JOURNAL OF BACrERIOLOGY, Mar. 1974, p. 1231-1239Copyright 0 1974 American Society for Microbiology

Vol. 117, No. 3Printed in U.S.A.

Inhibition of lacZ Gene Translation Initiation in trp-lac FusionStrains

WILLIAM S. REZNIKOFF, CORINNE A. MICHELS,' TERRANCE G. COOPER,2 ALLEN E.SILVERSTONE, AND BORIS MAGASANIK

Department of Biochemistry, College of Agricultural and Life Sciences, University of Wisconsin-Madison,Madison, Wisconsin 53706; Department of Bacteriology and Immunology, Harvard Medical School, Boston,Massachusetts 02115; Department of Biological Sciences, Columbia University, New York, New York 10027;

and Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139

Received for publication 9 October 1973

Different levels of ,-galactosidase are found in various trp-lac fusion strains.These levels of ,B-galactosidase fall within a 60-fold range. The amount ofthiogalactoside transacetylase activity detected in these same strains only varies10-fold and is found in amounts greater than those predicted from the0-galactosidase levels. The observation that the 0-galactosidase and thiogalacto-side transacetylase levels are not directly proportional, that the lacZ messengerribonucleic acid (mRNA) levels are not proportional to the 0-galactosidaseactivity, that, at least for the one fusion strain tested, the SuA polaritysuppressor does not affect the ,B-galactosidase level, and that, in all but onestrain, the f-galactosidase activity appears to reside in normal ,B-galactosidasemolecules suggests that the disproportionately low production of f-galactosidaseis due to a decrease in the frequency of translation initiation of lacZ mRNA inthese strains. Several mechanisms are proposed to explain this decrease. Somepossible bases for the disproportional production of ,B-galactosidase and thi-ogala.ctoside transacetylase are also described. The preferred explanation forthese disproportional enzyme levels is that only a fraction of the full complementof ribosomes need initiate translation at lacZ for the functional synthesis of lacmRNA to occur and that once the lac ribonucleic acid is made a full complementof ribosomes can bind at internal translation initiation sites at Y and A.

Gene expression is dependent in part uponthe process of translation. The initiation of thisprocess is known to require an appropriateinitiation sequence (or ribosome binding site) inthe messenger ribonucleic acid (mRNA) con-taining at least an AUG codon. Experimentswith several systems suggest that the existenceof a ribosome binding site sequence per se is notsufficient to result in initiation of translation.Rather, the ability of this sequence to berecognized by ribosomes is affected by thestructure of neighboring mRNA sequences. Forinstance, the ribosome binding site for the R17replicase gene does not appear to be functionalunless exposed by translation of the coat proteingene (5, 7, 12, 31), and there exist internaltranslation initiation sites within the lacI genethat only function in the presence of a precedingnonsense codon (20). Thus, it might be possible

'Present address: Department of Biology, Queens Collegeof the City University of New York, Flushing, N.Y. 11367.

2Present address: Department of Biochemistry, Univer-sity of Pittsburgh, Pittsburgh, Pa. 15261.

to find mutations in which the efficiency oftranslation initiation for a particular gene isaltered due to an alteration in the structure ofneighboring mRNA sequences.

Recently a large collection of trp-lac fusiondeletions have been isolated, all of which havethe same property; they fuse all of the lacstructural genes to the tryptophan (trp) operon.Many of these deletions end within the lac-con-trolling element region (between I and Z) (23,24; W. Jurgen Schrenk, Ph.D. thesis,Universitat zu Koln, 1972; and D. Mitchell, W.S. Reznikoff, and J. R. Beckwith, manuscript inpreparation) and therefore result in the pres-ence of new mRNA sequences near the lacZribosome binding site. The experiments de-scribed in this paper show that, although the lacstructural genes in these deletion strains are inall cases fused to the same transcription unitacting at the same level (the genetically dere-pressed trp operon), they manifest an apparent60-fold range in levels of 0-galactosidase pro-duction. It is proposed that most of the strains

1231

REZNIKOFF ET AL.

produce different levels of f-galactosidase be-cause initiation of lacZ translation occurs atdifferent efficiencies in the various strains.(This generalization may only apply to thosestrains in which the fusion deletions end withinthe lac-controlling element region. Recent stud-ies have shown that two of the strains studied inthese experiments [FlOl and F104] containtrp-lacI deletions in combination with the lacpromoter mutation L8 [Mitchell et al., manu-script in preparation]. In these strains the lowlevel of trp-controlled lacZ expression may bedue to two phenomena: the termination of someof the trp-initiated transcription at the end ofthe I gene, and inefficient initiation of lacZtranslation on the fused mRNA that is made.)

