Vol. 173, No. 15JOURNAL OF BACTERIOLOGY, Aug. 1991, p. 4725-47350021-9193/91/154725-11$02.00/0Copyright X 1991, American Society for Microbiology

Transcriptional Attenuation Control of ermK, a Macrolide-Lincosamide-Streptogramin B Resistance Determinant

from Bacillus licheniformisJIN-HWAN KWAK,1 EUNG-CHIL CHOI,1 AND BERNARD WEISBLUM2*

College ofPharmacy, Seoul National University, San 56-1, Sinlim-Dong, Kwanack-Gu, Seoul 151-742, Korea,1 andDepartment ofPharmacology, UnCversity of Wisconsin Medical School, Madison, Wisconsin 537062

Received 31 December 1990/Accepted 17 May 1991

ernK instructs bacteria to synthesize an erythtomycin-inducible 23S rRNA methylase that confers resistanceto the macrolide, lincosamide, and streptogramin B antibiotics. Expression,of ermK is regulated by transcrip-tional attenuation, in contrast to other inducible erm genes, previously described, which are regulatedtranslationally. The ermK mRNA leader sequence llas a total length of 387 nucleotides and encodes a 14-amino-acid leader peptide together with its ribbsome binding site. Additionally~the mRNA leader sequence canfold in either. of two mutually exclusive conformations, one of which is postulated to form in the absence ofinduction afid to contain two rho factor-independent terminators. Truncated transcription products ca. 210and 333 nucleotides long were synthesized in the absence of induction, both in vivo and in vitro, as predictedby the transcriptional attenuation model; run-off transcription in vitro with rITP favored the synthesis of thefull-length run-off transcript over that of the 210- and 333-nucleotide truncated products. Northern (RNA) blotanalysis of transcripts synthesized in vivo in the absence of erythroniycin indicated that transcriptionterminated at either of the two inverted compiementary repeat sequences in the leader that were postulated toserve as rho factor-independent terminators; -moreover, no full-length transcripts were detectable in theuninduced samples. In contrast, full-length (ca. 1,200-nucleotide) transcripts were only detected in RNAsamples synthesized in vivo in the presence of erythromycin. Full-length transcripts formed in the absence ofinduction from transcriptional readthrough past the two proposed transcription terminators would fold in away that would sequester the ribosome binding site together with the first two codons of the ErmK methylase,reducing its efficiency in translation. This feature could therefore provide additional control of expression in theabsence of induction; however, such regulation, if operative, would act only secondarily, both in time andplace, relative to transcriptional control. Analysis by reverse transcriptase mapping of in vivo transcripts fromtwo primers that bracket the transcription terminator responsible for the 210-nucleotide truncated fragmentsupports the transcriptional attenuation model proposed and suggests further that the synthesis of the ermKmessage is initiated constitutively upstream of the proposed terminator but completed inductively downstreamof this site.

Translational attenuation is utilized by gram-positive bac-teria for the regulation of resistance to the macrolide-lincosamide-streptogramin B (MLS) antibiotics (11, 13).From studies with an ermC model system, it has been shownthat during indfiction by eiythromycin, the arrested ribo-some binds at a specific location on the message (22, 23, 29).The ermC leader, contains inverted complementary repeatsequences that allow it to fold in three alternative conform-ations (11, 13). As a consequence of the erythromycin-stabilized binding between mRNA and the ribosome duringinduction, the nascent message assumes a translatioriallyactive conformation. Conformational isomerization of ermmessages because of stalled ribosomes could, in principle,mediate induction in both transcriptional and translationalattenuators; however, transcriptional attenuation control ofan erm gene has not yet been found. Several inducible ermgenes have been cloned and characterized, and translationalattenuation control has been proposed for their mode ofregulation. For reviews, see references 8 and 33.Of the inducible erm genes that have been characterized,

ermD from Bacillus licheniformis, described by Docherty etal. (7), is of particular interest because of its reported

* Corresponding author.

complex 354-nucleotide mRNA leader sequence, for which atranslational attenuation mechanism was conjectured (10).In the present study, we analyze ermK from an isolate of B.licheniformis found in Korea (5). The ermK leader sequencedescribed in the present study was found to contain 357(rather than 354) nucleotides and, over the portion of itssequence that overlaps ermD, to differ from the latter at only3 nucleotide positions. The two leader sequences appearotherwise functionally indistinguishable; both can be foldedin the same manner, and both contain an open reading framethat encodes the 14-amino-acid leader peptide, MTHSMRLRFPTLNQ. Ih a departure from translational attenuationcontrol models previously reported for inducible erm genes,we show that ermK (and most probably ermD, too) isregulated by transcriptional attenuation.

MATERIALS AND METHODS

Strains and plasmids. The strains and plasmids used in thisstudy are listed in Table 1.Cloning of ermK and selection of constitutive mutants. ermK

was initially cloned as a 3.2-kb fragment obtained by diges-tion of DNA from B. licheniformis EMR-1 with restrictionendonuclease EcoRI and ligation of the resultant digest withplasmid pBS42 DNA digested with EcoRI (see Fig. 1A). The

4725

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

4726 KWAK ET AL.

TABLE 1. Bacterial strains and plasmids

Strain or plasmid Description

StrainsB. subtilis UOTO-277 .......Transformable host for studying

expression of plasmid construc-tions; from H. Shimotsu; BacillusGenetic Stock Center code 1A509

B. licheniformis EMR-1 ..... Source of ermK (5)E. coli CSH-26 ................Transformable host for preparation

of B. licheniformis DNA library(25)

PlasmidspBS42 .................. Plasmid vector for clonling of ermK

from B. licheniformis DNA (1)pEC101 .................. Derived from pBS42,contains

3.2-kb EcoRI fragment of B. li-cheniformis DNA with ermK in-sert

pEC103 .................. Derived from pEC101; contains1.3-kb fragment of ermK insert ofpEC101 bounded by PvuII andHindIll sites (Fig. 1)

pEC68 .................. Derived from B. subtilis UOTO-277(pEC101) by selection with tylosinat 10 ,ug/ml

pEC301 .................. Derived from B. subtilis UOTO-277(pEC101) by selection with carbo-mycin at 10 ,ug/ml

pEC501 .................. Derived from B. subtilis UOTO-277(pEC101) by selection with tylosinat 10 ,ug/ml

pEC597 .................. Derived from B. subtilis UOTO-277(pEC101) by selection with tylosinat 10 p.g/ml

pEC712 .................. Derived from B. subtilis UOTO-277(pEC101) by selection with tylosinat 10 p.g/ml

resultant preparation was introduced into Escherichia coliCSH-26 by transformation, and selection was done forexpression of the chloramphenicol resistance marker on

plasmid pBS42. Plasmid DNA was prepared from the result-ant library in E. coli and used to transform Bacillus subtilisUOTO-277, and selection was done for both chlorampheni-col and erythromycin resistance. One of the transformantclones was selected for further study and used as source ofplasmid pEC101. A partial digest of the cloned fragmentobtained by digestion with PvuII and Hindlll was blunted bytreatment with S1 nuclease and ligated to plasmid pBS42DNA digested with BamHI and similarly blunted with S1nuclease, yielding plasmid pEC103. The resultant pEC103insert, 1.3 kb in length (coordinates 380 to 1670; see Fig. 1B),conferred inducible MLS resistance on transformants. Plas-mid pEC103 was introduced into B. subtilis UOTO-277 bytransformation, and regulatory mutants were selected byinoculation of the resultant strain onto solid medium con-

taining either carbomycin or tylosin (10 ,ug/ml). Resistantcolonies were analyzed as described previously (14).ermK run-off transcription. The 457-bp DNA fragment that

was used as a template for in vitro transcription was ob-tained by digestion of pEC101 DNA with PvuII (map coor-

dinate 380) and MspI (map coordinate 840), agarose gelelectrophoresis of the resultant digest, and extraction fromthe gel by electroelution. Transcription was performed as

described by Levin and Chamberlin (20). The reactionmixture contained, in 20 p.l, 20 mM Tris HCl (pH 7.9), 4 mM

