Proc. Natl. Acad. Sci. USA Vol. 81, pp. 6100-6104, October 1984 Genetics Demonstration of two operator elements in gal: In vitro binding studies (/gaLR /electrophoretic mobility/DNA-protcin complex) ALOKES MAJUMDAR AND SANKAR ADHYA Laboratory of , National Cancer Institute, National Institutes of Health, Bethesda, MD 20205 Communicated by Allan Campbell, June 18, 1984

ABSTRACT Genetic and DNA base sequence analyses of F-hsdS20 recA13 aral4 proA2 lacY1 gal2 rpoL20 xyl5 su'II, cis-dominant mutations that derepress the gal operon ofEsch- used for large-scale preparation of gal plasmids (a gift of G. erichia coli suggested the existence of two operator loci needed Gaitanaris); SA1984 = F-strR galRA XcIts857 (Xint- for gal repression. One (OE) is located immediately upstream XcIII)AXN' (Xcro-chlA)A, made during this study and used to the two overlapping gal promoters and the other (01) is in- to maintain a PL-galR fusion plasmid. side the first structural gene. We have investigated the ability Fig. 1 Upper shows the genetic structure of the E. coli gal of wild-type and mutant OE and 01 DNA sequences to bind to operon and its restriction endonuclease sites relevant to the gal repressor. The repressor has been purified from cells con- construction of the following plasmids (i-v), which were the taining a multicopy plasmid in which the repressor gene is source ofgal DNA fragments used in protein binding assays. brought under the control of phage A PL . The DNA- (i) pMI3 was generated by cloning the 260-bp gal DNA repressor interactions are detected by the change in electro- spanning sites b-e between the EcoRI and HindIII sites of phoretic mobility of labeled DNA that accompanies its com- pBR322 (7). The gal DNA (fragment A) contains a wild-type plex formation with repressor protein. The purified repressor OE (O) allele. shows concentration-dependent binding to both O and Ot (il) pAM230 was generated by restriction of gal DNA at but not to cEand 0( sequences. These results authenticate the site b, digestion with DNase exoIII and nuclease S1 to the proposed operator role of the two homologous gal DNA control position marked c, and addition of an EcoRI linker. The gal elements and thereby establish that the negative control of the DNA between sites c and f (fragment B) was subsequently gal operon requires repressor binding at both OE and 0, cloned between EcoRI and HindIII sites in pBR322. which are separated by >90 base pairs. pAM230 contains a wild-type 01 allele (Ot). Fragment B is 500 bp long. Three structural , E, T, and K, constitute the gal oper- (iii) p291 contains the wild-type alleles of both OE and 01 on of (1). They are transcribed from either gal DNA (O Ot). p291 was generated by cloning the 664-bp of two partially overlapping promoters PG1 and PG2, which segment between sites b and f in gal (fragment C) between are modulated by cAMP and its receptor protein (CRP) in the EcoRI and HindIll sites of pBR322 (6). The distance be- opposite ways (2, 3). cAMP'CRP complex activates PG1 and tween OE and 0° is 97 bp. inhibits PG2 by binding to a single site, cat (refs. 4 and 5; see (iv) pBdC6 contains a 