J. Mol. Biol. (1995) 253, 691–702

The Sixty Nucleotide OccR Operator Contains aSubsite Essential and Sufficient for OccR Binding anda Second Subsite Required for Ligand-responsiveDNA Bending

Lu Wang and Stephen C. Winans*

OccR is a transcriptional regulatory protein of Agrobacterium tumefaciens thatSection of Microbiologyactivates the occQ operon in response to octopine, an arginine derivativeCornell University, Ithacareleased from plant tumors. OccR binds to its operator with similar affinityNew York 14853, USAand the same stoichiometry in the presence or absence of octopine, butoctopine shortens the protein’s DNase I footprint and partially relaxes anOccR-incited DNA bend. In this study, resections and other alterations ofthe operator were used to demonstrate that 19 nucleotides near the end ofthe operator furthest from the occQ promoter were essential for high affinityOccR binding. This sequence, denoted the high affinity subsite, wassufficient for binding, provided that the deleted operator sequences werereplaced with vector-derived DNA. The same number of OccR monomersbound to resected operators as to the wild-type operator, and OccR was ableto protect vector-derived sequences adjacent to the high affinity subsite.Sequences at the promoter proximal end of the operator were requiredfor wild-type patterns of ligand-responsive DNA bending. A sequencealteration at the end furthest from the high affinity subsite caused a partiallylocked low angle DNA bend, while two more centrally localized mutationscaused fully or partially locked high angle bends. This suggests that thepromoter proximal half of the operator may contain at least two sitesrequired for wild-type ligand-responsive DNA bending. These mutationsalso provided evidence that the partial relaxation of this bend by octopinemay be essential for occQ activation.

7 1995 Academic Press Limited

*Corresponding author Keywords: OccR; operator; octopine; LysR-type regulator; DNA bending

Introduction

The activities of many prokaryotic transcriptionalregulators are directly modulated by low molecularmass ligands. Although most of these ligands causelarge changes in the affinity of the protein for itsoperator, some function in other ways. For example,the Escherichia coli MerR protein incites a mercury-responsive DNA twist and DNA bend, while theE. coli AraC protein causes an arabinose-sensitiveDNA loop that prevents transcription (Ansari et al.,1992, 1995; Lobell & Schleif, 1990). We have pre-viously shown that the OccR protein of Agrobac-terium tumefaciens causes a high angle bend at itsbinding site, and that this bend angle is decreased bythe addition of the plant released compound octo-pine (Wang et al., 1992). Octopine also shortens theregion protected from DNase I by one helical turn.However, octopine has a minimal effect on operator

affinity and no effect on the oligomeric state ofbound OccR.

OccR, a LysR-type transcriptional regulator, acti-vates transcription of the 14 kb occQ operon encod-ing the proteins required for octopine uptake andcatabolism (Valdivia et al., 1991), a putative oligo-peptide permease (Fuqua and Winans, unpublishedresults), and the regulatory protein TraR, whichactivates the transcription of Ti plasmid conjugaltransfer genes (Fuqua & Winans, 1994; Fuqua et al.,1994). OccR also represses the transcription of itsown gene, occR, in both the presence and absence ofoctopine (Habeeb et al., 1991; Wang et al., 1992).

The region of DNA protected by OccR fromDNase I digestion is 60 bases long in the absenceof octopine and 50 bases long in the presence ofoctopine, much longer than the sequences protectedby most other transcriptional regulators (Collado-Vides et al., 1991), but similar in length to the foot-

0022–2836/95/450691–12 $12.00/0 7 1995 Academic Press Limited

OccR Operator692

prints of several other LysR-type regulators (Schell,1993). By resecting this operator from each end andcreating internal alterations, we were able to identifya region essential for OccR binding and a secondregion dispensable for binding but required forligand-induced changes in DNA bending and foot-print length. By using altered operators to expressoccQ fusions, we obtained evidence that relaxation ofthe high angle DNA bend by octopine may be anessential step in transcriptional activation.

Results

Localization of minimal DNA sequencerequired for OccR binding

We previously subcloned an 89 nucleotide frag-ment containing the OccR binding site, creatingplasmid pLW103 (Figure 1). This fragment boundOccR with a Kd of 1.3 nM in the absence of octopineand 2.4 nM in the presence of octopine (Figures 1and 2A), affinities very similar to those of a muchlarger fragment (Wang et al., 1992). To furtherlocalize the DNA sequence required for binding, weused PCR amplification to resect the promoterproximal end of this fragment. Fragments of plas-mids pLW104, pLW113, pLW114, pLW105, pLW106,pLW112 and pLW120 bound OccR, although thefragment cloned in pLW120 bound OccR only

weakly (Figure 3B and C and data not shown).Operator DNA in these experiments is flanked by50 to 100 nucleotides of vector-derived sequences.These results (summarized in Figure 1A) indicatethat the sequence between bp −55 and bp −86upstream of the occQ transcription start site is suf-ficient for binding, while a sequence in the intervalbetween bp −55 and bp −59 is essential for highaffinity binding.

Similar resections were made from the promoterdistal end of this region (Figure 1B). Fragments fromthe longest of these, pLW107, still bound OccRprotein (Figure 3D), while two other deletion deriva-tives, pLW119 and pLW117, did not detectably bindOccR (Figure 3E and data not shown; summarizedin Figure 1B). These results indicate that the se-quence between bp +4 and bp −73 of occQ is suf-ficient for OccR binding, and that some sequencebetween bp −69 and bp −73 is essential.

The intergenic region cloned in pLW107 wasfurther resected, resulting in pLW108, pLW115,pLW116, pLW109, pLW110, and pLW111 (Figure 1C).Fragments from all these plasmids allowed OccRbinding (Figure 3F and data not shown). Thesmallest of these plasmids, pLW111, contains a dyadsymmetry between nucleotide −57 and nucleotide−71 (ATAAN7TTAT) with two additional nucleotideson each side. To determine whether this dyadsymmetrical sequence is sufficient for OccR binding,

Figure 1. PCR-generated deletions of the occQ–occR intergenic region. A, Deletions of pLW103 from the left end.B, Deletions of pLW103 from the right end. C, Deletions of pLW107 from the left end. The ability of OccR to bind andbend each fragment in the absence and presence of octopine is shown on the right. +, high affinity binding; +/−, lowaffinity binding; nd, non-detectable. Hill coefficients were calculated from gel shift experiments conducted in the absenceof octopine. DNase I footprints of the top strand of the wild-type operator in the absence (top) or presence (bottom) ofoctopine are indicated above the DNA sequences. DNase I protected and sensitive sites are indicated by horizontal barsand filled triangles, respectively. nt, not tested.