MgCl2, 140 mM KCl, 0.1 mM Na2EDTA, 150 ,uM rATP, 20p.M rGTP (or 150 puM rITP), 0.1 mM dithiothreitol, 20 ,ug ofbovine serum albumin per ml, 20 ng of DNA template, 10,uCi of [Ot-32P]GTP (3,000 Ci/mmol), and 3 U of B. subtilisRNA polymerase. The reaction mixture was incubated at37°C for 10 min. rCTP and rUTP, 150 puM each, were added,and incubation was continued for an additional 20 min. Thereaction was terminated with 5 p.l of stop solution (8 M urea,0.1% sodium dodecyl sulfate, 0.025% xylene cyanol, 0.025%bromophenol blue), the reaction mixture was fractionated bypolyacrylamide gel electrophoresis, and the resultant frag-ment pattern was visualized by autoradiography.

Determination of ermK transcription initiation by analysis ofin vivo transcripts with Si nuclease. For S1 nuclease map-ping, the nuclease protection method of Berk and Sharp (3)was used. The DNA probe, labeled at one end, was preparedby the polymerase chain reaction with oligonucleotide 20553as the upstream unlabeled primer (sense direction) andoligonucleotide 10178 as the downstream 5'-end-labeledprimer (antigense direction). Plasmid pEC101 DNA wasused as a template for the polymerase chain reaction step.The labeled probe was partially purified by phenol extractionand ethanol precipitation and annealed with total RNAextracted from B. subtilis UOTO-277(pEC101) in a mediumcontaining 40 mM 1,4-piperazine-diethanesulfonic acid (pH6.4), 1 mM Na2EDTA, 400 mM NaCl, and 80% deionizedformamide. Denatured DNA and RNA were mixed, heatedat 85°C for 10 min, and annealed at 45°C for 4 h. Theresultant heteroduplexes were digested with S1 nuclease.The same end-labeled probe was also used to preparecalibration standards by purine- and pyrimidine-specificchemical cleavages as described by Maxam and Gilbert (21).The resultant products were fractionated by polyacrylamidegel electrophoresis (8%, in the presence of 8 M urea) anddetected by autoradiography.Northern (RNA) blot analysis, reverse transcriptase map-

ping, and DNA sequence analysis. Northern blot analysis wasperformed as described by Davis et al. (6) with variousprobes and templates. Reverse transcriptase mapping byprimer extension was performed as described by Stern et al.(32) with various oligonucleotide primers. DNA sequencingwith dideoxy chain terminators was performed as describedby Sanger et al. (31) with ermK and its fragments cloned inphage M13 (24).DNA oligonucleotides. DNA oligonucleotides were synthe-

sized at the University of Wisconsin Biotechnology Centeras follows: oligonucleotide 20553, 5'-ACG AGA AAA TTCCAG CTG TTG (21-mer, DNA coordinates 368 to 388; seeFig. 2 [sense strand]); oligonucleotide 2168, 5'-T ATT ATGCAT CAA ATG CTG TCC GGA (25-mer, DNA coordinates859 to 835; see Fig. 2 [antisense strand, mRNA coordinates444 to 420; see Fig. 3a]); oligonucleotide 2462, 5'-GAA TCCGCG GTA TTT CCA TAA CCA A (25-mer, DNA coordi-nates 689 to 665; see Fig. 2 [antisense strand, mRNAcoordinates 275 to 251; see Fig. 3a]); oligonucleotide 4400,5'-ACA AGT TAC AAA TTT TAT ATG CAA ATT (27-mer,DNA coordinates 595 to 569; see Fig. 2 [antisense strand,mRNA coordinates 181 to 155; see Fig. 3a]); and oligonucle-otide 10178, 5'-TGG GAA ACG CAG TCG CAT TGA(21-mer, DNA coordinates 385 to 365; see Fig. 2 [antisensestrand, mRNA coordinates 66 to 46; see Fig. 3a]).

RESULTS

Cloning of ermK. B. licheniformis EMR-1 was isolatedfrom a soil sample collected in Seoul, Korea, and, after

J. BACTERIOL.

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

TRANSCRIPTIONAL ATTENUATION CONTROL OF ermK 4727

1 1 I1000 2000

ermK _

MsplDdeI (BspMODDdeI MspI

-- ANrAttenuator Methylase

1I1 500 1000 IS

NUCLEOTIDEFIG. 1. Physical map of ermK. (A) The 3.2-kb fr

licheniformis EMR-1 DNA that contains ermK is sh(with a summary of restriction sites that are pertinent tof ermK regulation. (B) Sequenced 1,841-nucleotideobtained by digestion with PvuII and HindIll. Thiscontains ermK in its entirety, together with its promoteand transcription terminator.

selection for resistance to erythromycin, was fpress erythromycin-inducible MLS resistance.this organism, digested with EcoRI, was cloned NpBS42 as a cloning vector. After transformatioltion for erythromycin resistance, plasmid pEC10a 3.2-kb insert was obtained. A physical map (Iobtained by analysis of pEC101 with restricticleases. The resistance determinant was furtherdigestion of the EcoRI insert with PvuII andblunting with Si nuclease. The resultant 1.3-I(Fig. 1B, coordinates 380 to 1670), inserted intoBamHI site of pBS42, also conferred inducibletance to B. subtilis UOTO-277 transformants.

Analysis of the ermK sequence. The nucleotidethe 1.3-kb PvuII-HindIII DNA fragment (Fig. 2,380 to 1670) as well as additional flanking seqidetermined. The sequence shown in Fig. 2additional 380 nucleotides upstream of PvuII (clto 380) to facilitate comparison with the erm

previously reported by Gryczan et al. (10) asadditional 171 nucleotides downstream of Hin(nates 1670 to 1841) to present the ermK transcrinator in its entirety.The ermK open reading frame shows amino ac

similarity to the erm methylases previously desc'particular, only minor differences from ermDreported [10]), the upstream boundary of whichposition corresponding to coordinate 342 (Fig.with an asterisk). The methylase open readipreceded by a 357-nucleotide leader sequenceencoding a 14-amino-acid peptide, MTHSMRLtogether with its ribosome binding site. The s;was also reported in ermD (10). A third open reappears to follow immediately after the transcrinator of ermK, with a ribosome binding site a1681 and an initiator Met codon at coordinatpredicted partial amino acid sequence wouldFLLPAYSFPVKNCASPRPKLQSASPFSVMIMPHSASS, and it has not yet appeared as sucibases that we have searched.

