Structural Analysis of the Purine Repressor, an Escherichia coli DNA-binding Protein*

The purine repressor protein, PurR, is a member of the lac repressor, LacI, family of Escherichia coli DNA-binding proteins that bind DNA via a highly conserved N-terminal helix-turn-helix motif. Addi-tionally, the members of this family display strong sequence homologies between their larger C-terminal effector binding/oligomerization domains. Analysis of the PurR primary and secondary structures reveals that its C-terminal corepressor binding domain is highly homologous to another group of E. coli-binding proteins, the periplasmic binding proteins, particularly to the ribose-binding protein (RBP). The high resolution x-ray structure of RBP allows this protein to serve as a template with which to model the predicted secondary structure of the corepressor binding domain of PurR. Similarly, the N-terminal DNA binding domain of PurR can be modeled using the NMR- determined structure of the corresponding region (res-idues 1-56) from LacI as a template. Combining the two, results in a description of the likely secondary structure topology of PurR and implicates residues important for corepressor binding and dimerization. CD spectroscopic studies on PurR, its corepressor binding domain and RBP result in secondary structure estimates nearly identical with those obtained by se- quence analyses, thereby providing further corroborating physical

PurR participates in the regulation of five other genes involved in pyrimidine biosynthesis, pyrimidine salvage, and the generation of one-carbon units (Zalkin and Dixon, 1992) and is autoregulated (Rolfes and Zalkin, 1990a). PurR is activated to bind its cognate DNA by binding the purine corepressors, hypoxanthine or guanine, thereby leading to repression of the Pur regulon (Rolfes and Zalkin, 1990b).
Homology between this family of DNA-binding proteins and another group of E. coli binding proteins, the periplasmic binding proteins, has been noted (Muller-Hill, 1983;Vartak et al., 1991;Mauzy and Hermodson, 199213;Weickert and Adhya, 1992). This homology exists despite lack of significant sequence identity. Of particular interest to this work is the recent study by Bowie et al. that suggested a striking structural homology between PurR and RBP (Bowie et al., 1991). The periplasmic binding proteins, including RBP, consist of a small N-terminal signal sequence, which targets these proteins to the bacterial inner membrane and is subsequently cleaved, and a larger C-terminal metabolite binding domain. The metabolite binding domain itself is further divided into two domains. Importantly, the structures of several of these proteins have been solved to high resolution by x-ray crystallography, showing them to have a highly conserved P/a type structure Mowbray and Cole, 1992). The homology between the periplasmic proteins and the LacI superfamily is found within the effector binding domains of the LacI members and the metabolite binding domains of the periplasmic proteins. Because the structures of several of the periplasmic proteins are known, they can potentially be used as scaffolds upon which t o model the structures of the effector binding domains of LacI members. To further our structural understanding of PurR and its structural relationship to RBP, we have carried out computer-aided sequence homology studies and secondary structure analyses and CD spectroscopic studies on PurR, its corepressor binding domain, and RBP. A complete secondary structure topology of PurR is presented and its functional ramifications discussed.

EXPERIMENTAL PROCEDURES
Sequence Homology-A sequence identity comparison program (Ohlendorf et al., 1983;Brennan et al., 1986) was used to analyze the primary structures of GBP-E, RBP, PurR, LacI, and COR. The secondary structure prediction of PurR was carried out using the method of Wilmot and Thorton (1988). Sequence alignments were made based on the algorithm of Feng and Doolittle (1990).
