Allosteric Mechanism of Induction of CytR-regulated Gene Expression

Transcription from cistrons of theEscherichia coli CytR regulon is activated by E. coli cAMP receptor protein (CRP) and repressed by a multiprotein complex composed of CRP and CytR. De-repression results when CytR binds cytidine. CytR is a homodimer and a LacI family member. A central question for all LacI family proteins concerns the allosteric mechanism that couples ligand binding to the protein-DNA and protein-protein interactions that regulate transcription. To explore this mechanism for CytR, we analyzed nucleoside binding in vitro and its coupling to cooperative CytR binding to operator DNA. Analysis of the thermodynamic linkage between sequential cytidine binding to dimeric CytR and cooperative binding of CytR to deoP2 indicates that de-repression results from just one of the two cytidine binding steps. To test this conclusion in vivo, CytR mutants that have wild-type repressor function but are cytidine induction-deficient (CID) were identified. Each has a substitution for Asp281or neighboring residue. CID CytR281N was found to bind cytidine with three orders of magnitude lower affinity than wild-type CytR. Other CytR mutants that do not exhibit the CID phenotype were found to bind cytidine with affinity similar to wild-type CytR. The rate of transcription regulated by heterodimeric CytR composed of one CytR281N and one wild-type subunit was compared with that regulated by wild-type CytR under inducing conditions. The data support the conclusion that the first cytidine binding step alone is sufficient to induce.

The transport proteins and enzymes required for nucleoside utilization in Escherichia coli are encoded by genes belonging to the CytR regulon (1). This gene family consists of nine unlinked transcriptional units whose expression is coordinately controlled by the interplay of two gene regulatory proteins. Transcription is activated in response to intracellular cAMP levels by CRP 1 and repressed by a three-protein, CRP⅐ CytR⅐CRP, complex. Transcription is induced when CytR binds cytidine. A central feature of this coordinate regulation is that CytR and CRP bind cooperatively to their respective operators (2). This is so despite the role of CytR as a functional antagonist of CRP. The critical role that cooperativity plays is highlighted by the fact that expression is induced, because this cooperative interaction is lost when CytR binds cytidine. Cytidine binding has no effect on intrinsic CytR binding to DNA.
CytR is a member of the LacI family of bacterial repressors (3). The gene regulatory activity of each of these proteins is modulated by binding a peripheral ligand, which functions as either inducer or co-repressor. The basic DNA binding unit of each of these proteins is a homodimer in which helix-turn-helix domains from both subunits combine to form the DNA binding interface. Since both subunits harbor identical ligand binding sites, the allosteric mechanism that couples inducer or corepressor binding to changes in the macromolecular interactions that regulate transcription is an important issue to this entire family of proteins.
For both PurR and LacI, conformational transitions that accompany ligand binding have been investigated by x-ray crystallography (4 -6). In these two cases, binding of co-repressor or inducer, respectively, causes a change in tertiary structure that alters substantially the dimer interface. In the non-DNA binding conformation, hinge helices that connect the helix-turn-helix motif to the ligand binding globular core domain are destabilized, and the helix-turn-helix motifs from the two subunits are thought to be out of register with successive DNA major grooves. In this manner, cooperative ligand binding (7,8) to the individual subunits controls a concerted quarternary conformational change of the dimer. These features are consistent with MWC allostery. While the structural mechanisms that couple ligand binding to tertiary conformation differ in the two proteins (4 -6), the tertiary and quarternary structural perturbations are remarkably similar.
The structures of the LacI family proteins, including CytR, appear to be highly conserved (5,6,9,10). Given the structural resemblance among family members plus the similarity of allosteric mechanism for LacI and PurR, a similar mechanism might be anticipated for CytR. Yet CytR differs from all LacI family members in that it is cooperativity that is allosterically controlled and not intrinsic DNA binding. Allostery thus appears to have a different structural basis in CytR than in other LacI/PurR proteins.
Understanding the allosteric mechanism is central to understanding coordinate regulation of the CytR regulon genes. Recently, we showed that CytR binds to multiple operators at one CytR regulated promoter, deoP2 (12). CytR binding to the operator responsible for repression interacts cooperatively with CRP binding to flanking CRP sites, CRP1 and CRP2. However, by binding to additional specific sites, CytR competes with CRP for binding to CRP1 and CRP2. The net result of cooperativity and competition is that while CRP recruits CytR to form the repression complex, there is no significant reciprocal recruitment of CRP by CytR. This effect has also been reported for the nupG promoter (13). These interactions presumably function to direct both a multistage activation of transcription, using both Class I and Class II CRP mechanisms (14) and also a similar multistage repression mediated by CytR. We have proposed that this might be a general feature of CytR-mediated gene regulation (12).
The unique mechanism of cytidine mediated induction also suggests a multistage process. The cooperativity to which cytidine binding is linked appears to be complementary pair wise in nature. This follows from the observation that the free energy change characterizing cooperativity in the three protein complex, CRP⅐CytR⅐CRP bound to DNA, is equal to the sum of free energy changes characterizing pair wise cooperativity between CytR and CRP bound either to CRP1 or to CRP2 (12). If cooperativity in the three-protein repression complex is pairwise, then it is easy to envision that the two subunits of the dimer might react independently to cytidine binding. This would result in sequential elimination of pairwise, CytR⅐CRP cooperativity, hence sequential relief from repression, in response to sequential cytidine binding to the subunits.