Alternate conformations, shown in Fig. 3a al

proposed for the leader region, which superficially resemblesBclI EcoRI other erm leader regions that have been studied and reported

to function as translational attenuators.The pattern of association shown in Fig. 3a would result in

33200 the pairing of complementary nascent segments as they

became available during transcription. The final set of in-verted repeats, 7 and 8, if formed, would sequester both themethylase ribosome binding site (SD-2) and the codons forthe first two amino acids, Met and Lys, of the methylase;

MspI Hi ndI however, as we show below, in the absence of induction,( transcription would be terminated much earlier. The alter-

nate conformation shown in Fig. 3b would free SD-2 and thecodons for Met and Lys, leaving them enclosed within a29-nucleotide loop which, because of its large size, would be

;oo 1941 functionally equivalent to unstructured mRNA and would

allow ribosomes to initiate protein synthesis. Once initiated,ragment of B. protein synthesis could progress as in any message thatown, together possesses secondary structure internal to its open readingo the analysis frame.subfragment Two sets of inverted repeats, 3 and 4 and 7 and 8, withinsubfragment the leader region resemble rho factor-independent transcrip-

.r, attenuator, tion terminators (Fig. 3a). The first, 3 and 4, at mRNAcoordinates 182 to 207 (DNA sequence coordinates 596 to621), is followed by a series of five uridylate residues thatwould serve to terminate the ermK message, yielding a ca.

Found to ex- 210-nucleotide fragment. A ribosome stalled in the ermKDNA from leader peptide coding region would be in position to disrupt

with plasmid the secondary structure of segments 1 and 2, favoring then and selec- alternate conformation of the leader region shown in Fig. 3b,'1 containing in which the conformation of the postulated rho factor-Fig. 1A) was independent terminator would be disrupted, favoring contin-ion endonu- ued transcription through this sequence. The second poten-localized by tial transcription terminator, 7 and 8, at mRNA coordinatesHindlIl and 308 to 332 (DNA sequence coordinates 722 to 746), iskb fragment followed by a series of four uridylate residues. Transcriptionthe blunted terminating in this region would yield a transcript ca. 333MLS resis- nucleotides long. As a result of induction, this region would

also undergo conformational isomerization and inactivationsequence of of its possible function as a transcription terminator. Thecoordinates remainder of segments 7 and 8, at mRNA coordinates 346 toaences were 362 (DNA sequence coordinates 762 to 779) (Fig. 3a), wouldcontains an appear to make resultant transcripts translationally inactive;oordinates 1 however, as discussed below, this may not be the case.iD sequence Determination of ermK transcription initiation by analysis ofwell as an in vivo transcripts with Si nuclease. The ermK transcription

dIII (coordi- start site was mapped experimentally by the hybird protec-iption termi- tion method of Berk and Sharp (3), and the results (Fig. 4)

indicated A at 415 or T at 416 as a site for ermK transcription;id sequence initiation. This assignment of + 1 places the start site ofribed and, in ermK transcription initiation 3 nucleotides upstream of G at(previously 418, which corresponds to the G residue at which transcrip-

i begins at a tion initiation was reported for ermD by Gryczan et al. (10).2, indicated Initiation of transcription at any of these three sites isng frame is consistent with the transcriptional attenuation model that wecapable of propose below.

,RFPTLNQ, Formation of truncated ermK transcripts in vitro. To deter-ame peptide mine whether inverted complementary repeat sequences in-ading frame the ermK message can function as rho factor-independentiption termi- terminators, we synthesized run-off transcripts with B. sub-Lt coordinate tilis RNA polymerase and the 457-nucleotide PvuII-MspIe 1693. The fragment (Fig. 1 and 2, DNA coordinates 380 to 837) as a

be MSLST template. Transcription termination at the proposed rho

)TMTSSGE factor-independent terminators would be expected to occur

h in the data around U at DNA coordinate 622, yielding a fragment of ca.

210 (i.e., in the range of 208 to 212) nucleotides as well as a

nd b, can be fragment of ca. 333 nucleotides further downstream, at DNA

AEcoRI PuuH

B

I MspIPvull (CfrIOI) Pvuf

VOL. 173, 1991

nlU

400- L

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

4728 KWAK ET AL. J. BACTERIOL.

100PvuII .CAGCTGTTTTTTCGGTTTCTATCATCATTGCTCTmMAGTATTCTCTCTGGTCTTACCTCTTCTTATAAGTGGGGTACACTCGGGAGCACCATCACATTGCTGCTAGTCGTATTTTTAA

200

TAGCGGGATTTATTATACAaAAGGCCGCTTTGCTGGTTAMAAGAACTCGAGATTAATGCACCTTGCTTCTTGATCTGATTATAATTCCGATTGATTAGTCTTCAAAAAGCCGCCAATCAA(CfrlOI) 300 ermD.MspI .. .. . . *

AACATGGCGGCTTTTTTGGGTCTGAAAACGACATGTCATCGCCGGCCGCTGATGAAGACG MCGTTGACTCCCGMCTTCACACATGCCTTGCCTGATGGCCTCCTGCGATATTCGTAPvuII 400 I+1

AGGGACGAGAAATTCCAGC1T'TTGACGTMCCAAAGAATGCCCTACAATGAGATCGTAACTTTAACTTTTCAGGAGGATTATTAAAAA ATG ACA CAC TCA ATG AGA CTG-35 -lo- SD-1 M T H S M R L

I ,2 LEADER PEPTIDE500~~~~. ..@*.

CGT TTC CCA ACT TTG AAC CAG TAA TTAAATACGTTCAAAGGCTCTGTTTGTGTATGCAGAGTAAACGGGATCAGTCTGTCCTTTTTTAATAATCTTCAATTTGCATATAAAAR F P T L N Q END

600 3,4 DdeI 5,6 700

* *** *78 AT AA AA AA* A * 800TTTGTAACTTGTAGAAAGGCGGATCGTTCTGCCTTTCTTTTTTGTTTAATCCTTTTTiTTCTAACGGCTGAGAAAGGCTTATTTGGTTATGGAAMTACCGCGGATTCATCCTTAAAGGATG.̂ ~~~~~~~~~~~~~~~~~800

GCTCTTCCCTTTACTCTGAATCACAGGCAGACCGCCTGTGATTTIIIATGATGAGAGGAAGAGGAAAC ATG AAG AAA AAA AAT CAT AAG TAC AGA GGA AAA AAG TTAM K K K N H K Y R G K K L(1) METHYLASE (10)

(BspMII) 900MspI .

