Proc. Nati. Acad. Sci. USAVol. 76, No. 10, pp. 5090-5094, October 1979Biochemistry

Interaction site of Escherichia coli cyclic AMP receptor protein onDNA of galactose operon promoters

(protein-DNA interactions/DNase protection/methylation protection)

TAKETOSHI TANIGUCHI*t, MICHAEL O'NEILLt, AND BENOIT DE CROMBRUGGHE*§*Laboratory of Molecular Biology, National Cancer Institute, National Institutes of Health, Bethesda, Maryland 20205; and tDepartment of Biological Sciences,University of Maryland, Baltimore County, Catonsville, Maryland 21228

Communicated by Philip Leder, July 31, 1979

ABSTRACT Cyclic AMP (cAMP) and its receptor protein(CRP) have a dual role in the regulation of the two promotersthat control the galactose (gal) operon of Escherichia coli. Onepromoter, P1, requires cAMP-CRP for activity; the other, P2, isinhibited by these factors. We have examined the interactionsite of cAMP'CRP on galDNA by using two types of protectionexperiments, involving DNase digestion and methylation bydimethyl sulfate. Our results indicate that cAMP-CRP binds togal DNA in a segment located between 50 and 24 base pairspreceding the P1 start point for transcription. Although the lo-cation of the cAMP-CRP interaction site is clearly different ingal and lac DNA, comparison of the DNA sequences suggestsa similar recognition sequence. The location of the cAMP-CRP-binding site in gal further suggests that protein-proteininteractions between RNA polymerase and cAMP-CRP play animportant role in transcription initiation at the gal and possiblyother cAMP-dependent promoters.

Bacterial or bacteriophage promoters can be subdivided intotwo groups. In one class of promoters the DNA sequence con-tains all the necessary information for the binding of RNApolymerase and the formation of a stable complex for pro-ductive initiation of transcription. In other promoters the se-quence information and hence the DNA structure is insufficientby itself to allow such stable complexes to be formed in theabsence of additional protein factors. Cyclic AMP-dependentpromoters fall in this second category. The additional infor-mation and interactions needed for the formation of a stablecomplex between RNA polymerase and DNA are provided bythe cyclic AMP (cAMP)-cyclic AMP receptor (CRP) complexand the CRP binding site on the DNA.The role of cAMP.CRP in the regulation of the galactose (gal)

operon of Escherichia coli is complex because this operon iscontrolled by two overlapping promoters, P1 and P2 (1). Piactivity requires cAMP and its receptor protein. P2 functionsin the absence of these factors but is inhibited by cAM'P-CRP.The start points for transcription of these two promoters areseparated by five base pairs or half a turn in the DNA helix.The existence of the two gal promoters probably reflects the

dual function of galactose in cellular metabolism. When ga-lactose becomes the principal carbon source in the medium, itis taken up by the cells, converted to glucose-i-P, and furthercatabolized to serve as a general energy source. One of the in-termediary products in this catabolic pathway, UDP-galactose,is also a precursor for cell wall biosynthesis. Even in the absenceof galactose, UDP-galactose continues to be generated fromUDP-glucose in a reaction catalyzed by the enzyme UDP-glucose 4-epimerase, specified by the promoter-proximal cistronof the gal operon. P1 probably controls the catabolic pathwaywhereas P2 regulates the anabolic or biosynthetic pathway of

galactose. It can be argued that the function of the two pro-moters is to ensure a constant basal level of synthesis of gal en-zymes, particularly UDP-glucose 4-epimerase, regardless ofthe physiological fluctuations in the intracellular concentrationsof cAMP (ref. 2; S. Adhya, personal communication).To try to understand how cAMP-CRP controls gene expres-

sion and more particularly how cAMP-CRP exerts its dual rolein gal regulation, we have determined the interaction site forcAMP.CRP on gal DNA. We have used two types of protectionexperiments. First, we have analyzed which DNA segment wasprotected by cAMP-CRP from DNase digestion (3). We havealso probed which purine residues could be protected bycAMP-CRP from methylation by dimethyl sulfate (4). We findthat cAMP-CRP protects a gal DNA fragment located between25 and 50 base pairs preceding the P1 mRNA start site. Hencethe location of the gal-CRP interaction site is clearly differentfrom the lac CRP-binding site located between -70 and -50.Examination of the DNA sequence suggests, however, thatcAMP-CRP may recognize similar sequences in gal and inlac.Our data indicate that positive factors for transcription can

activate genes by interacting with promoters in a segmentaround 35 base pairs preceding the transcription initiation site.The location of the gal CRP-binding site suggests that pro-tein-protein interactions between cAMP-CRP and RNApolymerase on gal DNA play an important role in transcriptioninitiation.