OccR Operator 693

Figure 2. Binding affinity of OccR for promoter de-letions. Plasmids pLW103, pLW107, and pLW111 (A to C)were digested and end-labelled as described in Materialsand Methods. 10−11 M of each 32P-labelled fragment wasused for OccR binding reactions. To plasmids pLW103 andpLW107 (A and B, respectively), OccR monomers wereadded at 0 nM (lanes 1 and 8), 0.25 nM (lanes 2 and 9),0.5 nM (lanes 3 and 10), 1 nM (lanes 4 and 11), 2 nM (lanes5 and 12), 4 nM (lanes 6 and 13), 8 nM (lanes 7 and 14).To plasmid pLW111 (C), OccR monomers were added at0 nM (lanes 1 and 8), 0.5 nM (lanes 2 and 9), 1 nM (lanes3 and 10), 2 nM (lanes 4 and 11), 4 nM (lanes 5 and 12),8 nM (lanes 6 and 13) and 16 nM (lanes 7 and 14). A 40 mmquantity of octopine was added to samples in lanes 8 to14, while no octopine was added in lanes 1 to 7.

Figure 3. Gel mobility shift assays of deletion deriva-tives of the occQ–occR intergenic region. PlasmidspLW103, pLW112, pLW120, pLW107, pLW119, pLW111,and pLW127 (A to G, respectively) were digested andend-labelled as described in Materials and Methods. OccRprotein was added at 0 nM (lane 1) and 200 nM (lanes 2and 3) in the presence (lanes 1 and 3) or absence (lane 2)of 40 mM octopine. The upper band is the linearized vectorpBend3.

explanation, that a decrease in stoichiometry mightbe precisely compensated by an increase in bendangle, is conceivable but improbable. This hypoth-esis is consistent with two other observations. First,the DNase I footprints of each operator were similarin length (described below). Perhaps the most con-clusive evidence was obtained by calculating theHill coefficients for binding to each operator. EachHill coefficient was approximately 2.0, in both theabsence of octopine (Figure 1) and its presence (datanot shown), indicating that, for each operator, theoligomeric state of bound OccR is 2-fold greater thanthat of unbound OccR.

OccR binding requires additional vectorDNA sequences

All the DNA fragments used for the gel mobilityshift experiments described above contained vector-derived sequences on both sides of the operators. Todetermine whether any vector-derived sequenceswere required for OccR binding, the two smallestplasmids, pLW110 and pLW111, were treated withdifferent restriction enzymes that cleave within ornear the tandemly repeated restriction endonucleasecleavage regions present in the vector. The resultingsingle-stranded termini were, in most cases, con-verted to blunt ends and radiolabelled as describedin Materials and Methods.

Vector sequences were first removed from theleft by digesting with either NruI, SspI or KpnI,each in combination with XhoI (Figure 4). Digestionof pLW110 with NruI and XhoI and digestion ofpLW110 and pLW111 with SspI and XhoI did notdemonstrably affect binding, while digestion ofpLW110 with KpnI and XhoI abolished detectablebinding (data summarized in Figure 4). Vectorsequences of pLW111 were also removed from theright using BglII or MluI individually and using acombination of SalI and EcoRI. All of the resultingfragments formed stable OccR complexes. We con-clude that a considerable length of vector-derivedsequence to the left of the high affinity subsite is

we constructed pLW127 (Figure 1C), which containsjust this sequence. Fragments derived from pLW127bound OccR only weakly (Figure 3G), indicatingthat this dyad symmetry is not sufficient for highaffinity binding. The 19 nucleotide sequence clonedinto pLW111, denoted the high affinity subsite, istherefore the smallest region demonstrated to besufficient for high affinity OccR binding.

The binding affinity of OccR for fragments clonedin plasmids pLW107, pLW108, pLW115, pLW116,pLW109, pLW110 and pLW111 were quantitated bydetermining the amount of OccR required to shift50% of the operator fragments. All tested operatorsbound OccR with similar affinities (Figure 2 anddata not shown; summarized in Figure 1C). Thus,the promoter proximal half of the operator makeslittle contribution to the affinity of OccR. In all cases,octopine caused an approximately 2-fold decrease inaffinity.

All complexes shown in Figure 3 migrated atapproximately the same rate. This provides sugges-tive evidence that the same number of OccR mono-mers were bound to each promoter. The alterative

OccR Operator694

Figure 4. Identification of vector-derived nucleotides required for OccR binding. Fragments generated by digestion ofplasmids pLW110 and pLW111 with the indicated restriction endonucleases were converted to blunt ends andradiolabelled as described in Materials and Methods. A 2500 cpm quantity of each 35S-labelled fragment was combinedwith 200 nM OccR monomers, and size-fractionated by gel electrophoresis. Operator-derived sequences are representedusing shaded boxes, while all other sequences were derived from the plasmid vector. The DNase I footprint of bothstrands of the wild-type operator in the absence of octopine is shown at the top of the figure.

required for OccR binding, while few nucleotides tothe right are required.

To identify the smallest fragment of DNA suf-ficient for OccR binding, we digested pLW110 andpLW111 with additional combinations of enzymes.Digestion of these two plasmids with SspI andMluI or with SspI and SalI did not affect binding(Figure 4). However, digestion of pLW110 with com-binations of either KpnI and MluI or KpnI and SalIabolished binding. Therefore, the 58 bp fragmentmade by treating pLW111 with SspI, SalI (and DNApolymerase) is the smallest fragment observed tobind OccR. We conclude that, while specific operatorsequences in the left half of the wild-type operatorare not required for OccR binding, non-specificDNA sequences throughout the entire protectedregion are essential for binding.