The final, "knowledge-based alignment of PurR was carried out in five steps. First, as a control, the secondary structures of GBP-E and RBP were predicted by using the above described methods and compared with their x-ray-determined secondary structures. The secondary structure of PurR was simultaneously predicted by this method. The three sequences were subsequently aligned employing the algorithm of progressive alignments of Feng and Doolittle. The final alignment of PurR was refined and confirmed by calculating the significance of the alignment via direct amino acid comparisons (DAAC) and minimum base change per codon (MBC/C) comparisons between eight variable-length stretches of RBP and CBD (Ohlendorf et al., 1983;Brennan et al., 1986) and by using the recent sequence alignment of RBP against the E. coli and Salmonella typhimurium GBP (Mowbray, 1992). To complete the topological analysis, the NMR-determined secondary structure of the LacI headpiece region (Kaptein et al., 1985) was used to model the secondary structure of the N-terminal region of PurR. CD Spectroscopy-CD spectra of CBD, PurR, and RBP were taken on a JASCO J-500A spectrophotometer. Measurements were made using a 0.1-mm path length cell (Helma) thermostatted cell at room temperature. The instrument was calibrated by using (+)-10-camphorsulfonic acid (Ae = +2.37 M" cm" at 290.5 nm and -4.95 at 192.5 nm). Data were collected on an IBM/PC-XT using the IF-500 interface and software provided by JASCO. Spectra and buffer base lines were the average of four to eight scans each recorded at 0.1-nm intervals, using a scanning rate of 5 nm/min and a 4-s time constant taken at room temperature. The buffer used for the spectral measurements of CBD and RBP was 10 mM potassium phosphate, pH 7.5, and because of decreased solubility at low ionic strength, 100 mM potassium phosphate, pH 7.5, for PurR. All proteins were purified as described previously (Choi and Zalkin, 1992; Mowbray and Cole, 1992). The protein concentrations were determined by amino acid analysis and were -0.5-1.0 mg/ml. Before spectral deconvolution for secondary structure analysis, the buffer base line was subtracted, and the resulting spectrum was smoothed using the smoothing program provided by JASCO. The CD spectra for each protein were deconvoluted for secondary structure content using the singular-value and variable selection methods described in detail elsewhere (Compton et al., 1987). For this analysis combinations of 11 of the 22 basis spectra are used to find those that result in the best fit using the criteria described elsewhere (Compton et al., 1987). All secondary structure values resulting from each combination which met these criteria were averaged to give the final secondary structure values for each experimental spectra.

RESULTS AND DISCUSSION
Corepressor Binding Domain-The analysis of PurR was undertaken to establish the extent of this protein's homology to the PBP and to determine which P B P family it most closely resembles. The periplasmic binding proteins contain several hallmark features that are highly conserved among its members (Spurlino et al., 1991). All mature PBP, i.e. after signal sequence cleavage, are divided into two structural domains, an N terminus-containing domain and a C terminuscontaining domain. Especially striking is the finding that in all periplasmic binding proteins, the x-ray structures of which have been determined, the first six secondary structural elements, which comprise approximately 100 residues, have the identical topological arrangement, ~A U & ( Y I I~C ( Y I I I Mowbray and Cole, 1992). The Nand C-terminal domains are connected by two or three peptide crossovers, which although not sequentially close, are structurally close. These peptide crossovers have been used to delineate members of the periplasmic proteins into two families and indicate the secondary structures being connected. The first family, the ABP family, includes RBP, ABP, GBP-S, and GBP-E and exhibits p -m crossovers for the first two crossovers and a P-$ crossover for the third. The second family, the SBP family, includes SBP and maltose-binding protein and exhibits p-p crossovers for the first two crossovers and an a+ a crossover as the last.
The structural analysis of PurR commenced by initially focusing on the corepressor binding domain. Primary and secondary structure analyses and comparisons were carried out against selected PBP, the high resolution x-ray structures of which are known, and followed by DAAC and MBC/C significance analyses ( Fig. 1 and Table I). Interestingly, the secondary structure predicted for CBD is that of the repeating P/a structure, with similarly located and sized helices and psheets as seen in the periplasmic binding proteins (Fig. 1). On the basis of all independent and corroborating methods used, it became evident that the first six secondary structural elements of CBD follow the ~A C Y I P B~I I P C~I I I topology of the PBP. These elements begin with residue 61 of PurR and end at residue 137. Thus, they are contained within 100 residues as observed in the periplasmic proteins. Secondary structural analysis reveals a second similar C-terminal motif which spans pFaVpGaVIpHaVII. An analogous p/a motif is found in the PBP as well. Interestingly, self-alignment of CBD sequences 60-183 against 184-318 reveals little sequence identity (-12%). This is consistent with similar internal comparisons made between domains within the periplasmic binding proteins in which sequence identities are only -14%. Indeed, sequence identity was shown to be greater between the corresponding domains of RBP and GBP-E (-24%) (Mowbray, 1992) and RBP and CBD (-24%) (this study).