For many operons in Escherichia coli, synthe-sis of detectable mRNA seems to be dependentupon the concomitant translation of the mRNAinto proteins (4, 11). Two general hypotheseshave been advanced to explain this phenome-non. Ribonucleic acid (RNA) polymerase mayrequire ribosomes to be actively reading themessage in order to progress (8) or the mRNAwhich is produced in the absence of translationmay be rapidly degraded by a combined en-donucleolytic and exonucleolytic attack (17,18). Whichever hypothesis is correct, it is possi-ble that only a small fraction of the wild-typeribosome complement is sufficient to permit thesuccessful propagation of mRNA. As will beshown in this paper, the trp-lac fusion strainsmanifest a second characteristic that suggeststhat this is in fact the case. 3-Galactosidase andthiogalactoside transacetylase are not producedin these strains in the same proportions as in thewild-type strain. Rather, ,-galactosidase is pro-duced in reduced amounts relative to the levelof thiogalactoside transacetylase (and lacZ and-A mRNA). The simplest explanation for thisobservation is that only a small fraction of thenormal ribosome complement must initiatetranslation at the beginning of lacZ for lacmRNA to be successfully propagated through tolacA where a normal complement of ribosomescould then initiate lacA translation.

MATERIALS AND METHODSStrains. The trp-lac fusion strains all have the

same genotype, except for the particular deletionends. All of these strains are F-, Su-. They carry thelac-pro A,B deletion XIII, a trpR allele, a strAmarker, a 080d1ac prophage, and the tonB deletionsshown in Fig. 1. The lac end of these deletions can becategorized as type 1 (those ending in lacI in which lacexpression results from transcription initiation atlacP), type 2 (those ending in lacI in which lacPactivity is reduced due to the presence of the L8promoter mutation; in these strains most of the lac

expression results from trp-initiated transcription),type 3 (those ending within the promoter), type 4(those ending between the promoter and the opera-tor), and type 5 (those deleting the operator butpresumably ending before Z). The isolation andcharacterization of these deletions has been describedpreviously (16, 23, 24; Schrenk, Ph.D. thesis; andMitchell et al., manuscript in preparation).The trpR- SuA+ and SuA- strains were con-

structed by mating an Hfr Cavalli containing thetrpR- marker with an F-SuA+, strA, A(lac-proA,B)XIII strain and selecting for 5-methyltryptophan-resistant, streptomycin-resistant progeny. The puri-fied recombinants were screened for their Lac charac-ter on Lac McConkey plates. The Lac- recombinantswere tested for their SuA character by introducing anFlac-proA,B episome carrying the early Z nonsensemutation U118. Recombinants which gave rise to amelibiose-positive phenotype at 42 C after introduc-tion of the episome were judged to be SuA+. SuA+trpR- and SuA- trpR- recombinants were pickedfrom among the Lac- recombinants. The F'W1 epi-some was provided to us by B. Beckett (University ofWisconsin). It was derived by crossing the Wl dele-tion onto an F, trp, tonB, lac episome isolated fromEC-8 (3) by D. Schwartz and J. Beckwith (HarvardMedical School).To look for the presence of the w peptide production

in trp-lac fusion strains, two F'lac, proA,B episomeswere introduced into these strains by the methoddescribed previously (24). One of these episomescarried the late lacZ nonsense mutation X90, which isknown to be an w acceptor (27, 28), and the otherepisome carried the early lac nonsense mutationU118, which served as a negative control.The other strains used in this study are: 3000, a

wild-type lac+ strain from the Luria collection, andX7026, an F- A(lac-proA,B) XIII strain.Media and chemicals. Media and chemicals are as

described elsewhere (14, 24, 26).Enzyme assays. These experiments derived from

independent observations made by C.A.M. andW.S.R. during the course of experiments described inreference 14 and by A.E.S. and B.M. made during thecourse of the experiments described in reference 26.Therefore, the exact procedures used in the assays foreach experiment depended upon the laboratory wherethe work was performed. These are specifically re-ferred to in each figure or table legend and aredescribed in detail elsewhere (15, 16, 24, 25). Theunits of enzyme activity were in all cases normalizedto the value for Wl found in Table 1 for ease ofcomparison.mRNA measurements. The lacZ and -A mRNA

determinations were performed by RNA-deox-yribonucleic acid hybridization as described by T. G.Cooper, P. A. Whitney, and B. Magasanik (manu-script submitted for publication).

