cAMP is an allosteric modulator of DNA-binding specificity in the cAMP receptor protein from Mycobacterium tuberculosis

Allosteric proteins with multiple subunits and ligand-binding sites are central in regulating biological signals. The cAMP receptor protein from Mycobacterium tuberculosis (CRPMTB) is a global regulator of transcription composed of two identical subunits, each one harboring structurally conserved cAMP- and DNA-binding sites. The mechanisms by which these four binding sites are allosterically coupled in CRPMTB remain unclear. Here, we investigate the binding mechanism between CRPMTB and cAMP, and the linkage between cAMP and DNA interactions. Using calorimetric and fluorescence-based assays, we find that cAMP binding is entropically driven and displays negative cooperativity. Fluorescence anisotropy experiments show that apo-CRPMTB forms high-order CRPMTB–DNA oligomers through interactions with nonspecific DNA sequences or preformed CRPMTB–DNA complexes. Moreover, we find that cAMP prevents and reverses the formation of CRPMTB–DNA oligomers, reduces the affinity of CRPMTB for nonspecific DNA sequences, and stabilizes a 1-to-1 CRPMTB–DNA complex, but does not increase the affinity for DNA like in the canonical CRP from Escherichia coli (CRPEcoli). DNA-binding assays as a function of cAMP concentration indicate that one cAMP molecule per homodimer dissociates high-order CRPMTB–DNA oligomers into 1-to-1 complexes. These cAMP-mediated allosteric effects are lost in the double-mutant L47P/E178K found in CRP from Mycobacterium bovis Bacille Calmette-Guérin (CRPBCG). The functional behavior, thermodynamic stability, and dimerization constant of CRPBCG are not due to additive effects of L47P and E178K, indicating long-range interactions between these two sites. Altogether, we provide a previously undescribed archetype of cAMP-mediated allosteric regulation that differs from CRPEcoli, illustrating that structural homology does not imply allosteric homology.

Signal transduction is an essential process that allows cells to cope and respond to changes in their environment (1). Many signaling pathways rely on small molecules to transduce external stimuli to one or more effector proteins inside the cell (2). cAMP is an ancient, ubiquitous small molecule that serves as a second messenger in many signal transduction pathways, including regulation of gene expression in response to changes in environmental conditions (3)(4)(5).
The cAMP receptor protein (CRP) is a homodimeric transcription factor targeted by cAMP (6)(7)(8). Each CRP subunit harbors a structurally conserved N-terminal cAMPbinding domain that is covalently linked to a DNA-binding domain located in the C-terminal portion of the protein (9)(10)(11)(12)(13)(14). Solution biophysical and structural studies have shown that cAMP binding to the CRP from Escherichia coli (CRP Ecoli ) stimulates a large conformational change in the DNA-binding domains (Fig. 1, A and B, top) (10,13,(15)(16)(17). In contrast, structures of the CRP from Mycobacterium tuberculosis (CRP MTB ) in the apo-form and cAMPbound form reveal smaller cAMP-induced conformational changes (Fig. 1, A and B, bottom) (11,14). The CRP Ecoli and CRP MTB display structural differences in both the apo-state and cAMP-bound state, most notably in their DNA-binding domain orientations relative to the cAMP-binding domains (Fig. 1C). Additional differences are observed in the homodimer symmetry. In the CRP Ecoli , the two subunits in the apo-state are symmetric, but the cAMP-bound state shows asymmetry between the DNA-binding domains' conformation (18) (Fig. 1D, top). Conversely, the subunits in the CRP MTB in the apo-state are asymmetric at the dimer interface helix (c-helix) and the DNA-binding domains, but the cAMP-bound state is highly symmetric (Fig. 1D, bottom). Finally, the CRP MTB harbors two additional α-helices, one at the N terminus and another at the C terminus, resulting in a slightly larger protein than the CRP Ecoli (Fig. 1E).
The CRP Ecoli and CRP MTB show structural differences, and their functional response to cAMP binding also differs. In the CRP Ecoli , cAMP binding enhances the affinity of the protein to tightly interact with pseudopalindromic gene promoter sequences involved in carbohydrate metabolism (18)(19)(20)(21)(22)(23). In the Figure 1. Structural comparison between the CRP Ecoli and CRP MTB . The CRP has a conserved structural organization with two identical subunits, each one harboring a cAMP-binding domain in the N terminus (pale cyan in the CRP Ecoli ; tan in the CRP MTB ) and a DNA-binding domain in the C terminus (dark teal in the CRP Ecoli ; dark brown in the CRP MTB ). A, structures of the CRP Ecoli and CRP MTB in the absence and presence of cAMP. B, alignment of apo-subunit (light teal and light brown) and cAMP-bound subunit for both the CRP Ecoli (top) and CRP MTB (bottom). C, alignment of the CRP Ecoli and CRP MTB homodimers. D, alignment of intraspecies monomers (left, apo-state; right, cAMP-bound state) for the CRP Ecoli (top) and CRP MTB (bottom). E, the CRP Ecoli and CRP MTB sequence alignment with mapped secondary structures (α-helices in blue; β-strands in light brown); (asterisk indicates residue identity; colon indicates similar residues, and dot indicates weakly similar residues). Differences in the sequence between the CRP MTB and CRP BCG are located at positions E178 (red spheres) and L47 (light blue spheres). cAMP is shown as brown spheres. See Experimental procedures for description of alignment and Protein Data Bank codes. CRP, cAMP receptor protein; CRP Ecoli , CRP from Escherichia coli; CRP MTB , CRP from Mycobacterium tuberculosis.