AAC CGC GGG GAA TAT CCG AAT TTT TCC GGA CAG CAT TTG ATG CAT AAT AAA AAA TTA ATU GAA GAA ATT GTG GAT CGG GCA AAT ATT AGCN R G E Y P N F T Q H L M H N K K L I E E I V D R A N I S

(20) (30) (40)

DdeI MspIATA GAC GAT ACG GTT TTA GAG TTA GGA GCG GGA MA GGT GCT TTG ACA ACT ATG CTA AGT CAA AAA GCC GGT AAG GTA TUG GCA GTG GAAI D D TV L E L G A G K G A L T T MTTT Q K K V L AY E

(50) (60) (70)1000

AAC GAT TCT AAA TTC GTT GCT ATA CTC ACA CGT AAA ACA GCA CAG CAT CCA MT ACG AAA ATT ATT CAT CM GAT ATC ATG MG ATT CATN D S K F V A I L T R K TA Q H P N T K I I H Q D I M K I H

(80) (90) (100)1100

TTA CCA AM GM MG TTT GTG GTG GTC TCT AAT ATT CCC TAT GCC ATC ACA ACT CCC ATC ATG MA ATG CTC TTG AAC AAT CCT GCA AGCL P K E K F V V V S N I P Y A I T T P I M K M L L N N P A S

(110) (120) (130)1200

GGA TM CM AM GGG ATC ATC GTA ATG GAA AAA GGG GCT GCT AAA CGT TUC ACA TCA AM TTC ATT AM AAT TCC TAT GUT TTA GCT TGGG F K G I I V M E K G A A K R F T S K F I K N S Y V L A W

(140) (150) (160)1300

AGA ATG TGG UTT GAT ATT GGC ATT GTC AGA GM ATA TCG AAA GAG CAT TUT TCT CCC CCT CCA AM GTG GAC TCG GCA ATG GTC AGA ATAR M W F D I G I V R E I S K E H F S P P -P K V D S A M V R I

(170) (180) (190)1400

ACA CGA AM AM GAC GCG CCT CTA TCA CAT AM CAT TAC ATT GCG TTT CTT GGG CTT GCC GM TAT GCG CTA AAG GAG CCG CAA GCC CCTT R K K D A P L S H K H Y I A F L G L A E Y A L K E P Q A P

(200) (210) (220)1500

TTC TGT GTT GCT TTA CGC GGA ATT TTT ACT CCG CGT CAA ATG AAA CAC TUA AGA AM AGT CTA AAA ATT MC AAT GAA AAA ACC GTT GGAF C V A L R G I F T P R Q M K H L R K S L K I N N E K T V G

(230) (240) (250)1600

ACG CTC ACC GM MC CAA TGG GCG GTT ATT TTT MC ACG ATG ACT CM TAT GTG ATG CAC CAC AAA TGG CCA AGA GCA MT MG CGA AAAT L T E N Q W- A V I F N T M T Q Y V M H H K W P R 'A N K R K

(260) (270) (280)

MspI . . * HindIII 1700CCC GGA GM ATA TM AGMAAAAGCTGCTGACGTCTCGTCAGCAGCTTAAGCTTTTTCTGGAGGGATTCAGATGTCCCTGTCMCATTCCTTTTGCCCGCTTACAGCTUWr a E I END

(287)1800 HindIII

rCCAGTTMAAACTGCGCCTCGCCGAGGCCGAAACTCCAATCGGCATCTCCATTTTCTGTGATGGACACCATGACATCTTCAGGTGAMTGCCGCATTCCGCTTCAAGCTT 1841

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

TRANSCRIPTIONAL ATTENUATION CONTROL OF ermK 4729

coordinate 746. The results, (Fig. 5, lane 2) showed that themajor transcription products synthesized in vitro were ca.210 and 333 nucleotides long. In addition, two weakerfragments ca. 135 and 155 nucleotides long were found.Their role as possible transcription pause sites is discussedbelow. When we used E. coli RNA polymerase, only the210-nucleotide fragment was detected (data not shown).

It is possible to test the dependence of hair pin looptranscription termination function on RNA secondary struc-ture in vitro by substitution of rITP for rGTP in the reactionmixture (20). This substitution serves as a form of quasi-induction, since IMP units are incorporated into RNA withthe specificity of G but only two hydrogen bonds are formedby pairing between guanosine and cytidine. Use of the samePvuII-MspI fragment as a template, under these conditions,would be expected to result in a shift in transcript lengthfrom the truncated 210- and 333-nucleotide terminationfragments to the full-length run-off transcript with a pre-dicted length of ca. 420 nucleotides. The results (Fig. 5, lane3) indicated that this was indeed the case; the use of rITPresulted in a shift to synthesis of a maximum-length productof the expected size.Northern blot analysis of ermK transcripts synthesized in

vivo. To test whether transcriptional attenuation controlplays a role in ermK expression in vivo, we analyzed totalRNA from B. subtilis UOTO-277(pEC101) by the Northernblot method (6). RNA extracted from intact cells was frac-tionated by agarose-formaldehyde gel electrophoresis, trans-ferred to nitrocellulose, and probed successively with 5'-end-labeled DNA oligonucleotides 10178, 2462, and 2168.At least three discrete DNA transcripts appeared to orig-

inate in the ermK region. The longest transcript, ca. 1,200nucleotides long, represents the full-length message frominduced cells. Because oligonucleotide 2168 is located far-thest downstream and complements part of the methylaseopen reading frame, it only hybridized to the full-lengthmessage and not to any of the proposed truncated fragmentsthat terminate in the leader (Fig. 6, lanes 5 and 6). Oligonu-cleotide 2462 hybridized to both the predicted 333-nucleotidetruncated fragment and the full-length message (lane 4),whereas oligonucleotide 10178 hybridized to both the pre-dicted 210- and 333-nucleotide truncated fragments and thefull-length message (lanes 1 and 2).

Appreciable synthesis of the 210-nucleotide fragment wasseen in the absence of induction (Fig. 6, lane 1), whereas thatof the two longer fragments (333-nucleotide fragment andfull-length message) was not. These observations support amodel according to which ermK transcription is initiatedconstitutively and, in the absence of induction, terminatedefficiently to yield the prominent 210-nucleotide fragment.Readthrough transcription that would form the 333-nucleo-tide fragment was not detected in the absence of induction.

Reverse transcriptase mapping of ermK transcripts synthe-sized in vivo. If transcriptional attenuation plays a role in theregulation of ermK, we would expect to see erythromycindependence for the synthesis of ermK mRNA beyond theproposed transcription stop at +210. Oligonucleotide 2462,

positioned downstream of the proposed rho factor-indepen-dent transcription terminator, was 5' end labeled and used asa reverse transcription primer. The results (Fig. 7) revealeda clear dependence of the full-length run-off transcript onerythromycin added for induction. In contrast, the similaruse of oligonucleotide 4400 as a primer positioned upstreamof the proposed transcription terminator revealed no depen-dence on erythromycin. These observations place the tran-scription termination site between oligonucleotide 4400 (up-stream limit) and oligonucleotide 2462 (downstream limit),mRNA coordinates 154 and 250, respectively, consistentwith the other determinations described above. Moreover,these observations suggest that during induction, erythromy-cin does not significantly affect the initiation of ermK mRNAsynthesis but does affect the completion (and possibly thestability) of ermK mRNA synthesis.ermK leader mutants. Mutations in the ermK leader region

that result in higher-level expression can be obtained bygrowth of cells carrying an inducible determinant in mediumcontaining a noninducing MLS antibiotic, in this case, ty-losin or carbomycin. Five mutants that were obtained ran-domly were found to cluster in part of segments 1 and 2,overlapping the part of the sequence that codes for theproposed leader peptide and pointing to the importance ofthe associated conformation in this region in maintaining therepressed state (Fig. 8). This finding was underscored by theexpectation that the mutations in pEC301 and pEC597 woulddestabilize the proposed secondary structure by altering thecomplementary members of a proposed GC base pair.

In summary, the analysis of in vivo and in vitro transcriptsof ermK by several independent analytical techniques sug-gests that this gene is regulated by a transcriptional attenu-ator that responds to erythromycin as an inducer. In theabsence of an inducer, transcription of ermK is abortedwithin the leader region at either of two rho factor-indepen-dent termination sites 210 and 333 nucleotides downstreamfrom the transcription start site of the message.