Heat inactivation of f-galactosidase. The cells ofthe control strain R-X8618 and of two or moreexperimental strains were harvested by centrifugationat 4 C from late-log-phase cultures grown in LB broth.The cell pellets were resuspended in one-tenth theoriginal volume in Z buffer and were sonically dis-

1232 J. BACTERIOL.

VOL. 117, 1974 lacZ GENE TRANSLATIOIS

rupted by using the microtip attachment of thesonifier cell disruptor (Heat Systems-Ultrasonics,Inc.). These crude extracts were clarified by centrifu-gation at 10,000 rpm for 15 min in an IEC centrifugemodel HR-1 with an 856 rotor. The supernatants werethen assayed for ,B-galactosidase and were dilutedwith a similar extract of X7026 so that they had thesame approximate specific activity. A 0.5-ml volumeof these diluted extracts was then pipetted into 2 ml ofprewarmed Z buffer and heated at 60 C. At theindicated time points (Fig. 3), 0.2 ml of the heatedextract was removed and pipetted into 0.8 ml ofice-cold Z buffer. These diluted samples were thenassayed for f8-galactosidase. Since the water bathtemperature might vary between and during experi-ments, a wild-type control was always included, andsamples for corresponding time points from differentextracts were removed within 1 min of each other.

TABLE 1. f-Galactosidase and thiogalactosidetransacetylase activities in trp-lac fusion strainsa

trp ~~~Thioga-cotrpl i-Galac- lactosideStrain ontrlaex tosidase trans- (3-Gz/TAofrlaciox (0-Gz) acetylase

pressionA(TA)

R-X8618 No 5775 9.95 580.4X7713 Yes 177 3.10 57.1Wi Yes 519 10.85 47.8W200 Yes 135 5.10 26.5W205 Yes 6805 9.25 735.7W209 Yes 2655 12.65 209.9W211 Yes 3290 15.70 209.6W227 Yes 204 5.45 37.4F23a Yes 561 9.15 61.3F36a Yes 114 1.80 63.3F101 Yes 146 2.05 71.2F104 Yes 95 2.45 38.8

a All assays were performed in duplicate. The exactprotocols and definitions of unit activities are asdescribed by Michels and Zipser (15). R-X8618 is acontrol trp-lacI deletion strain in which lac operonexpression is under the control of the lac promoter.

trpI,C, c 2, C,B, A

q INITIATION INHIBITION 1233

RESULTS,B-Galactosidase and thiogalactoside

transacetylase activities in trp4lac fusionstrains. Starting with a strain in which the lacoperon is transposed to a site near the trp op-eron (Fig. 1), it has been possible to isolate tonBfusion deletions that fuse the lac structuralgenes Z, Y, and A to the trp operon so that thelac genes are under the control of the trp regula-tory elements. Most of these deletions appear toend within the lac-controlling elements (Fig. 1)(16, 23, 24; Shrenk, Ph.D. thesis; and Mitchellet al., manuscript in preparation).Two striking characteristics of these strains

are revealed by the data in Tables 1 and 2, andFig. 2. First, the various trp-lac fusion strainsproduce dramatically different levels of f-galac-tosidase. For instance, the amount of 3-galac-tosidase found in F36a is more than 60-foldlower than that found in W205. Second, the,3-galactosidase production in these strains isnot proportional to the thiogalactoside trans-acetylase, lacZ mRNA, and lacA mRNA pro-duction if compared to the ratios found in thecontrol strains R-X8618 and 3000. (The twofolddifference in the Z enzyme-A enzyme ratio for3000 as compared to R-X8618 is presumablydue to differences in the genetic background ofthe two strains, and this difference has beennoticed before when comparing 3000 to strainssimilar to R-X8618 [26].) Rather, with theexception of W205, less f-galactosidase is foundthan would be predicted. As will be discussedsubsequently, these two properties of the trp-lacfusion strains are probably closely related; how-ever, this paper will first present observationswhich suggest a possible explanation for thediffering levels of 3-galactosidase in the variousstrains and then will discuss the observed dis-proportionate production of the two enzymes.Production of fl-galactosidase. The follow-

ton B Ott 80-L

Iocz y a

type 1 (X8618)

type 2 (F1Ol, F104) L8

type 3 (F36a, W227, F23a)

type 4 (Wl, W211)

type5 (X7713, W200, W209, W205)

typel 3. (W2) I!1

FIG. 1. Deletions fusing the trp and lac operons. The isolation and characterization of the deletions picturedare described in detail elsewhere (13, 23, 24; Schrenk, Ph.D. thesis; and Mitchell et al., manuscript inpreparation). The type 2 deletions are all present in combination with the promoter point mutation L8indicated by an X.

REZNIKOFF ET AL.

ing three models could be envisaged to explainthe differing levels of ,8-galactosidase in the trp-lac fusion strains. (i) Z gene expression in somestrains might be reduced due to a polar effectgenerated by a nonsense codon possibly locatedin the residual lacP-O region. (ii) The 3-galac-tosidase activity in some strains might reside inabnormal trp-3-galactosidase hybrid proteinmolecules which are only partially active. (iii)There may be an inhibition to varying degreesin different strains of lacZ mRNA translation,probably at the level of translation initiation.The observations described in Tables 1 and 2

and in Fig. 2 seem to rule out a polar effect onlacZ expression as being the primary reason forthe reduced levels of,-galactosidase. This isbecause polar effects resulting from nonsensecodons are known to have proportional effects