Role of cAMP in MTB transcription regulation
CRP MTB , the affinity for DNA promoters with and without cAMP appears to be similar (12,20,24). The small cAMPinduced conformational change in the CRP MTB provides a structural explanation by which this protein does not change its affinity to DNA upon binding to the cyclic nucleotide (6,12,16,25). It is therefore possible that the CRP MTB is not sensitive to cAMP, but previous studies have shown that cAMP interactions with the CRP MTB are important in the regulation of the gene whiB1 (12,26). It is therefore possible that the allosteric regulation triggered by cAMP in the CRP MTB may not be directly associated to large changes in protein conformation that enhance the affinity for specific DNA promoter sequences, as seen in the CRP Ecoli . To dissect the mechanisms by which cAMP allosterically regulates CRP MTB -DNA binding, in this study, we quantitatively characterize the linkage between cAMP and DNA interactions. We combine complementary solution biophysical approaches to measure cAMP-binding affinity and cooperativity, interactions with the DNA promoter SerC (6,27) as a function of cAMP concentration, and protein solution structure, assembly, and thermodynamic stability.
The results from this study reveal that the CRP MTB binds cAMP with moderate negative cooperativity. In agreement with previous reports (12), the affinity of the CRP MTB for promoter sequences is similar in the presence and in the absence of cAMP, indicating that the cyclic nucleotide does not regulate transcription at the level of affinities to specific DNA promoter sequences. We find that in the apo-state, the protein forms high-order CRP MTB -DNA oligomers. These oligomers are mediated by interactions between the CRP MTB and nonspecific DNA sequences, and by interactions between a CRP MTB -DNA complex and the free CRP MTB . Unexpectedly, the presence of cAMP decreases nonspecific interactions with DNA and reversibly dissociates high-order CRP MTB -DNA oligomers into stable, 1-to-1 stoichiometric complexes. We also investigated the double-mutant L47P/E178K, which is found in the CRP from the attenuated Mycobacterium bovis Bacille Calmette-Guérin strain (CRP BCG ) (mutation sites shown in Fig. 1A, bottom) and only differs from the CRP MTB sequence in those two amino acid residues (28)(29)(30). While the CRP BCG displays negative cAMP-binding cooperativity like the CRP MTB , we find that cAMP does not prevent the formation of high-order CRP BCG -DNA oligomers. These functional differences are not observed in the single mutants L47P (CRP MTB -L47P) and E178K (CRP MTB -E178K), indicating nonlinear contributions and long-range interactions between the two mutation sites. In agreement with nonlinear mutant contributions, the thermodynamic stability and dimerization constant of the CRP BCG are also different from the single mutants.
In combination, these results provide an archetype of cAMP-mediated regulation that is significantly different from those described previously in other CRPs, such as the wellcharacterized E. coli homolog, and illustrate that structural homology does not imply allosteric homology. In other words, two structures could be very similar but respond very differently to the same allosteric effector. The error corresponds to the SD from fitted parameters using a two-site binding model as described in Experimental procedures. The units of k 1 and k 2 are 10 4 M −1 and c = k 2 /k 1 .  from left to right: cacodylate, PBS, and Hepes buffer. The error bars correspond to the SD from three to four experimental repeats. We find that ΔH 1 in each buffer is not statistically indistinguishable (p = 0.23), but ΔH 2 shows statistical pairwise differences among all buffers (p = 3 × 10 −4 ). C, cAMP binding monitored by changes in ANS fluorescence. The solid line represents the fit using a two-site binding model (Equation 16 in Experimental procedures). CRP MTB , CRP from Mycobacterium tuberculosis; ITC, isothermal titration calorimetry; ΔH 1 , enthalpy change for the first cAMP-binding site; ΔH 2 , enthalpy change for the second cAMP-binding site; ANS, 8-anilino-1naphthalenesulfonic acid.

CRP MTB exhibits negative cooperativity between the two cAMP-binding domains
We first used isothermal titration calorimetry (ITC) to quantitatively determine the cAMP-binding affinity constants, cooperativity, and their underlying thermodynamic driving forces ( Fig. 2A, Table 1). To ensure full saturation of the CRP MTB , we conducted ITC experiments using up to a fivefold molar excess of cAMP to protein. By fitting the ITC data to various binding models, we found that a two-site sequential binding mechanism (19, 24) provided better fitting statistics than a set of independent binding sites ( Fig. S1 and Table S1). The site-specific binding constants we obtained were k 1 = (3.0 ± 0.9),10 4 M −1 and k 2 = (1.2 ± 0.2),10 4 M −1 (all errors represent the SD of the fit). The ratio between the two binding constants, c = k 2 /k 1 = 0.4 ± 0.2, indicates negative cooperativity between the two cAMP-binding domains (Table 1). ITC experiments using various buffers (Hepes, PBS, and cacodylate, Fig. S2 and Table S1) were performed to dissect a potential contribution of proton ionization to the observed cAMPbinding enthalpies and to determine whether the release or uptake of protons is associated with the cAMP-binding reactions (31,32). We find that both cAMP-binding events are endothermic (enthalpy change for the first cAMP-binding site = 4.7 ± 0.3 kcal mol −1 and enthalpy change for the second cAMP-binding site = 5.0 ± 1.0 kcal mol −1 ) and therefore entropically driven (TΔS 1 = 10.8 kcal mol −1 and TΔS 2 = 10.5 kcal mol −1 ). Moreover, we find that the first cAMPbinding event is independent of the buffer-ionization enthalpy, whereas the second one displayed a slope of 1.0 ± 0.3, indicating proton uptake by the protein (Fig. 2B). The asymmetry in proton uptake during cAMP binding may be a consequence of the asymmetry seen in the apo-CRP MTB structure (Fig. 1D, bottom) or asymmetric states in partially liganded conformations (19,22). Importantly, in all three buffers used in this study, a two-site sequential binding mechanism resulted in better fitting statistics (Table S2) and revealed negative cAMP-binding cooperativity ( Table 2). In addition to ITC experiments, we monitored cAMP binding via changes in 8-anilino-1-naphthalenesulfonic acid (ANS) fluorescence ( Fig. 2C) (19,24,33). The data were fitted to two binding polynomials, wherein a model allowing for cooperativity provided a statistically better fit than a model with independent binding sites ( Fig. S3 and Table S2). Moreover, the binding constants obtained from the ANS-based assay are in agreement with the results using ITC and support the observed negative cooperativity between the two cAMPbinding sites: k 1 = (3.5 ± 0.2),10 4 M −1 and k 2 = (2.2 ± 0.1), 10 4 M −1 and c = 0.6 ± 0.1 (Table 1).

CRP MTB -DNA interactions as a function of cAMP concentration
Previous structural studies have shown that binding of cAMP induces small conformational changes to the DNAbinding domain of the CRP MTB , allowing the protein to switch from an asymmetric structure to a symmetric, active Light purple and dark pink squares correspond to the apo-conformation and doubly cAMP-bound conformation, respectively. The solid line corresponds to a control experiment with the buffer added instead of the protein. E, the CRP MTB -DNA complex formation using a 20-bp-long SerC promoter (3 nM). Dark pink and light purple circles correspond to the cAMP titration curve for the apo-conformation and doubly cAMP-bound conformations, respectively. F, stoichiometric binding using 400 nM of DNA (20-bp-long SerC) for the apo-conformation (light purple) and doubly cAMP-bound (dark pink) conformations. The dashed line denotes the concentration of the CRP by which the formation of the CRP MTB -DNA complex is at a 1-to-1 molar ratio. Solid lines in panels A and E are the fit as described in Equation 18 in Experimental procedures. In all panels, error bars correspond to the SD of 3 to 5 experimental repeats. CRP MTB , CRP from Mycobacterium tuberculosis; CRP, cAMP receptor protein. Role of cAMP in MTB transcription regulation conformation (11,14) (Fig. 1D, bottom). However, the effect of this conformational transition on the affinity between the CRP MTB and DNA promoter sequences is not fully understood. Thus, we investigated the role that cAMP occupancy plays in the formation of the CRP MTB -DNA complex. The formation of the CRP MTB -DNA complex was monitored via changes in fluorescence anisotropy, normalized to the first protein concentration point (19,20). We used a 32-bp fluorescein-labeled SerC promoter, a well-characterized promoter targeted by the CRP MTB (6,14,27,34). These experiments were conducted using 0, 0.1, and 1 mM of cAMP. At these concentrations, the protein is in the apo-state, in a mix of singly and doubly cAMP-bound states (based on the binding affinity constants determined in this study) and in the doubly cAMP-bound state, respectively (Fig. S4). In all three cAMP concentrations, the anisotropy of the labeled promoter increased as a function of the protein concentration, indicating that the formation of the CRP MTB -DNA complex occurs even in the absence of cAMP (Fig. 3A), a result that is in agreement with Rickman et al. Altogether, these results indicate that the allosteric linkage initiated by cAMP binding is not associated with enhancing the binding affinity for specific DNA promoter sequences.
cAMP prevents the formation of high-order CRP MTB -DNA oligomers While the DNA-binding affinity constants were similar in all three concentrations of cAMP, we did observe important differences in the anisotropy signal toward the end of the titration (Fig. 3A, red arrow). In the absence of cAMP, the anisotropy signal gradually increased after the DNA-binding transition. In contrast, in the presence of cAMP (0.1 or 1 mM), the anisotropy signal remained nearly constant after the DNA-binding transition. This distinctive behavior suggests the formation of high-order CRP MTB -DNA oligomers in the apo-state that are prevented or reduced when the protein is bound to cAMP.
We confirmed these results by conducting stoichiometric DNA-binding assays using a concentration of the SerC promoter that is 10 to 20 times larger than the dissociation constant (K d ) ( Table 3). These experiments revealed a linear increase in the anisotropy signal that plateaus at a 1-to-1 molar ratio of protein to DNA, demonstrating that one molecule of the CRP MTB binds to one molecule of DNA (Fig. 3B). In the absence of cAMP, we observe an overlap with the titration curve obtained with cAMP until a protein-to-DNA molar ratio of 1. However, after the 1-to-1 molar ratio is reached, the anisotropy signal in the apo-state continues to rise steadily, indicating again the formation of high-order CRP MTB -DNA oligomers.
In agreement with previous reports (12), our results show that cAMP does not have a large effect on DNA-binding affinities. However, titrations with a molar excess of protein to DNA, either with or without the cyclic nucleotide, suggest a noncanonical role for cAMP in allosteric signaling. Namely, that cAMP binding to the CRP MTB prevents the formation of high-order DNA-protein oligomers. We therefore sought to determine what intermolecular interactions are involved in the formation of these oligomers and how cAMP binding prevents their formation.