DISCUSSION

Regulation of ermK and ermD. The sequences for ermDreported previously (10) and ermK reported above differ inonly minor respects; a leader sequence of 357 nucleotideswas found to precede the ermK methylase structural gene,whereas a 354-nucleotide leader sequence was reported forermD. The differences between the two sequences thatcomprise the ermK and ermD mRNA leaders, three innumber, are located at DNA sequence coordinates 471 (Gand T), 591 (C and T), and 686 (A and C) for ermK and ermD,respectively. In a study of the mechanism of regulation ofermD, Gryczan et al. (10) noted that the expression ofresistance remained inducible despite the removal of theresident promoter and transcriptional fusion of the residualermD sequence to a presumed upstream constitutive pro-moter. From this observation, they favored a model accord-ing to which ermD was regulated posttranscriptionally,probably by a translational attenuation mechanism similar to

FIG. 2. ermK DNA sequence and deduced amino acid sequence. The nucleotide sequence of ermK is shown, together with the deducedamino acid sequences of the ErmK methylase and its leader peptide. The proposed promoter sequences, -35 and -10, and the transcriptioninitiation site are also indicated. Sites labeled with closed circles are used to indicate four centers of symmetry in the proposed transcriptionalattenuator (see Fig. 3a and b for explicit details of the proposed base pairing). Numbers (1 to 1841) above the nucleotides and associated withdots enumerate the nucleotide sequence. Numbers (1 to 287) in parentheses below the nucleotides enumerate the amino acid residues deducedfrom the nucleotide sequence. Recognition sequences for several restriction endonucleases that play a role in the analysis of ermK are alsolabeled.

VOL. 173, 1991

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

4730 KWAK ET AL.

(a) * M'~AUUMU AG UA A

Q C CCG

N AU2N AU

GCUA

L 'UA +100UA /

a)_ CG l) UUGTco

AA GCJCUGU U Terminator 2- C GAGACG G

P O CC U UAU7- UAk 7,8F o UA

c *UA AGC( G C

R CG A CL GACUG GG CG

R A A- +127 GCM GUAAC CU GC

S * UA& AUCG CGAU AU

H C C CGAU ' UAPCG AU +333

T *AU AU /GC Terminator 1 5,6 CUCUG UUUUAU

M AU AUUUC AGAGUAAU. 3.4 c u 1 CG

Leader AU G U +300' CG rAU CG G A UA0SDU2

Peptide AU u u AU" UAAU AU AU CG.UA GC *AU UA

* UA *GU UGAG GG *CGAU CGI-+200 C U GGAUGG GAUA GC GGCA AU A AUA GC AU AAUUCC CAAU* AU *UG ' UA MGC AU CG* AU

SD-I GU AU UA CG+1 AU GC +210 UA UA K

I I GC AU I UA UA K K N I/pppATCGUAACUUUAACUUUUCAG AAUUUGCAUAUAAAAYUUGUAACUUQU UUUUUGUUUAAUCCUUUU U A C C G C G G A GAAMMAAU //** ~~~~~~~~~~~~~~0.

oligo 4400 oligo 2462 MethylaseFIG. 3. Proposed secondary structures and mechanism of induction of ermK. (a) Inactive conformation. The ermK leader region beginning

with, + 1 of the message is drawn showing four pairs of associated segments numbered 1 to 8. A part of the leader region comprising segments1 and 2 is capable of encoding a 14-amino-acid leader peptide. Ribosome binding sites, SD-1 for the leader peptide and SD-2 for the methylase,are also shown. According to the model proposed for the transcriptional control of ermK, ribosomes stalled in segment 1 of the leader peptideregion in response to erythromycin enhance the probability that the nascent ermK message will assume the conformation shown below inpanel b, with the resultant completion of transcription. It is proposed that the two sets of paired segments, 3 associated with 4 and 7 associatedwith 8, both function as transcription terminators. The locations of three DNA oligomers, labeled 10178, 4400, and 2462, used as reversetranscription primers and/or probes in various experiments are also indicated. (b) Active conformation. The active conformation of the ermKmessage beginning at coordinate +127 is shown. In this conformation, ribosomes stall during synthesis of the leader peptide encoded bysegment 1, allowing for sequential association of segment 2 with 3, 4 with 7, and 8 together with part of the methylase open reading frame,as they are synthesized, resulting in a reduced probability of transcription termination at the sites indicated in panel a. Note that segments5 and 6 remain associated as in panel a. The ribosome binding site and initiator Met codon for ErmK methylase are contained within aproposed 30-nucleotide loop whose function, together with its associated stems, remains unknown. With the initiation of methylase synthesis,the progress of the ribosome would probably be partially retarded, as in the case of other proteins whose messages have secondary structures.

that proposed for ermC from Staphylococcus aureus. The translational attenuation control models have been proposedfindings reported here suggest that ermK (and therefore for the mode of regulation used; these include ermA (27),ermD) is regulated by a transcriptional attenuation mecha- ermAM (12), ermC (11, 13), ermG (26), ermSF (16), andnism instead. A promoter substitution that leaves regulation ermSV (17). It was therefore surprising to find an example ofintact would also be consistent with transcriptional attenua- an erm gene whose regulation could be explained by tran-tion, translational attenuation, and translational repression. scriptional attenuation.Several inducible erm genes have been sequenced, and Preference for translational over transcriptional attenuation

J. BACTERIOL.

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

TRANSCRIPTIONAL ATTENUATION CONTROL OF ermK 4731

(b) 5,6c u

G U S -AUAU'AU AA AGA

*AU G AUGAGUGG G A MethylaseC U A CGGCA AU G A M

AU A U*UG G GCG U A KUA OA AUA G G

----- UA ----- U A KAAA uU UAC UA A

A UA UU CGC UA KU U CG. UAAU -CG UACG UA UA N

'GU AU UAUVA AU / UCA //UA UUU CAU *A U

A C G C \ GAAA U UUU UUC GU

UCUUC U UA UA(Go UA 8 GCAAUAAU U UA UA

UA CG / AGGAAAAAUG UG C AUA 3 4 UA 7 \ CCAAAUUGoVA ~ UU CGUA CG ......lGC

+127 CG CG' CGCG _ rc9GC CG

AUCAGUCUGU CGGAUCGUUCU UCUUCCCUUUACUCUGAAUCACAGGCAGA GGAAUAUCCGAAUUUUUCCGGACA.a,MneL

FIG. 3-Continued.

control of antibiotic resistance genes. It could be argued thattranslational and transcriptional attenuators might be suitedto regulating different classes of genes, e.g., that ermCtranscripts synthesized in the absence of induction andtherefore translationally inactive nevertheless remain capa-ble of becoming activated if needed at a subsequent time. Incontrast, a messenger whose expression is regulated bytranscriptional attenuation (as in the case of amino acidbiosynthetic operons; for a review, see reference 18 andsynthesized in the absence of induction would be truncatedand therefore would have no possibility of becoming acti-vated. In such a model, the preferential use of certain classesof regulatory elements to control expression of antibioticresistance genes could be rationalized in terms of differencesin the relative dangers posed by a toxic antibiotic on onehand and slow starvation owing to the absence of an essen-tial amino acid on the other hand.