METHODSIsolation of gal Promoter DNA. Plasmid pBCl is a deriv-

ative of pBR322 in which the 30-base pair EcoRI/HindIIIDNA fragment has been replaced by a DNA fragment con-taining the gal operator-promoter region and the promoter-proximal third of the galE cistron (unpublished experiments).The smaller gal subfragments (F-2 and F-3 in Fig. 1) can beconveniently isolated from pBCl by endonuclease digestionfollowed by preparative 5% polyacrylamide gel electropho-resis.Terminal 32P Labeling ofDNA Fragments. The 5' ends of

the DNA fragments were treated with bacterial alkalinephosphatase and were labeled with 32P by using phage T4polynucleotide kinase as described by Maniatis et al. (5). Toobtain fragment 1, (F-1) (Fig. 1) labeled at its Hha I 5' end("upper strand," see Fig. 5), F-2 was isolated, labeled at its 5'end, cleaved by endonuclease Hap II, and fractionated by 5%polyacrylamide gel electrophoresis. Similarly, to obtain F-1labeled at its Hap II 5' end ("lower strand"), F-S was isolated,

Abbreviations: cAMP, cyclic AMP; CRP, cAMP receptor protein.t Present address: Department of Biochemistry, Kouchi MedicalCollege, Nangoku-Shi, Kouchi, Japan.

§ To whom reprint requests should be addressed.5090

The publication costs of this article were defrayed in part by pagecharge payment. This article must therefore be hereby marked "ad-vertisement" in accordance with 18 U. S. C. §1734 solely to indicatethis fact.

Proc. Natl. Acad. Sci. USA 76 (1979) 5091

E T K

-80 -60 -40 -20 +1 +20 +40 =--_Emr I I , -_

Operator mRNA (+cAMP-CRP,-FRNA (-cAMP-CRP)