The most general possibilities for coupling between ligand binding and transcription initiation are presented in Fig. 1. We have combined biophysical chemical and molecular genetic approaches to investigate these possibilities. First, CytR binding to CRP-saturated deoP2 was analyzed to evaluate the total contribution from cooperativity. Subsequently, CytR binding titrations were conducted as a function of cytidine concentration. The shape of the transition characterizing loss of cooperativity as cytidine binds CytR indicates that induction is an all or nothing process that occurs concomitant with only one of the cytidine binding steps. Second, CytR mutants were isolated and characterized as fully functional repressors, but which do not induce. The only defect in these mutants is inability to bind cytidine. By co-expressing cytidine induction-defective subunits and wild-type subunits, we evaluated whether the resulting heterodimers would support induction with only one subunit capable of binding cytidine. The combined data from these studies indicate that induction results when cytidine binds to the first subunit of the CytR dimer. Table I lists the bacterial strains and plasmids used in this study. The CID cytR allele, cytRD281N, was transferred to the bacterial chromosome as described by Winans et al. (15). First, the cat gene was inserted into plasmid pCB071-161 at a position 44 bp 3Ј of the cytR termination codon, resulting in plasmid, pCB122. Second, E. coli strain VJS803 was transformed with linearized pCB122, and a recombinant strain, SS6140, was selected as chloramphenicol-resistant (Cm r ) and ampicillin-sensitive (Amp s ). The cytRD281N allele was subsequently transferred to other strains by P1 transduction. The presence of the cytRD281N allele was verified in each Cm r isolate by enzyme assays. The tsx-lac gene fusions carried by strains GP4, Tsx-lac500, Tsx-lac501, Tsx-lac502 and Tsx-lac503 (16) were transferred into strain SS6003 by P1 transduction. The cytR::Tn10dTet insertion was then moved from SS6018 into each SS6003 derivative, yielding strains SS6117 through SS6121 (Table I).

Bacterial Strains, Plasmids, and Growth Conditions-
To express CytR, the coding sequences of wild-type and mutant cytR alleles were subcloned as an NdeI/BamHI fragment downstream of the T7 promoter carried by plasmid pSS584. Strain BL21(DE3) (17,18) was transformed with each construct. The control plasmid for these experiments is pCB135, a pSS584 derivative devoid of cytR coding sequence.
Bacteria were collected from exponentially growing cultures for enzyme assays. The medium contained Vogel and Bonner salts (19) supplemented with vitamin B 1 at 5 g/ml, 0.02% casamino acids, and 0.4% glycerol (20). BL21(DE3) derivatives harboring CytR plasmids were grown in a 1% Bacto-tryptone, 0.4% glycerol medium containing Vogel and Bonner salts (TV medium). Either L-broth or 2 ϫ YT was used for transformations and plasmid preparations (21). The Lac Ϯ phenotypes of the various strains were determined on solid TTC-Lac medium as described previously (20). When used, the final cytidine concentration was 2 mM. Antibiotic concentrations used in the media were: ampicillin, 100 g/ml; tetracycline, 15 g/ml; chloramphenicol, 20 g/ml; and kanamycin, 25 g/ml in minimal medium or 50 g/ml in rich medium.
Generation, Identification, and Characterization of CytR Mutants-A FIG. 1. Schematic of CytR cooperativity states reflecting different allosteric models. CRP and CytR are shown bound to their respective operators, CRP1, CRP2, and CytO. Shading and object shape are used to distinguish differences in tertiary and quarternary states, respectively. Cooperativity is indicated by contact between adjacent proteins. Only the favored state is shown in each case, and symmetrically redundant configurations are not shown. A represents a concerted model featuring an equilibrium between pre-existing states. Cytidine binding shifts the poise of the equilibrium from favoring the cooperative state to favoring the noncooperative state. This is MWC allostery. B represents a sequential allosteric model in which cytidine binding induces a conformational change of only the subunit to which it binds. This eliminates cooperative interaction between that subunit and the adjacent DNA-bound CRP⅐cAMP and also affects interactions with the remaining subunit to affect its affinity for cytidine binding. This is KNF allostery. C represents sequential but concerted transitions between cooperativity states. Unliganded CytR interacts cooperatively with CRP⅐cAMP complexes bound to both flanking sites, CRP1 and CRP2. Binding of cytidine to the first subunit induces a state in which neither subunit interacts with adjacent bound CRP⅐cAMP. A similar model in which the obligatory transition to noncooperative binding occurs only when the second cytidine binds is also possible. mixture of mutagenic oligonucleotides complementary to cytR codons 276 through 284 was synthesized using an Applied Biosystems model 381A DNA synthesizer. The spiking protocol of Hutchison et al. (22,23) was used to create degeneracy in the oligonucleotide sequence. The mutagenic oligonucleotide mixture and a site-directed mutagenesis kit from Amersham Corp. was used to mutate the cytR gene on an M13mp19cytR10 template. Both single and multiple mutations were obtained, the frequency of single mutations being about 30%. Phage pooled from about 5000 mutagenized M13mp19cytR10 plaques was propagated in E. coli strain JM103 by incubation for 4 h in 2 ϫ YT medium. RF-M13 DNA was prepared as described (24). The cytR gene fragment containing the mutagenized sequence bounded by ApaI and BamHI cleavage sites was subcloned into pCB093 by fragment exchange (20).
The recombinant plasmid pool was transferred into SS6018 (cytR), which was grown on TTC-Lac-Kan medium containing 2 mM cytidine, to identify mutant CID repressors. The dominant negative phenotype of CytR mutants was established using CytR ϩ strain, SS6004, as described previously (20). The steady-state level of wild-type and mutant CytR was measured using a Western immunoblot analysis (20). Each cytR mutation that yielded a stable mutant protein was identified by DNA sequencing of the mutagenized cytR gene segment on a purified, double-stranded template (20).
Enzyme Assays-Bacteria used for enzyme assays were grown and cell extracts prepared as described previously (25). Cytidine deaminase (CDA) and uridine dephosphorylase (UDP) spectroscopic assays were performed as described previously (25,26) except that the CDA assay mixture contained 50 mM Tris-HCl (pH 7.5) and 0.5 mM cytidine. The ␤-galactosidase activity of exponentially growing bacteria was measured as described by Miller (21). To express the enzyme activity as specific activity of the cell extracts, the total protein concentration of the extracts was measured by the Bradford assay (27) using bovine serum albumin as a standard.