All resected operators are bent by OccR

We determined whether the plasmids describedin Figure 1C could be bent by OccR and whetheroctopine could relax bent DNA. We first digestedeach plasmid individually with BglII, XhoI, EcoRV,SmaI, KpnI, and BamHI, and identified the fastest-migrating and slowest-migrating complexes (datanot shown). The fastest-migrating and slowest-

migrating complexes of all plasmids were thensize-fractionated on a single gel in the absence orpresence of octopine (Figure 5). All complexesshowed detectable position-dependent gel mobility(Figure 5A), indicating that each operator is bentby OccR. The center of each bend was localized (byidentifying the center of the slowest migratingfragment) just to the left of the high affinity subsite(data not shown). For most plasmids, this differencein mobility was decreased by octopine (Figure 5B),indicating a decrease in bend magnitude. Octopinehad little, if any, effect on the bend angles ofplasmids pLW109 and pLW111. The data for pLW110may be misleading, as deleted sequences may havebeen fortuitously replaced by similar vector-derivedsequences (data not shown).

Figure 5 could be interpreted to suggest a gradualloss of DNA bending as progressively greateramounts of operator DNA are removed. However,interpretation of these data is complicated by the factthat the permuted fragments liberated from oneplasmid differ from those of another plasmid bothin length and in the position of the bend centers withrespect to the termini. It is, therefore, very difficultto make comparisons between the apparent bendangles of different resected operators. We concludethat OccR bends all resected operators and that most

OccR Operator 695

Figure 5. Circular permutation assays of resected pro-moters. A 100 mM quantity of OccR protein in the absence(A) or presence (B) of 200 mM octopine was incubated withpermuted DNA fragments and size fractionated using 6%native polyacrylamide gels containing no octopine (A) or100 mM octopine (B). The DNA fragments used are, fromlane 1 to lane 16: pLW103 (BglII) and pLW103 (SmaI),pLW107 (BglII), pLW107 (SmaI), pLW108 (BamHI),pLW108 (SmaI), pLW115 (BamHI), pLW115 (EcoRV),pLW116 (BamHI), pLW116 (EcoRV), pLW109 (BamHI),pLW109 (EcoRV), pLW110 (SmaI), pLW110 (XhoI), pLW111(BglII) and pLW111 (XhoI).

binds to it asymmetrically, since it protects at least14 vector-derived nucleotides to the left, but protectsonly four vector-derived nucleotides to the right.There must be some asymmetrical sequence (eitherat the middle of this sequence or flanking it) whichis important for binding, and which positions OccRasymmetrically.

We previously reported that, in the absence ofoctopine, OccR protects a 60 bp region of DNA andthat this footprint contains two DNase I hyper-sensitive intervals. Octopine shortens the footprintto 50 bp, and attenuates hypersensitivity withinthis region (Wang et al., 1992). Plasmids pLW107,pLW108 and pLW115 showed properties very simi-lar to wild-type, while plasmids containing smallerfragments of the operator differed from wild-type,and also differ from each other in several ways(Figures 6 and 7). We conclude that some sequencebetween bp −25 and −29 is essential for wild-typeligand-responsive changes in the DNase I footprint,while no sequence downstream of −25 is required.This same interval contains a site required for wild-type high angle DNA bending in the absence ofoctopine (see below), suggesting that both responsesto octopine require this site.

Pleiotropic operator mutations preventrelaxation of the high angle DNA bend byoctopine and block transcriptional activation

We previously hypothesized that relaxation of theOccR-incited DNA bend in the occQ–occR intergenicregion might be essential for the transcriptionalactivation of occQ. Since most of the resectedoperators described above lack the occQ promoter,they were not suitable for testing this hypothesis. Wetherefore synthesized a 259 bp fragment containingthe entire intergenic region as a template for site-directed mutagenesis. We mutated three differentfour nucleotide intervals within the promoter prox-imal half of the operator (Figure 8). The wild-typefragment and these three mutated fragments werecloned into pBend3 (Kim, 1989), resulting in plas-mids pLW132, pLW142, pLW143, and pLW134.

The mobilities of the fastest-migrating andslowest-migrating complexes were used to calculateapparent bend angles (Thompson & Landy, 1988).These data probably understate the true bendangles, since none of these complexes has its bend atthe fragment’s center or at its terminus (Figure 9A).Even so, pLW132 (Figure 9B) exhibited a greaterapparent bend angle than pLW103, probably due toits having a larger insert (Thompson & Landy, 1988).All three mutations affected DNA bending (sum-marized in Figure 8). Plasmid pLW142 showed awild-type low angle bend in the presence of inducer,but failed to attain the wild-type high angle bend inthe absence of inducer (Figure 9C). In contrast,pLW143 showed a wild-type high angle DNA bendin the absence of inducer, but failed to relax thisbend in response to inducer (Figure 9D). PlasmidpLW134 showed properties similar to pLW143, inthat its bend was only partially relaxed by octopine

of these bends are detectably relaxed by octopine.Little, if any, octopine-induced relaxation wasdetected for pLW109 or pLW111.

DNase I footprinting of resected OccRbinding sites

DNase I footprinting assays were conducted withOccR and plasmids pLW107, pLW108, pLW115,pLW116, pLW109, pLW110, and pLW111. Each ofthese footprints has identical promoter distal ends(at the bottom of the gel in Figure 6; summarized inFigure 7). The promoter proximal ends show someheterogeneity, but are sufficiently similar to providesuggestive evidence that the same number of OccRmonomers bind to each fragment, a finding sup-ported above. This also indicates that OccR is fullyable to protect vector-derived sequences adjacent tothe high affinity subsite.

The plasmid containing the smallest operatorfragment (pLW111) contains only the 19 nucleotidehigh affinity subsite, which includes the dyad sym-metry ATAAN7TTAT. It is noteworthy that, despitethe dyad symmetry of this site, OccR apparently

OccR Operator696

Figure 6. DNase I protection assays of resected OccR binding sites. Plasmids pLW107, pLW108, pLW115, pLW116,pLW109, pLW110, and pLW111 (A to G) were end-labelled on the top strand and incubated without (lane 1) or with (lanes2 and 3) 50 mM OccR, in the absence (lane 2) or presence (lanes 1 and 3) of 150 mM octopine, prior to treatment withDNase I.