To determine which family of peripl-asmic proteins PurR most closely resembles, the potential crossover regions need to be identified. This presented a problem as the secondary structure prediction data were ambiguous in the regions in which the crossovers were likely to occur. This problem was first addressed by studying the alignment data. Here it was noticed that, whereas the sequence of PurR aligns well with SBP family members within the N-terminal subdomain (data not shown), it aligns well with ABP family members, especially GBP-E and RBP, throughout its entire sequence, including the crossover regions ( Fig. 1). For example, the helix in the first P-a crossover in RBP and GBP-E contains several highly conserved residues.
In particular, residues Gly'09, Gly113, and Ile'16 of RBP correspond directly to Gly116, Gly'", and IlelZ3 of GBP-E. PurR has the corresponding glycines at positions 166 and 170 and a conservative change at position 173, a leucine for isoleucine ( Fig.  1). Further comparison of PurR with RBP reveals even more striking identity within this region in that Gly"j5, Ala'69, and Tyr17' of PurR correspond directly to Gly108, Ala"', and Tyr115 of RBP. Additionally, several other conservative substitutions within the RBP, GBP-E, and PurR sequences are found within this region (Fig. 1). Together, these identities strongly implicate this region as being the first crossover helix in PurR. Further corroborating this assignment is the secondary structure analysis which strongly predicts residues 162-176 of PurR to PurR:

-G E --C A K V I E L P G 1 A C T S A A R Q R G E G F Q O A V A A H --K F Y V L A S O P A D F D M l K G L N v M a~L~
GEP-E:135   ' The DAAC disagreement score for two random sequences is 0.94.

*The MBC/C disagreement score
for two random sequences is 1.44. be helical (Fig. 1). On the basis of this assignment, PurR belongs to, or is closely related to, members of the ABP family. Therefore, a second ,f3+a crossover should be located near residue 292 in PurR after &strand J (Fig. 1). In accordance with this supposition CBD residues 300-311 are predicted to be helical.
The analysis of PurR's third crossover was aided greatly by the recent x-ray structure of RBP (Mowbray and Cole, 1992).
Corresponding PurR residues 318-321 and 324-326 are predicted to be p-strands by alignment and sequence identity data (Fig. 1). Additional evidence implicating PurR residues 318-326 as the third crossover region is the near identity between the sequences of residues 289-291 and 324-326 of PurR and 232-234 and 266-268 of RBP, respectively. In RBP, residues 232-234 are Thr-Ile-Ala and residues 266-268 are Lys-Leu-Val, whereas in PurR the corresponding residues are Thr-Ile-His and Arg-Leu-Ile. The finding that PurR exhibits such similarity to RBP in these two distant regions, which form the only antiparallel p-sheet in RBP, suggests strongly that it contains the same secondary structure.
Further support for CBD's striking similarity to RBP and other ABP family members comes from the recent structural comparison of the high resolution structures of RBP, GBP-S, and GBP-E (Mowbray, 1992). This study identified several conserved residues as being key structural elements, whereby the regions of highest sequence identity are those involved in forming the hydrophobic cores of these proteins. Remarkably, these same regions are also the most conserved between RBP and PurR (Fig. 1). For example, PurR residues Ile6', Leug2, Leu"', Metz3', and correspond directly to RBP residues Ile4, Leu34, Leu6', Met'73, and Leu'77, respectively. Importantly, these identities span both the Nand C-terminal domains. Conserved residues in RBP and GBP-E were also noted as forming unusual and highly specific interactions within these proteins. For example, Asp"' in RBP and Aspz1' in GBP-E, located within helices adjacent to a ligand binding site residue, are buried within the protein. Critical to the stabilization of this buried aspartate are hydrogen bonds to main chain amide nitrogens from a nearby loop and to the side chain of ThrZ3' in RBP and T h P 3 in GBP-E. The presence of this interaction in PurR is strongly implicated by the presence of the corresponding pair Asp24s and ThrZS9. Furthermore, Aspz4' is predicted to be in a helix (Fig. 1).
Other striking amino acid conservations include several glycines that in RBP and GBP-E are C-CAP residues (Richardson and Richardson, 1988). In PurR two of these glycines are Gly" and G1yz6' and are predicted to be near the Cterminal end of helices (Fig. 1). An unusual interaction in RBP is a 1-3 hydrogen bonding interaction involving residues 88-90, which contains a central aspartate residue, AspSg, that is also involved in sugar binding. The equivalent position in PurR is pointing out again the extraordinary conservation between PurR and RBP in residues known to be structurally important in RBP, despite the lack of significant global sequence identity between these two proteins. The main differences between RBP and GBP-E have been noted as occurring in regions in which GBP-E binds Ca2+ (Mowbray et al., 1990;Vyas et al., 1991). Since RBP does not bind Ca2+ the corresponding sequences in RBP contain deletions within the alignment (Mowbray and Cole 1992). Gaps also occur in these regions for PurR, indicating, as expected, that there are no Ca2+ binding sites within PurR (Fig. 1).
