Proc. Natl. Acad. Sci. USAVol. 93, pp. 11919-11924, October 1996Microbiology

The embAB genes of Mycobacterium avium encode an arabinosyltransferase involved in cell wall arabinan biosynthesis that is thetarget for the antimycobacterial drug ethambutolAIMEE E. BELANGER*, GURDYAL S. BESRA, MICHAEL E. FORD*, KATARINA MIKUSOVA&t, JOHN T. BELISLE,PATRICK J. BRENNAN, AND JULIA M. INAMINEtDepartment of Microbiology, Colorado State University, Fort Collins, CO 80523

Communicated by Arnold Demain, Massachusetts Institute of Technology, Cambridge, MA, August 5, 1996 (received for review April 30, 1996)

ABSTRACT The antimycobacterial compound ethambu-tol [Emb; dextro-2,2'-(ethylenediimino)-di-1-butanol] is usedto treat tuberculosis as well as disseminated infections causedby Mycobacterium avium. The critical target for Emb lies in thepathway for the biosynthesis of cell wall arabinogalactan, butthe molecular mechanisms for drug action and resistance areunknown. The cellular target for Emb was sought using drugresistance, via target overexpression by a plasmid vector, as aselection tool. This strategy led to the cloning of the M. aviumemb region which rendered the otherwise susceptible Myco-bacterium smegmatis host resistant to Emb. This region con-tains three complete open reading frames (ORFs), embR,embA, and embB. The translationally coupled embA and embBgenes are necessary and sufficient for an Emb-resistantphenotype which depends on gene copy number, and theirputative novel membrane proteins are homologous to eachother. The predicted protein encoded byembR, which is relatedto known transcriptional activators from Streptomyces, isexpendable for the phenotypic expression of Emb resistance,but an intact divergent promoter region between embR andembAB is required. An Emb-sensitive cell-free assay for ara-binan biosynthesis shows that overexpression of embAB isassociated with high-level Emb-resistant arabinosyl trans-ferase activity, and that embR appears to modulate the in vitrolevel of this activity. These data suggest thatembAB encode thedrug target ofEmb, the arabinosyl transferase responsible forthe polymerization of arabinose into the arabinan of arabi-nogalactan, and that overproduction of this Emb-sensitivetarget leads to Emb resistance.

Ethambutol [Emb; dextro-2,2'-(ethylenediimino)-di-1-butanol] is a synthetic compound that has been known for itsantimycobacterial activity since it was initially described in1961 (1). It is one of the first-line drugs recommended for thetreatment of disease caused by Mycobacterium tuberculosis (2)as well as opportunistic infections of AIDS patients caused bythe Mycobacterium avium complex (3), and so it has broaderapplication than isoniazid, the other widely used mycobacte-ria-specific drug which is effective only against M. tuberculosis.The genetic basis for Emb resistance has not been determined,and definition of the cellular target(s) has been complicated bythe wide variety of disparate cellular processes that are dis-rupted by this drug (reviewed in ref. 4). However, the synergyagainst M. avium that is achieved when Emb is used incombination with drugs such as rifampicin (discussed in ref. 5)is most readily explained if the site of action of Emb is inarabinogalactan (AG) biosynthesis: disruption of the biosyn-thesis of AG would destroy the macromolecular assembly ofthe mycolyl-AG-peptidoglycan complex of the cell wall (6, 7),

permitting drugs with intracellular targets (such as rifampicin)to enter the cell more easily.

Recent studies support the early work of Takayama andKilburn (8) in defining the biosynthesis of the arabinan ofAGas the critical target for Emb. It has been shown that Embspecifically inhibits the polymerization of cell wall arabinan(9), and that treatment of Emb-susceptible (Embs) Mycobac-terium smegmatis cells with Emb leads to the accumulation of,B-D-arabinofuranosyl-l-monophosphoryldecaprenol (DPA,or decaprenol phosphoarabinose) (10), a likely intermediate inarabinan biosynthesis. These findings led to the developmentof a cell-free assay to show that DPA is an arabinosyl donor forarabinan biosynthesis and that Emb inhibits arabinosyl transfer(11). The results also suggest that multiple arabinosyl trans-ferase targets with varying sensitivities to Emb are present inmycobacteria (9, 11) and that these targets reside in thepathways for the biosynthesis of both AG and lipoarabino-mannan (9). However, the biogenesis of AG appears to bemore important for the antimycobacterial activity ofEmb sincea subinhibitory concentration of Emb still inhibits lipoarabi-nomannan biosynthesis, but not AG biosynthesis, in an Emb-resistant (Embr) mutant of M. smegmatis (9).

In this report, we describe the cloning and sequencing of theemb region from M. avium that confers resistance to Emb whenit is expressed in M. smegmatis on a multicopy vector. Geneticand biochemical evidence are presented to support our con-clusion that the embA and embB genes encode for the arabi-nosyl transferase that is the primary cellular target for Emb.