Preparation of Cell-free Extracts for Cytidine Binding Assays-BL21(DE3) derivatives harboring T7 expression plasmids for either wild-type or mutant CytR were grown at 37°C in nucleoside-free TV medium with 0.4% glycerol to A 600 Ϸ 1.0. CytR expression was induced by adding 1% lactose and 2 mM isopropyl-␤-D-thiogalactopyranoside and incubating for 60 min before harvesting the cells. Cell pellets were washed, resuspended in 20 mM MOPS (pH 6.8), 2 mM MgSO 4 , 1 mM Na 4 EDTA, 200 mM NaCl, and treated with lysozyme, added to 10 g/ml. The cells were frozen at Ϫ20°C, thawed at 23°C, and broken by sonication. The final cell extract was the clear supernatant remaining following centrifugation at 100,000 ϫ g for 1 h at 4°C.
Purification of CytR-CytR was purified using a simpler protocol than that reported several years ago (2) but which yielded a higher yield of CytR with similar purity. All purification steps were carried out at 4°C. Pellets from cells harvested 165 min postinduction were resuspended in 20 mM MOPS (pH 6.80), 2 mM MgSO 4 , 1 mM Na 4 EDTA, 1 mM dithiothreitol (R-buffer) supplemented with 0.3 M NaCl. The resuspended cells were lysed using two passes through a French press and centrifuged at 50,000 ϫ g for 3 h. The supernatant was adjusted to 0.2 M NaCl and 10% glycerol in R-buffer. Polyethyleneimine was added to a final concentration of 0.04% to precipitate nucleic acids and some proteins. The supernatant from a low speed centrifugation was adjusted to 0.1 M NaCl and chromatographed on two Bio-Rad EconoPac Q cartridges (5 ml each) connected in series using a Pharmacia FPLC. The pooled CytR containing the flow-through peak was loaded on two Bio-Rad Econo-Pac S cartridges connected in series. After washing the column with 0.2 M NaCl R-buffer until the A 280 of the wash returned to the buffer base line, the column was eluted using a 0.2-0.6 M NaCl gradient. CytR elutes in a broad peak between 0.3 and 0.4 M NaCl.
CytR concentration was estimated from an extinction coefficient of 0.30 Ϯ 0.03 cm Ϫ1 mg Ϫ1 ml at 280 nm. This value was calculated from the average extinction coefficients for amino acid residues in a protein (28 -31). The unusually low extinction is due to the fact that CytR contains no tryptophan. Based on the calculated extinction, the yield of CytR using this expression and purification protocol is 1.5-2 mg/g of cell paste. CytR is stored at a concentration of 2-3 mg/ml in the S-column elution buffer in a liquid nitrogen dewer after flash freezing as ϳ25-l beads in liquid nitrogen. The DNA and cytidine binding activities re-  main stable for at least several years when stored in this manner.
Sedimentation equilibrium analysis shows this material to be homogeneous dimer, with no evidence for either dissociation to monomer or association to higher order polymers over the concentration range, 0.1-10 M. 2 More recent analysis of gel mobility shift assays of CytR binding to DNA has been interpreted to indicate that CytR remains as stable dimer over the range of concentrations at which it binds DNA operators (11). Based on these data, the binding experiments were analyzed according to the simplest model in which CytR exists only as dimer.
Cytidine Binding Assays-Binding of [ 3 H]cytidine to purified wildtype CytR and to both wild-type and mutant CytR containing cell-free extracts was measured using a filter binding assay. Binding reaction mixtures contained either 18 -50 nM purified CytR dimer or 15-30 g of cell extract protein in a 100-l volume containing 0.04 -11.0 M [ 3 H]cytidine (NEN Life Science Products). Two different buffers were used: 1) 20 mM MOPS (pH 6.8), 2 mM MgSO 4 , 1 mM NaEDTA, 200 mM NaCl and 2) 10 mM bis-Tris (pH 7.0), 100 mM NaCl, 0.5 mM MgCl 2 , 0.5 mM CaCl 2 . Both buffers contained 100 g/ml bovine serum albumin and 1 g/ml calf thymus DNA. Following a 5-min incubation at 23°C, the CytRbound [ 3 H]cytidine contained in 80 l of assay mix was collected on a prewashed nitrocellulose filter (Millipore HAWP 02500; Millipore Corp., Bedford, MA). The filters were washed once with 500 l of assay buffer, air-dried, and then dissolved in 3.5 ml of Packard Filter-Count LSC mixture (Packard Instrument Co.). Radioactivity was measured using a Packard model 1900TR scintillation counter.
For determination of nucleoside binding constants, binding assays were conducted as titrations by varying the nucleoside concentration at constant CytR concentration. The data were analyzed according to a simple Langmuir binding model as described below (see Equation 1) using a nonlinear least squares curve fitting program (32). The CytR concentration used in some titrations was not negligible. To analyze data under these conditions, the conservation polynomials for total cytidine and total CytR were solved for the concentration of free cytidine for each data point and at each iteration of the nonlinear least squares analysis. The program NONLN (33) was used for this purpose.
In experiments to compare [ 3 H]cytidine binding by CytR mutants in cell-free extracts and wild-type CytR, the CytR content of the cell-free extracts was estimated using Western immunoblots as described above. For each extract, 25-200 ng of extract protein was electrophoresed on SDS-16.5% acrylamide gels. Proteins were electrotransferred to Immobilon-P membranes (Millipore Corp.) as described previously (20). CytR was complexed with anti-CytR antibody and 125 I-protein A. 125 I-Protein A in the complex was quantitated using a Molecular Dynamics Phos-phorImager (Molecular Dynamics, Sunnyvale, CA).
Competition by other nucleosides for [ 3 H]cytidine binding to wildtype CytR in extracts was measured in the MOPS assay buffer. [ 3 H]Cytidine and competitor concentrations were 1 M and 1 mM, respectively. Cytidine deaminase and adenosine deaminase present in the cell extracts were inhibited by addition of 1 M tetrahydrouridine and 10 M deoxycoformycin, respectively. These nucleoside deaminase inhibitors did not affect cytidine binding by purified, wild-type repressor.
CRP Protein Used in DNA Binding Studies-CRP was the kind gift of James C. Lee (University of Texas Medical Branch, Galveston, TX). This protein preparation shows no evidence of any contaminating material by Coomassie staining of overloaded SDS-polyacrylamide gels, from which we conservatively estimate at least 98% purity.