1100-bp gal DNA piece (fragment Fig. 1 Upper). The gal promoters are also regulated nega- D) spanning the segment a-f, between the EcoRI and Hin- tively by a gal repressor protein, the product of an unlinked dIII sites of PBR322 (11). Fragment D carries a mutation in galR gene (8-10). We have isolated and characterized cis- the OE allele (OE81 O'). dominant constitutive mutations in the gal operon and locat- (v) pAM401 was generated by cloning the gal DNA of an ed them by DNA base sequence analysis (3, 6, 11). These 01 mutant between sites d and f into the BstEII and HindIII studies have revealed the existence of two homologous 17- sites of pMI3. The resulting 664-bp-long gal DNA (fragment base-pair (bp) sequences in gal DNA, mutation in either one E) in pAM401 contains a mutant 01 allele (O 0i'6). of which causes derepression of PG, and PG2 in the presence (vi) pBR322 was obtained from Bethesda Research Labo- of gal repressor (Fig. 1). One of the sites (OE) is located im- ratories and digested with HindIII and BamHI to generate mediately upstream to the promoter region-i.e., around the 346-bp fragment F. -60 bp from the start site of PGl-promoted . (vii) pAM2 and pAM3. pAM2 contains a complete gal re- The other (0r) is located downstream to the promoter and pressor gene (galR') within a HincII and EcoRI DNA frag- inside the first structural gene, galE. Because of the cis-act- ment (12) fused to the PL promoter of bacteriophage X in the ing nature of the two sites and of the sequence homology plasmid pKC30 of R. N. Rao (26). The HincII-EcoRI gaiR between them, we have proposed that the gal operon is neg- DNA fragment was generated by restriction digestion of a atively regulated by two operators, OE and 0° (6). Both of XgalR-transducing phage (13). The HincII site is upstream of them are required for repression ofthe two gal promoters. In the NH2 terminus and the EcoRI site is downstream of the this paper, we report on the binding of purified gal repressor COOH terminus of the gaIR gene. The PL-galR fusion was to DNA fragments that contain either wild-type or mutant made by the gaiR HincII site and the X Hpa I site located OE and 01 base sequences. downstream to PL. pAM3 contains the COOH-terminal seg- ment of the gaIR gene cloned in a similar fashion but by fus- MATERIALS AND METHODS ing a different HincII site, located within the gaiR structural The E. coli strains used in this paper are SA1796 = F-strR gene (12) to the X Hpa I site in pKC30. These two plasmids galRs8, used to test the presence of a gal operator allele (7); are maintained in the host strain SA1984 lysogenic for X. The RW1401 = C600 galA, used for cloning wild-type and mutant details of construction of pAM2 and pAM3 will be published gal DNA fragments (a gift of R. Weisberg); HB101 = elsewhere. Microbiological Technics. Standard microbiological media, agar plates, and methods, including transformation of cells The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. Abbreviations: CRP, cAMP receptor protein; bp, base pair(s). 6100 Downloaded by guest on October 2, 2021 Genetics: Majumdar and Adhya Proc. NatL Acad Sci. USA 81 (1984) 6101