cAMP reverses preformed CRP MTB -DNA oligomers
Our previous experiments show that cAMP prevents the formation of high-order CRP MTB -DNA oligomers, yet to be determined is whether cAMP can reversibly dissociate such oligomers in a preformed state. To address this question, we preformed high-order CRP MTB -DNA oligomers and monitored changes in anisotropy as a function of the cAMP concentration ( Fig. 3C). In these experiments, we used a concentration of CRP MTB = 230 nM and the 32-bp fluoresceinlabeled SerC promoter = 3 nM. At these concentrations of protein and DNA, we obtained the highest normalized anisotropy value that is experimentally accessible in the absence of cAMP, around 1.25 (Fig. 3A). Figure 3C shows that upon titration of cAMP, the anisotropy signal of preformed CRP MTB -DNA oligomers decreases systematically, whereas in the absence of cAMP, the anisotropy remained constant. Importantly, the change in normalized anisotropy (0.09) upon cAMP binding is consistent with the difference in the normalized anisotropy signals seen between titration curves of the protein in the apo-state (1.25) and cAMP-bound states (1.15) (Fig. 3A). This quantitative agreement indicates that the decrease in anisotropy during the titration of cAMP corresponds to the reversible dissociation of high-order CRP MTB -DNA oligomers into a 1-to-1 complex. Furthermore, we fitted the changes in anisotropy as a function of cAMP to a single-site binding isotherm, which reflects the affinity of the preformed CRP MTB -DNA complex for cAMP. The apparent binding affinity constant was (6.3 ± 1.5), 10 4 M −1 , a value that is three times higher than the affinity of the first cAMP-binding site in the absence of DNA (Table 1). A two-site binding model did not improve the residuals of the fit (data not shown), suggesting that only one cAMP molecule per CRP dimer is sufficient to reversibly dissociate high-order CRP MTB -DNA oligomers. Given that apo-CRP MTB binds cAMP with modest negative cooperativity, it is possible that the anisotropy assay cannot detect small differences in affinity between one or two cAMP-binding events.

CRP MTB binds nonspecifically to DNA in the absence of cAMP
To begin uncovering the molecular interactions that stabilize high-order CRP MTB -DNA oligomers, we first studied nonspecific DNA interactions using a 32-bp fluoresceinlabeled scramble sequence. Because the scramble sequence lacks the conserved SerC-binding site (6), any changes in anisotropy would reflect nonspecific binding of the CRP MTB to Role of cAMP in MTB transcription regulation DNA. Figure 3D shows the titration curve of the CRP MTB to the 32-bp scramble DNA sequence in the absence and presence of 1 mM cAMP. The titration shows identical anisotropy values in both conditions (i.e., apo-state and cAMP-bound state) up to a concentration of protein of 80 nM. At higher protein concentrations, we observe an increase in anisotropy values only in the absence of cAMP, indicating that the protein in the apo-state is binding to DNA in a nonspecific manner. The control titration (solid black line, Fig. 3D), where only the buffer was added instead of the protein, shows negligible changes in anisotropy. These results suggest that the formation of high-order CRP MTB -DNA oligomers in the absence of cAMP can be driven by interactions with nonspecific DNA sequences.
Formation of CRP MTB -DNA complexes with shorter promoter sequences Next, we explored the nature of nonspecific DNA binding. We reasoned that the increase in anisotropy fluorescence in the absence of cAMP could arise from (1) apo-proteins interacting nonspecifically to flanking sequences outside the DNA footprint region or (2) binding of free proteins to preformed CRP MTB -DNA complexes. To distinguish between these two scenarios, we used SerC promoter sequences of decreasing lengths, down to the DNA footprint of the CRP MTB based on the high-resolution structure, 18 bp (11,14).
First, we determined the shortest DNA fragment that stably binds to the CRP MTB . We performed EMSAs using six different lengths of the SerC promoter (18,20,22,24,26, and 32 bps) (Fig. S5). Our data show that 20-bp is the minimum base pair length required for the CRP MTB to bind robustly to DNA. A 20-bp SerC promoter sequence would only have 1-bp of overhang on each side, thereby minimizing potential protein interactions to DNA flanking regions.
The DNA binding constant for the 20-bp SerC promoter was quantitatively determined by fluorescence anisotropy. Figure 3E shows the titration curve of the CRP MTB in the absence and presence of 1 mM cAMP. The DNA binding constants for the apo-state and cAMP-bound state were k DNA(apo) = (0.12 ± 0.02),10 8 M −1 and k DNA(cAMP-2) = (0.5 ± 0.1),10 8 M −1 , respectively. These values are 10-fold lower than the binding constant for the 32-bp-long SerC promoter (Table 3). Importantly, stoichiometric binding assays shown in Figure 3F demonstrate that even in the almost complete absence of DNA flanking regions, there is still formation of high-order CRP MTB -DNA oligomers when cAMP is absent. We interpret this result as the free CRP MTB binding to preformed CRP MTB -DNA complexes. Together with our previous results using the 32-bp scramble DNA sequence (Fig. 3D), our data indicate that these oligomers can be mediated by both nonspecific interactions between the protein and DNA and the binding of the free CRP MTB to a preformed CRP MTB -DNA complex.