Alternatively, a single stalled ribosome can lead to thetruncation of a nascent message, whereas continuous occu-pancy of the stall site in a translational attenuator would berequired for maintaining the ermC message in its inducedconformation. The results obtained in our study of ermKsuggest that such differences may have little bearing on themechanism of regulation and that both types of attenuationmechanism can be used to regulate the expression of ermgenes.

It will be of interest to see whether a purely translationalattenuation mechanism can be utilized for the regulation ofamino acid biosynthesis. Kadam (15) has shown that ermC

can be induced by pseudomonic acid, an antibiotic thatinhibits the catalytic function of isoleucine tRNA syn-thetase. The action of pseudomonic acid results in Ile tRNAdeficiency and would thus be expected to produce a corre-spondingly reduced rate of incorporation of Ile at criticalposition Ile-9 in the ermC leader peptide. It will be of interestto see whether a purely translational attenuation mechanismcan be utilized for the regulation of processes other thanantibiotic resistance.

Secondary structure of the ermK leader region. The second-ary structure that we proposed for ermK shows features notpreviously seen in other attenuators that regulate erm geneexpression. The proposed uninduced, translationally inac-tive form of the ermK attenuator shown in Fig. 3a should notbe present at any appreciable level, since transcriptionwould be terminated at +210 or, failing that, at +333, andthe stem segments that sequester SD-2 and the initial twocodons (Met and Lys) of ermK are downstream of this site,beginning at + 343. What relevance does the secondarystructure around the methylase ribosome binding site there-fore bear to regulation? The translationally inactive confor-mation of the ermK message shown in Fig. 3a would onlyresult, because of transcriptional readthrough, in the ab-sence of induction, past the two rho factor-independentterminators that produce the 210- and 333-nucleotide trun-cated fragments. Our analysis of the ermK leader sequencetherefore points to two transcriptional barriers to the expres-sion of ermK in the absence of induction. The selectiveadvantage to organisms that possess this mode of regulation

VOL. 173, 1991

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

4732 KWAK ET AL.

ACGG

GA

T

TGTGT

TATCA

1 2 3

5,TAA

GCA

TTG_

AA |*A_T

FIG. 4. S1 nuclease mapping of the transcription start site of theermK message. The 110-bp DNA fragment obtained by the poly-merase chain reaction was denatured, annealed to RNA from B.licheniformis, and treated with Si nuclease. A sample of the labeled110-mer probe was also treated by the chemical sequencing methodof Maxam and Gilbert (21). Samples were loaded onto a 20% gel andfractionated in parallel. Lanes: 1, A+G cleavage; 2, T+C cleavage;3, S1 nuclease-treated ermK message-DNA probe heteroduplex.

of erm remains unclear, but the apparent presence of thesebarriers to the expression of ermK makes it appear that B.licheniformis is "determined" not to express this gene in theabsence of induction.

Also noteworthy is the conformation that the proposedermK leader may assume upon induction by erythromycin(Fig. 3b). The ribosome binding site and initiator Met codonof the ErmK methylase both appear to be contained within a29-nucleotide loop structure. The lack of a secondary struc-ture and the large size of the 29-nucleotide loop may allowrelatively efficient translation initiation which, once started,would permit ribosomes that are active in translation todisrupt the remaining secondary structure shown.The impediment to translation that results from six se-

quential AU base pairs and, further on, by four sequentialGC base pairs may only present a relative barrier to trans-lation already in progress. Pavlakis et al. (28), for example,have demonstrated experimentally that the respective mes-sages of both the mouse alpha- and beta-globin chainspossess extensive secondary structure in their coding re-gions. Despite this secondary structure, the alpha- andbeta-globin chains are, in fact, synthesized. Alternatively, a

ribosome may be able to bind to the nascent ermK messageand form an initiation complex faster than the time neededfor the synthesis of the complementary leader segment.The apparent constitutive transcription of ermK when

tested with oligonucleotide 4400 and inducible transcriptionof ermK when tested with oligonucleotide 2462 could beexplained by a simple mechanism according to which ermKtranscription is initiated constitutively but completed induc-tively; however, other factors that may contribute to theinduction of erm genes include stabilization of the message

(2) and the direct action of methylase as a translationalrepressor (4). ermA, sequenced by Murphy (27), has recentlybeen studied (29, 30) with respect to the contribution ofstabilization to induction; these studies indicated thatmRNA stabilization did not contribute to increased expres-

sion.Estimating the transcriptional traffic through the ermK

FIG. 5. Demonstration of the truncated attenuation fragmentfrom transcriptional attenuation in vitro by run-off transcription.The PvuII-MspI (BspMII) DNA fragment of ermK (coordinates 380to 837; Fig. 2) was used as a template for the synthesis of RNA invitro with B. subtilis RNA polymerase. Transcription products were

analyzed by polyacrylamide gel electrophoresis (6% gel) and auto-radiography. Lanes: 1, standards (32P-5'-end-labeled MspI digest ofplasmid pBR322 DNA); 2, transcription with rATP, [at-32P]rGTP,rGTP, rCTP, and rUTP; 3, transcription with rATP, [a-32P]rGTP,rITP, rCTP and rUTP. The numbers at the sides of the gel indicatethe fragment lengths (in nucleotides).

leader region. The Northern blot analysis shown in Fig. 6allowed us to estimate the transcriptional traffic through theermK leader region. The truncated 210-nucleotide transcriptappears to be formed with high efficiency irrespective ofinduction, i.e., the major component in the faster-movingband consists of the 210-nucleotide transcript. It is at thislevel (rather than at the level of the 333-nucleotide tran-script) that most of the transcriptional attenuation occurs.Despite induction, most of the RNA that is synthesized (Fig.6, lane 2) appears to correspond in mobility to the truncatedbands. The faster-moving band seen in Fig. 6, lane 2,probably contains both 210- and 333-nucleotide truncatedtranscripts, predominantly the former. This is inferred by theuse of oligonucleotide 2462 (Fig. 6, lanes 3 and 4), which candetect only the full-length and 333-nucleotide truncatedtranscripts. The results suggest that equimolar amounts of333-nucleotide and full-length transcripts are formed duringinduction. Since the combined 210-nucleotide and 333-nucle-otide transcript seen in lane 2 is much stronger than thefull-length transcript, we must infer that most of the materialpresent in the faster-moving band consists of the 210-nucleotide truncated transcript. This interpretation could beaffected by mechanisms ofmRNA degradation that differen-tially degrade the three different forms of the ermK message;however, such differential degradation might be expected toact on the truncated fragments preferentially.