on distal gene expression and on detectablemRNA synthesis (1, 4, 11, 19, 30). An experi-ment which further examines the possibilitymentioned above is described in Table 3. Theeffect of the polarity suppressor, SuA (2), on thesynthesis of d-galactosidase coded for by onetrp-lac fusion was tested by introducing an F'episome carrying the Wl trp-lac fusion intoSuA+ and SuA- strains that were otherwise iso-genic and measuring the amount of ,8-galac-tosidase produced. The level of 0-galactosidaseactivity was not affected by the presence of theSuA suppressor (Table 3).A trp-lac fusion strain might produce an

abnormal hybrid trp-f3-galactosidase protein ifthe fusion deletion began in a trp structuralgene and ended in phase within the lac-control-ling element region or within the sequence

TABLE 2. Lac enzyme and mRNA production in trp-lac fusion strainsa

trpStrain control Zenzyme!Z mRNA Aenzyme!A mRNA Z enzyme/ A enzyme/ Z enzyme/ Z RNA/

of lac ex- Z mRNA A mRNA A enzyme A RNApression

3000 No 3,147 7,860 12.15 3,340 0.40 3.69 x 10' 259.0 2.3Wi Yes 519 5,810 10.85 2,480 0.089 4.38 x 10' 47.8 2.3W2 Yes 236 3,210 3.41 2,010 0.074 1.70 x 10-' 69.2 1.6X7713 Yes 100 1,510 2.06 930 0.066 2.06 x 10' 48.5 1.6

a The assays were performed as follows: 0-galactosidase as in Silverstone et al. (25) and thiogalactosidetransacetylase as in Miller et al. (16). The enzyme activities were normalized to the values for Wl found inTable 1. The lac mRNA was isolated from cells which had been exposed to 5 mM IPTG for 20 min and thenlabeled for 4 min with ['H Juracil, and the hybridization assays were performed as described (Cooper et al., man-uscript submitted for publication), with activity shown as counts per minute of hybridized RNA.

b ,B -Galactosidase. 0

c Thiogalactoside transacetylase.

1,000 2,000 3,000/3-Galoctosidose

4,000 5,000 6p00 700C

FIG. 2. fB-Galactosidase and thiogalactoside transacetylase activities in trp-lac fusion strains. The datapresented in Table I are plotted with the dashed line indicating the expected enzyme activity values forproportional Z and A expression.

14

121

4) 10c

cnrc

*W211

W*209

.*2

* F23a R-RX68

0 *227*W 200

*X7713 _,,*F104

Q,JOF1cSY-F 36a _ -

4

2

1234 J. BACTERIOL.

lacZ GENE TRANSLATION INITIATION INHIBITION

TABLE 3. Effect of the SuA suppressor onf,-galactosidase production by the Wl trp-lac fusiona

Strain I-Galactosidase

F'Wl/SuA+ 519F'Wl/SuA- 507

An F' episome carrying the Wl trp-lac fusion wasintroduced into SuA+ and SuA- strains, and theBi-galactosidase activity produced by the merozygoteswas measured by using the techniques described byReznikoff et al. (24). The enzyme activities werenormalized to the values for Wl found in Table 1.

corresponding to the amino terminus of,B-galac-tosidase. To test whether abnormal ,B-galactosi-dase is made in any of the fusion strains, thethermostability of the 0-galactosidase producedin these strains when incubated at identicalspecific activities in Z buffer at 60 C wasexamined. Representative inactivation curvesfor the control strain R-X8618 and for thefusions, Wl and W209, are shown in Fig 3. All ofthe strains tested (Wl, W2, W200, W205, W211,W227, F23a, F36a, F101, and F104) except forone (W209) produced 3-galactosidase that wasindistinguishable from that of the control strainby this test. The 3-galactosidase of W209 fromthree independent cultures has been tested. Inall cases it was approximately 2.7 times morethermolabile than the wild-type enzyme. Theseresults are consistent with the preliminary re-sults of sodium dodecyl sulfate (SDS)-acryli-mide gel analysis of the crude extracts fromfusion strains to be reported subsequently. Ex-tracts of all of the fusion strains except for W209(and possibly W200) manifest a single band atthe position expected for 0-galactosidase in anamount approximately proportional to the 0-galactosidase activity found in the strain.Therefore, with the possible exception of W209,it can be concluded that the decreased levels of,B-galactosidase activity do not result fromstructural alterations in the enzyme itself butfrom a decreased rate of formation of theenzyme.