PG1 PG2 OE cat EN I-60 t -35 + 1 +27 det f

0+ A 260

I I B 500

c OI C I- 664 b C D t 1100 a 0c E I 664

t f F 346

I I FIG. 1. (Upper) Genetic map of the initial part of the gal operon. O' and O. are the two control elements (6) shown to be operators in this paper; PG1 is the cAMP-dependent promoter and PG2 is the cAMP-independent promoter; cat is the cAMP CRP complex binding site; EN is the start site of translation of the first structural gal gene. The lettered arrows pointing upward indicate the following natural or engineered restriction nuclease sites in gal DNA: a, a natural EcoRI site present in bacterial DNA brought closer to gal by a deletion (A320); b, an engineered EcoRI site (7); c, an engineered EcoRI site (see text); d, a natural BstEII site; e, a natural Hpa II site engineered into a HindIII site (7); f, a natural HindIII site. The numbers indicate the corresponding distance in base pairs from the start site of transcription initiation from PG1. (Lower) DNA fragments A-F generated by restriction nucleases from their corresponding plasmids and used in repressor binding assays of Fig. 3. The following lettered arrows indicate restriction sites of pBR322: g, a natural HindIII site; h, a natural BamHI site. The size of each fragment is indicated in base pairs at its right.

with plasmid DNA, were used (14, 15). sponding plasmids and labeled at their 5' ends with [y- 32P Labeling of DNA. DNA fragments generated by restric- 32P]ATP by polynucleotide kinase. The size and the relevant tion nucleases was labeled at the 5' end with [y32P]ATP as genetic makeup of the fragments are diagramed in Fig. 1 described by Maxam and Gilbert (16). The labeled fragments (Lower). These fragments were tested for their ability to bind were then purified by elution from gels (16). to gal repressor, as shown later. DNA-Protein Binding Assays. DNA complexed with pro- Purified gal Repressor. The gaiR gene encodes the repres- tein characteristically shows slower electrophoretic mobility sor of gal operon and maps between lysA and thyA genes at on acrylamide gels than free DNA. This property of DNA 60 min on E. coli chromosome (8, 9). We have fused the has been utilized successfully by Garner and Revzin (17) and wild-type gaiR gene to a strong promoter of bacteriophage X by Fried and Crothers (18) to demonstrate specific binding of in a hybrid plasmid vector pKC30, creating the plasmid proteins to DNA. We used the same principle to study the pAM2 (Materials and Methods). Because high-level tran- binding of gal repressor to 5'-end-labeled [32P]DNA frag- scription of the PL promoter in pAM2 is deleterious to its ments. An assay mixture of 20 ,td contained 20 mM Tris HCl maintenance, the plasmid was maintained in a cryptic X lyso- (pH 8.0), 40 mM KCl, 10 mM MgCl2, 1 mM dithiothreitol, gen (SA1984). The chromosomally located prophage in 0.1 mM EDTA, 50 jig of bovine serum albumin per ml, and SA1984 contains the cI repressor gene, the N gene, and no 10-100 ng of 5'-end-labeled [32P]DNA (about 20,000 cpm). other X genes. The prophage repressor gene carries a heat- The reaction was started by adding gal repressor (see legend sensitive mutation (c1857). In this host, the PL promoter of of Fig. 3), which was diluted 1:100 in the above buffer imme- plasmid pAM2 is repressed at 32°C but is induced by a tem- diately before assay. The reaction mixture was incubated at perature shift of the cells to 42°C. After 2 hr of induction, the room temperature (22°C) for 10 min. After addition of load- cells were lysed with lysozyme (19), followed by adjustment ing dye a portion of the reaction mixture was loaded onto 5% to 0.5 M KCl and 0.5 M NH4Cl and centrifugation at 70,000 polyacrylamide gels containing 0.125% bisacrylamide and x g. The repressor has been purified from the supernatant electrophoresed for 2-3 hr with 1.5 mA/cm of gel at room by a low-salt (0.1 M KCl) precipitation, followed by a phos- temperature. The gel was then autoradiographed. phocellulose chromatography (in 25 mM Tris'HCl, pH 8.0/1 mM EDTA/1 mM dithiothreitol/15% glycerol); the final RESULTS preparation is about 80% pure (Fig. 2). The details of the gal DNA Fragments. DNA fragments A-E carrying various purification of the gal repressor will be communicated later. segments of wild-type and mutant OE and OI alleles and The authenticity of the gal repressor in our preparation DNA fragment F of pBR322 were separated from the corre- was verified by the following two criteria. (i) The size of the Downloaded by guest on October 2, 2021 6102 Genetics: Majumdar and Adhya Proc. NatL Acad ScL USA 81 (1984)