Effect of mutations L47P and E178K on cAMP-binding affinity and cooperativity
The CRP BCG only differs in two amino acids at positions L47P and E178K relative to the CRP MTB , which are located in the cAMP-binding domain and the DNA-binding domain, respectively (Fig. 1A, bottom). These mutations, which are not present in other CRP orthologs found in M. bovis, M. tuberculosis, or Mycobacterium leprae, have been implicated as potential contributing factors to the attenuation of BCG strains (28,30). However, it remains unclear how the CRP BCG differs from the CRP MTB in its interaction mechanisms with cAMP or what are the contributions of each individual mutation toward cAMP-binding affinities and cooperativity.
To answer these questions, we placed the individual mutations on the CRP MTB (termed CRP MTB -L47P and CRP MTB -E178K) or the two together (CRP BCG ) and determined their cAMP-binding affinities and cooperativity by monitoring changes in ANS fluorescence. Our data show that the affinity constants for the first cAMP-binding site (k 1 ) are similar among the three CRP mutants, but the affinity for the second site (k 2 ) was significantly lower for the CRP BCG (Fig. 4A, Table 4). As a result, the cAMP-binding cooperativity ranges from neutral for CRP MTB -E178K and CRP MTB -L47P (c = 1) to negative for the CRP BCG (c = 0.3) (Fig. S6). These results suggest that the cAMP-binding mode of the CRP BCG is not attributed to a single mutation or a simple linear addition between the effects of the two individual mutations.

Nonlinear effects of BCG mutations on CRP-DNA interactions
Given the nonlinear contributions of the individual BCG mutations over cAMP binding, we investigated the role of each mutation on DNA interactions using the 32-bp fluoresceinlabeled SerC promoter. In the absence of cAMP, we found that CRP MTB -L47P has a DNA-binding affinity similar to that of the CRP MTB . However, CRP MTB -E178K binds to the promoter sequence with a 10-fold enhancement (Fig. 4B, Table 4). Because E178K is located at or is near the DNAinteraction surface (Fig. 1, bottom) and the mutation involves a change from a negatively to a positively charged amino acid side chain, it was not unexpected to observe a higher DNAbinding affinity than the CRP MTB or CRP MTB -L47P. The unexpected result was that the CRP BCG binds DNA with an affinity similar to that of the CRP MTB or CRP MTB -L47P, indicating that the enhancing DNA-binding affinity effect of E178K is largely reduced by the presence of L47P.
CRP MTB -L47P and the CRP BCG revealed an important difference in the formation of high-order CRP-DNA oligomers. In the absence of cAMP, both proteins did not form oligomers as pronouncedly as the CRP MTB and CRP MTB -E178K. For example, at a concentration of the CRP MTB of 50 nM (with [DNA] = 3 nM) the presence of oligomers becomes very evident and pronounced for the CRP MTB and CRP MTB -E178K (red arrow in Figs. 3A and 4B, respectively). Neither CRP MTB -L47P nor the CRP BCG forms noticeable CRP-DNA oligomers (Fig. 4B). These results again highlight nonlinear contributions of each BCG mutation toward both DNA-binding affinities and reduction in the formation of high-order CRP-DNA complexes. By comparison, the functional phenotype of the CRP BCG is dominated by the contributions of the L47P mutation. Interestingly, the location of L47P is in the cAMPbinding domain, but its dominant effect over DNA interactions indicates long-range allosteric communication between cAMP-and DNA-binding domains.

Role of cAMP in MTB transcription regulation
Effect of cAMP on CRP BCG -DNA interactions We showed that the CRP MTB and CRP BCG have similar cAMP-binding affinity constants and display negative cooperativity (Tables 1 and 4). Here, we examined the linkage between cAMP and DNA binding for the CRP BCG . We monitored changes in anisotropy upon the formation of the CRP BCG -DNA complex (using the 32-bp fluorescein-labeled SerC promoter) at three cAMP concentrations: 0, 0.1, and 1 mM. At [cAMP] = 0.1 mM, 60% of the CRP BCG population corresponds to the singly bound conformation, whereas the other 40% corresponds to doubly bound conformation. At 1 mM of cAMP, 90% of the population corresponds to a doubly cAMP-bound state, thus almost reaching a saturated state (Fig. S7). The anisotropy data revealed indistinguishable DNA-binding constants in all cAMP concentrations: k DNA(apo) = (5.6 ± 3.0)⋅10 8 M −1 , DNA-binding affinity constant in the singly cAMP-bound state = (6.0 ± 5.3)⋅10 8 M −1 , and k DNA(cAMP-2) = (7.5 ± 3.4)⋅10 8 M −1 (Fig. 4C, Table 4). This result is consistent with titrations with the CRP MTB that shows small cAMP effects over the interaction with the specific promoter-binding site. In contrast with the CRP MTB , the effect of cAMP binding in reducing high-order CRP-DNA oligomers was negligible for the CRP BCG . This was also observed in stoichiometric binding assays (Fig. 4D). We therefore explored the effect of cAMP on DNA interactions for each individual mutant (Fig. 4, E and F). We find that both the CRP MTB -L47P and CRP MTB -E178K form high-order CRP-DNA oligomers in the absence of cAMP, which are significantly reduced in the presence of intermediate (0.1 mM) or saturating (1 mM) amounts of cAMP (Fig. 4, E and F). Altogether, our DNA binding data are consistent with cAMP-binding studies that indicate asymmetric contributions of individual mutations to the CRP BCG homolog. In this case, their influence on DNA interactions (specific or nonspecific) does not follow a simple linear combination, an analogous observation to results obtained in cAMP-binding assays (Fig. 4A).
Solution structure and stability of the CRP BCG differs from single mutants and the CRP MTB The functional differences observed between the CRP BCG and the single mutants CRP MTB -L47P and CRP MTB -E178K or the CRP MTB may arise from differences in their native solution structure, assembly state, or stability. The protein solution structure and assembly were evaluated by using three biophysical methods: intrinsic protein fluorescence, CD, and analytical ultracentrifugation (AUC). The intrinsic fluorescence emission spectra were similar for all four CRPs, indicating that the tertiary structures surrounding the tryptophan residues (with excitation wavelength at 295 nm) are largely unaffected by the mutations (Fig. 5A). Spectra obtained using an excitation wavelength of 280 nm that includes the contribution of one tyrosine residue per subunit show no differences between all four CRPs (data not shown). Similarly, the CD spectra for all CRPs overlapped, indicating that the global native fold and secondary structure content of the proteins are the same (Fig. 5B).
The degree to which CRP is a stable homodimer was assessed by AUC (i.e., sedimentation velocity [SV]). Experiments conducted at monomer concentrations ranging between 1 and 40 μM showed, with exception of the CRP BCG , a constant sedimentation coefficient corresponding to the molecular mass of the homodimer (52.2 kDa) and a dimerization K d lower bound of 10 nM. This result indicates that the CRP MTB , CRP MTB -L47P, and CRP MTB -E178K were stable homodimers at concentrations of protein used throughout these studies (Fig. 5C). In contrast, the CRP BCG showed monomer-dimer association processes (Fig. 5C, cyan). Nonlinear square fitting of the SV data indicate that the double mutant has a significantly lower dimerization K d of 17.5 μM. This result suggests that the CRP BCG was in a monomeric state in the DNAbinding assays conducted with a protein concentration in the nanomolar range. However, stoichiometric DNA-binding data (Fig. 4D) show a plateau when the DNA and CRP BCG dimer concentrations reach a 1-to-1 ratio. This suggests that at equilibrium each CRP BCG dimer forms a stable 1-to-1 complex with DNA. The alternative scenario, in which the CRP BCG monomers were to form stable complexes with DNA, would reach a plateau in stoichiometric DNA-binding assays at molar ratios lower than 1. Thus, our results suggest that the CRP BCG readily dimerizes when it binds DNA, but the dissection and quantification of the linkage between CRP BCG dimerization and DNA interactions (with and without cAMP) remains unknown and is currently being investigated.
To determine mutational effects on protein stability, we monitored changes in tryptophan fluorescence (Fig. 5D) and CD (Fig. 5E) as a function of guanidine hydrochloride. Although all CRPs have indistinguishable tryptophan emission and CD spectra in their native state, the unfolding titrations  (Table 5). It is possible that the lower stability is due to mutational perturbations of the tryptophan residue (W202) located in the DNA-binding domain, which is in close proximity to E178K. Experiments monitored with CD showed a different pattern: CRP MTB -L47P has the lowest unfolding free energy (ΔG 3.9 kcal mol −1 ) and the smallest m-value (m) (-2.5 kcal mol −1 M −1 ) compared with the other proteins (Table 5). L47 is a fully buried residue (0% accessible surface area), and therefore, mutations in this position may contribute to a global destabilization of the protein. Interestingly, the unfolding curve of the CRP BCG monitored by CD displayed two clear unfolding transitions (Fig. 5E). All the other CRPs displayed one unfolding transition, which was analyzed using a two-state unfolding model. These results again illustrate the nonlinear contributions and effects of the individual mutants to the CRP BCG . Given that unfolding experiments were conducted using 5 μM of protein, which is three times lower than its dimerization constant, it is possible that the two transitions observed for the CRP BCG represent unfolding transitions of monomers and dimers in the solution.