Individual-Site Binding Experiments-A series of DNase I footprint titrations of CytR binding to deoP2 was conducted to assess the effect of cytidine binding to CytR on heterologous cooperativity between CRP⅐cAMP and CytR (Fig. 3). An 879-bp NotI/HincII restriction fragment containing the deoP2 regulatory sequence was isolated from a derivative of pSS1332 (Table I) in which an 8-bp NotI linker was inserted into the SmaI site of the polylinker. The fragment contains the E. coli deo sequence from ϩ151 to Ϫ801 from the P2 start site for transcription. The fragment was labeled with 32 P at the NotI site by using the Klenow fill-in reaction and purified as described (34).
Each separate titration of CytR binding to this fragment ( Fig. 4) was conducted at a different, fixed concentration of cytidine ranging from 1 nM to 10 mM and at a saturating concentration of CRP⅐cAMP. CytR was the titrating ligand. DNase I footprint titrations were conducted as described (12) in the bis-Tris (pH 7) cytidine binding buffer described above. CRP and cAMP were added to final concentrations of 0.10 M (as dimer) and 100 M, respectively. Each titration was analyzed using NONLN (33) according to Equation 1 to obtain the apparent free energy change for cooperative binding of CytR at one cytidine concentration (⌬G app ϭ ϪRTlnk app ), where Y in Equation 1 is the fractional saturation of the deoP2 CytR operator, L is the free CytR concentration, and k app is the apparent association constant for CytR binding to deoP2. The fractional protection end points were also included as adjustable parameters (35). Parameter confidence limits reported are worst case joint limits corresponding to approximately one standard deviation. 3 Analysis of ⌬G app as a function of cytidine concentration (Fig. 3) used models described in the text. In this analysis, the confidence limits to ⌬G app were used to calculate normalized weighting factors for the individual data points (36).

RESULTS
Identification and Characterization of CID Mutants-We previously described two CytR mutants, D281N and D281Y, that are repression-competent but nonresponsive to induction by cytidine (20). We denote such cytidine induction-defective mutants as CID. Degenerate oligonucleotide-directed mutagenesis of cytR codons 276 through 284 was used to identify other residues that are critical for cytidine induction. This mutagenesis yielded 21 mutant cytR genes that produced stable CytR. Twelve of these mutant genes encoded inactive repressors. Eleven of these inactive repressors have amino acid substitutions for either Ile 277 or Asp 274 and are recessive. The other inactive repressor (G279R) has an inactive, dominant negative phenotype. The remaining nine of these mutant genes produced CID phenotypes. Three of these are previously identified mutants, and six are newly identified mutants. Four of the newly identified mutants have single amino acid substitutions, three for Asp 281 (Table II) and one for Gly 279 . The others carried double substitutions (F280I/D281A and F280S/N282I).
The CID phenotype of the four new single substitutions was verified by measuring enzyme synthesis from two CytR controlled genes, cdd and udp, in the presence and absence of cytidine. As was found with the original CID mutants, expression of these CytR-controlled genes is repressed by the mutant CytR proteins and is unaffected by cytidine (Table II). Only D281A was a significantly less active repressor than the wildtype protein. However, like all CID mutants, D281A did not a Strain SS6018 (cytR) harboring plasmids expressing each allele was grown in minimum glycerol minimum medium as described previously (25,26). Extracts were prepared and assayed for CDA and UDP activity as described under "Materials and Methods." Specific activity is in units of enzyme activity per mg of total extract protein. CDA units are mol of cytidine deaminated/min; UDP units are mol of uridine dephosphorylated/min. respond to cytidine. All CID cytR alleles were expressed identically and all produced steady-state cellular CytR concentrations equal to that of the wild-type protein based on Western blot analysis of soluble extracts from the mutant strains. As found for other CytR mutants (20), the CytR controlled enzyme levels and the phenotype observed for bacteria expressing these CID mutants reflect directly the change in repressor function.
Functional repressor might show the CID phenotype if repression was no longer dependent on heterologous cooperative interactions with CRP⅐cAMP. To determine whether CID mutants require such interactions for repression, we compared the ability of each repressor to regulate transcription from wildtype and mutant tsx-lac reporter gene fusions. The tsx-lac gene fusions were constructed and characterized by Gerlach et al. (16). The tsx DNA of each mutant reporter gene fusion contains a single bp substitution in CRP2, which greatly reduces its affinity for CRP⅐cAMP binding. These point mutations have no direct affect on intrinsic CytR binding; CytR binding is affected only indirectly via loss of CytR⅐CRP cooperativity. This loss of cooperativity prevents repression of mutant tsxP2 promoters by wild-type CytR in vivo, even when expressed at high levels from a multicopy plasmid (37). Thus, expression from tsx-lac gene fusions provides a specific, direct assessment of CytR⅐CRP cooperativity. It avoids the use of bacteria having either cyaϪ or crpϪ mutations and their pleiotropic effects.
Both wild-type and CID CytR repress ␤-galactosidase synthesis from wild-type tsxP2. Regulation of mutant tsx-lac fusions by wild-type and CID repressors was also identical (Table  III). Neither protein represses expression from the mutant promoters. Therefore, repression by mutant CID CytR is dependent on cooperative interactions with operator-bound CRP⅐cAMP, the same as repression by wild-type CytR. Thus, the only identifiable defect in function of the CID CytR proteins is their failure to respond to cytidine induction.