Disproportionate production of,B-galacto-sidase and thiogalactoside transacetylase. Apossible basis for the observed disproportionate,8-galactosidase and thiogalactoside transace-tylase activities in the fusion strains might bethat the ,8-galactosidase activity observedunderestimates the actual amount of Z genetranslation occurring in these strains. One couldassume that two types of Z gene translation areoccurring. For instance, strains which produceapproximately normal levels of thiogalactosidetransacetylase but low levels of 0-galactosidase

might contain in-phase deletions so that theyproduce both normal ,B-galactosidase and ahybrid inactive trp-,-galactosidase protein. Thetotal translation ofZ might be normal, althoughthe amount of normal, enzymatically active,B-galactosidase would be reduced. If this modelwere correct, a fused protein containing the,B-galactosidase w peptide fragment should bemade in these strains. This might be detectedby introducing an episome coding for an wacceptor into the fusion strains and looking forincreased 0-galactosidase activity (the totalactivity would correspond to normal ,B-galac-tosidase plus w donor plus w acceptor hybridmolecules). Partial diploids were constructed byintroducing either an F' which carries the X90w-accepting mutation or a control F' carryingthe U118 mutation into several of the fusionstrains. These partial diploids were assayed for,B-galactosidase activity. No increase in ,B-galac-tosidase activity was detected in the merozy-gotes carrying the F'X90 (Table 4). This result

100 '

50

c)

>10-> Wi<

)

10 20 30Minutes at 600

FIG. 3. Heat inactivation of,-galactosidase. Crudeextracts of the control strain (R-X8618) and twoexperimental strains (Wl and W209) were diluted in acrude extract of a lac- strain (X7026) to the samespecific activity. These extracts were diluted intoprewarmed Z buffer, and samples were removed atthe indicated times and assayed for ,3-galactosidase asdescribed in Reznikoff et al. (24).

1235VOL. 117, 1974

REZNIKOFF ET AL.

TABLE 4. f3-Galactosidase w peptide production bytrp-lac fusions

Strain F' lacZ- X90 F' lacZ- U118

W209 847 846Wi 376 519W227 316 288W200 255 302F104 186 160

a An F' lac episome coding for the X90 u-acceptingfragment or a control F' lacZ- U118 episome wasintroduced into various trp-lac fusion strains, and the,B-galactosidase activity produced by the merozygoteswas measured by using techniques described before(24), except that the cells were grown by using 0.4%glycerol as the carbon source. The activities werenormalized to values for Wl found in Table 1.

does not rule out the above model, since themaximum level of 3-galactosidase activity re-sulting from such complementation is only 25%of the wild-type level (or approximately 600units in this experiment) and frequently isconsiderably lower (27, 28) and since an unsta-ble w peptide might be made in these strainsand not be detected by our assay. But otherconsiderations to be presented in the Discussionsuggest that this model is not correct.

DISCUSSIONThe first unusual feature of the trp-lac fusion

strains examined in this paper is the differencein the levels of s-galactosidase made in the var-ious strains. The evidence indicates that thesediffering levels of W-galactosidase are not due topolarity generated by a nonsense codon pre-ceding lacZ. The strains did not have similarlydepressed levels of transacetylase activity; thelevels of lacZ mRNA were not lowered to thesame degree as the ,B-galactosidase activity; andthe polarity suppressor SuA did not alter thelevel of 3-galactosidase generated by at leastone fusion. The fact that the level of lacZmRNA was greatly in excess of that expectedfrom the amount of 3-galactosidase activityfound also indicates that the lowered levels of0-galactosidase do not result from defects intranscription. Another possibility, which hasnot been completely ruled out yet, is that thesestrains produce abnormal ,B-galactosidase pro-teins (probably fused to trp proteins). However,the heat-inactivation studies suggest that, withthe exception of W209, all of the strains pro-duced normal 0-galactosidase. Furthermore, arecent sequence analysis of the lac operatorregion indicates that there is an in-phase non-sense codon (UGA) within the operator (W.Gilbert, N. Maizels, and A. Maxam, Cold

Spring Harbor Symp. Quant. Biol., in press),which would suggest that no such partiallyactive fused protein could be made in strainsWi, W2, W211, W227, F23a, F36a, F101, andF104, all of which contain deletions endingbefore lacO (see Fig. 1). Additional evidencethat normal 0-galactosidase is produced inthese strains has come from preliminary SDS-acrylimide gel analyses of crude extracts ofthese fusion strains.What other phenomenon could explain the

varying amounts of f-galactosidase found inthese fusion strains? We propose that theamount of 3-galactosidase found in these strainsis a measure of the frequency of initiation oftranslation at the start of lacZ. One couldhypothesize three different means by which thisinitiation might be depressed in these strains.(i) The deletions may have removed somenucleotides that normally code for part of thelacZ ribosome binding site. Presumably this siteincludes the AUG codon at the start of lacZ andperhaps 15 or 20 nucleotides on either side. Thispossibility cannot explain the lowered ,B-galac-tosidase production in the type 2, 3, and 4strains, since these strains contain deletionsending before lacO (24; Schrenk, Ph.D. thesis;and Mitchell et al., manuscript in preparation)and therefore must end at least 35 nucleotidesaway from the lacZ AUG (6).