Hap ii Hha Hinf Hap I I Hha Hi(F-i ) -135bp

' ' ~~~~~~235bp'I| (F-2) @

~~~~~~240bp(F-3) 2

iO~~~1 5 bpfe

(F -4)365 bp

(F-5)

0.2-0.5 Mg of labeled fragment in 200 Ml ofmM Tris-HCI, pH7.9/0.1 M KCI/10 mM MgCI2/0.1 mM EDTA/1 mM di-thiothreitol/21 Mg of CRP per ml/500,uM cAMP. The mixturewas incubated at 370C for 10 min. Sonicated chicken bloodDNA (20 Mg/ml) was then added. After 1-min incubation, 1

Ml of dimethyl sulfate (10.7 M) was added and the mixture wasincubated at 370C for 1 min. The sample was precipitated anddepurinated at 90'C for 15 min at pH 7.0 and hydrolyzed at90'C for 30 min with alkali in a sealed capillary. The hydro-lyzed sample (20 Ml) was mixed with 20Ml of 10 M urea/0.05%xylene cyanol/0.05% bromphenol blue and electrophoresed on20% and 10% polyacrylamide/7 M urea gels.CRP. CRP prepared as described (7) was a generous gift of

J. Krakow.

FIG. 1. Schematic representation of the gal operon and the rel-evant restriction fragments used in this work. Lengths of fragmentsare given in base pairs (bp). The start site for cAMP-CRP-dependenttranscription is represented as +1.

labeled at its 5' end, and cleaved by endonuclease Hha I. F-4labeled at its Hinfl 5' end can be generated by labeling F-5 andthen cleaving by Hap II endonuclease. Conversely, to obtainF-4 labeled at its Hap II end, we isolated F-3, labeled thisfragment, and cleaved by Hinfl endonuclease.DNase Protection. The DNase protection method of Galas

and Schmitz (3) was used. The principle of this technique is asfollows. A DNA fragment, specifically labeled at one 5' end,is digested by pancreatic DNase I under conditions of partialdigestion in the presence or absence of a DNA-binding protein.The products are fractionated according to size by electro-phoresis on high-resolution polyacrylamide gels and the gelsare autoradiographed. The patterns of bands corresponding tothe reactions with and without the DNA-binding protein are

compared. A decrease in the intensity of a band is attributedto protection of the cleavage site by the DNA-bound protein.In our experiment the reaction mixture consisted of 0.2 Mig ofa labeled fragment in 10OMl of 10 mM Tris-HCl, pH 7.9/10mM MgCI2/5 mM CaCl2/100 ,uM dithiothreitol/21 ,ug of CRPper ml and various concentrations of cAMP. The reactionmixture was incubated at 250C for 10 min and then bovinepancreatic DNase I (Worthington) was added to give a finalconcentration of 0.13 ug/ml. The digestion was carried out at250C for 30 sec. The products were precipitated with ethanoland dried under reduced pressure. The samples were dissolvedin 10 Ml of 0.1 M NaOH/1 mM EDTA and -10 Ml of 0.05% xy-

lene cyanol/0.05% bromphenol blue/10 M urea solution andapplied on both a 10% and a 20% polyacrylamide/7 M urea gelfor electrophoretic fractionation.

Methylation Protection. Dimethyl sulfate methylates thepurine residues in DNA at the N-7 of guanines and at the N-3of adenines (6). The methylation method of Maxam and Gilbert(6) used for DNA sequencing can also be used to study the in-teractions of DNA-binding proteins with a DNA fragment (4).This procedure includes the following steps. A DNA fragment,selectively labeled at one 5' end, is partially methylated in thepresence or in the absence of a DNA-binding protein. Themethylated guanines and adenines are removed by heating atneutral pH and then the phosphodiester bonds of DNA are

cleaved at the depurinated sites by heating with alkali. Thefractionation products are separated on a high-resolutionpolyacrylamide/7 M urea gel and autoradiographed to producea pattern of bands. The binding of DNA by a protein can bothdecrease and increase the level of methylation of purines in thecontact region (4). Decrease (protection) seems to be due tosteric hindrance. Conditions of methylation were as describedby Gilbert and Maxam (6). The reaction mixture consisted of

RESULTSA Segment of gal Promoter DNA Is Protected by cAMP