MATERIALS AND METHODSBacterial Strains and Growth Conditions. M. smegmatis

strain mc2155 (12) and its transformants were grown in 7H11(Difco) broth and agar medium unless otherwise noted. Esch-erichia coli strain X2764 (13) or strains SURE, XL1-Blue orXL2-Blue (all from Stratagene) and their transformants weregrown in Lennox L broth (GIBCO/BRL) and agar medium.All strains were incubated at 37°C except X2764 which wasgrown at 30°C. Antibiotics (Sigma) were added to media at thefollowing concentrations: kanamycin (Kan) at 10 ,ug/ml, tet-racycline at 12.5 ,ug/ml, chloramphenicol at 100 ,ug/ml, andampicillin at 100 ,tg/ml.

Cloning Procedures. A genomic library (14) of M. aviumserovar 2, strain 2151, constructed by cloning 35- to 40-kb

Abbreviations: AG, arabinogalactan; DPA, decaprenol phosphoarabi-nose, or ,B-D-arabinofuranosyl-l-monophosphoryldecaprenol; Emb,ethambutol; Embr, Emb-resistant; Embs, Emb-susceptible; Kan, ka-namycin; MIC, minimal inhibitory concentration.Data deposition: The sequence reported in this paper has beendeposited in the GenBank data base (accession no. U66560).*Present address: Department of Biological Sciences, University ofPittsburgh, Pittsburgh, PA 15260.

tPresent address: Department of Biochemistry, Faculty of NaturalSciences of Comenius University, Bratislava, Slovakia.:To whom reprint requests should be addressed.

11919

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020

11920 Microbiology: Belanger et al.

partial Sau3AI fragments of chromosomal DNA into theshuttle cosmid pYUB18 (15), was the source of cosmidspAEB1-pAEB22. Cosmid pAEB45, which was the target forTnS mutagenesis by standard procedures (16) using ATnS5(provided by Linda Lee, University of Texas Health SciencesCenter, San Antonio), was constructed by trimolecular ligationof the 34-kb HpaI-DraI insert fragment from pAEB4, the402-bp HincII fragment from pYUB18 (15) that contains thelambda cos site, and Bstl 107-linearized shuttle plasmidpYUB56 (17) (pYUB18 and pYUB56 were provided by Wil-liam R. Jacobs, Jr., The Albert Einstein College of Medicine,Yeshiva University). The ligation reaction was packaged withthe Gigapack Plus Kit (Stratagene) and the resulting phageparticles were transduced into E. coli X2764.

Subclones pAEB109-pAEB148 were generated by partialdigestion of pAEB1 with Sau3AI and ligation of the 10- to12-kb fragments into the BamHl site of shuttle plasmid pMD31(18), while subclones pAEB202 and pAEB203 were con-structed by cloning the 8.3-kb SstI fragment from pAEB148 inboth orientations into the SstI site of the mycobacterialintegrating vector pMH94 (19) (these vectors were providedby Graham F. Hatfull, University of Pittsburgh). PlasmidspWRH8 and pWRH10 were produced by treating the 8.3-kbSstI fragment from pAEB1 with mung bean nuclease (Strat-agene) prior to ligation with the 6.4-kb PvuII fragment ofpMD31 (referred to as pMD31APvu) in both orientations.Deletion derivatives were constructed by digesting pWRH8and pWRH10 with SphI and religating the purified 14-kbfragments. E. coli transformations were carried out by theone-step procedure (20), electroporation (Bio-Rad), or byusing commercially-available competent cells (strains SUREor XL2-Blue from Stratagene). Plasmid DNA was isolatedfrom E. coli by alkaline lysis (20) or Qiagen (Chatsworth, CA)columns, and characterized by restriction analysis prior totransformation of M. smegmatis by electroporation (15). Plas-mids were recovered from M. smegmatis transformants byelectroduction into E. coli (21), or by isolating plasmid DNAvia alkaline lysis (20) of cells disrupted with 0.5 mm zirconiumbeads followed by transformation of E. coli.Drug Resistance Assays. M. smegmatis mc2155 transfor-

mants were plated in duplicate on medium containing Kanwith or without Emb at 0.75 ,ug/ml. Isolates were scored asEmbr if growth appeared on both types of plates after a 3-dayincubation. The minimal inhibitory concentration (MIC), de-termined by plating on medium containing Kan plus 0-2.5,ug/ml of Emb in increments of 0.25 ,ug/ml, was defined as thefirst concentration of Emb that inhibited 99-100% of growthafter 3-5 days of incubation. Control strains contained theappropriate cloning vector.Disk susceptibility assays were performed with mc2155

transformed with either pAEB1 or the cloning vectorpYUB18. Overlays consisting of 200 ,lI of a saturated cultureand 5 ml of soft agar were poured onto 100-mm plates of agarmedium and allowed to harden. Paper disks containing thefollowing drugs or antibiotics were fixed to the agar surface: 25jig of capreomycin, 10 gg of Emb, 25 ,ug of ethionamide, 15 ,gof isoniazid, S ,ug of rifampin, and 5 ,ug of streptomycin. Thezones of inhibition were examined after 3 days of incubation.DNA Hybridization. Colony blot hybridizations of E. coli

transformants were carried out on nitrocellulose disks bystandard procedures (20). For Southern blot analysis (20), -2,g of mycobacterial chromosomal DNA was digested withEcoRI. DNA from Mycobacterium leprae and M. tuberculosiswere obtained from National Institutes of Health ContractsAI55262 and AI25147, respectively, at Colorado State Uni-versity, while the others were isolated by the modified beadbeater method (14). Restriction fragments were nonradioac-tively labeled with digoxigenin using the Genius Kit (Boehr-inger Mannheim). Hybridization and washes were performedunder stringent conditions, and the results were visualized by