Cytidine Binding by Wild-type and CID CytR-The induction-defective phenotype of CID CytR could result either if the protein fails to bind inducer or if it binds but fails to respond to inducer. To assess these options, cytidine binding was studied directly. Titration of purified wild-type CytR yielded a single binding transition (Fig. 2). When analyzed according to a simple binding model (Equation 1), an apparent K d value of 0.17 (0.11,0.33) 3 M was obtained. The limiting ratio of cytidine retained to CytR dimer is 0.82. While this is similar to a stoichiometry of one per dimer, it doesn't account for CytR protein that isn't retained by the nitrocellulose filters and so only represents a lower limit to the stoichiometry. Cytidine binding is unaffected by addition of cell free extracts from an E. coli strain that is deleted of the cytR gene [K d ϭ 0.16 (0.11,0.30) M]. Cytidine binding was also the same whether using purified CytR or cell free extracts from an E. coli strain containing the wild-type cytR gene (data not shown). Ligand binding is also specific. In competition assays conducted in the presence of a 1000-fold excess of nucleoside competitor (1 mM), only adenosine competed with cytidine (K I ϭ 22.5 Ϯ 1.5 M). Uridine, 2Ј-deoxycytidine, 2Ј,3Ј-dideoxycytidine, cytidine arabanoside, and both 2Ј-and 3Ј-O-methylcytidine had no effect on cytidine binding in vitro, consistent with the inducer specificity in vivo (38 -40).
Since addition of CytR-free E. coli extracts has no effect on cytidine binding to purified CytR, cytidine binding to wild-type and mutant CytRs in cell free extracts can be compared directly. Extracts were prepared from isogenic strains expressing either wild-type CytR, CytR from one of the different classes of mutants, or no CytR. There was no detectable cytidine binding in extracts lacking CytR. Of the mutant CytR containing extracts, only that containing the CID repressor (CytRD281N) showed a significant decrease in cytidine binding (Table IV) as compared with the extract containing wild-type CytR. At a cytidine concentration (11 M) that essentially saturates wildtype CytR (0.98 saturation), this CID CytR bound less than 1% as much cytidine as wild-type. This indicates a K d for the mutant CID CytR of 1 mM or greater, an affinity at least 2000-fold reduced from that of wild-type CytR. An effect of such magnitude can only be readily explained by differences between CytR in the different cell-free extracts. By contrast to this result, CytR with amino acid substitutions in domains proposed previously (20) to function in DNA binding (CytRV15A) and in signal transduction (CytRM151I and CytRM151V) bound roughly the same amount of cytidine as wild-type CytR, thus indicating no effect on cytidine binding affinity.
Effect of Cytidine Binding to CytR on Heterologous Cooperative Binding of CytR to deoP2-DNase I footprint titration was used to determine the apparent affinity of CytR for deoP2 at a Each fusion contains a single point mutation in CRP-2 of the tsx-p2 promoter (16). The mutation yielding tsx-501 was Gly 310 3 Thr, the tsx-502 change was Gly 304 3 Ala, and that for tsx-503 was C309 3 T.
b Expressed from plasmids pCB123 (wild-type) or pCB124 (cytRD281N). The control strain carried the vector lacking the cytR gene.
c The bacteria were grown and extracts prepared as described under "Materials and Methods." One ␤-galactosidase unit represents hydrolysis of 1 nmol of ONPG/min/mg of extract protein at 25°C. CytR (37 nM, dimer). Points plotted are averages of three measurements. Curves are: solid, analysis according to a simple binding model (Equation 1). This yields a binding constant, K ϭ 5.9(3.0, 9.1) ϫ 10 6 M Ϫ1 and a maximum ratio of cytidine bound per dimer, C max ϭ 0.82; dashed, analysis assuming sequential binding to two sites; broken dashed, further assumes stepwise binding constants equal to values obtained by analysis of the data in Fig. 3. These analyses are described in the text (see "Discussion"). different, fixed concentrations of inducer cytidine. The concentrations of cAMP and of CRP used yield 97 and Ͼ99% saturation of CRP1 and CRP2, respectively (12). Fig. 3 presents the apparent Gibbs free energy changes (⌬G app ) for binding of CytR dimer to the deoP2⅐(CRP⅐cAMP) 2 complex as a function of cytidine concentration. Each ⌬G app was determined from a separate CytR titration experiment such as that shown in Fig. 4. Under the conditions of these experiments, heterologous cooperativity increases the apparent affinity of CytR for deoP2 by 100-fold ( Fig. 3 and Ref. 12). Thus, the ⌬G app values plotted reflect a sum of two contributions. One contribution arises from intrinsic CytR binding (Ϫ10.4 Ϯ 0.4 kcal/mol; Fig. 5 and Ref. 12); the other is from cooperative interactions between CytR and CRP⅐cAMP. In the absence of cytidine, this sum of contributions yields ⌬G app ϭ Ϫ13.1 Ϯ 0.2 kcal/mol ( Fig. 3 and Ref. 12). The contribution from cooperativity is reduced when CytR binds cytidine, as indicated by the smooth transition in Fig. 3. The data in Fig. 5 confirm no effect on intrinsic binding. At cytidine concentrations above 1 mM, the apparent free energy change reaches an asymptotic limit of about Ϫ10.7 kcal/mol. The cooperative effect is thus reduced from 100-fold to less than 2-fold.

FIG. 2. Binding of radiolabeled cytidine by purified wild-type
The equilibria between CytR, cytidine, and the deoP2⅐ (CRP⅐cAMP) 2 complex are shown in Scheme I. An implicit simplifying assumption is that CytR binds to only a single site near deoP2. The effect of CytR binding to additional sites (12) is negligible under the conditions employed in these assays. Scheme I accounts for one cytidine binding site on each subunit of homodimeric CytR. K 1 and K 2 are macroscopic stepwise association constants, which describe the binding of the first and second cytidine ligands to free CytR dimer. K 3 , K 4 , and K 5 are association constants for binding of CytR, CytR⅐ cytidine, and CytR⅐(cytidine) 2 to the deoP2⅐(CRP⅐cAMP) 2 complex. K 6 and K 7 are macroscopic stepwise association constants for binding of the first and second cytidine ligands to deoP2⅐(CRP⅐cAMP) 2 -bound CytR. The apparent equilib-rium constant for CytR binding to deoP2⅐(CRP⅐cAMP) 2 at any cytidine concentration ([cyt]) is as follows.