CRP from DNase Digestion. We first examined the effects ofcAMP-CRP on the digestion pattern by DNase I of the 135-basepair gal promoter DNA fragment 1 (F-1) depicted in Fig. 1.This fragment extends from 93 base pairs preceding the startsite for PI to 43 base pairs following this start site and includespart of the first structural gene galE. The nucleotide sequenceof this fragment has been determined (1). Fig. 2A analyzeswhich bases are protected on the upper strand of the gal pro-moter; for this experiment F-1 was specifically labeled at itsHha I 5' end. Similarly, the experiment reproduced in Fig. 2Bexamined which bases are protected on the lower strand. Thelocation of the protected bases is determined by comparing theposition of the gel bands derived from the DNase reaction withthe position of G and A bands obtained in a purine-specificreaction (Fig. 2A, lane 1, and Fig. 2B, lane 1) and the position

.4 11 2 3 4 5 6

"Nw____~~~~x2 --m 'W" _ij:

vli

.- - *04 or...

mE|||

B 1 2 3_

-60-

-50 *_.-*

m

-40 * a -

&mm

* _m-30

-20 * a

w_- a"

--10 0

-Wm mmm._aON. -1 * -

FIG. 2. DNase footprint analysis of cAMP-CRP protection ofgalDNA. Fragment F-1 (see Fig. 1) was end-labeled at one or the other5' end and digested under partial reaction conditions with DNase inthe presence or absence of cAMP-CRP. The reaction products were

fractionated on a 20% polyacrylamide/7 M urea gel. (A) Analysis ofthe upper strand. Lane 1, G + A reaction; lane 2, no cAMP*CRP; lanes3-6, CRP (20 g/ml) and cAMP at 500 MM (lane 3), 50 MM (lane 4),5 jM (lane 5), 0.5,M (lane 6). (B) Analysis of the bottom strand. Lane1, G + A reaction; lane 2, no cAMP-CRP; lane 3, CRP (20 Mg/ml) and5 MM cAMP.

QP7PI,

7

i - - I I

i

Biochemistry: Taniguchi et al.

.... am& 40 &-.*

amw 0.- It=

mw*w ovooft SW

;7, .,-! a -Oman so .-

:Raw -,.!-. mm&Am at&

.

5092 Biochemistry: Taniguchi et al.

of C and T bands obtained in a pyrimidine-specific reaction (notshown). The pattern of bands corresponding to a sample thatcontained CRP but no cAMP is very similar to the pattern de-rived from a reaction without CRP and cAMP. Hence, CRPitself does not affect the DNase digestion pattern and the pro-tection that is seen is specifically due to the cAMP-CRP com-plex.The concentration range of cAMP that ensures protection

of segments of the gal promoter by cAMP-CRP is very similarto the concentration of cAMP needed for PI stimulation andP2 repression as measured in an in vitro transcription assay (8).The concentration of cAMP needed for half-maximal stimu-lation of PI (5 jM) is the same as the cAMP concentration thatinhibits 50% of P2 activity (8). If the concentration dependencyof CRP is examined, we find that the same concentrations ofCRP that in vitro activate transcription at P1 or inhibit tran-scription at P2 protect gal DNA from DNase digestion.The results of Fig. 2 are summarized in Fig. 5. In the top

strand protection occurs at the 3' side of the following bases: Aat -50, A at -46, A at -41, T at -40, G at -39, T at -38, C at-37, A at -36, and C at -35. On longer exposure we see thatthe bases between -33 and -25 on the upper strand are alsoprotected. To examine the full extent of protection longer ex-posures are needed because the patterns of digestion in theabsence of cAMP-CRP are not uniform, DNase being muchmore active at certain specific sites in the DNA sequence thanat others. On the bottom strand fewer sites are protected: G at-42 and A at -38. The finding that fewer bases are protectedon the lower strand is mostly due to the absence of digestion atmany sites on this strand between -50 and -35 in the absenceof cAMP-CRP. Interestingly, DNase digestion is enhanced at

1 2 3 4 5 6 7

+10. _i

+1*- *|s *11--

-10'

_-qMom Mw

-30*

_ _ ........

--_ mo

-400 0

-42 C, -34 A, -23 C (less pronounced), -16 A, and -14 A onthe upper strand, and at -49 A, -40 A, and -30 A on the lowerstrand. These enhancements occur at concentrations of cAMPof 5MgM or higher and thus similar to those that cause protection.(We noted that the enhanced sensitivity to DNase at -14 ismore pronounced at a cAMP concentration of 50,tM.) We at-tribute the increased sensitivity to DNase digestion at these sitesto a change in DNA conformation resulting from the interactionof cAMP-CRP with gal DNA.