T o provide statistical verification for the remarkable similarity between RBP and CBD, DAAC and MBC/C analyses were carried out by comparing various stretches of RBP against the entire sequence of CBD (Ohlendorf et al., 1983;Brennan et al., 1986). The results of DAAC and MBC/C comparisons between eight variable length segments of RBP and the entire CBD are shown in Table I and reveal overall 27.3% amino acid sequence identity. These segments were chosen around the few gaps and insertions that were necessary to align the two sequences optimally. Overall, in both DAAC and MBC/C analyses, the better scores are those obtained from comparisons between the N-terminal regions, with the lower scores corresponding to the C-terminal region. The significance scores (Table I, fourth column) for the DAAC analysis for the eight aligned pairs of sequences of RBP versus CBD are 4.21, 4.10, 4.56, 3.96, 5.07, 5.23, 3.86, and 3.99 and underscore the very strong homologies between these segments. Remarkably, the best scores were found in regions in which the predicted secondary structure of PurR matches exactly the known secondary structure of RBP. The values of the given MBC/C also support this homology with significance values of 4.72, 3.62, 3.78, 4.21, 3.00, 2.66, 3.39 and 2.88 (Table I, sixth column). However, in the MBC/C analyses, two segments of RBP (residues 191-230 and 251-258) score better with regions of PurR other than the aligned stretches shown in Fig. 1. The former segment, which contains the most divergent residues between RBP and PurR, residues 218-230 and 275-287, respectively, corresponds to a region in LacI and CytR which has been implicated in dimerization (Daly and Matthews, 1986;Chakerian and Matthews, 1991;Barbier and Short, 1992;Weickert and Adhya, 1992) (Fig. 2). This being the case, one might expect this region to diverge structurally between PurR and the monomeric PBP. In ac-cordance with this supposition, PurR residues 275-287 are predicted to be aperiodic unlike the corresponding regions in RBP and GBP-E which form helix VIII.
The above analyses have been combined to generate the alignment seen in Fig. 1. It should be emphasized that this is the best alignment achievable based on predictive and comparative methods, and therefore the beginning and ends of each predicted structural element are subject to error. Only with the x-ray structure in hand will the exact secondary, tertiary, and quaternary structures be known. However, it can be seen that the predicted secondary structural elements of CBD align very well with those observed in the x-ray structures of RBP and to a significant but lesser extent with those of GBP-E, thereby providing strong support that these proteins are structurally very similar.
Ligand-binding Residues-A remarkable feature displayed by RBP and GBP-E, as well as other PBP, is the conservation of those residues involved in ligand binding, despite the differences in ligand specificity Mowbray and Cole, 1992). Specificity is attained almost entirely by direct protein-ligand hydrogen bonding. However, the RBP and GBP-E ligand complexes are further stabilized by stacking interactions (Vyas, 1991;Mowbray and Cole, 1992). This type of interaction is expected to be even more significant in PurR's interaction with its planar, aromatic ligands, hypoxanthine and guanine. The residues of the PBP involved in ligand binding are widely dispersed throughout their primary structures making it more difficult to predict the functionally analogous residues in PurR. However, the alignment of PurR with RBP and GBP-E clearly suggests residues that form, in part, the purine binding pocket. That is to say, that although specific PurR-ligand contacts cannot be predicted, by virtue of locating probable loop locations, the site of PBP-ligand interaction, certain PurR residues can be implicated in ligand binding.