Discussion
Elucidating the role of cAMP signaling in M. tuberculosis is a biomedically important topic because cAMP plays an important role in virulence and host interactions (4). Despite the relevant role that the CRP MTB plays in cellular processes, there is limited information regarding its mechanism of allosteric regulation of transcription by cAMP. This is in contrast to the well-studied CRP Ecoli , which shares high sequence and structural similarity with the CRP MTB : 53% sequence similarity and an r.m.s.d = 2.5 Å between all atoms in the cAMPbound structures (9,11,35). In this study, we use several biophysical approaches to investigate the linkage between cAMP binding and DNA interactions in the CRP MTB .

Physiological role of cAMP and the CRP MTB in M. tuberculosis
To survive the host's defense mechanisms, M. tuberculosis has developed a number of strategies that include the following: (1) interfering with phagosomal acidification and trafficking, (2) blocking autophagy and apoptosis-mediated killing, (3) perturbing calcium signaling, and (4) inhibiting inflammatory responses by modulating the host cytokine defenses and quenching the production of reactive oxygen and nitrogen species (5,36). Some of these strategies can be accomplished by elevating levels of cAMP inside the host cell. Elevated levels of cAMP can suppress innate immune functions by modulating protein expression of inflammatory mediators, dampening the phagocytic response, and reducing intracellular killing of ingested pathogens (5). The best studied microbial strategy for elevating cAMP levels inside the host is by producing toxins that include adenylyl cyclases themselves (M. tuberculosis has 17, compared with E. coli that has 1). One such adenylyl cyclases is Rv0386, which is linked to the production and secretion of cAMP within macrophages and whose deletion decreases M. tuberculosis virulence and pathology in mice (37,38). By using 14 C-radiolabeled M. tuberculosis, Agarwal et al. showed that the increase in cAMP was mediated by the bacteria rather than by the host macrophages, and it was dependent on Rv0386 (37).
While the intracellular concentration of cAMP in E. coli has been well determined (1-40 μM) (39)(40)(41), reports on the concentration of the cyclic nucleotide in M. tuberculosis and M. bovis show variation between 0.5 and 7 pmoles per 10 8 bacteria, depending on the growth media (42). M. tuberculosis has an irregular shape, ranging between a length of 2 and 4 μm and a width of 0.2 and 0.5 μm (43). Using this information and assuming a rod-shape morphology (43), the concentration of cAMP has lower and upper boundaries of 6.3 μM to 1.1 mM. Our cAMP-and DNA-binding studies indicate that at the lowest cAMP concentrations, CRP MTB -DNA oligomers formed via nonspecific interactions will be the dominant species, whereas at the highest concentration range, these oligomers will be lowly populated. Interestingly, the levels of cAMP inside M. tuberculosis from infected macrophages were reported to be 20 pmoles per 10 8 bacteria, which by a similar calculation as above result in an intracellular cAMP concentration range between 0.26 and 3.2 mM (42). At those concentrations, the reduction of CRP MTB -DNA oligomers will be almost complete. Thus, the modulation of the cAMP concentration before and after macrophage infection will be accompanied by direct effects over the interactions between the CRP MTB and DNA.