(ii) In several (but probably not all) of thesestrains, the fusion deletion extends from a trpstructural gene to the lac-controlling elements.It would be expected that there might beribosomes and associated translation machineryproceeding through from the trp structural geneacross that section of the lac mRNA usuallyreserved for ribosome binding. This read-through translation might therefore competewith the correct binding of new ribosomes. Sucha possibility has been postulated by Platt et al.(20) to explain the necessity for a chain termi-nation codon prior to an initiation codon inorder for initiation to occur. There are twostrains in which the fusion deletions apparentlydo not end in a trp structural gene (W205 andF36a) but rather after trpA (Mitchell et al.,manuscript in preparation). Presumably read-through translation cannot occur, and yet the,B-galactosidase level in F36a is very low. Also,this model does not take into consideration theexistence of two nonsense codons within theoperator region (Gilbert et al., in press). One ofthese, which has been mentioned before, is anin-phase UGA codon which would terminatetranslation 33 nucleotides prior to the lacZAUG. The second nonsense codon is a UAA inanother reading frame, 25 nucleotides prior to

J. BACTrERIOL.1236

lacZ GENE TRANSLATION INITIATION INHIBITION

the lacZ AUG. One would therefore expect thattrp translation would read through the lacZribosome binding site only in strains in whichthe fusion deletion extends into 0 beyond thetwo nonsense codons or in which the deletionends so that the third reading frame is utilized.The deletions which end in 0 can be detected(23; Schrenk, Ph.D. thesis; and Mitchell et al.,manuscript in preparation), but there is no wayof knowing yet which reading frames are repre-sented by the various deletions. In conclusion, itis possible that in some strains (those whosedeletions end in 0 or in the third reading frame)this is the mechanism by which lacZ translationinitiation is inhibited, although it is very un-likely in the specific case of F36a.

(iii) A general feature of all of these fusions isthat abnormal RNA sequences have been addedonto lac mRNA close to the lacZ ribosomebinding site. It is possible that these sequencesalter the configuration of the lacZ ribosomebinding site or make it structurally inaccessible,thereby reducing the affinity of ribosomes forthe site. That such a phenomenon can occur isknown from the studies on R17 RNA transla-tion. The RNA sequences close to the R17replicase gene ribosome binding site dramati-cally affect its accessibility to incoming ribo-somes (5, 7, 12, 31). This mechanism may wellexplain the depressed level of lacZ translationin some of the fusion strains. It may be possibleto test this model directly. trp-lac fusion mRNAshould bind ribosomes to the lacZ ribosomebinding site at a lower efficiency than normallac mRNA, but the sequence of the site shouldbe the same.The second striking feature of these trp-lac

fusion strains is the obvious disproportionateproduction of B-galactosidase and thiogalacto-side transacetylase. A similar observation hasbeen made in another system. Voll (29) andRechler et al. (21, 22) have described a frame-shift mutation located late in the D gene of thehis operon. This mutation reduces expression ofthe immediate next gene, hisC, 98 to 99% butreduces the expression of subsequent genes only90%. The reduction in hisC expression and thedisproportionate polar effect observed was notaltered by introducing a nonsense codon imme-diately prior to the frame-shift mutation in hisDand therefore was not due to read-throughtranslation competing for normal initiation.The protein produced by the immediate nextgene was normal in all respects.The experiments described in Table 2 indi-

cate that the disproportionate production of,B-galactosidase and thiogalactoside transacety-lase does not seem to be reflected in the

amounts of lacZ mRNA and lacA mRNA foundin these strains. Rather, these experimentssuggest that it is specifically due to decreasedamounts of 3-galactosidase. Some theorieswhich would explain these apparent decreasedlevels of 3-galactosidase relative to thiogalacto-side transacetylase and lacZ mRNA rely on thenotion that the f-galactoside activity found inthese strains is not an accurate indicator of thelevel of Z gene translation. For instance, bothnormal f-galactosidase and fused proteins orrestart fragments might be synthesized. Suchfused proteins or restart fragments might bedetectable as w peptides, which we were unableto find. Again, the presence of the two nonsensecodons in the operator suggests that only thefew strains that contain deletions extendinginto lacO could synthesize such fused proteinsand only strains with deletions either ending inthe operator or in the third possible readingframe could give rise to restart fragments.Furthermore, in F36a the deletion ends aftertrpA. Therefore, this strain probably does nothave any of the required read-through transla-tion. These theories also cannot explain thesimilar observation made in the his system.One possible explanation for the dispropor-

tionate production of f-galactosidase and thi-ogalactoside transacetylase can be derived fromthe following considerations. Let us assume thatthere is some minimum number of ribosomebinding events (translation initiation events)necessary at the start of any given cistron toensure propagation (or protection) of mRNAdown to the next cistron. Any number belowthis minimum will result in a polar effect on theexpression of the subsequent cistrons. Thiscould either be due to a combined endonu-cleolytic-exonucleolytic degradation of themRNA or to the fact that the RNA polymerasedoes not progress in the absence of an adequatesupply of ribosomes (8, 17, 18). Presumablyribosomes can bind at the start of internalcistrons as long as mRNA is present. Thisassumption is supported by the fact that it hasbeen possible to isolate lacO-Z deletions whichremove the lacZ translation initiation site yetstill permit expression of the lacY gene (6).Thus, as long as some minimum number ofribosomes translates the Z gene, the whole lacmessage will be produced and available fornormal translation initiation at Y and A.