(Figure 9E). We conclude that particular sequenceswithin the left half of the operator are essential forwild-type levels of ligand-responsive DNA bending.

These fragments were also introduced into apromoter probe plasmid to create occQ-lacZ trans-lational fusions, and the resulting plasmids(pLW135, pLW146, pLW147, and pLW136, respect-

ively) were introduced into strains containing orlacking OccR. Alteration of bp −36 to −39 (pLW147)or of bp −44 to −47 (pLW136) abolished induction byoctopine (Table 1). This establishes a correlationbetween low angle DNA bending and transcrip-tional activation, since the two mutants that failed torelax their DNA bend by octopine also failed to

Figure 7. Summary of DNase I protection assays. The sequence of each operator deletion is shown. N indicatesvector-derived DNA sequence. DNase I cleavage pattern of free DNA, DNA plus OccR, and DNA plus OccR and octopineare shown in the first, second, and third lines, respectively. 3 represents normal cleavage by DNase I, while R representsenhanced DNA cleavage and R represents strongly enhanced DNA cleavage compared to unbound DNA.

OccR Operator 697

Figure 8. Mutations of the occQ–occR intergenic region. Altered bases are represented using black boxes. OccR-incitedbend angles are shown on the right.

be transcriptionally activated by octopine. Alterna-tively, it is possible that these transcription defectsare attributable to altered recognition of the −35region. Alteration of bp −24 to −27 upstream of theoccQ transcription start site (pLW146) had little, ifany, effect on transcriptional induction by octopine(Table 1). Therefore, the failure to attain a wild-typehigh angle bend in the absence of octopine did notlead to constitutive expression of occQ. Either theresidual high angle bend in this plasmid in theabsence of octopine was sufficient to prevent consti-tutivity, or octopine causes other conformationalchanges in OccR that cannot be simulated bymutations at its binding site.

Discussion

We set out to determine whether the exceptionallylong DNA sequence contacted by OccR couldbe divided into two or more functionally distinctsubsites. By resecting this operator from each end,we were able to divide the OccR operator into twofunctionally distinct halves. The promoter distalhalf, designated the high affinity subsite, contributesessentially all of the nucleotides required for highaffinity OccR binding. This is consistent with foot-printing experiments using hydroxyl radicals andmethidium propyl EDTA (Wang & Winans, 1995),which show that OccR protects the high affinity sub-site far more strongly than the rest of the operator.

Several dyad symmetrical sequences are found inthe OccR operator, one of which (ATAAN7TTAT) iscentered at nucleotide −64 in the center of the highaffinity subsite. This sequence contains the TN11Amotif found in the binding sites of virtually allLysR-type proteins (Goethals et al., 1992), andresembles a sequence (ATAGN7CTAT) proposed tobind the homologous OxyR protein (Toledano et al.,1994). The major grooves of both arms of thissequence are localized on the helical face contacted

by OccR (Wang & Winans, 1995). We propose thatthis sequence could play an important role in highaffinity binding. However, additional sequenceswithin this region must be involved, since, asnoted above, OccR appears to bind this regionasymmetrically.

The promoter proximal half is neither sufficientnor essential for OccR binding, but is required for thefull range of ligand-responsive DNA bending. DNAresections and nucleotide substitutions in this regionhave provided data which demonstrate that at leasttwo sequences in this region are essential forwild-type OccR–DNA interactions. The importanceof the first sequence (between −25 and −29) can beillustrated by comparing pLW115, which has thesite, to pLW116, which lacks it. These plasmids differin three respects. (1) OccR affinity is 2-fold greaterfor pLW115 than for pLW116. (2) pLW115 showedtwo DNase I hypersensitive regions in the absenceof octopine, while pLW116 showed only one. (3) TheOccR footprint of pLW115 was shortened by octo-pine, while the footprint length of pLW116 wasunaltered. The conclusion that a site between −25and −29 is required for these phenomena is stronglysupported by the properties of a four nucleotidealteration within this region (bp −24 to −27), whichalso decreased high angle DNA bending in theabsence of octopine. Other properties of thesemutated binding sites are currently being tested.

While the alteration of bp −24 to −27 decreasedhigh angle bending in the absence of octopine, twoother alterations had the opposite effect. Nucleotidechanges at bp −36 to −39 caused a fully locked highangle bend, and changes at bp −44 to −47 resultedin a partially locked high angle bend. Together,the three mutants indicate that different sites arerequired for high angle DNA bending in the absenceof octopine than those required for low anglebending in the presence of octopine. Relaxation ofthe high angle bend by octopine is therefore unlikelyto be due simply to loss of contacts with DNA, but

OccR Operator698

Figure 9. Effect of four-base changes at differentpositions of the left half of OccR operator on OccR-incitedDNA bend. A, Schematic representation of circularpermuted DNA fragments used. Thick lines represent theoccQ–occR intergenic region cloned in pBend3 plasmids.Thin lines represent vector DNA. The arrow points to theestimated bend center in each fragment. BglII, XhoI,EcoRV, or BamHI fragments (lanes 2 to 5 and lanes 7 to 10)generated from pLW132, pLW142, pLW143, and pLW134(B to E, respectively) were incubated with 100 mM OccRprotein in the absence (left panels) or presence (rightpanels) of 200 mM octopine, and size fractionated usinggels lacking (left panels) or containing (right panels)100 mM octopine.

allow OccR contacts to be altered in response tooctopine. We recently proposed a model (adaptedfrom Toledano et al., 1994) in which one dimer ofOccR binds to the high affinity subsite, while asecond dimer shifts between two binding sites in theproximal half of the operator (Wang & Winans, 1995;see Figure 10). The sequence ATTCN7TTCA (cen-tered at nucleotide −33; see Figure 10) shows weaksimilarity to the ATANN7TTAT motif. The majorgrooves of both arms of this sequence lie on thehelical face that is contacted by OccR. Furthermore,altering the left end of this sequence prevented OccRfrom forming highly bent complexes. Exactly onehelical turn to the right is found another copy of thesequence ATTC, which could conceivably act asa half site for OccR binding. Disruption of thissequence caused OccR to form highly bent lockedcomplexes. It seems plausible that OccR might bindthe ATTCN7TTCA in the absence of octopine andthat it might translocate 10 bp to the right in thepresence of octopine, binding the upstream copy ofthe sequence ATTC. We are currently altering thesesequences so that they more closely resemble theproposed OccR binding motif, and will determinewhether OccR binds irreversibly to the alteredsequences.