At much higher concentrations of cAMP than needed fortranscription activation, many additional sites are protected(Fig. 2A, lane 3). We had observed previously that, at highconcentrations of cAMP, f3-galactosidase synthesis was inhibitedin a cell-free DNA-dependent S30 system (9). It is possible thatat higher concentrations cAMP induces a nonspecific cooper-ative binding of CRP to DNA. It is obvious from our results thatat concentrations of cAMP that activate transcription from P1no bases are protected between -70 and -55. In this segmentthe gal sequence presents considerable analogies both in se-quence and in symmetry with a similarly located sequence inthe promoter of the lac operon (10). This sequence contains thelac CRP-binding site (refs. 11 and 12; J. Majors, personalcommunication).

It was formally possible that cAMP.CRP might first recognizethe sequence similar to the lac CRP-binding site at -70 to -55and then bind to the -50 to -25 segment in a second step. Totest this possibility we asked whether the 105-base pair fragmentF-4 of Fig. 1, which lacks the bases to the left of -59, was ca-pable of interacting with cAMP-CRP in a similar fashion as thelarger fragment. The results presented in Fig. 3 indicate that

1 2 3 4

-_ .-20a-mo ow

-309t

.- ...X

gm* - -

anAA- -m a*-40* *-30

. 20

'0-10

--50.* :M_mp - * a +1

.....

-6 -0_-

-70'

-100gmw Ivmw 0

-- -_m

ammmw

FIG. 3. Effect of cAMP-CRP on DNase footprint analysis offragment F-4 (see Fig. 1). F-4 was labeled at its Hinfl 5' end. Thepartial DNase digestion was conducted as in Fig. 2 with or withoutcAMP-CRP. Reaction products were fractionated on a 20% polyac-rylamide/7 M urea gel. Lane 1, G + A reaction; lane 2, no cAMP-CRP;lanes 3-7, CRP (20 ,g/ml) and cAMP at 500,gM (lane 3), 50MM (lane4), 5 ,gM (lane 5), 0.5 ,M (lane 6), 0MM (lane 7).

FIG. 4. Effect ofcAMP-CRP on methylation ofgal DNA. Frag-ment F-1 (see Fig. 1) was end-labeled at one or the other 5' end andtreated with dimethyl sulfate (6). The fractionation products wereseparated on a 20% denaturing polyacrylamide gel. Lanes 1 and 2,analysis of the upper strand; lane 1, no cAMP-CRP; lane 2, CRP (21jg/ml) and cAMP (500 ,M). Lanes 3 and 4, analysis of the lowerstrand; lane 3, no cAMP-CRP; lane 4, CRP (21 ,ug/ml) and cAMP (500AM).

Proc. Natl. Acad. Sci. USA 76 (1979)

*s 1 0

Proc. Natl. Acad. Sci. USA 76 (1979) 5093

cAMP-CRP protects the same residues from DNase digestionas in the larger 135-base pair fragment (F-i). cAMP-CRP alsoproduces an enhancement of DNase digestion at the same sitesas with the larger fragment except at -34 on the upper strand.Similar concentrations of cAMP are needed to obtain DNaseprotection as with the larger fragment. Thus a DNA fragmentthat is missing part of the sequence that is similar to the lacCRP-binding site is capable of interacting with cAMP-CRP aswell as a larger fragment that contains these sequences.

Methylation Protection. We next examined which specificbases in gal DNA made contact with cAMP-CRP by probingwhich purine residues could be protected from methylation bycAMP-CRP (4). Again, DNA fragment F-1 (see Fig. 1) wasspecifically labeled at one or the other of its 5' ends. This frag-ment was then allowed to react with dimethyl sulfate in eitherthe presence or the absence of cAMP-CRP. As can be seen inFig. 4, one residue, G -35 on the lower strand, was stronglyprotected by cAMP.CRP. Weaker protection occurred at G -37on the lower strand and at G-39 on the upper strand.