Although the data presented in this paper donot allow an exact determination of this mini-mum number of lacZ ribosome binding events,they do suggest that this number is considerablylower than the normal complement of ribo-somes. This is indicated by the observation that

1237VOL. 117, 1974

13REZNIKOFF ET AL.

strains Wl and F23a have low levels of ,-galac-tosidase but near normal levels of thiogalacto-side transacetylase and lacZ and -A mRNA.The question as to how many ribosomes are

necessary for successful propagation of mRNAhas not been resolved by work previously re-ported. Yanofsky and Ito (30) have found that,for the case of suppressed nonsense mutationsin the trp operon, the relief of polarity isapproximately proportional to the level of sup-pression. However, earlier work by Newton etal. (19) using the lac system did not find such asimple relationship. There have been otherstudies examining the production of mRNA,either in temperature-sensitive translation-defective mutants or in the presence of antibiot-ics inhibiting translation (see 9, 10, 13). Thedata in these papers are insufficient to deter-mine whether detectable trp or lac mRNA isdependent upon a minimum number of ribo-somes or is directly proportional to the numberof ribosomes present.

If the model presented in this paper is correct,one would predict that a mutation in the lacZribosome binding site should have the followingphenotype. It should reduce lacZ expression butit should not affect lacY or -A expression unlessthe level of fully induced 0-galactosidase madeby the mutant is less than that found in Wl. Inother words, the effects should not be propor-tionate.

ACKNOWLEDGMENTSThis work was supported by Public Health Service post-

doctoral fellowships 1-FO2-GM-30970 (to W.S.R.) and5-FO2-GM-42815 (to C.A.M.), career development award5-KO4 GM-30970 (to W.S.R.), and grants 1-ROl GM-19670(to W.S.R.), GM-13017 (to J. Beckwith), and GM-14426 (toB.M.), all from the National Institute of General MedicalSciences, grant AM-13894 (to B.M.) from the NationalInstitute of Arthritis, Metabolism and Digestive Diseases,and PH 5-ROl AI-08902 (to C. Levinthal) from the NationalInstitute of Allergy and Infectious Diseases, and by NationalScience Foundation grants GB-20462 (to W.S.R.) andGB-5322 (to B.M.). T.G.C. was supported by postdoctoralfellowships from the National Research Council and theAmerican Cancer Society (PF-613). A.E.S. was supported bya Public Health Service microbiology training grant(GM-00602) from the National Institute of General MedicalSciences to the Department of Biology at M.I.T.

The authors thank J. Beckwith and D. Mitchell for theirhelpful criticisms of the manuscript.

LITERATURE CITED

1. Ames, B. N., and P. E. Hartman. 1963. The histidineoperon. Cold Spring Harbor Symp. Quant. Biol.28:349-356.

2. Beckwith, J. 1963. Restoration of operon activity bysuppressors. Biochim. Biophys. Acta 76:162-164.

3. Beckwith, J. R., and E. R. Signer. 1966. Transposition ofthe lac region of E. coli. I. Inversion of the lac operonand transduction of lac by 080. J. Mol. Biol.19:254-265.

4. Contesse, G., S. Naono, and F. Gros. 1966. Effet desmutations polaires sur la transcription de l'operonlactose chez Escherichia coli. C.R. Acad. Sci. (Paris)263:1007-1010.

5. Engelhardt, D. L., R. E. Webster, and N. D. Zinder. 1967.Amber mutants and polarity in vitro. J. Mol. Biol.29:45-58.

6. Eron, L., J. R. Beckwith, and F. Jacob. 1970. Deletion oftranslation start signals in the lac operon of E. coli, p.353-358. In J. R. Beckwith and D. Zipser (ed.), Thelactose operon. Cold Spring Harbor Laboratory, ColdSpring Harbor, N.Y.

7. Gussin, G. N., M. R. Capecchi, J. M. Adams, J. E.Argetsinger, J. Tooze, K. Weber, and J. D. Watson.1966. Protein synthesis directed by RNA phage mes-sengers. Cold Spring Harbor Symp. Quant. Biol.31:257-271.

8. Imamoto, F. 1970. Evidence for premature termination oftranscription of the tryptophan operon in polaritymutants of Escherichia coli. Nature (London)228:232-235.