Few, if any, proteins homologous to OccR functionexactly the way OccR does. Most LysR-type proteinsprotect relatively long regions of DNA and activatetheir target genes in response to a diffusible induc-ing ligand (Schell, 1993), although at least twoproteins, MetR (Urbanowski & Stauffer, 1989) andNAC (Bender, 1991), protect a much shorter regionand at least one is not activated by any inducingligand (Bender, 1991). The affinity of most of theseproteins for their promoters is changed only slightlyby these inducers. In some cases, inducing ligandscause a 2-fold increase in footprint length and asevere decrease in gel mobility, indicating a changein binding stoichiometry. For example, TrpI ofPseudomonas putida protects a region from −77 to −52in the absence of its inducer, but binds from −77 to−32 in its presence (Gao & Gussin, 1991). Similarly,the CatR protein of P. putida protects its promoterfrom −79 to −53 in the absence of its inducer, but pro-tects a region from −79 to −22 in its presence (Parseket al., 1992). Like OccR, CatR incites a ligand-relaxedDNA bend at its operator (Parsek et al., 1995). In bothcases, the upstream ends of these binding sites wererequired for binding, while the downstream regionwas much less important (Gao & Gussin, 1991;Parsek et al., 1992). These high affinity binding sitesare centered at approximately nucleotide −70, andhave a dyad symmetrical sequence including theconserved TN11A motif. Similarly, the CysB proteinof Salmonella typhimurium protects the cysJIH pro-moter from −76 to −60 in the absence of inducingligand, while considerably more downstream DNAsequences are protected in the presence of ligand(Hryniewicz & Kredich, 1995).

In contrast to the proteins described above, at leastfour LysR-type proteins do not undergo a change inbinding stoichiometry when activated. These are

rather to require alternate contacts within the lefthalf of the operator.

It is unclear what sequences OccR recognizes inthe promoter proximal half of the operator. How-ever, it is plausible that these sequences may besimilar to those recognized in the high affinitysubsite. If so, the similarity should be weak, so as to

OccR Operator 699

Table 1. Induction of different fusions of occQ–lacZ by octopineb-Galactosidase specific activitya

Without WithPlasmids octopine octopine Induction ratio

pLH506b, pLW135c 0.9 21.8 24pSW208, pLW135 1.2 1.6 1.3pLH506, pLW146c 1.0 18.5 18.5pSW208, pLW146 1.3 1.5 1.3pLH506, pLW147c 0.1 0.1 1.0pSW208, pLW147 0.1 0.1 1.0pLH506, pLW136c 0.6 0.7 1.2pSW208, pLW136 0.7 0.6 0.9

a Miller units (Miller, 1972).b pLH506 is a derivative of pSW208 containing occR.c pLW135 contains the wild-type operator, while pLW146, pLW147, and

pLW136 contain the same inserts as pLW142, pLW143, and pLW134, respect-ively (Figure 8).

OccR itself, the closely related NocR protein ofA. tumefaciens (Von Lintig et al., 1994), the IlvYprotein of E. coli (Wek & Hatfield, 1988), and theOxyR protein of S. typhimurium (Storz et al., 1990).Of these, IlvY protects the same region in thepresence and absence of inducer. Like other LysRmembers, IlvY binds the upstream end of its bindingsite with high affinity, and binds the downstreamend with much lower affinity (Wek & Hatfeld, 1988).In contrast, OccR, NocR, and OxyR bind longerregions of DNA when inactive than when active.OxyR protects a region from −80 to −26 nucleotidesupstream of the oxyS promoter under noninducingconditions, and protects a region from −80 to −36under inducing conditions (Toledano et al., 1994).The longer footprint is associated with a high angleDNA bend, while the shorter footprint is associatedwith a low angle bend. These properties are ex-tremely similar to those of OccR, and the model thatwe propose (Figure 10) was adapted from a modeldescribing OxyR (Toledano et al., 1994). However,

in other respects, OxyR is quite unusual. First,OxyR binds two other target promoters onlyunder activating conditions. Second, binding ofOxyR to the oxyS–oxyR region requires the entireoperator, rather than just an upstream high affinitysubsite.

Protein-incited DNA bending is a general themeof gene regulation (Van Der Vliet & Verrijzer, 1993;Perez et al., 1994). For example, E. coli CAPprotein bends its operators (Wu & Crothers, 1984;Lichenstein et al., 1987; Schultz et al., 1991; Perez &Espinosa, 1991, 1993), and the CAP binding sitecan be functionally replaced by an intrinsicallybent DNA fragment both in vitro and in vivo, orby another protein-incited DNA bend in vivo(Gartenberg & Crothers, 1991; Bracco et al., 1989;Perez & Espinosa, 1993). Alterations in bend anglecan influence transcription initiation (Leidig et al.,1992). The mammalian transcription factor E2F andthyroid hormone receptor bend their target pro-moters, and the bend angle is partially relaxed by

Figure 10. A model for OccRactivation of the occQ promoter(adapted from Toledano et al., 1994).In the absence of octopine, OccRbinds four non-consecutive majorgrooves (top), resulting in a 60nucleotide footprint and a high anglebend. Octopine (represented byfilled dots) causes one dimer to shiftposition by one helical turn (bot-tom), shortening the DNase I foot-print and relaxing the DNA bend.OccR binds at all times to the highaffinity subsite, possibly using theATAAN7TTAT dyad symmetricalsequence. Filled bars refer to nucleo-tide pairs that lie on the helical facecontacted by OccR. Filled arrowsindicate the strong dyad symmetryat the high affinity subsite, whilebroken arrows indicate a possibleweak binding site at the left end ofthe operator. Shaded boxes indicatethe occQ promoter.

OccR Operator700

accessory regulatory proteins (Huber et al., 1994;Nevins, 1992; King et al., 1993; Lu et al., 1993).