DISCUSSIONThe results presented here (see Fig. 5 for summary) indicatethat cAMP-CRP interacts with a gal DNA segment locatedbetween 50 and 24 base pairs preceding the start site of PImRNA. Many sites in this segment are protected by cAMP-CRPfrom DNase digestion. This protection by CRP is strictly de-pendent on cAMP at concentrations of the cyclic nucleotide thatare identical to those needed for CRP-dependent activation ofPI or inhibition of P2. Within this protected segment a numberof sites exhibit an enhanced sensitivity to DNase as a result ofcAMP-CRP interaction with gal DNA. They appear at regu-larly spaced intervals in the protected sequence at -42, -34,and -23 on the upper strand (and -49, -40, and -30 on thelower strand). In addition, two other sites at -16 and -14 onthe upper strand exhibit an enhanced sensitivity to DNase.These latter sites lie within the Pribnow "heptamer" for P2 andprecede by a few bases the "heptamer" for P1. These heptamersare located about 5 to 6 bases preceding the start site of tran-scription in different prokaryotic promoters and show a highdegree of conservation (13, 14). Bases within these heptamerscould participate in the localized DNA melting associated withthe formation of a stable RNA polymerase-promoter complex(13). The interactions of cAMP-CRP with the -50 and -25region could thus cause a conformational change in the DNAstructure in the P2 heptamer. This structural change, in additionto the protein-protein interactions between cAMP-CRP andRNA polymerase, might favor formation of a stable complexat PI or inhibit the formation of such a complex at P2.Our results clearly indicate that cAMP-CRP does not bind

to the -70 to -50 region in gal, a segment that presentshomologies both in sequence and in symmetry with theCRP-binding site in lac located between -70 and -50. Fur-

thermore, this gal segment is not needed for CRP binding at-50 to -25, because removal of the sequences to the left of -59does not alter the cyclic AMP concentration dependency ofCRP binding. Earlier results had, in fact, strongly suggested thatthe -70 to -50 segment was not the gal CRP-binding site (8).Indeed several galOc mutants map in this segment between-66 and -55 (ref. 8; M. Irani and S. Adhya, personal commu-nication). These Oc mutants exhibit the same cAMP-CRP de-pendent stimulation of PI and repression of P2 as wild-type galDNA. One such gal mutant results in an identical base change(G-C to A-T) as a lac CRP-binding site mutant and is locatedthe same distance from the cAMP-CRP-dependent initiationsite in gal and in lac (8, 11). This mutation severly reduces thecAMP-CRP response in lac but leaves it unchanged in gal.

In other experiments the cAMP-CRP interaction site on galDNA was also functionally mapped by examining whethersequences to the left of the HinfI cleavage site at -59 wereneeded for transcriptional control of P1 and P2 activity.Transcription of two hybrid DNA fragments in which thesegment to the left of -59 was replaced by two different un-related DNA sequences exhibits a cAMP-CRP response for P1stimulation or P2 repression identical to that of wild-type galDNA (unpublished data). Hence, the DNA sequence to theright of -60 contains all the specific interaction sites neededfor the cAMP-CRP-dependent activation of P1 and repressionof P2.

It is clear that the location of the cAMP-CRP interaction sitein gal is very different from the CRP-binding site in the lacoperon. The identity of this site in lac has been firmly estab-lished by DNA sequencing of CRP-binding site mutants andby chemical protection experiments (ref. 11; J. Majors, personalcommunication). In gal the -50 to -70 segment covers the sitewhere the repressor interacts with gal DNA.We wished to determine more precisely the sequence within

these CRP regulatory regions that is recognized by CRP. In thisanalysis, we employed the following criteria: (i) the commonCRP recognition sequence should be altered by the CRP-binding site mutations known in lac and should include thebases protected by or from dimethyl sulfate in lac and in gal;(ii) the sequence should be consistent with the size of CRP (7),spanning 10-15 base pairs; (iii) the sequence should accom-modate a dimer, presenting similar sequence elements to eachsubunit; (iv) the region of the DNA protected against DNaseby CRP should be approximately coextensive with the occur-rences of the CRP recognition sequence.A computer was utilized to search for homologies in the lac