colorimetric detection (for colony blots) or by exposure tox-ray film following incubation with Lumi-Phos 530 (for South-ern blots), according to the manufacturer's instructions(Boehringer Mannheim).DNA Sequencing. Overlapping restriction fragments that

spanned the Emb resistance region were subcloned frompAEB1 into pBlueScript SK- (Stratagene), and then nesteddeletions were generated using the Exo/Mung DNA Sequenc-ing System (Stratagene). Double-stranded DNA templateswere sequenced using the Sequenase version 2.0 Kit (UnitedStates Biochemical) with pBlueScript reverse and -20 primers(Stratagene). Custom primers were synthesized (Macromolec-ular Resources Facility, Colorado State University) as neces-sary to resolve sequence ambiguities and possible frameshifterrors. Contiguous DNA sequences were constructed with theSequence Assembly Manager (Molecular Biology InformationResource Program, Baylor College of Medicine) and potentialframeshifts in the sequence were identified with Large ScaleSequence Analysis Suite (Keith Robison, Harvard University;made available to us by Douglas Smith). The DNA sequencewas analyzed by the Staden (22), PCGENE (IntelliGenetics), andGenetics Computer Group (GCG, version 10.1) sequenceanalysis programs using default parameters. ORFs were iden-tified with codon usage tables derived from the sequence ofmycobacteriophage L5 (23). Predicted amino acid sequenceswere analyzed with FASTA against the Protein IdentificationResource, SwissProt, and Genpept data bases.

Protein Analysis. M. smegmatis containing either pMD31 orpAEB148 was grown in glycerol-alanine-salts broth containingKan or Kan plus 0.75 ,ug/ml Emb, respectively. Approximately5 g of stationary phase cells was disrupted in a Bead Beater(Biospec, Bartlesville, OK) with 0.1 mm zirconium beads using10 cycles of 10-sec pulses and 30-sec cooling periods. Followingcentrifugation at 500 x g to remove beads and any unbrokencells, the supernatant was centrifuged at 28,000 x g for 30 min.Protein concentrations of the resuspended pellet, containingcell wall and membrane proteins, and the supernatant, con-taining membrane and cytosolic proteins, were determinedusing the BCA assay (Pierce). Equivalent amounts of eachsample (5 jig) were analyzed by SDS/10% PAGE.Whole-Cell Radiolabeling Experiments. Duplicate cultures

of mc2155 containing pMD31 or pAEB148 were grown to anoptical density of 0.6 at 600 nm in glycerol-alanine-salts brothcontaining 0.05% Tween and Kan. Emb was then added to oneset of flasks at a concentration of 1 gg/ml while all flasksreceived 1 ,uCi/ml of D-[14-C]U-glucose (ICN, 250 mCi/mM;1 Ci = 37 GBq). After 4 more hours of incubation, the cellswere harvested and washed with buffer. The lipoarabinoman-nan-free mycolyl-AG-peptidoglycan complex, prepared byethanol reflux of delipidated cells as described (9), was hy-drolyzed for 2 h at 120°C with 2 M CF3COOH prior toextraction with CHCl3 to remove fatty acids. The monosac-charides released by acid treatment were analyzed by HPLC(Dionex, Sunnyvale, CA) using a CarboPack PAl anion-exchange column (4 x 250 mm) and isocratic elution at 1ml/min with 100 mM NaOH. The stream was split between apulsed amperometric detector and a }3-RAM detector (InsusSystems, Tampa, FL). A mixture of unlabeled standards(arabinose, galactose, glucose, mannose, rhamnose, and ri-bose) was coinjected with a separate sample to determine peakidentities.

Cell-Free Arabinosyl Transferase Assays. The incorpora-tion of [14C]-D-arabinose from DPA into arabinan was assayedas described previously (11). Briefly, cell sonicates preparedfrom cultures grown in glycerol-alanine-salts broth were cen-trifuged at 27,000 x g, and the supernatant was recentrifugedat 100,000 x g to yield the membrane pellets containing theenzyme activity (0.4-0.5 mg of protein per reaction). Theacceptors for the reaction, provided by the particulate cell wallfraction (0.2-0.3 mg of protein per reaction), were prepared by

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020

Proc. Natl. Acad. Sci. USA 93 (1996) 11921

fractionating the 27,000 x g pellet on a 60% Percoll gradient,while the radiolabeled arabinose donor (20,000 cpm, 9,M perreaction) was chemically synthesized. The products of thereaction were separated by decending paper chromatography,and the radioactivity retained at the origin was counted todetermine incorporation (typically 30% in the absence ofEmb) into polymer.