cAMP is an allosteric modulator of DNA-binding specificity
We provide evidence for a previously unrealized role of cAMP signaling, in which cAMP regulates the specificity of CRP MTB -DNA interactions. This is in contrast to its structurally conserved CRP Ecoli homolog, wherein cAMP controls the binding affinity to sequence-specific promoters (18)(19)(20)44). This new role of cAMP in the CRP MTB activation is supported by four experimental observations: first, fluorescence anisotropy experiments quantitatively show that the difference in CRP MTB -DNA affinities in the presence and absence of cAMP are marginal, a result that is in agreement with previous studies (7). This indicates that the bound cyclic nucleotide does not regulate transcription at the level of affinity to specific DNA promoter sequences (Fig. 3A, Table 3). Second, the observed anisotropy at high protein concentrations (relative to the concentration of DNA) is significantly higher in the absence of cAMP than in its presence. This difference is related to the formation of high-order CRP MTB -DNA oligomers that are prevented in the presence of cAMP or reversibly dissociated by adding cAMP after high-order CRP MTB -DNA oligomers are formed (Fig. 3, C and D). Third, from stoichiometric DNA-binding assays (Fig. 3B), we conclude that high-order CRP MTB -DNA oligomers only appear after 1-to-1 CRP MTB -DNA complexes have formed Role of cAMP in MTB transcription regulation (i.e., one CRP dimer and one DNA promoter). This result indicates that the absence of cAMP increases the affinity for nonspecific interactions between a preformed CRP MTB -DNA complex and apo-CRP MTB , between two preformed CRP MTB -DNA complexes, or both. Future AUC experiments will address the size distribution of these oligomers to determine their stoichiometries and relative populations. Fourth, we provide experimental evidence showing that a single cAMP molecule per CRP MTB dimer prevents nonspecific DNA interactions and reverse the formation of high-order CRP MTB -DNA oligomeric complexes.
It is important to note that the aforementioned role of cAMP in regulating DNA-binding specificity occurs at concentrations of the CRP MTB higher than 0.1 μM (Fig. 3A and Fig. 4, B, C, and E). If we consider the intracellular concentration of the CRP MTB similar to that reported for the CRP Ecoli , 2.5 μM (40), then CRP MTB -DNA oligomers will readily form when the concentration of cAMP is in the low micromolar range (<50 μM). From previous reports (42), we estimated the concentration of cAMP ranging between 6.3 μM and 1.1 mM. CRP MTB -DNA oligomers will be present at the lowest levels of cAMP, while dissociating at the highest level of cAMP from the range provided. These protein and cAMP concentration estimates therefore underscore the biological relevance of cAMP in regulating the CRP MTB specificity toward DNA sequences.
The ability of cAMP to modulate DNA-binding specificity is lost in M. bovis BCG The CRP ortholog of the attenuated M. bovis BCG (CRP BCG ), whose sequence only differs in two amino acids at positions L47P and E178K relative to the CRP MTB , exhibits significant differences in gene regulation (29). Furthermore, previous studies have shown that the CRP BCG has slightly higher DNA-binding affinities than the CRP MTB for the same promoter sequences (29). Studies dissecting the role of each mutation site in the CRP BCG showed that L47P, located at the cAMP-binding domain, had a greater effect in decreasing the ability of the protein to repress gene expression than E178K, located at the DNA-binding domain (30). Given these results, it has been proposed that the mutations observed in the CRP BCG play a significant role in the attenuation of M. bovis BCG (30). Here we investigated the mechanisms by which the CRP BCG differs from the CRP MTB in its interactions with cAMP and DNA.
We find that the CRP BCG binds the SerC promoter with a slightly higher affinity (2-fold) than the CRP MTB (Table 4). However, the presence of cAMP for the CRP BCG has no effects on the prevention of formation of high-order CRP-DNA oligomers (Fig. 4C). This result indicates that the allosteric control exerted by cAMP is largely reduced by the mutations L47P and E178K found in the CRP BCG . Which of these mutations is responsible for this new behavior? When the two mutations were investigated individually, we found that CRP MTB -L47P and CRP MTB -E178K behave similarly to the CRP MTB , namely, that the presence of cAMP reduced the formation of high-order CRP-DNA oligomers (Fig. 4, E and  F). Thus, our results suggest that the functional characteristics of the CRP BCG are not the consequence from a linear contribution of each individual mutant; rather, it is the result of cooperative interactions between the two mutation sites. For instance, the change in ANS emission due to cAMP binding is twice as large as compared with the single mutants. This is due to higher initial ANS-protein complex emission for the CRP BCG (data not shown), indicating that ANS has a different mode of interaction or that more ANS molecules bind to the CRP BCG , or both. In agreement with this conclusion, the thermodynamic stability and dimerization K d of the CRP BCG are different from that of the CRP MTB or the single mutants, indicating long-range interactions between the two mutation sites that give rise to unique functional and biophysical characteristics. Potential biological role for negative cAMP-binding cooperativity and DNA-binding specificity in transcription regulation mediated by the CRP MTB What are the functional consequences of having negative cAMP-binding cooperativity and regulation of DNA-binding specificity in the CRP MTB ? The observed negative cooperativity during cAMP-binding dictates that the probability of binding a second cAMP molecule to a singly cAMP-bound CRP MTB dimer is lower than the probability of binding an apo-CRP MTB dimer, resulting in a larger population of the singly cAMP-bound state. Given our results indicating that a single cAMP molecule bound to a CRP MTB dimer reverses high-order CRP MTB -DNA oligomeric complexes, we hypothesize that negative cAMP-binding cooperativity may maximize the pool of available cAMP to favor sequence-specific interactions between the CRP MTB and DNA. Thus, instead of requiring a doubly cAMP-bound CRP MTB to reduce nonspecific interactions or reverse high-order CRP MTB -DNA oligomeric complexes, our results indicate that one cAMP molecule is sufficient to achieve the same outcome.
The results from this study provide a previously unrecognized archetype of cAMP-mediated regulation of transcription that is different from previously described models for other CRPs. Figure 6 shows a model by which cAMP allosterically regulates CRP MTB interactions with DNA: (I) In the absence of cAMP, the CRP MTB binds to both specific (i.e., promoters) and nonspecific (i.e., intragenic) sequences of DNA or forms high-order CRP MTB -DNA complexes. (II) When cAMP levels increase inside the cell after macrophage infection, cAMP binds to a preformed CRP MTB -DNA complex. Because of the negative cooperativity of the CRP MTB , the first cAMP-binding event dominates over the second; thus, cAMP is preferably bound to a single subunit within homodimers. (III) While singly cAMP-bound proteins dissociate from nonspecific DNA sequences or dissociate high-order CRP MTB -DNA oligomers into 1-to-1 complexes, proteins that were interacting with specific DNA sequences remain bound to its promoter. (IV) Finally, as the cAMP concentration increases, the doubly cAMP-bound state is reached with mostly specific CRP MTB -DNA interactions taking place.
This model offers three scenarios by which the CRP MTB may regulate transcription and underscores its role as a global regulator. First, like removing roadblocks along the DNA structure (45,46), cAMP will trigger the dissociation of the CRP MTB from nonspecific DNA sequences. Second, CRP MTB -cAMP will remain bound to specific promoter sequences, facilitating the recruitment of the transcription machinery such as other transcription factors or RNA polymerase. Third, and a less studied role attributed to the CRP MTB , is chromosome organization (26). The CRP MTB binds to >900 sites in the M. tuberculosis genome, 83% of which are intragenic regions (47). This type of binding resembles that of nucleoidassociated proteins and suggests that the CRP MTB might regulate the global architecture of the mycobacterial chromosome (48). The ability of the CRP MTB to bend DNA (6) could also alter the interaction mode of other factors that interact to DNA proximal to CRP sites. In the context of our results, the property of CRP MTB -DNA to form oligomeric complexes and then dissociate as a function of cAMP concentration might be a strategy for regulating gene expression via chromosomal organization. Altogether, the interplay between these three mechanisms results in the expression of genes involved in virulence, such as ESX-1 type VII secretion system (T7SS), espACD-Rv3613c-Rv3612c operon, Rv3616c-Rv3612c genes, espA operon, to name a few, all of which are associated to CRP MTB activity (26). The CRP MTB also activates expression of rpfA and whiB1 genes that encode proteins that are thought to be involved in reviving dormant bacteria (12,25,26,49). Within this model, our results indicate that the variant CRP BCG has difficulties dissociating from nonspecific DNA sequences or reverse the formation of high-order CRP-DNA oligomers, possibly obstructing transcription.