Equilibrium constants, K 1 -K 5 are independent parameters. Since K 1 , K 3 , K 4 , and K 6 comprise a thermodynamic cycle, K 6 is defined by K 6 ϭ K 1 K 4 /K 3 . Similarly, K 7 ϭ K 2 K 5 /K 4 . Equation 2 was used to analyze the values in Fig. 3 (⌬G app ϭ ϪRTlnK app ) to estimate the Gibbs free energy changes corresponding to K 1 -K 5 . Results of this analysis are indicated by the broken curve in Fig. 3. The maximum likelihood parameter estimates indicate an affinity for binding of the first cytidine, K 1 ϭ 2.7 ϫ 10 7 M Ϫ1 . In this model, this binding step is accompanied by only a negligible effect on K app (1.5-fold decrease). The second subunit binds cytidine with significantly lower affinity (K 2 ϭ 1.0 ϫ 10 6 M Ϫ1 ) and is accompanied by the major reduction in K app (40-fold decrease). (Note: K 2 is a stepwise macroscopic equilibrium constant; it corresponds to a microscopic constant for cytidine binding to the remaining cytidine binding site equal to 2.0 ϫ 10 6 M Ϫ1 .) This result suggests that cytidine binding is negatively cooperative and that the major allosteric switch occurs at only one of the two cytidine binding steps.
An alternative possibility that is also consistent with this conclusion is that the allosteric switch occurs at the first cytidine binding step. If so, the transition in Fig. 3 represents cytidine binding to only the first of the two CytR subunits. To test this possibility, the data were analyzed using a simplified version of Scheme I, in which K 2 and K 5 were set equal to zero. Results of this analysis, indicated by the solid curve in Fig. 3, yielded K 1 ϭ 2.0 ϫ 10 6 M Ϫ1 . Thus, both models estimate the same cytidine affinity as being allosterically coupled to cooperative binding of CytR to the deoP2⅐(CRP⅐cAMP) 2 complex.  Goodness of fit criteria such as the variances and distributions of residuals provided no basis for discriminating between the two models. By contrast, all other models tested, such as the model in which the two subunits have independent (noninteracting) cytidine binding sites, which are either separately or together coupled to cooperative operator binding, failed to provide an adequate fit to the data.

Inducibility of Wild-type-CID Heterodimers in Vivo-
The existence of well characterized CID alleles provides the means to test, in vivo, the conclusion that cooperativity is coupled primarily to only one cytidine binding step and also to determine whether the first or second step. To do so, steady-state expression from CytR-regulated promoters was examined in bacteria that co-express both wild-type, inducer-responsive CytR subunits and CID, nonresponsive CytR subunits. Wild-type and CID subunits were expressed by cistrons that had identical promoter and operator regions. Since both wild-type and CID CytR are fully functional repressors, their co-expression should result in heterodimeric CytR with one wild-type subunit and one CID subunit.
To promote heterodimer formation, CID allele cytRD281N was first recombined in single copy into the chromosome. As shown in Table V, CytRD281N that is expressed from the bacterial chromosome is indistinguishable from wild-type CytR in its ability to repress CDA and UDP synthesis, but retains its CID phenotype. Second, isogenic strains that differ only in their chromosomal cytR allele (either wild-type or CID) were transformed by a plasmid that expresses the wild-type cytR gene. A low copy number plasmid (41,42) containing a P15A origin was used. These constructions yielded bacteria with the same cytoplasmic CytR level (see "Materials and Methods") as one another. This is as expected, since expression of all cytR alleles is identically regulated and both wild-type and CID proteins are functional, stable repressors. The response to induction was compared using these two strains. To ensure that steady-state transcriptional activity was being compared as opposed to steady-state enzyme levels governed by protein turnover, the differential rate of CytR-controlled ␤-galactosidase synthesis from a udp-lac fusion was measured in exponentially growing cells.
The differential rate of CytR regulated ␤-galactosidase synthesis in bacteria that co-express wild-type and CID CytR subunits should depend on the response of CID/wild-type heterodimers to cytidine. There are three possibilities. First, if the heterodimer is induction-defective, then both CID/CID homodimers and CID/wild-type heterodimers would repress udplac expression. Because the wild-type allele is plasmid-encoded, while the CID allele is chromosomal, these bacteria have excess wild-type subunits, hence excess wild-type homodimers. However, under inducing conditions, these homodimers no longer bind cooperatively to the promoter; their affinity is reduced by 2 orders of magnitude (Fig. 3). Consequently, even a severalfold excess of wild-type homodimers would be insufficient to compete effectively with induction refractive heterodimers and the small proportion of CID homodimers. 4 The differential rate of ␤-galactosidase synthesis would approximate that of CID homozygous bacteria, even under inducing conditions. Second, if CID and wild-type subunits do not associate to form heterodimers, then all CID subunits expressed must be present as CID homodimers. Again, the level of noninducible CytR, now comprised of the CID/CID homodimers, would be sufficient to yield a noninducible phenotype. Third, if (and only if) cytidine binding by the wild-type subunit of CID/wild-type heterodimers is sufficient for induction, then the differential rate of ␤-galactosidase synthesis under inducing conditions would be much greater than that observed for CID homozygous bacteria, even approaching that observed for wild-type CytR homozygous bacteria. Full wild-type levels of induction would not be expected because of competition by the nonresponsive CID/ CID homodimeric CytR, even when present as a small fraction of the total repressor.
The rates of bacterial growth and ␤-galactosidase synthesis for the individual strains are shown in Fig. 6. These rates were constant throughout the entire experiment, demonstrating that cytidine induction is at equilibrium and that the enzyme levels represent steady-state production in all cultures. The differential rates of ␤-galactosidase synthesis were calculated from these curves. Under inducing conditions, the rate for bacteria expressing both wild-type and CID CytR subunits was found to be 5-fold greater than that measured for CID homozygous bacteria (2543 units versus 499 units). It was slightly greater than half of that found for wild-type homozygous bacteria (4827 units). These results can be explained if wild-type and CID subunits do associate to form heterodimers, which in turn do respond to induction by cytidine. DISCUSSION CRP and CytR mediate coordinate regulation of the unlinked genes that encode the proteins necessary for nucleoside utili-zation in E. coli. The interplay between these regulatory proteins, comprised of both cooperative and competitive interactions, appears to direct both multistage activation and multistage repression of transcription of individual cistrons (12). The critical role of CytR⅐CRP cooperativity is highlighted by the mechanism of cytidine-mediated induction. This gene regulatory mechanism relies on loss of cooperativity, not on reduction in DNA binding affinity as found with other LacI family members. Understanding the allosteric coupling between cytidine binding and CytR⅐CRP cooperativity is necessary to understanding the coordinate regulation of this gene family.