Specific examples include the three polar and planar groups that have been shown to be critical in protein-ligand hydrogen bonding interactions on GBP-E and RBP (Vyas, 1991;Mowbray and Cole, 1992). In RBP these are Asp", Arg14', and Asp2I5, and Amg1, Arg15', and in GBP-E (Table  11). The corresponding PurR residues are Asp'46, Arglg6 and Asp275. In RBP and GBP-E these amino acids make extensive cooperative and bidentate H-bonds with other protein residues as well as the ligand. Stacking interactions have also been shown to be important in RBP and GBP-E, acting to

LaCI:
1 CyfR:   "sandwich" the sugar ligands. For the purine repressor with its planar aromatic ligands, hypoxanthine and guanine, such stacking interactions will likely be of greater importance in the formation of a stable protein-ligand complex. Indeed, PurR residues Tyr73 and PheZz1 clearly correspond to stacking residues Phe" and TrpIE3 in GBP-E and Phe" and Phe'64 in RBP (Fig. 1). Phe" of RBP is also involved in stacking and a analogous stacking interaction could be made by the homologous PurR residue, Phe74. Another important PBP-ligand interaction is the hydrogen bond formed between Ar? (RBP) and Lysg2 (GBP-E) and their respective carbohydrates. Interestingly, the corresponding residue in PurR is It is possible that in PurR this Trp provides an additional stacking interaction or is involved in ligand H-bonding through its indole ring N,H.

PurR:305 L L D R I V N K R E E P P S I E V H P R L l E R R S V A D G P F R D V R R Lacl:SO4 L L O L S P G P A V -K G Y P L L P V S L V K R K T l L A P N ------l O l A S P R A L A D S L M O L A R P V S R L E S G
Other residues in PurR, which correspond to ligand binding residues in GBP-E and RBP, include Ala7' and Gly247. In GBP-E and RBP, the corresponding residues are polar and planar, either aspartates or aspargines, and function in Hbonding and van der Waals interactions. It is possible that in PurR, these small, nonpolar residues exist in regions within the binding cleft where small residues are needed to accommodate the larger, more hydrophobic, planar purines. A summary of possible ligand binding residues is presented in Table  11. DNA Binding Domain-Upon binding corepressor, the DNA binding domain of PurR is activated to bind to its cognate DNA (Rolfes and Zalkin, 1990b). As described above, members of the LacI family display a high degree of sequence homology throughout their sequences with their DNA binding domains particularly well conserved (Weickert and Adhya, 1992). Sequence alignment of PurR with LacI reveals almost 50% identity within this region (residues 1-60) (Fig. 2). NMR studies on a LacI DNA binding domain (residues 1-51) confirm the presence of a helix-turn-helix within residues 6-25 and also identify a third helix from residues 34 to 45 after an extended loop (Kaptein et al., 1985). The corresponding residues of the DNA binding domain of PurR, residues 4-23 and 32-43, respectively, display greater than 55% sequence identity and 68% similarity when considering conservative changes, strongly suggesting the presence of a similar helixturn-helix-loop-helix motif. An additional 9-residue stretch of PurR, residues 52-60, is predicted to contain a small helix. However, no structural data have been presented for this final stretch in any LacI family member.
Subunit Interactions-PurR, like most members of the LacI family and unlike members of the periplasmic proteins, exists as a dimer (Choi and Zalkin, 1992). Those residues of PurR involved in the subunit interface can be surmised, in part, by previous mutagenesis studies on LacI and CytR, which implicate residues 269-291 and residues 288-310, respectively, as critical in subunit interaction (Daly and Matthews, 1986;Chakerian and Matthews, 1991;Barbier and Short, 1992). The sequences in these regions diverge most dramatically from the PBP and are predicted to be loops instead of a helix as observed in the PBP Mowbray, 1992). However, this region does display significant sequence identity among all LacI family members, over which PurR, LacI, and CytR are -60% identical (Fig. 2). The sequence homology and predicted structural differences between the PBP and LacI proteins are consistent with this region's involvement in oligomerization.
Support for this assertion is provided by the observation that one residue within this region, TyrzEz in LacI and the directly analogous residue in CytR, Cys2", has been shown to be essential for dimerization (Chakerian and Matthews, 1991;Barbier and Short, 1992). Mutation of LacI Tyr2'* to any amino acid other than phenylalanine or leucine abolishes dimerization, implying that a large hydrophobic residue is required in this position for effective subunit interaction (Chakerian and Matthews, 1991). In PurR, the corresponding residue, Phe283, is also a large hydrophobic residue and presumably plays a like role. Similarly, substitutions for CysZE9 in CytR results in the production of only monomeric protein (Barbier and Short, 1992). In LacI, Cys281 has also been implicated as important in subunit interaction. Substitution of Cyszs' with virtually any amino acid does not affect dimerization but does influence inducer affinity and cooperativity (Chakerian and Matthews, 1991). It is interesting to speculate that the corresponding residues in PurR and CytR, Tyr282 and Phe2=, respectively, play analogous roles in ligand binding and subunit cooperativity and that their aromatic nature reflects their aromatic ligands.