RESULTS

Cloning and Characterization of the Emb-Resistance Re-gion. The cellular target for Emb was sought using drugresistance, via target overexpression by a plasmid vector, as aselection tool. A genomic library of DNA from a moderatelyEmbr (MIC of 14 ,ug/ml) strain of M. avium was screened inthe Embs (MIC of 0.25 ,ug/ml), electroporatable host, Msmegmatis strain mc2155, for clones that produced an Embrphenotype. Direct selection of transformants on mediumcontaining Kan and 2.5 ,ug/ml ofEmb led to the identificationof six clones that conferred a 10-fold increase in Emb MIC.Comparison of two overlapping clones, pAEB1 and pAEB4(Fig. 1A), defined a resistance region of 22.5 kb. Disk suscep-tibility assays with other antimycobacterial drugs (includingcapreomycin, ethionamide, isoniazid, rifampin, and strepto-mycin) indicated that resistance was specific for Emb (data notshown).The region required for Emb resistance was further exam-

ined by transposon mutagenesis of pAEB45, a Kan-susceptiblederivative of pAEB4. The resulting composite map of the Tn5insertion sites that localized the Emb resistance region towithin 9.8 kb, based on flanking insertions which did not affectdrug resistance, is shown in Fig. 1B. At the same time, the

A Resistance

pAEB4c E E EB S E V B S E5B d. is ii iii U

dB VE E E e EB S E V B cpAEBI M- I H

dB S E V B S ESB B B_~~~~~~~~~ .ne .pAEB22

pAEB21

B

pAEB138

pAEB148

pAEB109

B VE E E E EB S E VC

r-

E B T K K KS E K TVI I l

<0R I A N B >

x K E B T KK KS E K kI I I i 1 1 II

iI kb

region required for Emb resistance was also defined by over-lapping clone analysis. Cosmids from the genomic library wereidentified by colony blot hybridization using a probe con-structed from the internal 7.1-kb EcoRI fragment common topAEB1 and pAEB4 (Fig. 1A) and then tested for Embresistance in M. smegmatis. The region needed to produceEmbr transformants was narrowed down to the 10.3 kb ofDNAshared by clones pAEB21 and pAEB22 (Fig. 1A); however,these cosmids conferred only about half the Emb MIC (1.25and 1.0 ,ug/ml, respectively) obtained with the original clones.The resistance regions identified by Tn5 mutagenesis andoverlapping clone analysis thus coincided, but the latter resultssuggested that flanking sequences might affect the level ofEmb resistance. This possibility was addressed by screening a10- to 12-kb sublibrary of pAEB1, and Embr clones pAEB138and pAEB109 defined a smaller essential region of 7 kb (Fig.1C). Since all the Embr subclones had a resistance level of 2.5,g/ml of Emb, the same as that obtained with pAEB1, it wasconcluded that the lower MICs obtained with overlappingclones pAEB21 and pAEB22 were possibly due to the insta-bility of those particular inserts in M. smegmatis.The Emb-Resistance Region Is Common to Other Myco-

bacteria. The ubiquity of the Emb-resistance region wasinvestigated by Southern blot analysis with genomic DNA fromM. avium, M. smegmatis, M. tuberculosis, and M. leprae. Usingthe 7.1-kb EcoRI fragment (Fig. 1) as a probe, hybridizingbands were detected in all of the mycobacterial species tested,although there were variations in signal intensity (Fig. 2).These results indicated that the Emb-resistance region (emb)was not unique to Embr mycobacteria and thus did not definean Emb-resistance determinant per se. This conclusion wassupported by the cloning of emb homologs from a number ofEmbr or Embs strains, including M. smegmatis mc2155, Mavium serovar 2 strain 724, and M tuberculosis strains H37Rvand CSU24 (data not shown).

Sequence of the emb Region. The sequence of the 9543-bpEcoRI-EcoRV fragment encompassing emb from M aviumstrain 2151 was determined. The G+C content of 70.4% wasconsistent with mycobacterial DNA (24), and the regioncontained three complete ORFs, designated embR, embA, andembB (indicated in Fig. 1B). The GTG start codon for embRwas chosen by its proximity to a potential AGGAGG ribo-some-binding site with homology to the 3'-OH end of 16Sribosomal RNA from M leprae (25). The product of embR ispredicted to be a 384-aa protein with a calculated Mr of 41,267.

1 23 45 6 7 1 2 3 4 5 6 7

+

kE B T K K K S E K TV x

x K K K S E K TV2

k7.1 kb- 7.1 kb-

+.

x K K K SII

k K E B TpAEB121 I 1

K K K SH-H--H-

E K TV kI I -1

FIG. 1. Genetic map of the M. avium Emb-resistance region. (A)Restriction maps of the inserts from representative cosmids identifiedby direct selection for Emb resistance or by overlapping clone analysis.(B) Transposon Tn5 insertion map relative to a restriction map and thepredicted ORFs (the direction of transcription of embR, embA, andembB is denoted by arrows). TnS insertions which abolish (0) or haveno effect (0) on Emb resistance are indicated. (C) Restriction mapsof the inserts of representative Sau3AI subclones of pAEB1. Resis-tance levels of the clones are given on the right: +, MIC >1.0 ,ug/mlEmb; -, no resistance at 0.25 ,ug/ml Emb; B, BglII; E, EcoRI; V,EcoRV; K,KpnI; S, ScaI; T, SstI; c, vector ClaI; d, vectorDraI; k, vectorKpnI; x, vector XbaI.