and gal sites that might satisfy these criteria. The regulatoryregion between araC and araBAD is also known to be involvedin complex interactions with cAMP-CRP (15, 16) and wastherefore included in this search. The result pointed to a se-quence appearing twice in the CRP site of lac, symmetricallylocated about the dyad axis centered at -60/-61 (11, 17), ap-

-80 -70 -60G 0C T ^ ^ A T TC T T a T a ^ ^ ^ C a 0GGCTAAATTCTTGTGTAAACGAT piCCGATTTAAGAACACATTTOCTAA

-50 -40 -30 j-20 -10 +10T C C A C T ^AT T TOT T C C0)MZJ TbT)T C T T TGT A CT G G T TAT T T CAT A CC A T A A G

AGO T OAT TA AAT A A GT CAGTT G AAG C G TAG A A AC A:A TAC A TA CC A A T AAT G TAT GOT AT V C* 0 '--~~~~~~~~~~~~~

FIG. 5. Summary of protection data. Circles indicate protection against DNase digestion by cAMP-CRP at the 3' side of the indicated base.Arrows indicate enhancement of digestion by DNase at the 3' side of the indicated bases. Square indicates base protected by cAMP-CRP frommethylation by dimethyl sulfate. P1 and P2 refer to the respective initiation sites for the P1 and P2 promoter (1). Boxed sequences indicate thePribnow heptamers for each of the two promoters. The upper strand reads 5' - 3' from left to right. -

Biochemistry: Taniguchi et al.

I I

5094 Biochemistry: Taniguchi et al.

pearing three times in the gal CRP region, and five (or more)times in the ara regulatory region (18, 19). These sequences arelisted as follows:

lac -73 T-T-A-A-T-G-T-G-A-G-T -63-50 A-A-T-G-A-G-T-G-A-G-G -60

gal -21 A-A-G-A-T-G-C-G-A-A-A -31-30 A-A-A-G-T-G-T-G-A-C-A -40-44 T-C-C-A-T-G-T-C-A-C-A -34

ara -29 A-A-A-G-C-G-T-C-A-G-G -39-82 A-A-A-G-T-G-T-G-A-C-G -92-141 A-A-A-G-T-G-T-C-T-A-T -131-148 A-A-A-G-C-G-C-T-A-C-A -138-158 A-A-A-G-C-G-C-T-A-C-A -148

The consensus sequence obtained from this list is 5' A-A-A-G-T-G-T-G-A-C-A 3'. This sequence appears withoutchanges in gal from -30 to -40. The underlined region of thesequence seems to be most important for recognition. The twoGs in this region are strongly implicated by methylation ex-periments both in the lac (J. Majors, personal communication)and in the gal sites. The first G and the A are totally conservedwithin this list. A transition of the second G to A constitutes thesymmetric CRP mutations that Dickson et al. (11) have de-termined in the lac CRP site. This sequence is generally dis-tinguished from rather similar sequences that appear in nu-merous promoters in the -35 region (20) and in the lac, gal, andX operators (8, 20, 21) by differences in the underlined regionof the sequence.The consensus sequence also satisfies criteria i and iii above

because it is the right size and can, according to our preliminaryanalysis, accommodate a dimer of identical subunits in eithera direct repeat or an inverted repeat arrangement.We find that cAMP.CRP protects a gal segment that is larger

than the sequence presenting homologies with the lac CRP sitesin gal. This could be due to binding of CRP to adjacent weakersites. The binding of CRP to these weaker sites might be stabi-lized by protein-protein interactions. In fact, two additionalsites with partial homology with the proposed recognition se-quence were found within the protected sequence (-21 to -31and -44 to -34).How can the difference in location between the CRP-binding

sites in lac and in gal be reconciled with their similar effects onactivation of transcription? One possible explanation would bethat the interaction between CRP and RNA polymerase occursaround -35 in both gal and lac. In lac the primary binding sitefor CRP is obviously at -70 to -50, but the binding of CRP toa secondary binding site around -35 might also be stabilizedby CRP-CRP or CRP-RNA polymerase interactions. In fact,the gal sequence

-36 -30A-C-A-C-T-T-TT-G-T-G-A-A-A

is also found in lac from -34 to -28. This sequence constitutespart of what we think is the CRP recognition sequence ingal.The interaction of an activator for gene transcription such