Similar structures, different allosteric activation mechanisms
It is intriguing that the CRP MTB and CRP Ecoli share high sequence and structural similarity (11,35) but differ in their cAMP-mediated activation mechanisms. Although highresolution structures indicate that these two homologs are cAMP-dependent transcription factors, it is more difficult to infer from the structures alone that these proteins would have very different cAMP-binding modes and cAMPdependent DNA interactions. A close inspection of the two CRP structures reveals small differences that may be associated with their unique allosteric properties. For instance, the carboxy-terminal residues of the CRP dimerization helix in the apo-CRP Ecoli are not well structured, whereas in the CRP MTB they are (Fig. 1A). These residues are part of the hinge that connects the cAMP-and DNA-binding domains and have been shown to contribute to the allosteric communication in the CRP Ecoli (50)(51)(52). Moreover, cAMPinduced domain motions in the CRP MTB originate at the hinge that connects the cAMP-binding domain and the dimerization helix. Instead, in the CRP Ecoli domain, motions originate at the hinge that connects the DNA-binding domain and the dimerization helix (Fig. 1, B and D). The interplay between the sequence composition and the location of these domain motions may help further dissect the unique allosteric behavior in the CRP MTB and how cAMP reduces nonspecific DNA interactions.
Given that the structures of the CRP MTB in the apo-state and cAMP-bound state are similar, it is plausible that protein dynamics (16,18,21,(53)(54)(55)(56) also play an important role in how cAMP allosterically reduces nonspecific DNA interaction. Our cAMP-binding studies show differences in ANS fluorescence between apo-state and cAMP-bound state (Fig. 4A), despite their similar structures (Fig. 1, A and B, bottom). This change in ANS fluorescence indicates dynamic transitions or protein fluctuations associated with cAMP binding that are not captured in static high-resolution structures. Future studies with high-resolution techniques such as hydrogen-deuterium exchange mass spectrometry (57, 58) will help elucidating the residue networks involved in cAMP-mediated allostery and communication.
It is well documented that the CRP Ecoli exhibits positive cooperativity between the two cAMP-binding sites, wherein the first binding reaction is exothermic and the second is endothermic (18,19,24,33). In contrast, by using two orthogonal techniques (ITC and fluorescence), we find that the CRP MTB displays negative cAMP-binding cooperativity (Fig. 2, Table 1), where binding of the cyclic nucleotide is endothermic for both sites.
A previous study with ITC by Stapleton et al. (12) reported that the two cAMP-binding events in the CRP MTB are independent from each other. Our cAMP-binding studies using ITC and ANS fluorescence and the underlying statistical analysis of the data (Table S2) do not agree with their results. A potential source for the difference is that we used a cAMPto-CRP MTB molar ratio of up to 5-fold for ITC experiments, whereas Stapleton et al. reached a maximum ratio of 2.5. Given the expected 2-to-1 binding stoichiometry between cAMP and the CRP MTB , we used a higher molar ratio to ensure full saturation of the protein at the end of the titration. Alternatively, the buffer type and composition could be a source of differences between the two studies. We conducted ITC experiments at a constant pH using three buffers with different ionization enthalpies, including PBS, which was used by Stapleton et al. In all three experiments, we maintained a cAMP-to-CRP MTB molar ratio up to 5-to-1 and obtained the same degree of negative cooperativity between the two cAMPbinding sites ( Table 2, Figs. S1 and S2, and Tables S1 and S2). Altogether, the three ITC experiments and fluorescence measurements used in this study are consistent with each other and strongly suggest that cAMP binds with negative cooperativity to the CRP MTB .

Concluding remarks
In this study, we begin to dissect the linkage between cAMP binding and DNA interactions in the CRP MTB . Importantly, this study indicates that the linkage operates at a level of DNA regulation that is substantially different to that from the CRP Ecoli . Figure 7 illustrates the functional differences between the CRP MTB and CRP Ecoli at the level of DNA interactions. Although cAMP enhances the affinity of the CRP Ecoli for DNA promoter sequences and promote intersubunit communication (10,13,19,23,24,59,60), our results from this study show that DNA-binding affinity to the CRP MTB is not sensitive to cAMP occupancy. Instead, cAMP plays a significant role on the specificity of DNA interactions and the reduction or prevention of high-order CRP-DNA oligomers. The exact functional consequences of such a mode of action will likely depend on the specific organization of regulatory elements for a particular gene (12) and the degree of energetic coupling between the four binding sites in CRP MTB , i.e., binding at any site has the potential to alter the binding affinity of the other three sites.

Cloning, expression, and purification of the CRP MTB
The DNA sequence of WT CRP from M. tuberculosis (CRP MTB ) was used in the present study. CRP flanked by NdeI and BamHI restriction sites was synthesized by PCR with PfuUltra Polymerase (Agilent Technologies). The amplicon was digested with NdeI and BamHI according to manufactured directions (New England Biolabs). To generate the Histag fusion construct, the resultant digested fragment was inserted into a pET-3a expression vector (Addgene) previously digested with the same restriction enzymes. The resultant expression vector was named the CRP MTB .
CRP MTB mutants (E178K, L47P, and BCG [E178K/L47P double mutant]) were generated following the QuikChange II Site-Directed Mutagenesis protocol (Agilent Technologies). All proteins were purified from E. coli strain T7 Express pLysS competent cells (New England Biolabs). The bacteria were grown overnight, and protein expression was induced with 1-mM IPTG for 2 h. The bacteria were resuspended in an ice-cold lysis buffer (20- ketone, 10-μM 3-isobutyl-1-methylxanthine, 1-mM PMSF). The bacterial suspension was homogenized with a glass homogenizer and lysed with an M-110P Microfluidizer at 10,000 psi (Microfluidics). The lysate was centrifuged at 15,000 rpm for 45 min at 4 C in a JA 25.50 rotor (Beckman Coulter). The supernatant was mixed with the His60 Nickel Superflow Resin (Clontech) and allowed to bind overnight at 4 C with constant shaking. The supernatant was supplemented with 30-mM imidazole to compete for nonspecific binding. The next day, the resin-supernatant mix was transferred to a prewashed column with the lysis buffer. The flow-through sample was collected, and the resin was washed twice with lysis buffer supplemented with 3-mM imidazole. 500-mM imidazole was added in the lysis buffer, the resin was incubated for 30 min, and the elutes were collected. Samples corresponding to the CRP MTB were pooled together, concentrated, and run through sizeexclusion chromatography. All proteins in the apo-state had elution profiles displaying a single peak at elution volumes consistent with a dimeric state (Fig. S8). Aliquots were taken at every step of the purification protocol and loaded on to 10% SDS-PAGE gels to follow the purification process and assess the quality of the purified protein. The purified proteins were >95% homogenous as judged by SDS-PAGE. The CRP MTB were stored at −80 o C in the storage buffer (50-mM Hepes, pH 7.6, 150-mM KCl, 1-mM EDTA, pH 7.2). Protein concentrations throughout this study were determined with the dimer extinction coefficient at 280 nm: 25,480 cm −1 M −1 .

CD
Measurements were performed in an Aviv Model 202-01 spectrometer with 5-μM protein in a buffer containing 150-mM KCl, 50-mM Hepes, and 1-mM EDTA, pH 7.2, over a range of 195 to 260 nm. For each sample, three scans with three different protein samples preparations were performed, averaged, and baseline-corrected.

Chemical denaturation with guanidine hydrochloride
Protein unfolding was monitored by changes in fluorescence (λ ex = 280 nm or 295 nm, and λ em = 340 nm) and CD absorption at 222 nm. In both set of experiments, we used 5 μM of protein in a buffer containing 150-mM KCl, 50-mM Hepes, and 1-mM EDTA, pH 7.2. At least two independent titrations were performed for each protein and corrected for buffer contributions to the signal. Data were fitted according to the linear extrapolation method (61). For the WT CRP MTB and single mutants CRP MTB -E178K and CRP MTB -L47P, the data were fitted to a two-state unfolding model (61): where S T is the total observed signal, S N and S D correspond to the native and denatured state signals, respectively, and f N and f D are the fractions of native and denatured protein, respectively. f N and f D are related to the equilibrium constant between folded and unfolded states: where: And: Here, ΔG o H 2 O is the free energy of unfolding in the absence of a denaturant, m is the m-value or the slope of the linear dependence of ΔG o on the denaturant concentration as described by the linear extrapolation method (61), and d is the denaturant concentration. Combining Equations 1-5 yields the data fitting equation: The unfolding data for the CRP BCG displayed two transitions, and therefore, the total signal, S T , was fitted to a threestate unfolding model (62): where S and f are the signals and fractions of native (N), intermediate (I), and denatured (D) states, respectively. Expressing the fractions of species in terms of equilibrium constants yields: To better compare the unfolding data of the four CRPs studied here, we plotted the fraction of folded protein. All fitting procedures of unfolding data were performed in Sig-maPlot (Systat Software).