FIG. 2. Agarose gel of EcoRI-digested mycobacterial chromo-somal DNAs and corresponding Southern blot using the 7.1-kb EcoRIfragment from pAEB1 as the probe. Lanes: 1 and 6, pAEB1; 2, M.avium serovar 2, 2151; 3, M. smegmatis mc2155; 4, M. tuberculosisH37Rv; 5, M. leprae; 7, 1-kb DNA marker (GIBCO/BRL).

pAEBl44

E KxI Il

Microbiology: Belanger et aL

d El

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020

11922 Microbiology: Belanger et al.

It shows good similarity over its entire length with members ofa family of transcriptional activators in Streptomyces, includingRedD (26), ActII-orf4 (27), the N-terminal domain of AfsR(28), and Dnr I (29); the highest score obtained by FASTAanalysis was with the S. peucetius Dnr I protein, with 33.5%identity over a 254-aa region (data not shown). The embRcoding region is separated from embA and embB by a 178-bpdivergent promoter region. The GTG start codon ofembA wasassigned by its proximity to a potential ribosome-binding site,GAG, while embB lacks a potential ribosome-binding site andthus appears to be translationally coupled to embA: the TG ofthe TGA stop codon of embA is shared by the proposed ATGstart codon of embB (data not shown). The respective deducedamino acid sequences of embA and embB, 1108 aa (Mr =117,372) and 1065 aa (Mr = 114,642), show no similarity toother proteins in the data base but they are related to eachother, with 44.8% identity (63.2% similarity) along their entirelengths (Fig. 3). The products ofembA and embB are predictedto have 8 and 10 potential membrane-spanning domains,respectively (see Fig. 3).The Embr Phenotype Requires Only embA and embB. Align-

ment of the ORFs with the maps of the subclones suggestedthat only embA and embB were required for Emb resistance.Comparison of the inserts of pAEB138 and pAEB121 (Fig.1C), together with the locations of Tn5 insertions in embB thatresulted in Emb-susceptibility (Fig. 1B), clearly indicated thatembB was essential. Similarly, embR was shown to be dispens-able by comparison of pAEB148 and pAEB109, while theEmb-susceptibility of pAEB144 indicated that an intact diver-gent promoter region was required. Since transposon inser-tions in embA could have polar effects on the expression ofembB, an in-frame deletion was constructed to determine ifembA was needed for resistance. The 8.3-kb SstI fragment

bAVMHDGKQR80RL. .. W Nrg P 0 A IPC 75

hbB ___ .OOUUVUNLIM PC 72

SLVII;____ [ __ _ A 147

oWbA n=MFCAA cII U. _ ___219

222

ObA _ S00_0_ _ _SR'h lAR 294

f"U .DbB RLIP R.P..... AN,, ...... =mum" m 277

g _ _~ ~~~______________ 36

VARDRG INVW7 P_ . _SM_ . . .AVSJ 349

444

EbbRPAIILl TlV LI _NRSRLP_ wUMA3IZDMJA 424

LY V519

" 3A;ILRZ ILRLLwRRHRAr.nNFz 499

a"~ ~ ~ ~ ~ x* 11 -1 11 1 1- - 11 1 11111111 --- -.: 11 11 11 :11 II:::

EbB _ __WR

IMbA

1. 1 1 I ::I1 * -1 - --11...: 1 . . :..11 .:.:a" w.......

1:01111-.11-.l. : 11.1: :11 11111 -I.I I.:. .:::".11.1...a"E.M NL_=owyTp _ _ ............................. RDT140

I- -1-1 *1 l1 *III....1 III.--II ... 11111- I 1D" YDWIAT-*UNM W

(shown on the map of pAEB148 in Fig. 1C) was cloned frompAEB1 into pMD31APvu in both orientations to producepWRH8 and pWRH10, both of which confer MICs of 2.5,ug/ml of Emb. Deletion of the 660-bp SphI fragment, whichremoved 220 internal amino acids (aa positions 544-763 in Fig.3) but maintained the reading frame of embA, resulted in theloss of Emb resistance. Together these results indicated thatboth embA and embB are required for the Embr phenotype.Copy Number Determines the Level ofEmb Resistance. The

Emb resistance level of 10-fold over background correlatedwith the predicted copy number of 5 to ten for the cloningvectors (15), and a strong linear relationship was noted be-tween the concentration of Emb used for selection of mc2155transformants and the number of Embr colonies obtained(data not shown). To see if resistance was influenced by copynumber, embAB was inserted into the nonreplicating, integra-tive vector pMH94 which contains attP and int from myco-bacteriophage L5 (19). When both orientations of the 8.3-kbSstI fragment were cloned from pAEB148 into the SstI site ofthe vector (to produce pAEB202 and pAEB203), and theconstructs were integrated into the attB site of the M. smeg-matis chromosome (data not shown), the MIC dropped to2-fold over background. Since the same SstI fragment pro-duced a 10-fold increase in Emb resistance when it was clonedin a replicating plasmid (pWRH8 and pWRH10, above), theseresults indicated that the level of Emb resistance mediated bythe emb region was not an intrinsic property of the clonedDNA but depended on gene copy number.EmbA and EmbB Are Overexpressed in M. smegmatis. The

above observation that gene copy number determined theEmb resistance level suggested that the embAB gene productsmight be overexpressed in M. smegmatis. Crude pellet andsupernatant fractions of mc2155 containing either pAEB148 orthe cloning vector pMD31 were compared by polyacrylamidegel electrophoresis. As shown in Fig. 4, the cell wall-membranefraction from cells harboring pAEB148 contains a significantincrease in a band of protein(s) above the 97.4-kDa marker,consistent in size with those predicted for the EmbA andEmbB proteins (117- and 115-kDa, respectively), while thisincrease was not observed in the vector control.