Materials and Methods

Strains, plasmids, and reagents

E. coli strain MC4100 (F−, thiA, araD139, DargF–lacU169,rspL150, relA1, flb5301, deoC1, pstF25, rbsR) was obtainedfrom C. Manoil, University of Washington. PlasmidpSW208, previously constructed in our laboratory, con-tained the a-complementation group of pTZ18R and thecat gene and the origin of replication of pACYC 184(Habeeb, 1991; Chang & Cohen, 1978). Plasmid pBend3was obtained from S. Adhya (National Cancer Institute,Bethesda, MD). Bacteriophage T7 DNA polymerase,Sequenase, and plasmid pTZ18R were purchased fromU.S. Biochemical (Cleveland, OH). Restriction endo-nucleases, T4 DNA ligase, [a-35S]dATP and [a-32P]dATPwere purchased from New England Nuclear ResearchProducts (Boston, MA). Octopine was purchased from theAldrich Chemical Co. (Milwaukee, WI). Syntheticoligonucleotides were obtained from the Cornell Bio-technology Center. Antibiotics, X-gal and o-nitrophenylb-D-galactopyranoside were purchased from SigmaChemical Co. (St Louis, MO).

Construction of resection mutants and site directedmutants of the occQ promoter

Serial deletions of the occQ–occR intergenic region wereconstructed by PCR amplification using the followingoligonucleotides: 5'-GCATCTAGATTTATAGGGTGAGC-TTTAC-3';5' -GCATCTAGAATATGCATTCGGTCAAATT-CATAAT-3';5'-GCATCTAGACATTCGGTCAAATTCATA-AT - 3'; 5' - GCATCTAGAGGTCAAATTCATAATGACCG -3'; 5' - GCATCTAGAAATTCATAATGACCGGGCAAGAA-TA-3'; 5' -GCATCTAGAGGCAAGAATAAGCAGATGTTA-TGGT - 3'; 5' - GACTCTAGAGAATAAGCAGATGTTATGG -3'; 5' - GCAGTCGACTCAATTCATAAACGCACCA - 3'; 5' -GCAGTCGACCCATAACATCTGCTTATTC-3';5'-GCAGT-CGACACATCTGCTTATTCTTGC-3'; and 5'-GCAGTCG-ACATTCTTGCCCGGTCATTATG-3'. Each PCR productcontained an XbaI site at its left end and a SalI site at itsright end. These DNA fragments were digested with XbaIand SalI and ligated to plasmid pBend3 digested with thesame enzymes. The sequences of the inserts in theresulting plasmids were verified by DNA sequencing.

To make site-directed mutants of the OccR operator,a 259 bp PCR amplification product containing thefunctional occQ promoter was synthesized using oligo-nucleotides 5'-GCAGTCGACTCAATTCATAAACGCAC-CATAACATCTG-3' and 5'-GCATCTAGAAGCATATCATA-GCCCCA - 3'. This PCR product contains an XbaI site atits left end and a SalI site at its right end. It was cleavedwith these two enzymes, and cloned into pTZ18R cleavedwith the same enzymes. The resulting plasmid, pLW131,was introduced into strain RZ1032 (dut−, ung−) and isolatedin a single-stranded, circular, uracil-containing form(Kunkel et al., 1986). Site-directed 4 bp changes in pLW131were made using the oligonucleotides 5'-TCAAATTCAT-AAGCTTCGGGCAAGAATA-3';5'-TACATTCGATATCGC-GTCGGTCAAATTC-3'; and 5'-CATTCGGTCAACGGG-ATAATGACCGG-3'. The resulting mutants, pLW133,pLW144, and pLW145, respectively, were checked by DNAsequencing. pLW131 and these three derivatives werecleaved with XbaI and SalI, and cloned into pBend3cleaved with the same enzymes, resulting in pLW132,

pLW134, pLW142, and pLW143, respectively. pLW131 andits derivatives were also cleaved with HincII and BamHIand cloned into plasmid pMC1403 cleaved with SmaI andBamHI, resulting in pLW135, pLW136, pLW146, andpLW147, respectively. These latter four plasmids containedoccQ–lacZ translational fusions.

Gel mobility shift assays and DNase Ifootprinting assays

Plasmids used for gel mobility shift assays (Fried &Crothers, 1981) and DNase I footprinting assays (Galas &Schmitz, 1978) were digested with restriction enzymes,and the 3' single-stranded termini of the resulting frag-ments were converted to blunt ends using Sequenase andcombinations of radiolabelled and nonlabelled dNTPs.DNA fragments were phenol extracted, ethanol precipi-tated, and resuspended in Tris (pH 7.8)-EDTA buffer.

To test the ability of OccR to bind to resected operators,plasmids were digested with EcoRI and HindIII, andlabelled with 35S, as described above. A 2500 cpm quantityof DNA was incubated with 200 nM OccR mono-mers in the absence or presence of 40 mM octopine at afinal volume of 10 ml (Figure 3). To measure the bindingaffinity of OccR for the resected operators, plasmids weredigested with EcoRI and HindIII and labelled with 32P asdescribed above. Binding reactions contained 10−11 MDNA fragments, 0 mM or 40 mM octopine, and varyingconcentrations of OccR in a final volume of 10 ml. Con-ditions for OccR binding reaction and electrophoresis wereas previously described (Wang et al., 1992). Radioactivityof free DNA and OccR–DNA complexes was quantitatedby scanning gels using a Betascope 603 Blot Analyzer(Betagen Corp). Binding affinities were determined bydetermining the amount of protein required to retardthe electrophoretic mobility of 50% of the DNA. Hillcoefficients (Edelstein et al., 1986) were calculated usingthe equation:

log([OccRn DNA]/[DNA]) = n(log[OccR]) + log K

and estimating n from the slope of plotted data.To localize vector DNA sequences required for OccR

binding, plasmids pLW110 and pLW111 were digestedwith the restriction enzymes indicated in Figure 4 andlabelled with 35S, as described above. A 2500 cpm quantityof each resulting fragment was incubated with 200 nMOccR monomers, and subjected to electrophoresis using6% (w/v) native polyacrylamide gels.