The specific question addressed herein is whether induction is a concerted process coupled to the quarternary state of the CytR dimer or a sequential process coupled to the ligation state of the individual subunits. The former holds for LacI and PurR, which undergo an MWC-type allosteric transition between quarternary T and R states (4 -6). CytR presumably does not experience the same global conformational change upon ligand binding as LacI and PurR, since inducer binding is not coupled to DNA binding. Moreover, CytR⅐CRP cooperativity appears to be pairwise and complementary, in nature (12). A multistage induction such as would result from a sequential coupling a Each strain was grown in minimum glycerol minimum medium as described previously (25,26). Extracts were prepared and assayed for CDA and UDP activity as In the first phase, each strain was inoculated into 37°C complete minimal medium and grown until A 600 reached ϳ0.1. At this point cytidine (CR) was added to a final concentration of 2 mM, and growth was continued. When the A 600 reached ϳ0.7, an aliquot of the culture was diluted into the same medium containing 2 mM cytidine to an initial A 600 of ϳ0.1. In this second phase, growth of the freshly diluted culture was continued until the A 600 again reached ϳ0.7. The specific growth rates for all strains during both growth phases varied by less than 10%. During each growth phase, samples were taken for determination of ␤-galactosidase synthesis as described by Miller (21). The differential rate of enzyme synthesis for each strain is shown both before (B) and after (C) dilution into fresh medium. These rates were unchanged by dilution for the each of the different and are as follows: SS6094(pCB095), 4827 units of ␤-galactosidase/A 600 (circles); SS6141(pCB095), 2543 units of ␤-galactosidase/A 600 (squares); SS6141(pGLP4), 499 units of ␤-galactosidase/A 600 (triangles). Strain SS6094 expresses wild-type CytR from the chromosome. Strain SS6141 expresses CID CytRD281N from the chromosome. pCB095 expresses wild-type CytR and pBGLP4 is a control plasmid that lacks the cytR gene. between binding of cytidine to an individual subunit, and a tertiary conformational change affecting only that subunit's cooperative interaction, could play a significant role in differential expression of the unlinked genes.
Three allosteric mechanisms can be considered: first, a classic MWC mechanism featuring an equilibrium between two symmetric quarternary states, one that interacts cooperatively with CRP and one that does not (Fig. 1A); second, a strictly sequential KNF mechanism in which the tertiary conformation of each subunit switches to a noncooperative state concomitant with cytidine binding to that subunit (Fig. 1B); third, a sequential but concerted mechanism in which distinct quarternary states are formed as each cytidine site is filled (Fig. 1C). Elimination of CytR⅐CRP cooperativity and induction of transcription might occur when the first cytidine binds, when the second binds, or in part when both bind in proportion to the overall fractional saturation.
We investigated the cytidine-mediated transition from cooperative to noncooperative CytR binding to CRP-saturated deoP2 to distinguish among these possibilities. The data were analyzed according to a general formulation (Scheme I; Equation 2) that encompasses all three allosteric mechanisms. Only two numerical solutions of Equation 2 were found to be consistent with the data. Two common features of these solutions both point to the third allosteric mechanism (Fig. 1C) and are inconsistent with MWC and KNF mechanisms (Fig. 1, A and  B). First, the analysis suggests that the transition from cooperative to noncooperative CytR binding is coupled to a single cytidine binding event. Second, cytidine binding is characterized by negative cooperativity.
That loss of cooperativity is coupled to only a single cytidine binding step is reflected by the characteristic shape of the transition curve. Other mechanisms, such as a cooperative transition between pre-existing states (MWC; Fig. 1A) or sequential coupling to each cytidine binding step (e.g. KNF; Fig.  1B) yield either sharper or shallower transitions. These are inconsistent with the data. The finding of negative cooperativity in cytidine binding to CytR is also inconsistent with an MWC allosteric mechanism. A concerted transition between pre-existing states necessarily yields positive cooperativity in ligand binding.
The two numerical solutions to Equation 2 do differ in detail regarding the negative cooperativity. When both cytidine binding steps are considered in the analysis, the cytidine binding constants estimated indicate approximately a 7-fold effect. When Equation 2 is truncated to consider only one cytidine binding step, this is equivalent to the assumption that linkage reflects only the first cytidine binding step. This would mean that binding of the second cytidine is insignificant over the concentration range investigated, suggesting a much higher degree of negative cooperativity.
We cannot distinguish between these possibilities, even based on the titration data for cytidine binding to free CytR (Fig. 2). These data were reanalyzed by considering cytidine binding as comprising two steps. This analysis (Fig. 1) estimated an intrinsic cytidine binding affinity equal to 0.4 M, nearly identical to that identified as being coupled to loss of cooperativity in DNA binding (0.5 M; Fig. 3) and negative cooperativity accounting for greater than an 80-fold affect. It also yielded a limiting ratio of cytidine retained to CytR dimer equal to 1.60, consistent with a stoichiometry of two per dimer. According to this model, the transition in Fig. 1 corresponds primarily to the first cytidine binding step; the plateau continues sloping upward, reflecting the second binding step that occurs at higher cytidine concentration. These features mirror the result obtained from the DNA binding data in Fig. 3, when the latter are analyzed simply by assuming linkage to only the first cytidine binding step.
However, the cytidine titration data in Fig. 2 are also reasonably described by the alternative analysis of the DNA binding data. The data in Fig. 2 were analyzed using the stepwise cytidine binding constants, K 1 and K 2 , obtained from analysis of the DNA binding data in Fig. 3 as fixed input parameters. According to this interpretation, cytidine binding in Fig. 2 looks like a single transition, because the negative cooperativity is too moderate to produce separate binding transitions. This analysis estimates a limiting ratio of cytidine retained to CytR dimer equal to 0.9. While this is quite low compared with the model's stoichiometry of two per dimer, the discrepancy could reflect poor CytR retention efficiency in the filter assay.