-PHOSPHO CELLULOSE FRACTIONS 500-bp DNA fragment B, which contains the wild-type O0 . DNA sequence (Fig. 3B, lanes a-j). These results clearly LL show that the formation of the slower moving DNA bands is w uia EL dependent upon the presence of repressor and is specifically I inhibited by D-. We conclude that the characteristic ) mobility change of DNA fragments A and B is associated tcc 60. with their to -j 1 3 5 7 9 11 13 15 18 binding repressor. a BLi. 0 0 The 664-bp fragment C contains wild-type alleles of both OE and 01. The repressor binding pattern of this fragment was different from that of fragments A and B (Fig. 3C, lanes a-j). Though fragment C moved fast in the absence of re- pressor, two new bands were seen in the presence of repres- sor-one with intermediate mobility and the other with slow mobility (lanes a-e). The intermediate mobility band ap- peared only at lower repressor concentrations. At higher re- pressor concentrations, this was replaced by the slow mov- ing band. If D-galactose was present during incubations the two new species were not observed (lanes f-j). Since frag- ment C contains both of the 0 alleles, we believe that the intermediate mobility band is formed by binding one repres- f aft sor molecule, whereas the slow moving band is the result of binding two repressor molecules. At very high repressor concentrations, a very slow moving band is occasionally seen for fragments A and C. In the presence of D-galactose, this band remains, although its mobility is somewhat en- hanced. We believe that it is the result of binding of addition- al repressor molecules to these fragments in a way that is insensitive to D-galactose (data not shown). Whether there is FIG. 2. NaDodSO4/PAGE of various fractions of cellular ex- any base sequence specificity for this binding is not known. tracts for gal repressor purification after Coomassie blue staining. When a 346-bp 5'-end-labeled pBR322 DNA piece (frag- The principles of the purification are described in the Results. The ment F) was incubated with gal repressor, the latter did not most intense band in the eluate from the phosphocellulose column change the mobility of fragment F even when present at high (0.2-0.5 M KQl has a Mr = 38,000; it is absent in the lane with concentrations (Fig. 3D, lanes a-j). This shows that gal re- galR- extracts. Fractions 3-7 were pooled for the DNA-protein pressor does not bind to pBR322 DNA and is specific for gal binding assays of Fig. 3. EXCT, extract; PPT, precipitate. DNA. The following experiments further demonstrate that the sequence specificity of repressor binding to gal DNA re- purified protein by NaDodSO4/PAGE mobility is Mr 38,000, sides in the two OE and 01 homologous sequences. which is identical to that calculated from the DNA base se- gal Repressor Binding Is Specific for OE and 01. To investi- quence of the gaiR gene as determined by von Wilcken- gate the specific role of the OE and 01 sequences in repressor Bergman and Muller-Hill (12). We have shown previously binding, we examined the interaction of repressor with gal that gal repressor sediments in glycerol gradient centrifuga- DNA fragments containing mutant OE and 01 alleles. Fig. 3 tion as a Mr 72,000 protein (20). These results suggest that E and F, respectively, show the binding pattern of fragment the repressor is a dimer. (i) The Mr 38,000 protein is absent D, which contains a mutant OE and an wild-type 0, sequence in an induced extract of a strain carrying plasmid pAM3, a (O81 Oi ), and fragment E, which has the opposite genotype similar construct to pAM2 but one in which most of the gaiR (O Oc). We found that repressor does bind to the two mu- gene is deleted (Fig. 2). tant DNA fragments but in each case mainly generates a sin- Gel Electrophoresis of DNA-Repressor Complexes. The in- gle additional band and very little, if any, of a second slower teraction of regulatory proteins with specific DNA base se- band. At high repressor levels fragments D and E, like frag- quences has been studied extensively by the use of nitrocel- ments B and C, sometimes show a very slow moving band lulose membrane filters, which retain DNA-protein com- that is insensitive to D-galactose. plexes but not free DNA. Recently, an alternative method of When the relative repressor concentration was very high, studying DNA-protein interactions has been reported (17, as was the case in Fig. 3E, lanes e and j, some of the DNA 18). It involves separation of free DNA from DNA-protein remained at the top of the gel, which is characteristic of non- complexes based on the difference in their electrophoretic specific binding (17). As expected, the binding of repressor mobility in acrylamide gels. We have used the method to to fragments D and E is not seen in the presence of D-galac- detect a specific complex of repressor and DNA. tose, whereas the nonspecific repressor-DNA complexes Different amounts of gal repressor protein were incubated remain insensitive to the (Fig. 3 E and F, lanes f-j). with purified 5'-end-labeled [ 2PJDNA pieces in the absence These experiments suggest that fragments D and E bind only and in the presence Of D-galactose, an inducer of the gal op- one repressor molecule. In other words, the mutation pres- eron, and electrophoresed on polyacrylamide gels. Gel auto- ent in each fragment has caused loss of one of the two re- radiographs for gal repressor binding to the DNA fragments pressor binding sites present in the wild-type fragment C. A-F are shown in Fig. 3. The 260-bp DNA fragment A, Repressor binding to single operator fragments (OE or Or), which contains the wild-type 0' DNA sequence, had the containing an Oc mutation, was also tested. We have found expected gel mobility for a fragment of this size (Fig. 3A, that only a very small amount of Of or Oc DNA molecules, lane a); the mobility of fragment A was reduced after incuba- unlike their wild-type counterparts, shows repressor-depen- tion with repressor (lanes b-e). Note that the slower moving dent mobility change, suggesting that the Oc mutations cause band disappeared and the faster band reappeared if the incu- loss of affinity toward repressor (data not shown). We be- bations were carried out in the presence Of D-galactose lieve these results provide strong support for the idea that (lanes f-j) but not in the presence Of D- or D-fructose repressor binds to gal DNA by recognizing the two homolo- (data not shown). Identical results were obtained with the gous O0 and 0 base sequences. Downloaded by guest on October 2, 2021 Genetics: Majumdar and Adhya Proc. NatL Acada ScL USA 81 (1984) 6103

A B C D a b c d e f g h jIa b c d e f g h i a b c d e f g h i j Ia b c d e f g h i j