Cell Wall Arabinan Biosynthesis Proceeds Normally inEmb-Treated Cells Containing embAB. It was shown in pre-vious studies that the treatment of 14C-glucose-labeled M.smegmatis mc2155 cells with Emb severely diminished theincorporation of radiolabeled arabinose into AG (9). Usingsimilar techniques, the effect of Emb was examined usingstrains of mc2155 transformed with either pMD31, the cloning

594

573

669

649

744

715

319

779

693

*51

966

926

kDa200

- 97.4

- 68

wranvrvwro____^_ 10411 11111 *1 -11 11 11 1-' *111 -1 . 11-1:1 :11 11 ... ..: 1---I1a"e_S 1001

FIG. 3. Alignment of the embA and embB gene products using theBESTFIT program (Genetics Computer Group). |, Identical amino acids;(:), conserved amino acids; (.), semiconserved amino acids. Theputative membrane-spanning domains identified by the programRAOARGOS (IntelliGenetics) are underlined.

FIG. 4. Protein profiles of the cell wall-membrane fractions of M.smegmatis mc2155 containing the cloning vector pMD31 (lane 1) orpAEB148 (lane 2) relative to size standards (lane 3; GIBCO/BRL)visualized by silver staining of a 10% polyacrylamide gel. The over-expressed protein band (lane 2) is denoted by an arrow.

II.: I::I 1:1-41 : : 1-1: 1.11 IIIIIII: I::- --I:II -1--I:a"

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020

Proc. Natl. Acad. Sci. USA 93 (1996) 11923

vector, or pAEB148, a pMD31-derived plasmid containing theembAB genes. Emb treatment caused a dramatic decrease inthe amount of 14C-labeled arabinose in AG when pMD31 waspresent in M. smegmatis, while Emb did not have an effectwhen the host cells contained pAEB148 (Fig. 5).

Association of embAB with Embr Arabinosyl TransferaseActivity and Effect of embR on the Level of Emb ResistanceAchieved in a Cell-Free Assay. It was previously shown thatEmb inhibits the incorporation of radiolabeled arabinose fromthe lipid carrier, DPA, into a polymer of arabinan (11). Thissame cell-free assay was used to determine if the above effectsof Emb on AG biosynthesis in whole cells could be directlyrelated to arabinosyl transferase activity in vitro. Maximalinhibition of arabinosyl transferase activity was determinedfrom dose-response curves with increasing concentrations ofEmb, and three types of Emb resistance were observed.Cell-free preparations ofM. smegmatis mc2155 with or withoutthe cloning vector pMD31 were the most sensitive to Emb,with only 26-29% residual arabinosyl transferase activity inthe presence of 50 ,tg/ml Emb (representative results areshown in Fig. 6); these results are consistent with previousstudies that indicated that this assay supports a number ofarabinosyl transferases and that one or more are intrinsicallyresistant to Emb (11). In contrast, preparations from cellscontaining pAEB1 or pAEB148 were highly resistant to Emb,with 63-65% of the arabinosyl transferase activity remainingEmbr, while that obtained with pAEB109, as well as an Embrmutant of mc2155 (9), displayed intermediate levels (37-40%)of Embr arabinan biosynthesis (representative results areshown in Fig. 6). As was noted earlier, the major differencebetween pAEB148 and pAEB109, which both contain embAB,is the presence or absence of embR, respectively (see Fig. 1C).A priori, it might be expected that the amount of the

Emb-sensitive target present in each cell-free preparationwould dictate the concentration of Emb required for inhibi-tion, and that all extracts would eventually achieve the samesaturated background activities, assuming that the amount ofarabinose acceptors remains constant. Since the initial levelsof endogenous arabinose acceptors are about the same for allthe preparations tested, as judged by the similar amounts of

pMD31 pAEB148

Elution Time

FIG. 5. Effect of Emb on the incorporation of 14C from [14C]glu-cose into the sugar components arabinose (Ara) and galactose (Gal)of AG. The hydrolysates (ca. 25,000 cpm of each) were from M.smegmatis mc2155 cells, containing either the cloning vector pMD31or pAEB148, that were untreated (-) or treated (+) with 1 ,g/ml ofEmb. The radiolabeled glucose (Glc) arises from some residual glucan.

1H

ArabinosylTranferaseActivity

0 5 10 15 20 25 30 35 40 45 50Concentration of Ethambutol (ug/ml)

FIG. 6. Effect of increasing concentrations of Emb on arabinosyltransferase activity in a cell-free assay. The incorporation of 14C-arabinose from DPA into a polymer of arabinan was measured in thepresence of 0, 1, 5, 10, 15, 25, and 50 ,ug/ml of Emb with extractsprepared from: *, mc2155, the Embs host; A, mc2155 containing thecloning vector pMD31 (Embs); *, EMB-R, an Embr mutant of mc2155(ref. 9); C1, mc2155 containing pAEB109 (Embr), a subclone ofembABthat is missing embR; 0, mc2155 containing pAEB148 (Embr), asubclone of embR and embAB; *, mc2155 containing pAEB1, one ofthe original Embr cosmid clones.

radiolabel that each can incorporate in the absence of Emb(see Materials and Methods), the observed differences inmaximum levels of inhibition likely reflect the differentialabilities of the cell-free extracts to support the synthesis ofintermediate arabinose acceptors, as well as final arabinanproducts, in the presence of Emb.