Despite uncertainty in details, these analyses support three conclusions: first, that cytidine binding to CytR is negatively cooperative; second, that cooperative binding of CytR to deoP2⅐(CRP⅐cAMP) 2 is primarily coupled to only one of the two cytidine binding steps; and third, the intrinsic cytidine binding affinity in free CytR that is coupled to this transition is 0.2-0.5 M. Thus, cytidine binding must switch CytR between three conformational states, one corresponding to each cytidine ligation state, as represented by Fig. 1C. However, these data do not identify which conformational change eliminates CytR⅐CRP cooperativity. To address this issue, it is necessary to evaluate the behavior of the intermediate state with only one subunit liganded. For this, wild-type and CID CytR alleles were coexpressed, thus allowing assembly of hybrid CytR dimers, which have only one subunit capable of binding cytidine. The behavior of these hybrids in vivo was used to assess the induction competency of the intermediate ligation state.
For this approach, it was necessary to find a CytR mutant whose only defect is inability to bind cytidine. We focused the search on a region of the CytR sequence in which CID mutants had been identified previously (20). The newly identified CID mutants expressed wild-type levels of protein, and most had repressor activity equal to wild-type (Table II). One allele, cytRN281N, was found independently both in the previous screen following random mutagenesis of the entire CytR gene (20) and in the present screen following targeted mutagenesis. CytRD281N was shown to require cooperative interaction with CRP for repression as does wild-type CytR (Table III). From this we infer that repression remains coupled to the mechanism that underlies induction. Cytidine binding assays demonstrated the only defect found, a 2000-fold or greater reduction in cytidine binding affinity (Table IV).
The finding that cytidine binding affinity is reduced as a result of amino acid substitutions for Asp 281 is consistent with observations on other LacI proteins. This is a conserved aspartate in the sequences of many of the LacI repressors as well as the E. coli periplasmic binding proteins (3). The equivalent aspartate is essential for ligand binding in LacI, PurR, and the periplasmic sugar-binding proteins for glucose, ribose, and arabinose (43)(44)(45). The substitutions for CytR Asp 281 that yield the CID phenotype (Asn, Ala, Glu, and Ile) are understandable if Asp 281 participates directly in cytidine binding as a hydrogen bond partner to sugar hydroxyls. Perhaps, the decreased affinity of CID mutants for cytidine is functionally equivalent to the inability of 2Ј-or 3Ј-deoxycytidine to compete with cytidine for binding to wild-type CytR. The substitutions that yield the CID phenotype suggest that charge (D281N) and side chain size (D281E) as well as hydrogen bonding potential (D281A) are important to this interaction. Whether this interaction provides only affinity and specificity, or is also important to the coupling between cytidine binding and CytR⅐CRP cooperativity, cannot be determined from our data. However, we note that in LacI, only affinity is affected (46).
Evaluation of the capacity of the hybrid dimers to support induction in vivo assumes that the steady-state levels of wildtype and CID subunits are proportional to the gene copy number. Since the wild-type and CID alleles are identically regulated, this assumption is supported by the finding of comparable CytR levels in extracts made from bacteria expressing either wild-type or mutant CytR. Thus, for bacteria expressing both alleles and assuming a plasmid copy number of ϳ6 -8 per cell (41,42) and random assortment of subunits, then CytR dimer is proportioned about 75% as wild-type homodimer, 23% as heterodimer, and the remainder as CID homodimer. Uncertainty in plasmid copy number has a negligible effect on this distribution. If heterodimers respond to cytidine, this distribution yields a 50-fold excess of inducible over noninducible dimers. When offset by the higher affinity of noninduced dimers, some reduction of the extent of de-repression is expected, perhaps in line with what was observed (Fig. 6). If heterodimers do not respond to cytidine binding, then only a 3-fold excess of inducible over noninducible dimers results. This is insufficient to compete effectively for DNA binding under inducing conditions and de-repression should not be observed. Similarly, if CID and wild-type subunits do not assemble as heterodimers, then the CID subunits must be assembled as CID homodimers to account for the fact that CID CytR is functional repressor. Again, the excess of inducible over noninducible dimer would not be sufficient to support significant de-repression. Therefore, the in vivo data support the conclusion that hybrid dimers respond to cytidine. Presumably, half-saturated wild-type dimers behave in the same manner.
These results, obtained by monitoring the effect of cytidine on CytR-regulated gene expression in vivo, are remarkably consistent with those obtained by in vitro investigation of allosteric coupling between cytidine binding to CytR and cooperative interaction of CytR and deoP2⅐(CRP⅐cAMP) 2 . This consistency supports a molecular model in which cytidine binding switches CytR between three states and binding of cytidine to the first subunit of CytR dimer yields complete induction. Because cytidine binding to CytR is negatively cooperative, the third CytR state, that generated by ligation of both subunits by cytidine, forms only at very high cytidine concentrations. Such concentrations are not normally attained in E. coli nor were they attained in our experiments. We note the similarity to CRP, which also has three distinct conformational and functional states corresponding to its cAMP ligation states (47).
Why does CytR differ from other LacI family members for which ligand binding affects the poise of an equilibrium between DNA binding and non-DNA binding conformations? We envision two possibilities. First, perhaps the CytR inducer binding core domain does undergo a similar conformational change when cytidine binds, but this is uncoupled from the structure of the DNA binding domain. Comparing the sequences of the LacI repressors, we note that where other family members have a pair of conserved alanines in the hinge helix sequence, CytR has proline (Pro 57 ) and glycine (Gly 59 ) (3). Perhaps, instead of hinge helices, CytR has an extended coil that only loosely tethers the Core and DNA binding domains, such that DNA binding is unaffected by the conformation of the Core dimer. Second, perhaps cytidine binding is uncoupled from the T-state to R-state transition, but is instead coupled to transitions between subconformations belonging to the quarternary R-state. With this scenario, it is interesting to speculate whether saturation by cytidine might induce a T-state transition and whether this would have an effect on DNA binding.