Ab so to aft"6046

lototo . .. m NW

E F a b c d e f g h i la b c d e f g h i j w~~

FIG. 3. PAGE analysis of 5'-end 32P-labeled DNA fragments A- F described in Fig. 1 Lower. (i) Fragment A (OE); (ii) fragment B (Ct); (iil) fragment C (O' Ot); (iv) fragment F (pBR322); (v) frag- ment D (O16 OCt); and (Vi) fragment E (O 0 ). The repressor (3 mg/ml) was diluted 1:100 and added to the 20-Al reaction mixtures to give the following final concentrations: lanes a and f, 0 nM; lanes b and g, 20 nM; lanes c and h, 60 nM; lanes d and i, 100 nM; lanes e andj, 200 nM. Lanes f-j contained D-galactose (10 mM) during incu- bations, which was added before repressor. The DNA concentration in each reaction mixture was about 5 nM. DISCUSSION one site is inactivated by mutation. We have shown recently the binding of gal repressor to two regions of gal DNA by Expression of the gal operon is regulated by two loci: OE DNase I protection experiments. The protected segments and 01. OE is located upstream to the gal promoters (extra- correspond to positions of about -75 to -50 and +43 to +67 genic) and 01 is downstream to them and within the galE bp of gal DNA and encompass the OE and O1 loci, respec- protein coding sequence (intragenic). Mutations at these two tively (unpublished data). A dimeric repressor molecule is loci have been identified in two ways. First, gal constitutive expected to bind to each site, but this needs to be confirmed mutations (QC) with a cis-dominant phenotype map in one of by determining the DNA-to-protein ratio in the various com- these two loci (6, 11). Second, the presence of a multicopy plexes. plasmid that contains OE or 01 (or both) in a cell induces the The demonstration of in vitro binding of repressor to OE chromosomal gal operon, presumably by binding all of the and Ot DNA but not to OE and 0c DNA together with the in gal repressor (6, 21). Plasmid mutations (Oa) that reduce the vivo results clearly establish an operator role for each ele- ability of DNA to "titrate" gal repressor also map at the OE ment. Note that the two operator elements are separated by and 01 loci (6, 22). From previous genetic and biochemical a segment ofDNA that is >90 bp long and contains the entire studies, the extragenic OE site was already assigned an oper- ensemble of gal promoters (see Fig. 1). How repressor bind- ator role (3, 8, 11, 20, 23). Because of the presumptive "titra- ing to the operator sites inhibits the promoters and why the tion" of repressor by the wild-type allele of both OE and 0° two operators are needed for each promoter remains un- sites and because of strong sequence homology of 0° to OE, known. We consider the following three models for the we have proposed that 01 defines an additional element of mechanism of gal repression. the gal operator-i.e., the gal operator is composed of two (i) Steric hindrance. In this model only one of the two elements, OE and 01 (6). Because mutation at either site de- sites is the actual operator. Repressor binding at this site ste- represses the gal promoters, we conclude that both of the rically hinders RNA polymerase binding. Because of its elements are needed for repression. To test this hypothesis close proximity to the gal promoters, the extragenic OE site and eliminate other more complex models that explain the is the more attractive candidate for the operator. According location of the intragenic O mutations, we have examined to this model the association between repressor and OE is the capacity of gal repressor to bind to the putative operator weak and the 01 site is responsible to increase the amount of sequences. The simplest interpretation of the results pre- complex. This might happen in one oftwo ways. sented here is that DNA carrying either O0 or O' sequences Binding of repressor to 01 could make more repressor avail- bind gal repressor. When both O0 and O+ are present in able for binding to OE through an intramolecular transfer DNA, only one of the two sites is occupied at low repressor (Fig. 4A). Alternatively, the repressor*0X complex could in- concentrations. We do not know whether one site is pre- teract with the repressor0E complex and prevent the disso- ferred over the other under these conditions. At higher re- ciation of the latter (Fig. 4B). pressor concentrations both of the sites are occupied, unless (il) DNA conformation change. A complex is formed be- Downloaded by guest on October 2, 2021 Genetics: Majumdar and Adhya Proc. NatL Acad Sci USA 81 (1984) 6104 A Howard Nash and Paul Kitt for their critical and helpful review of OE the manuscript and to Allan Campbell for interesting suggestions. PG2 PG1 A I~~~~ I 1. Buttin, G. (1963) J. Mol. Biol. 7, 164-182. 