Analytical ultracentrifugation
SV experiments were performed in a Beckman Optima XLA with absorbance optics in 12-mm cells at 280 nm, 50,000 rpm, and 19.7 o C. The buffer density was measured in an Anton Paar DMA 5000. Extinction coefficients at 280 nm (12,740 M −1 per monomer or 0.4855 ml/mg) and vbar (0.735639) are estimated from amino acid sequence in Sednterp (63). Three samples were run at approximately 0.2, 0.4, and 0.6 absorbance at 280 nm (equivalent to dimer CRP concentrations of 7.9, 15.8, and 23.7 μM) in 150-mM KCl, 50-mM Hepes, and 1-mM EDTA, pH 7.2. Data were analyzed with DCDT+ (64) to generate g(s) distributions and plotted versus s* (Fig. 5C). Superposition of the WT CRP MTB and single mutants CRP MTB -E178K and CRP MTB -L47P g(s) curves is consistent with no concentration dependence in the concentration regime tested. Data were then fit with SedAnal (version 7.14) (65) to determine global S values and for the CRP BCG dimerization constants. The WT CRP MTB and single mutants CRP MTB -E178K and CRP MTB -L47P have an average S 20,w of 3.587 s ± 0.082 or 2.3% consistent with an estimate using HullRad (66), 3.65 s. The SV data for the CRP BCG were fit to a monomer-dimer model constraining the dimer S2 value to the individual and average values for the dimeric constructs or float S1 values which constrain the ratio of S2/S1 to 1.5. The best value for the CRP BCG dimerization is 5.7⋅10 4 M −1 or a K d of 17.5 μM (binding free energy -6.47 ± 0.73 kcal mol −1 ).

cAMP binding monitored by ITC
Experiments were performed in a Nano-ITC (TA instruments) using three different buffers (pH 7.2): Hepes, PBS, and cacodylate. Each buffer was supplemented with 150-mM NaCl, 1-mM EDTA, and 0.2-mM TCEP. All solutions were filtered and degassed thoroughly prior use. The protein and cAMP concentrations were 16 to 20 μM and 1 mM, respectively. The cAMP solution was prepared in the buffer from the last step of protein dialysis to minimize artifacts due to differences in the buffer composition. The reaction cell contained 0.35 ml of the protein solution. The injection syringe was filled with cAMP, and the titration experiment consisted of 18 injections. The first injection was of 0.5 μl and was discarded from the analysis step. The other 17 injections were of 2 μl. A separate reference titration of the cAMP into each buffer was performed to determine the heat of dilution of the ligand which was then subtracted from the cAMP titration to the protein solution. Raw data were analyzed using the software NITPIC (67) and MicroCal Origin using two different models: independent and sequential cAMP-binding events. The incremental heat (Q i ) of the titration was fitted using Equation 12: where V i and V 0 are the initial and active volumes, respectively. For the independent binding model, the total heat (Q) is the following: where n is the number of binding sites, [M] t is the bulk protein concentration, ΔH is the ligand-binding enthalpy, [x] t is the total ligand concentration, and K represents the binding constant. The total heat for the sequential binding model is calculated via Equation 14.
where [x] is Here, K 1 and K 2 correspond to the microscopic binding constants 2,k 1 and 1/2,k 2 , respectively.
We performed one-and two-way ANOVA tests to determine differences between ΔH and the buffer used in ITC experiments. To compare pairwise differences between buffers, we used the post hoc Tukey test with a significance level of 0.05. p-Values resulting from these tests are indicated in the legend of Figure 2. These tests were performed in Mathematica (Wolfram Research, Inc).
cAMP binding monitored by ANS fluorescence cAMP binding to the CRP MTB was measured by the quenching of the fluorescent signal from the CRP-ANS complex (λ ex = 350 and λ em = 480 nm), using a PTI spectrometer (Horiba). Normalized intensity counts as a function of the cAMP concentration were fitted to a cooperative two-site binding model as described (19) and an independent two-site binding model. The cooperative model is shown in Equation 16.
where F 480 nm is the observed signal; F 0 , F 1 , and F 2 represent the fluorescent signal of the apo, singly liganded, and doubly liganded states of the protein, respectively; k 1 and k 2 corresponds to the microscopic binding affinity constants of the first and second cAMP, respectively, and x is the concentration of ligand. In the independent binding model, k 2 = k 1 which assumes that the ligand binding sites are identical (i.e., no cooperativity). The ANS-based fluorescence data is normalized to the initial fluorescence value in the absence of cAMP. DNA biding curves were fitted with Sigma Plot (Systat Software).

Statistical analysis of cAMP binding data
We used the f-ratio function to determine which cAMP binding model is statistically more robust in fitting the data from ITC and ANS experiments. First, we determine f obs , which is the f-ratio calculated from the fitted parameters of cAMP binding data: where SSR is the sum of square of the residuals and ν correspond to degrees of freedom. The subindexes 1 and 2 refer to model 1, which in this case is identical and independent binding sites with no cooperativity, and model 2, which corresponds to identical binding sites with cooperative interactions. The values for SSR, ν and f obs are listed in Table S2. In its application, f obs is compared to the cumulative distribution of the f-ratio function using the corresponding degrees of freedom for models 1 and 2. This comparison provides a means to reject the hypothesis that model 1 is statistically equivalent to model 2 with a given confidence interval (68). Additionally, we calculated the probability that a value selected randomly from the f-ratio probability distribution exceeds f obs , i.e., the probability that model 1 provides a better fit than model 2. This is achieved by integrating the cumulative distribution of the f-ratio function from f obs to infinity. The results are listed in Table S2. These tests were done in Mathematica (Wolfram Research, Inc).

DNA binding monitored by fluorescence anisotropy
Measurements were collected with a PTI spectrometer using a 32-bp SerC promoter (5 0 -GCGCGTAGTGTGAA CAAGCTCACATGCA-AGCC-3 0 ), 20-bp SerC promoter (5 0 -AGTGTGAACAAGCTCACATG-3 0 ) and Scramble 32-bp DNA (5 0 -AGATCCGCAACATGTGTCGAACACACGCGGT A-3 0 ) covalently linked to a fluorescein molecule (IDTDNA). The excitation and emission wavelengths were 480 nm and 518 nm, respectively. The reaction mixture contained either 3 nM, 200 nM or 400 nM of fluorescein-labeled DNA and various concentrations of cAMP (0, 100 and 1000 μM). Fluorescence anisotropy measurements were collected with a PTI spectrometer (Horiba Scientific). Data was normalized to the first experimental anisotropy value, and analyzed as described previously by Heyduk and Lee (20) and Lanfranco et al. (19) with minor modifications. Briefly, we removed experimental data points displaying anisotropy values with 2 standard deviation higher that the plateau overserved after the first DNA-binding phase (indicated by the red arrow in Fig. 3A). The data were fitted according to Equation 18, where A obs is the observed anisotropy, A DNAF and A P-DNA are the anisotropy values for free DNA and the protein-DNA complex, respectively, [DNA T ] is the total DNA concentration, [P T ] is the total protein concentration, and K Role of cAMP in MTB transcription regulation represents the association constant for the protein and DNA.

Data availability
Data not contained in this article are available upon request to Rodrigo Maillard (rodrigo.maillard@georgetown.edu, Georgetown University).
Supporting information-This article contains supporting information.