskuil MI, Tompa M, Schoolnik GK,

Sherman DR (2003) Mol Microbiol 48:833–843.24. Shalel-Levanon S, San KY, Bennett GN (2005) Biotechnol Bioeng 92:147–159.25. Levanon SS, San KY, Bennett GN (2005) Biotechnol Bioeng 89:556–564.26. Agarwal N, Raghunand TR, Bishai WR (2006) Microbiology 152:2749–2756.27. Banaiee N, Jacobs WR, Jr, Ernst JD (2006) Infect Immun 74:6449–6457.28. Kumar A, Toledo JC, Patel RP, Lancaster JR, Jr, Steyn AJC (2007) Proc Natl

Acad Sci USA 104:11568–11573.

Singh et al. PNAS � July 10, 2007 � vol. 104 � no. 28 � 11567

BIO

CHEM

ISTR

Y