2. Musso, R. E., diLauro, R., Adhya, S. & de Crombrugghe, B. (1977) Cell 12, 847-854. 3. Adhya, S. & Miller, W. (1979) Nature (London) 279, 492-494. 4. Taniguchi, T., O'Neill, M. & de Crombrugghe, B. (1979) Proc. B Natl. Acad. Sci. USA 76, 5090-5094. 5. Busby, S., Irani, M. & de Crombrugghe, B. (1982) J. Mol. Biol. 154, 197-209. 6. Irani, M. H., Orosz, L. & Adhya, S. (1983) Cell 32, 783-788. 7. Irani, M., Orosz, L., Busby, S., Taniguchi, T. & Adhya, S. (1983) Proc. Nadl. Acad. Sci. USA 80, 4775-4779. 8. Buttin, G. (1963) J. Mol. Biol. 7, 183-205. 9. Adhya, S. & Echols, H. (1966) J. Bacteriol. 92, 601-608. 10. Saedler, H., Gullon, A., Feithen, L. & Starlinger, P. (1968) Mol. Gen. Genet. 102, 79-88. 11. diLauro, R., Taniguchi, T., Musso, R. & de Crombrugghe, B. (1979) Nature (London) 279, 494-500. 12. von Wilcken-Bergmann, B. & Muller-Hill, B. (1982) Proc. FIG. 4. Proposed mechanism for gal repressor operator interac- Natl. Acad. Sci. USA 79, 2427-2431. tion. (A) A repressor dimer binds independently to OE and 0I. (B) A 13. Parks, J. S., Gottesman, M., Shimada, K., Weisberg, R. A., repressor dimer binds to OE and 0j; the two repressorDNA com- Perlman, R. L. & Pastan, I. (1971) Proc. Natl. Acad. Sci. USA plexes then associate. This could either stabilize the repressors 68, 1891-1895. complex or alter the conformation of DNA in the resulting loop. It Molecular be out that there are alternate ways to draw the 14. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) should pointed loop A Labora- structure shown in B. It has been shown that , which is Cloning: Laboratory Manual (Cold Spring Harbor tory, Cold Spring Harbor, NY). a tetramer, can bind simultaneously to two lac operator DNA frag- 15. Miller, J. H. (1972) Experiments in (Cold ments (24, 25). Another possibility, not shown here, is that a gal Spring Harbor Laboratory, Cold Spring Harbor, NY). repressor dimer binds simultaneously to OE and 0° generating a 16. Maxam, A. & Gilbert, W. (1980) Methods Enzymol. 65, 497- DNA loop. 559. 17. Garner, M. M. & Revzin, A. (1981) Nucleic Acids Res. 9, tween repressorOE and the repressor 01 complexes as dis- 3047-3060. cussed above. Such an interaction forms a DNA loop con- 18. Fried, M. & Crothers, D. M. (1981) Nucleic Acids Res. 9, taining the promoter region (Fig. 4B). In this model, the loop 6505-6523. formation changes the conformation of promoter DNA to a 19. Arai, K.-I., McMacken, R., Yasuda, S.-I. & Kornberg, A. (1981) J. Biol. Chem. 256, 5281-5286. state that is inadequate to RNA polymerase activity. 20. Nakanishi, S., Adhya, S., Gottesman, M. E. & Pastan, I. (iii) RNA polymerase entry site. RNA polymerase entry (1973) Proc. NatI. Acad. Sci. USA 70, 334-338. sites for the gal promoters reside outside the OE=°I DNA 21. Saint-Girons, I., Fritz, H.-J., Shaw, C., Tillmann, E. & Star- segment. The molecule enters the DNA at these linger, P. (1981) Mol. Gen. Genet. 183, 45-50. sites and then gains access to the promoters. In this model, 22. Fritz, H.-J., Bicknase, H., Gleumes, B., Heibach, C., Roshl, occupation of both OE and 01, which do not overlap with the S. & Ehring, R. (1983) EMBO J. 2, 2129-2135. promoters, by repressor (Fig. 4A) blocks rapid access of 23. Feithen, L. & Starlinger, P. (1970) Mol. Gen. Genet. 108, 322- RNA polymerase to the promoters. 330. 24. O'Gorman, R. B., Dunaway, M. & Matthews, K. S. (1980) J. Biol. Chem. 255, 10100-10106. We thank our colleagues in the laboratory for assistance and en- 25. Culard, F. & Maurizot, J. C. (1981) Nucleic Acids Res. 9, couragements, Kathleen Matthews for valuable discussions, and 5175-5184. Annette Kuo for typing the manuscript. We are specially indebted to 26. Rao, R. (1984) Gene, in press. Downloaded by guest on October 2, 2021