1 2 3

622-_o527-*

404 --_

309-_

4- 424

.4- 333

4- 210

-4- 155

242_238*217-_-201-4-190-*180-0

160-0

147-*

-4- 135

123-_-

110-40

J. BACTERIOL.

I

a

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

TRANSCRIPTIONAL ATTENUATION CONTROL OF ermK 4733

1 2 3 4 5 6 AT G C A1 2 34 56 78 9

BT G CA1 2 34 56 78 9

-23S- 16S

- 5S

FIG. 6. Analysis of ermK transcripts by the Northern blotmethod. Total RNA from B. subtilis UOTO-277(pEC101) was frac-tionated by agarose-formaldehyde gel electrophoresis and blottedonto nitrocellulose for use as a target. The same nitrocellulosemembrane was probed successively with end-labeled DNA oligonu-cleotides 10178, 2462, and 2168, with intervening washes to removehybridized probe. Conditions were as follows: lanes 1 and 2,oligonucleotide 10178 as the probe and target RNA from cells grownwithout and with erythromycin, respectively; 3 and 4, oligonucleo-tide 2462 as the probe and target RNA from cells grown without andwith erythromycin, respectively; 5 and 6, oligonucleotide 2168 as

the probe and target RNA from cells grown without and witherythromycin, respectively. (See Fig. 3a for the relationship ofprobes 10178 and 2462 to the attenuator sequence; see Fig. 2 for thelocation of oligonucleotide 2168, which overlaps the codons foramino acid residues 22 to 30, SGQHLMHN, of the ermK sequence.)The mobilities of rRNA markers are indicated to the right; nucleo-tides indicated to the left were assigned interpretatively on the basisof experiments described in Fig. 4 and 5.

The data shown in Fig. 7 provide independent support forthe presence of a critical regulatory sequence betweenoligonucleotides 2462 and 4400, namely, the transcriptionterminator that generates the 210-nucleotide truncated frag-ment of the ermK message. Moreover, the analysis shows astrong erythromycin dependence of transcription detectedby oligonucleotide 2462 but not by oligonucleotide 4400.

Transcriptional pausing in the ermK leader region. Inaddition to the truncated 210- and 333-nucleotide (major)fragments synthesized in vitro, two other fragments of ca.135 and 155 nucleotides are clearly seen (Fig. 5) that possiblyrepresent transcriptional pause sites in the ermK message. Inrelation to attenuators, such pause sites would serve toensure close synchrony between transcription and transla-tion (9). As a consequence, inverted complementary repeatsequences upstream of the pause sites would associate in atemporally ordered fashion, reflecting the time order inwhich they were synthesized, before additional pairing op-tions were offered by newly synthesized downstream se-quences. In terms of ermK, these considerations wouldensure the preferential association of attenuator segments 1and 2 over the association of attenuator segments 2 and 3,thus minimizing the level of background expression.The 210- and 333-nucleotide fragments shown in Fig. 6

could reflect mRNA processing rather than regulation ofmRNA synthesis. The in vitro transcription studies help toexclude the possibility ofRNA degradation or processing. Itcould be argued that the RNA polymerase preparation thatwas used contained trace amounts of nuclease that wasresponsible for the degradation. Such nuclease activitywould have to possess the additional attribute that it couldbe inhibited by the rITP preparation.

Analysis of in vivo ermK transcription by reverse tran-scriptase mapping yielded different findings, depending on

the primer that was used. Transcription up to oligonucleo-

FIG. 7. Reverse transcription analysis of RNA synthesized invivo. Total RNA from B. subtilis UOTO-277(pEC101) was used as a

template for the synthesis of cDNA. (A) Lanes 1 to 4, sequencingreactions terminated with ddATP, ddCTP, ddGTP, and ddTTP,respectively, but with the complementary nucleotide'shown; lanes 5to 9, reverse transcription of the mRNA template from cells treated,respectively, with erythromycin at 0, 0.01, 0.10, 1.0, and 10.0 p.g/ml.Synthesis was primed with 5'-end-labeled oligonucleotide 4400. (B)Same as panel A, except that synthesis was primed with 5'-end-labeled oligonucleotide 2462. (See Fig. 3a for the relationship ofprimers to the attenuator sequence.)

tide 2462, the primer located downstream of the proposedtranscription termination site, showed a dependence on theconcentration of erythromycin added to the culture mediumin which the cells were grown (Fig. 7). Such a concentrationdependence could be explained by alternative mechanismssuch as transcriptional repression and/or mRNA stabiliza-tion; however, recent studies of ermA (27) indicated thatmRNA stabilization did not contribute to increased expres-sion (29). The apparent lack of erythromycin dependence oftranscription up to oligonucleotide 4400 could be explainedby a simple mechanism according to which ermK transcrip-tion is initiated constitutively but completed inductively andthe resultant transcripts are not stabilized in the presence oferythromycin.

Mutations in the ermK attenuator region. Despite theapparent complexity of associated segments 1 and 2, the

1260-333-210-

VOL. 173, 1991

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

4734 KWAK ET AL.

AAUUAA* U A

G UA A

Q C CpEC501 CG-* pEC68A->C AU G->A

N AUpEC712) GC pEC597G->A UA C->ApEC301 L 1 VUA 2G->T UACG < +100

T AA GCUCUGUFIG. 8. Summary of sequence alterations in constitutively resis-

tant ermK mutants. Constitutively resistant ermK mutants wereobtained from B. subtilis UOTO-277(pEC101) and sequenced asdescribed in Materials and Methods.

proposed conformation is supported by the pattern of amutation to constitutive expression. The involvement ofcomplementary members of a proposed base pair in amutation to constitutive expression is consistent with theircomparable functional roles in maintaining the inactive con-formation of the ermK transcript. In addition, the penulti-mate attenuator segment has been found to provide a targetfor mutation to constitutive expression, e.g., segment 3 (of 4in the ermC attenuator (14) and segment 9 (of 10) in theermSF attenuator (16). In the conformations proposed forthe ermK attenuator, nucleotides 270 to 300 in segment 7 arethose that complex with SD-2 and the first two codons of themethylase (Fig. 3a). Moreover, despite analysis of fiveconstitutive mutants, none that directly destabilized seg-ments 7 and 8 was found (Fig. 3a).The proposed inactive form of the ermK message (Fig. 3a)

also appears to sequester both the methylase ribosomebinding site and the Met Lys codon sequence at the aminoterminus of the methylase. What useful function would thishighly associated segment of ermK mRNA serve if synthesisof the message under noninducing conditions is terminatedupstream at the proposed transcription terminator? Lee andYanofsky (19) reported that readthrough transcription be-yond the trp transcriptional attenuators of E. coli andSalmonella typhimurium occurred with efficiencies of 5 and30%, respectively. Transcriptional readthrough of the ermKattenuator would produce mRNA with the leader conforma-tion shown in Fig. 3a. The terminating inverted repeats 3associated with 4 and 7 associated with 8 would form underthese conditions without termination of the message, but theresultant transcript would be translationally inactive. To theextent that a stalled ribosome could disrupt segments 1 and2, an ermK mnessage inactive in this way might becometranslationally activated under conditions of induction. Inany case, the regulation of ermK by attenuation appears tobe primarily at the transcriptional rather than the transla-tional level.

ACKNOWLEDGMENTS

Oligonucleotides and peptides were synthesized at the Universityof Wisconsin Protein Sequence-DNA Synthesis Facility supportedby funds from the Public Health Service, National Institutes ofHealth (shared equipment grant S10-RR01684, National CancerInstitute continuing support grant CA-07175, and a general researchsupport grant to the University of Wisconsin Medical School), fromthe National Science Foundation Biological Instrumentation Pro-gram (Division of Molecular Biosciences grant DMB-8514305), andfrom the University of Wisconsin Graduate School. This work wassupported by research grant AI-18283 from the Public Health

Service, National Institutes of Health. E.-C.C. was also supportedby the Korea Science and Engineering Foundation, by the SeoulNational University Daewoo Research Fund, and by the GeneticsEngineering Program of the Ministry of Education, Republic ofKorea.We thank Dick Burgess and Dale Hager for RNA polymerase

from B. subtilis, Terry Stewart and Tony Chang for preparation ofthe figures, and Barb Don (Genetics Computer Group) for assistancewith the DNA sequence analysis.