DNase I footprinting assays were conducted as pre-viously described (Wang et al., 1992). Plasmids weredigested with HindIII and SacII, and end-labelled using32P as described above. 30,000 cpm of labelled DNA wasincubated with 50 mM OccR in the absence or presence of150 mM octopine at a final volume of 10 ml for 10 minutesat room temperature before being treated with 0.1 ng ofDNase I.

Circular permutation assays

Plasmids used to measure the degree of OccR-incitedDNA bend on resected operators were cleaved individu-ally using BglII, XhoI, EcoRV, SmaI, KpnI, and BamHI. Theresulting fragments were incubated with 100 mM OccRprotein in the absence or presence of 200 mM octopine at afinal volume of 10 ml, fractionated using native 6% poly-acrylamide gels lacking or containing 100 mM octopine,and stained with ethidium bromide. The fastest andslowest migrating complexes for each operator were thensize fractionated a second time using a single polyacryl-amide gel, thus avoiding errors arising from using

OccR Operator 701

multiple gels. The angle of DNA bends was estimatedusing the equation mM/mE = cos(a/2), where mM and mE

represent the gel mobilities of molecules with bends attheir centers and ends, respectively, and a is the angle bywhich the DNA departs from linearity (Thompson &Landy, 1988).

Plasmids pLW132, pLW134, pLW142, and pLW143 weretreated individually with BglII, XhoI, EcoRV, and BamHI,and the resulting fragments were subjected to circularpermutation assays. Experimental conditions were thesame as those described above, except that 5% (w/v)polyacrylamide gels were used. The reported bend anglesfor each altered operator are the average of three assays.

Induction of occQ–lacZ fusions

Derivatives of E. coli strain MC4100 containing the indi-cated plasmids were cultured overnight at 37°C in 2 ml ofAB minimal glucose medium (Chilton et al., 1974) supple-mented with 100 mg/ml of carbenicillin and 50 mg/ml ofchloramphenicol. The saturated cultures were diluted100-fold into fresh AB minimal glucose medium withoutantibiotics, and cultured at 37°C to an A600 of 0.2. Cultureswere then divided into two tubes, and octopine was addedinto one tube to a final concentration of 400 mM. b-Galac-tosidase specific activities (Miller, 1972) were measuredafter two additional hours of incubation.

AcknowledgementsWe thank Clay Fuqua, Dong Cho Han, Kyungyun Cho

and Gisela Storz for helpful discussions and suggestions.Financial support from U.S.D.A. grant 9301084, N.I.H.grant GM42893, and the NSF/DOE/USDA Plant ScienceCenter are gratefully acknowledged.

ReferencesAnsari, A. Z., Chael, M. L. & O’Halloran, T. V. (1992).

Allosteric underwinding of DNA is a critical step inpositive control of transcription by mercury-MerR.Nature, 355, 87–89.

Ansari, A. Z., Bradner, J. E. & O’Halloran, T. V. (1995).DNA-bend modulation in a repressor-to-activatorswitching mechanism. Nature, 374, 371–375.

Bender, R. A. (1991). The role of the NAC protein inthe nitrogen regulation of Klebsiella aerogenes. Mol.Microbiol. 5, 2575–2580.

Bracco, L., Kotlarz, D., Kolb, A., Diekmann, S. & Buc, H.(1989). Synthetic curved DNA sequences can act astranscriptional activators in Escherichia coli. EMBO J.8, 4289–4296.

Chang, A. C. Y. & Cohen, S. N. (1978). Construction andcharacterization of amplifiable DNA cloning vehiclesfrom the p15A cryptic miniplasmid. J. Bacteriol. 134,1141–1156.

Chilton, M. D., Currier, T. C., Farrand, S. K., Bendich, A. J.,Gordon, M. P., and Nester, E. W. (1974). AgrobacteriumDNA and PS8 bacteriophage DNA not detected incrown gall tumors. Proc. Natl Acad. Sci. USA, 71,3672–3676.

Collado-Vides, J., Magasanik, B. & Gralla, J. D. (1991).Control site location and transcriptional regulation inE. coli. Microbiol. Rev. 55, 371–394.

Edelstein, S. J., Poyart, C., Blouquit, Y. & Kister, J. (1986).Self-association of hemoglobin olympia: a human

hemoglobin bearing a substitution at the surface ofthe molecule. J. Mol. Biol. 187, 277–290.

Fried, M. & Crothers, D. M. (1981). Equilibria and kineticaof lac repressor-operator interactions by poly-acrylamide gel electrophoresis. Nucl. Acids Res. 9,6505–6525.

Fuqua, W. C. & Winans, S. C. (1994). A LuxR–LuxI typeregulatory system activates Agrobacterium Ti plasmidconjugal transfer in the presence of a plant tumormetabolite. J. Bacteriol. 176, 2796–2806.

Fuqua, W. C., Winans, S. C. & Greenberg, E. P. (1994).Quorum sensing in bacteria: The LuxR–LuxI familyof cell density-responsive transcriptional regulators.J. Bacteriol. 176, 269–275.

Galas, D. J. & Schmitz, A. (1978). DNase I footprinting:A simple method for the detection of protein DNAbinding specificity. Nucl. Acids Res. 5, 3157–3170.

Gao, J. & Gussin G. N. (1991). Mutations in TrpI bindingsite II that differentially affect activation of trpBApromoter of Pseudomonas aeruginosa. EMBO J. 10,4137–4144.

Gartenberg, M. R. & Crothers, D. M. (1991). SyntheticDNA bending sequences increase the rate of in vitrotranscription initiation at the Escherichia coli lacpromoter. J. Mol. Biol. 219, 217–230.

Goethals, K., Van, M. M. & Holsters, M. (1992). Conservedmotifs in a divergent nod box of Azorhizobiumcaulinodans ors571 reveal a common structure inpromoters regulated by LysR-type proteins. Proc.Natl Acad. Sci. USA, 89, 1646–1650.

Habeeb, L. F., Wang, L. & Winans, S. C. (1991). Tran-scription of the octopine catabolism operon of theAgrobacterium tumor-inducing plasmid pTiA6 is acti-vated by a LysR-type regulatory protein. Mol. PlantMicrobe Interact. 4, 379–385.