DISCUSSIONThese studies suggest that the embAB genes of M. aviumencode for the primary cellular target for Emb, namely thearabinosyl transferase III responsible for the polymerization ofarabinose into the arabinan of AG (9). This conclusion wassupported by the ability of the embAB genes to specificallyconfer an Emb-resistant phenotype that depends on gene copynumber, consistent with drug resistance resulting from over-expression of the drug target. In addition, overexpression ofthese genes permits normal incorporation of arabinose intoAG in the presence of Emb. Although these results could notrule out the possibility that the observed Emb resistance wasdue to a novel mechanism for drug exclusion or efflux, thecell-free assays for arabinan biosynthesis clearly showed thatembAB are associated with high-level Embr arabinosyl trans-ferase activity. These genes are ubiquitous amongst mycobac-teria regardless of the strains' Emb susceptibility levels, aswould be expected for genes involved in cell wall biosynthesis.Since recent-studies indicate that multiple-drug resistance inM. avium and M. tuberculosis usually results from the accu-mulation of mutations in genes encoding the drug targets(reviewed in ref. 30), it can be predicted that Embr clinicalisolates will have up-regulatory mutations in embR or thedivergent promoter region, or have altered embA or embBstructural genes.A working hypothesis that takes into account the potential

translational coupling of embAB, the mechanism of action ofEmbA and EmbB, and the possible role of EmbR, has beenformed by using secondary metabolite biosynthesis as a model.

Microbiology: Belanger et aL

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020

11924 Microbiology: Belanger et al.

Although secondary metabolites vary in composition andfunction, the synthesis of these complex molecules alwaysoccurs from the condensation of single identical starter units.Mycobacterial arabinan may be considered similar in thisregard since this elaborate homopolysaccharide is generatedfrom multiple D-arabinofuranose units (7). The translationalcoupling that is predicted for embAB is usually associated withgenes whose protein products are required in equimolaramounts, such as those functioning in multienzyme complexes(31), and this feature, together with the amino acid sequencesimilarity of EmbA and EmbB, is found among enzymesinvolved in secondary metabolite biosynthesis. For example,the antibiotics actinorhodin and granaticin are synthesizedfrom heterodimeric enzyme complexes which are composed oftwo translationally coupled proteins sharing 49% aa similarity(32, 33). Likewise, the erythromycin gene cluster encodes forthree ORFs, two of which are translationally coupled, with64% or higher amino acid similarity (34). The target of Embmay thus be a heterodimeric enzyme complex derived fromEmbA and EmbB. The identification of embR, with its highdegree of similarity to transcriptional activators which regulatesecondary metabolite synthesis, is compatible with this hy-pothesis.A role for embR in modulating expression of embA and

embB was originally suggested by the characteristic geneticorganization of regulatory protein-divergent promoter region-structural gene (35), and this model is supported by thedecrease in the level of Embr arabinosyl transferase activityobserved when embR is missing. EmbR lacks a helix-turn-helixmotif but is similar (27.1% identity over a 129 aa region; datanot shown) to ToxR, a transcriptional activator for choleratoxin production (36). The region of similarity encompasses a96-aa domain of the N terminus of ToxR that associates withdirectly repeated sequences in the promoter region of the ctxoperon (36), and the divergent promoter region between embRand embA embB contains two direct DNA repeats ofCGGGCGGA and one related sequence (GCGGCGGA)(data not shown). Curiously, embR does not appear to influ-ence the phenotypic expression of Emb resistance by wholecells, although the intact divergent promoter region is re-quired. It is possible that the host cells, which presumablycontain an embR homolog, provide sufficient complementa-tion in trans to produce an Embr phenotype that is notdistinguishable from that of embR-containing cells until ex-amined at the cellular level. Future studies should resolve thesequestions and provide further insights into the regulation ofmycobacterial cell wall biosynthesis.

We thank Douglas Smith and Graham Hatfull for assistance withDNA sequence analysis; Richard Lee for the synthesis of radiolabeledDPA; Michael McNeil for assistance with the Dionex HPLC; MichaelSonnenberg for assistance with protein analysis; and William Howeand Caroline Morehouse for excellent technical assistance. This workwas supported by Grants A130189 and A138087 of the NationalCooperative Drug Discovery Group-OI, National Institute of Allergyand Infectious Diseases (P.J.B., Program Primary Investigator); PublicHealth Service Grants AI33706 (Michael R. McNeil, Primary Inves-tigator) and AI01185 (to J.M.I.); the Colorado Agricultural Experi-ment Station, and Animal Health and Disease Formula funds (toJ.M.I.).

1. Thomas, J. P., Baughn, C. O., Wilkinson, R. G. & Shepherd,R. G. (1961) Am. Rev. Respir. Dis. 83, 891-893.

2. Centers for Disease Control and Prevention (1993) Morb. Mortal.Wd&y. Rep. 42, 1-8.

3. Masur, H. (1993) N. Engl. J. Med. 329, 898-904.4. Winder, F. G. (1982) in The Biology of the Mycobacteria, eds.

Ratledge, C. & Stanford, J. (Academic, London), Vol. 1, pp.353-438.