CD Analysis-The above described primary and secondary structure analyses strongly suggest that the corepressor binding domain of PurR has a @/cy type structure and fold very similar to that of RBP. When combined with NMR studies on the LacI DNA binding domain, these analyses also clearly indicate that PurRs N-terminal DNA binding domain has a closely related helix-turn-helix-loop-helix structure. However, they provide no direct physical evidence. Perhaps one of the best physical methods available for the examination of protein structure, exclusive of x-ray crystallography and NMR, is circular dichroism spectroscopy. By measuring primarily the amide chromophore, CD spectroscopy is exquisitely sensitive to a protein's secondary structure (Johnson, 1990). If, indeed, the CBD and RBP have similar secondary structures, their CD spectra should reveal this. Additionally, the intact form of PurR should also produce a similar spectrum, albeit with a greater helical content than the CBD and a slightly lower helical content than RBP. This difference will be reflected by a more positive absorbance at 192 nm and a more negative absorbance at 220 nm, two wavelengths that are particularly sensitive to a protein's helix content. Therefore, CD spectroscopic studies were undertaken on RBP and both the intact form and the CBD of PurR.
The CD spectra of RBP, CBD, and PurR are shown in Fig.  3. The known or calculated percentages of the various secondary structural elements present in all three proteins, taken from either the x-ray crystallographically determined structure (RBP) or the optimum alignment (CBD and PurR) are presented in Table 111 as are those values for each protein calculated after deconvolution of its CD spectrum. These

RBP
results correlate extremely well with the predicted values for CBD given that the correlation coefficients for proteins of CD versus x-ray are 0.97 for a-helix, 0.76 for @-sheet, 0.49 for @-turn, and 0.86 for other (Johnson, 1990). The close match between the amount of known secondary structure of RBP as determined from the x-ray structure and that calculated by analysis of the CD spectra indicates that CD spectral analysis is a very reliable method for calculating the secondary structure content of proteins with similar PBP folds and bolsters the sequence analyses of CBD and PurR. Further support for the PurR structural prediction was provided from CD data for the intact PurR protein, which shows 40% helix compared with the predicted value of 39%. As discussed above PurR should show greater helix content than the CBD because of the four additional helices that are predicted to lie within the N-terminal DNA binding domain.
Although the amount of a-helix predicted for PurR agrees very well with that determined by CD analysis, the match for @-sheets and aperiodic structures is weaker. A plausible explanation for this latter discrepancy can be offered in light of the nature of PurR's tertiary and quaternary structures. It is possible that the tertiary contacts between the DNA binding and corepressor binding domains as well as quaternary contacts between subunits of the intact PurR dimer lead to a more ordered structure (p-sheet perhaps) in the region encompassing the domain-domain or monomer-monomer contact points than is predicted by the algorithm employed in this study, which only examines the linear sequence of the chain.

CONCLUSIONS
A diagram describing the purine repressor protein's deduced secondary structure topology is presented in Fig. 4. Both predictive and physical methods indicate that the structure of PurRs corepressor binding domain is very similar to the fila structure observed for the metabolite binding domains of the PBP, especially RBP, and that PurRs DNA binding domain assumes the structure of the helix-turn-helix-loophelix DNA binding domain of LacI. A similarly conserved topology is anticipated for all members of the LacI family. Significantly, although the intact LacI protein (Pace et al., 1990) and the LacI core domain (Steitz et al., 1980) have been crystallized, no inducer binding domain of the LacI superfamily has been solved. The high resolution x-ray structure determination of the corepressor binding domain of PurR (in progress) could serve as a potential model for other inducer binding sites of proteins within the LacI family (Schumacher et al., 1992). Furthermore, this structure will allow the comparison of the predicted secondary structure of PurRs corepressor binding domain from sequence analyses and the CD analysis with that found in CBD's crystal structure. It is anticipated that CBD will have a structure very similar to GBP-E and RBP. However, certain regions, in particular the purported dimerization domain, will likely be different but conserved among the LacI family members. Knowledge of those subunit interactions that affect dimerization are essential to understanding not only how and where dimerization occurs, but should shed considerable light on the biochemistry and dynamics of cooperativity in ligand binding by PurR.