REFERENCES1. Band, L., and D. J. Henner. 1984. Bacillus subtilis requires a

stringent Shine-Dalgarno sequence for gene expression. DNA3:17-21.

2. Bechhofer, D. H., and D. Dubnau. 1987. Induced mRNA stabil-ity in Bacillus subtilis. Proc. Natl. Acad. Sci. USA 84:498-502.

3. Berk, A. J., and P. A. Sharp. 1977. Sizing and mapping of earlyadenovirus mRNAs by gel eletrophoresis of S1 endonuclease-digested hybrids. Cell 12:721-732.

4. Breidt, F., and D. Dubnau. 1990. Identification of cis-actingsequences required for translational autoregulation of the ermCmethylase. J. Bacteriol. 172:3661-3668.

5. Choi, E. C., B. K. Kim, M. J. Shim, K. S. Chung, J. W. Woo,H. R. Kim, and C. K. Lee. 1982. Studies on resistance toantibiotics in bacteria: induced resistance to antibiotics inBacillus sp. Yakhak Hoeji 26:169-175. (In Korean.)

6. Davis, L. G., M. D. Dibner, and J. F. Battey. 1986. Basicmethods in molecular biology, p. 75-78. Elsevier, New York.

7. Docherty, A., G. Grandi, T. J. Gryczan, A. G. Shivakumar, andD. Dubnau. 1981. Naturally occurring macrolide-lincosamide-streptogramin B resistance in Bacillus licheniformis. J. Bacte-riol. 145:129-137.

8. Dubnau, D. 1984. Translational attenuation: the regulation ofbacterial resistance to the macrolide-lincosamide-streptograminB antibiotics. Crit. Rev. Biochem. 16:103-132.

9. Fisher, R. F., A. Das, R. Kolter, M. E. Winkler, and C.Yanofsky. 1985. Analysis of the requirements for transcriptionalpausing in the tryptophan operon. J. Mol. Biol. 182:397-409.

10. Gryczan, T., M. Israeli-Reches, M. Del Bue, and D. Dubnau.1984. DNA sequence and regulation of ermD, a macrolide-lincosamide-streptogramin B resistance element from Bacilluslicheniformis. Mol. Gen. Genet. 194:349-356.

11. Gryczan, T. J., G. Grandi, J. Hahn, R. Grandi, and D. Dubnau.1980. Conformational alteration of mRNA structure and theposttranscriptional regulation of erythromycin-induced drug-resistance. Nucleic Acids Res. 8:6081-6097.

12. Horinouchi, S., W.-H. Byeon, and B. Weisblum. 1983. A com-plex attenuator regulates inducible resistance to macrolides,lincosamides, and streptogramin type B antibiotics in Strepto-coccus sanguis. J. Bacteriol. 154:1252-1262.

13. Horinouchi, S., and B. Weisbium. 1980. Posttranscriptionalmodification of messenger RNA conformation: mechanism oferythromycin inducible resistance. Proc. Natl. Acad. Sci. USA77:7079-7083.

14. Horinouchi, S., and B. Weisblum. 1981. The control region forerythromycin resistance: free energy changes related to induc-tion and mutation to constitutive expression. Mol. Gen. Genet.182:341-348.

15. Kadam, S. K. 1989. Induction of ermC methylase in the absenceof macrolide antibiotics and by pseudomonic acid A. J. Bacte-riol. 171:4518-4520.

16. Kamimiya, S., and B. Weisbium. 1988. Translational attenuationcontrol of ermSF, an inducible resistance determinant encodingrRNA N-methyltransferase from Streptomyces fradiae. J. Bac-teriol. 170:1800-1811.

17. Kamimiya, S., and B. Weisblum. Unpublished data.18. Kolter, R., and C. Yanofsky. 1982. Attenuation in amino acid

biosynthesis operons. Annu. Rev. Genet. 16:113-134.19. Lee, F., and C. Yanofsky. 1977. Transcription termination at the

trp operon attenuators of Escherichia coli and Salmonellatyphimurium: RNA secondary structure and regulation of ter-mination. Proc. Natl. Acad. Sci. USA 74:4365-4369.

20. Levin, J. R., and M. J. Chamberlin. 1987. Mapping and charac-

J. BACTERIOL.

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

TRANSCRIPTIONAL ATTENUATION CONTROL OF ermK 4735

terization of transcriptional pause sites in the early geneticregion of bacteriophage T7. J. Mol. Biol. 196:61-84.

21. Maxam, A. M., and W. Gilbert. 1980. Sequencing end-labeledDNA with base-specific chemical cleavages. Methods Enzymol.65:499-560.

22. Mayford, M., and B. Weisblum. 1989. Conformational alter-ations in the ermC transcript in vivo during induction. EMBO J.9:4307-4313.

23. Mayford, M., and B. Weisbium. 1989. ermC leader peptide:amino acid sequence critical for induction by translationalattenuation. J. Mol. Biol. 206:69-79.

24. Messing, J. 1983. New M13 vectors for cloning. MethodsEnzymol. 101:20-78.

25. Miller, J. H. 1972. Experiments in molecular genetics, p. 17.Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y.

26. Monod, M., S. Mohan, and D. Dubnau. 1986. Cloning andsequence analysis of ermG, a new macrolide-lincosamide-strep-togramin B resistance element from Bacillus sphaericus. J.Bacteriol. 169:340-350.

27. Murphy, E. 1985. Nucleotide sequence of ermA, a macrolide-lincosamide-streptogramin B (MLS) determinant in Staphylo-coccus aureus. J. Bacteriol. 162:633-640.

28. Pavlakis, G. N., R. E. Lockard, N. Vamvakopoulos, L. Rieser,U. L. RajBhandary, and J. N. Vournakis. 1980. Secondarystructure of mouse and rabbit alpha- and beta-globin: differentialaccessibility of alpha and beta initiator AUG codons towardnucleases. Cell 19:91-102.

29. Sandler, P., and B. Weisblum. 1988. Erythromycin-inducedstabilization of ermA messenger RNA in Staphylococcus aureusand Bacillus subtilis. J. Mol. Biol. 203:905-915.

30. Sandler, P., and B. Weisbium. 1989. Erythromycin-inducedribosome stall in the ermA leader: barricade to 5'-to-3' nucleo-lytic cleavage of the ermA transcript. J. Bacteriol. 171:6680-6688.

31. Sanger, F., S. Nicklen, and A. R. Coulson. 1980. DNA sequenc-ing with chain-terminating inhibitors. Proc. Natl. Acad. Sci.USA 74:5463-5467.

32. Stern, S., D. Moazed, and H. F. Noller. 1988. Structural analysisof RNA using chemical and enzymatic probing monitored byprimer extension. Methods Enzymol. 164:481-489.

33. Weisbium, B. 1985. Inducible macrolide-lincosamide-strepto-gramin resistance: a review of the resistance phenotype, itsbiological diversity, and structural elements which regulateexpression. J. Antimicrob. Chemother. 16(Suppl. A):63-90.

VOL. 173, 1991

on January 4, 2020 by guesthttp://jb.asm

.org/D

ownloaded from