as cAMP-CRP with promoter DNA around-35 is not uniqueto gal. Indeed, recent experiments on the stimulation of the XPRM promoter by the X repressor support this notion. Althoughthe X repressor preferentially binds to a site 59 to 75 base pairspreceding PRM, its binding to a segment between -35 and -51is essential for PRM stimulation. Interaction of the X repressorwith the -59 to -75 site induces a cooperative binding of therepressor to the -35/-51 site (B. Meyer and M. Ptashne, per-sonal communication).

Our results establish that factors that activate transcriptioninitiation can interact with promoters between -50 and'-25.The sequences around -35 are one of two regions of homologyin prokaryotic promoters (20), the other being the heptamercentered at -10 (7, 13, 14). In the lac UV5 promoter RNApolymerase makes contact with a G residue at -32 (4). Thesegment around -35 is also the site of several promoter muta-tions. These mutations either decrease or increase promoteractivity. They alter the binding of RNA polymerase or changethe rate of initiation of transcription at these promoters (20,22-25). The proposed CRP recognition sequence in the sameregion in gal probably provides some of the necessary inter-actions for productive initiation of transcription at P1. The lo-cation of the CRP-binding site strongly suggests that initiationof transcription includes protein-protein interactions betweencAMP-CRP and RNA polymerase in gal DNA and maybe alsoin other cAMP-dependent promoters.We are grateful to Joe Krakow for CRP. We thank Linda Harris for

secretarial assistance and Ray Steinberg for the illustrations.

1. Musso, R. E., DiLauro, R., Adhya, S. & de Crombrugghe, B.(1977) Cell 12,847-854.

2. de Crombrugghe, B. & Pastan, I. (1978) in The Operon, eds.Miller, J. H. & Reznikoff, W. S. (Cold Spring Harbor Laboratory,Cold Spring Harbor, NY), pp. 303-324,

3. Galas, D. & Schmitz, A. (1978) Nucleic Acids Res. 5, 3157-3170.

4. Johnsrud, L. (1978) Proc. Natl. Acad. Sci. USA 75, 5314-5318.

5. Maniatis, T., Jeffrey, A. & Kleid, D. G. (1975) Proc. Nati. Acad.Sci. USA 72, 1184-1188.

6. Maxam, A. M. & Gilbert, W. (1977) Proc. Nati. Acad. Sci. USA74,560-564.

7. Anderson, W. B., Schneider, A. B., Emmer, M., Perlman, R. L.& Pastan, I. (1971) J. Biol. Chem. 246,5929-5937.

8. DiLauro, R., Taniguchi, T., Musso, R. & de Crombrugghe, B.(1979) Nature (London) 279,494-500.

9. Emmer, M., de Crombrugghe, B., Pastan, L. & Perlman, R. (1970)Proc. Nati. Acad. Sci. USA 6 480-487.

10. Musso, R., DiLauro, R., Rosenbeg, M. & de Crombrugghe, B.(1977) Proc. Natl. Acad. Sci. USA, 74, 106-110.

11. Dickson, R. C., Abelson, J., Johnson, P., Reznikoff, W. S. & Barnes,W. M. (1977) J. Mol. Biol. 111, 65-75.

12. Majors, J. (1975) Nature (London) 256, 672-674.13. Pribnow, D. (1975) Proc. Nati. Acad. Sci. USA 72,784-788.14. Schaller, H., Gray, C. & Herrman, K. (1975) Proc. Natl. Acad.

Sci. USA 72,737-741.15. Greenblatt, J. & Schleif, R. (1971) Nature (London) New Biol.

233, 166-170.16. Casadaban, M. (1976) J. Mol. Biol. 104,557-566.17. Gilbert, W. (1976) RNA Polymerase, eds. Losick, R. & Cham-

berlin, M. (Cold Spring Harbor Laboratory, Cold Spring Harbor,NY), pp. 193-205.

18. Greenfield, L., Boone, T. & Wilcox, G. (1978) Proc. Natl. Acad.Sci. USA 75,4724-4728.

,19. Smith, B. & Schleif, R. (1978) J. Biol. Chem. 253,6931-6932.20. Reznikoff, W. S. & Abelson, J. N. (1978) in The Operon, eds.

Miller, J. & Reznikoff, W. (Cold Spring Harbor Laboratory, ColdSpring Harbor, NY), pp. 221-243.

21. Maniatis, T., Ptashne, M., Backman, K., Fleid, D., Flashman, S.,Jeffrey, A. & Maurer, R. (1975) Cell 5, 109-113.

22. Meyer, B. J., Kleid, D. G. & Ptashne, M. (1975) Proc. Natl. Acad.Sci. USA 72,4785-4789.

23. Kleid, D., Humayun, Z., Jeffrey, A. & Ptashne, M. (1976) Proc.Natl. Acad. Sci. USA 73,293-297.

24. Maquat, L. E. & Reznikoff, W. S. (1978) J. Mol. Biol. 125,467-490.

25. Calos, M. P. (1978) Nature (London) 274,762-765.

Proc. N.tl. Acad. Sci. USA 76 (1979)