9. Imamoto, F. 1973. Diversity of regulation of genetictranscription. I. Effect of antibiotics which inhibit theprocess of translation on RNA metabolism in Esche-richia coli. J. Mol. Biol. 74:113-136.

10. Imamoto, F., and Y. Kano. 1971. Inhibition of transcrip-tion of the tryptophan operon in Escherichia coli by ablock in initiation of translation. Nature N. Biol.232:169-173.

11. Imamoto, F., and C. Yanofsky. 1967. Transcription of thetryptophan operon in polarity mutants of Escherichiacoli. II. Evidence for normal production of trp-mRNAmolecules and for premature termination of transcrip-tion. J. Mol. Biol. 28:25-35.

12. Lodish, H. F. 1968. Independent translation of the genesof bacteriophage f2 RNA. J. Mol. Biol. 32:681-685.

13. Mackie, G., and D. B. Wilson. 1972. Polarity andtranscription in the galactose operon of E. coli.. Bio-chem. Biophys. Res. Commun. 48:226-234.

14. Michels, C. A., and W. S. Reznikoff. 1971. The gradientof polarity of Z gene nonsense mutations in trp-lacfusion strains of Escherichia coli. J. Mol. Biol.55:119-122.

15. Michels, C. A., and D. Zipser. 1969. The non-linearrelationship between the enzyme activity and struc-tural protein concentration of thiogalactoside transace-tylase of E. coli. Biochem. Biophys. Res. Commun.34:522-527.

16. Miller, J. H., W. S. Reznikoff, A. E. Silverstone, K.Ippen, E. R. Signer, and J. R. Beckwith. 1970. Fusionsof the lac and trp regions of the Escherichia colichromosome. J. Bacteriol. 104:1273-1279.

17. Morse, D. E., and P. Primakoff. 1970. Relief of polarity inE. coli by "SuA". Nature (London) 226:28-31.

18. Morse, D. E., and C. Yanofsky. 1969. Polarity and thedegradation of mRNA. Nature (London) 224:329-331.

19. Newton, W. A., J. R. Beckwith, D. Zipser, and S.Brenner. 1965. Nonsense mutants and polarity in thelac operon of Escherichia coli. J. Mol. Biol. 14:290-296.

20. Platt, T., K. Weber, D. Ganem, and J. H. Miller. 1972.Translation restarts: AUG reinitiation of a repressorfragment. Proc. Nat. Acad. Sci. U.S.A. 69:897-901.

21. Rechler, M. M., C. B. Bruni, R. G. Martin, and W. Terry.1972. An intercistronic region in the histidine operon ofSalmonella typhimurium. J. Mol. Biol. 69:427-452.

22. Rechler, M. M., and R. G. Martin. 1970. The intercis-tronic divide: translation of an intercistronic region inthe histidine operon of Salmonella typhimurium. Na-ture (London) 226:908-911.

23. Reznikoff, W. S., and J. R. Beckwith. 1969. Genetic evi-dence that the operator locus is distinct from the z

1238 J. BACTERIOL.

VOL. 117, 1874 lacZ GENE TRANSLATION

gene in the lac operon of Escherichia coli. J. Mol. Biol.43:215-218.

24. Reznikoff, W. S., J. H. Miller, J. G. Scaife, and J. R.Beckwith. 1969. A mechanism for repressor action. J.Mol. Biol. 43:201-213.

25. Silverstone, A. E., M. Goman, and J. G. Scaife. 1972.ALT: a new factor involved in the synthesis of RNA byEscherichia coli. Mol. Gen. Genet. 118:223-234.

26. Silverstone, A. E., and B. Magasanik. 1972. Polycistroniceffects of catabolite repression on the lac operon. J.Bacteriol. 112:1184-1192.

27. Ullmann, A., and D. Perrin. 1970. Complementation in,B-galactosidase, p. 143-172. In J. R. Beckwith and D.Zipser (ed.), The lactose operon. Cold Spring Harbor

INIATION INHIBITION 1239

Laboratory, Cold Spring Harbor, N.Y.28. Ullmann, A., D. Perrin, F. Jacob, and J. Monod. 1965.

Identification par complementation in vitro et purifica-tion d'un segment peptidique de la ,B-galactosidased'Escherichia coli. J. Mol. Biol. 12:918-923.

29. Voll, M. J. 1967. Translation and polarity in the histidineoperon. III. The isolation of prototrophic polar mu-tants. J. Mol. Biol. 30:109-124.

30. Yanofsky, D., and J. Ito. 1966. Nonsense codons andpolarity in the tryptophan operon. J. Mol. Biol.21:313-334.

31. Zinder, N. D., D. L. Engelhardt, and R. E. Webster. 1966.Punctuation in the genetic code. Cold Spring HarborSymp. Quant. Biol. 31:251-256.