Hryniewicz, M. M. & Kredich, N. M. (1995). Hydroxylradical footprints and half-site arrangements of bind-ing sites for the CysB transcriptional activator ofSalmonella typhimurium. J. Bacteriol. 177, 2343–2353.

Huber, H. E., Goodhart, P. J. & Huang, P. S. (1994).Retinoblastoma protein reverses DNA bending bytranscription factor E2F. J. Biol. Chem. 269, 6999–7005.

Kim, J., Zweib, C., Wu, C. & Adhya, S. (1989). Bending ofDNA by gene-regulatory proteins: construction anduse of a DNA bending vector. Gene, 85, 15–24.

King, I. N., De Soyza, T., Catanzard, D. F. & Lavin, T. N.(1993). Thyroid hormone receptor-induced bendingof specific DNA sequence is modified by an accessaryfactor. J. Biol. Chem. 268, 495–501.

Kunkel, T. A, Roberts, J. D. & Zakoor, R. A. (1986).Rapid and efficient site-directed mutagenesiswithout phenotypic selection. Methods Enzymol. 154,367–382.

Leidig, F., Shepard, A. R., Zhang, W. G., Stelter, A. &Caffni, P. A. (1992). Thyroid hormone responsivenessin human growth hormone-related genes. J. Biol.Chem. 267, 913–921.

Lichenstein, H. S., Hamilton, E. P. & Lee, N. (1987).Repression and catabolite gene activation in theara-bad operon. J. Bacteriol. 169, 811–822.

Lobell, R. B. & Schleif, R. F. (1990). DNA looping andunlooping by AraC protein. Science, 250, 528–532.

Lu, X. P., Eberhard, N. L. & Pfahl, M. (1993). DNAbending by retinoid × receptor-containing retinoidand thyroid hormone receptor complexes. Mol. Cell.Biol. 13, 6509–6519.

Miller, J. H. (1972). Experiments in Molecular Genetics, ColdSpring Harbor Laboratory Press, Cold Spring Harbor,NY.

OccR Operator702

Nevins, J. R. (1992). E2F: A link between the Rb tumorsuppressor protein and viral oncoproteins. Science,258, 424–429.

Parsek, M. R., Shinabarger, D. L., Rothmel, R. K. &Chakrabarty, A. M. (1992). Roles of CatR and cis-cismuconate in activation of the catBC operon which isinvolved in benzoate degradation in Pseudomonasaputida. J. Bacteriol. 174, 7798–7806.

Parsek, M. R., Kivisaar, M. & Chakrabarty, A. M.(1995). Differential DNA bending introduced by thePseudomonas putida LysR-type regulator, CatR, atthe plasmid-borne pheBA and chromosomal catBCpromoters. Mol. Microbiol. 15, 819–828.

Perez, M. J. & Espinosa, M. (1991). The RepA repressor canact as a transcriptional activator by inducing DNAbends. EMBO J. 10, 1375–1382.

Perez, M. J. & Espinosa, M. (1993). Protein-inducedbending as a transcriptional switch. Science, 260,805–807.

Perez, M. J., Rojo, F. & De Lorenzo, V. (1994). Promotersresponsive to DNA bending: A common themein prokaryotic gene expression. Microbiol. Rev. 58,268–290.

Schell, M. A. (1993). Molecular biology of the LysR familyof transcriptional regulators. Annu. Rev. Microbiol. 47,597–626.

Schultz, S. C., Shields, G. C. & Steitz, T. A. (1991). Crystalstructure of a CAP DNA complex: The DNA is bentby 90 degrees. Science, 253, 1001–1007.

Storz, G., Tartaglia, L. A. & Ames, B. N. (1990). Tran-scriptional regulator of oxidative stress-induciblegenes: Direction activation by oxidation. Science, 248,189–194.

Thompson, J. F. & Landy, A. (1988). Empirical estimationof protein-induced DNA bending angles: Appli-cations to lambda site-specific recombination com-plexes. Nucl. Acids Res. 16, 9687–9705.

Toledano, M. B., Kulik, I., Trinh, F., Baird, P. T., Schneider,T. D. & Storz, G. (1994). Redox-dependent shift ofOxyR–DNA contacts along an extended DNA bindingsite: A mechanism for differential promoter selection.Cell, 78, 897–909.

Urbanowski, M. L. & Stauffer, G. V. (1989). Genetic andbiochemistry analysis of the metR activator-bindingsite in the metE–metR control region of Salmonellatyphimurium. J. Bacteriol. 171, 5620–5629.

Valdivia, R. H., Wang, L. & Winans, S. C. (1991). Char-acterization of a putative periplasmic transportsystem for octopine accumulation encoded by Agro-bacterium tumefaciens Ti plasmid pTiA6. J. Bacteriol.173, 6398–6405.

Van Der Vliet, P. C. & Verrijzer, C. P. (1993). Bending ofDNA by transcription factors. Bioessays, 15, 25–32.

Von Lintig, J., Kreusch, D. & Schroder, J. (1994). Opine-regulated promoters and LysR-type regulators in thenopaline (noc) and octopine (occ) catabolic regions ofTi plasmids of Agrobacterium tumefaciens J. Bacteriol.176, 495–503.

Wang, L. & Winans, S. C. (1995). High angle and ligandinduced low angle DNA bends incited by OccR lie inthe same plane with OccR bound to the interior angle.J. Mol. Biol., 253, 32–38.

Wang, L., Helmann, J. D. & Winans, S. C. (1992). TheAgrobacterium tumefaciens transcriptional activatorOccR causes a bend at a target promoter, which ispartially relaxed by a plant tumor metabolite. Cell, 69,659–667.

Wek, R. C. & Hatfeld, G. W. (1988). Transcriptionalactivation at adjacent operators in the divergent-overlapping ilvY and ilvC promoters of Escherichiacoli. J. Mol. Biol. 203, 643–663.

Wu, H. M. & Crothers, D. M. (1984). The locus ofsequence-directed and protein induced DNAbending. Nature, 308, 509–513.

Edited by R. Schleif

(Received 7 June 1995; accepted 15 August 1995)