5. Heifets, L. B. (1992) in Drug Susceptibility in the Chemotherapy ofMycobacterial Infections, ed. Heifets, L. B. (CRC, Boca Raton,FL), pp. 179-199.

6. Draper, P. (1982) in The Biology of the Mycobacteria, eds.Ratledge, C. & Stanford, J. (Academic, London), Vol. 1, pp.9-52.

7. Brennan, P. J. & Nikaido, H. (1995) Annu. Rev. Biochem. 64,29-63.

8. Takayama, K. & Kilburn, J. 0. (1989) Antimicrob. Agents Che-mother. 33, 1493-1499.

9. Mikusova, K., Slayden, R. A., Besra, G. S. & Brennan, P. J.(1995) Antimicrob. Agents Chemother. 39, 2484-2489.

10. Wolucka, B. A., McNeil, M. R., de Hoffman, E., Chojnacki, T. &Brennan, P. J. (1994) J. Biol. Chem. 269, 23328-23335.

11. Lee, R. L., Miku§o&d, K., Brennan, P. J. & Besra, G. S. (1995) J.Am. Chem. Soc. 117, 11829-11832.

12. Snapper, S. B., Melton, R. E., Kieser, T., Mustafa, S. & Jacobs,W. R., Jr. (1990) Mol. Microbiol. 4, 1911-1919.

13. Jacobs, W. R., Jr., Barrett, J. F., Clark-Curtiss, J. E. & Curtiss, R.,III (1986) J. Bacteriol. 52, 101-109.

14. Belisle, J. T., Klaczkiewicz, K., Brennan, P. J., Jacobs, W. R., Jr.,& Inamine, J. M. (1993) J. Biol. Chem. 268, 10517-10523.

15. Jacobs, W. R., Jr., Kalpana, G. V., Cirillo, J. D., Pascopella, L.,Snapper, S. B., Udani, R. A., Jones, W., Barletta, R. G. & Bloom,B. R. (1991) Methods Enzymol. 204, 537-555.

16. de Bruijn, F. J. & Lupski, J. R. (1984) Gene 27, 131-149.17. Mills, J. A., McNeil, M. R., Belisle, J. T., Jacobs, W. R., Jr., &

Brennan, P. J. (1994) J. Bacteriol. 176, 4803-4808.18. Donnelly-Wu, M. K., Jacobs, W. R., Jr., & Hatfull, G. F. (1993)

Mol. Microbiol. 7, 407-417.19. Lee, M. H., Pascopella, L., Jacobs, W. R., Jr., & Hatfull, G. F.

(1991) Proc. Natl. Acad. Sci. USA 88, 3111-3115.20. Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seid-

man, J. G., Smith, J. A. & Struhl, K. (1996) Current Protocols inMolecular Biology (Greene & Wiley, New York).

21. Baulard, A., Jourdan, C., Mercenier, A. & Locht, C. (1992)Nucleic Acids Res. 20, 4105.

22. Staden, R. (1986) Nucleic Acids Res. 14, 217-231.23. Hatfull, G. F. & Sarkis, G. J. (1993) Mol. Microbiol. 7, 395-405.24. Clark-Curtiss, J. E. (1990) in Molecular Biology of the Mycobac-

teria, ed. McFadden, J. (Surrey Univ. Press, London), pp. 77-96.25. Smida, J., Kazda, J. & Stackebrandt, E. (1988) Int. J. Lepr. 56,

449-453.26. Narva, K. E. & Feitelson, J. S. (1990) J. Bacteriol. 172, 326-333.27. Fernandez-Moreno, M. A., Cabellero, J. L., Hopwood, D. A. &

Malpartida, F. (1991) Cell 66, 769-780.28. Horinouchi, S., Kito, M., Nishiyama, M., Furuya, K., Hong, S.-K.,

Miyake, K. & Beppu, T. (1990) Gene 95, 49-56.29. Stutzman-Engwall, K. J., Otten, S. L. & Hutchinson, C. R. (1992)

J. Bacteriol. 174, 144-154.30. Musser, J. M. (1995) Clin. Microbiol. Rev. 8, 496-514.31. Normark, S., Bergstrom, S., Edlund, T., Grundstrom, T., Jaurin,

B., Lindberg, F. P. & Olsson, 0. (1983) Annu. Rev. Genet. 17,499-525.

32. Fernandez-Moreno, M. A., Martinez, E., Boto, L., Hopwood,D. A. & Malpartida, F. (1992) J. Biol. Chem. 267, 19278-19290.

33. Sherman, D. H., Malpartida, F., Bibb, M. J., Keiser, H. M. &Hopwood, D. A. (1989) EMBO J. 8, 2717-2725.

34. Donadio, S., Staver, M. J., McAlpine, J. B., Swanson, S. J. &Katz, L. (1991) Science 252, 675-679.

35. Beck, C. F. & Warren, R. A. J. (1988) Microbiol. Rev. 52,318-326.

36. Miller, V. L., Taylor, R. K. & Mekalanos, J. J. (1987) Cell 48,271-279.

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020