Crystal Structure of a Cyclic AMP-independent Mutant of Catabolite Gene Activator Protein*

Escherichia coli NCR9l synthesizes a mutant form of catabolite gene activator protein (CAP) in which alanine 144 is replaced by threonine. This mutant, which also lacks adenylate cyclase activity, has a CAP* phenotype; in the absence of cAMP it is able to express genes that normally require CAMP. CAP91 has been purified and crystallized with cAMP under the same conditions as used to crystallize the wild type CAPOCAMP complex. X-ray diffraction data were measured to 2.4-8L resolution and the CAP91 structure was determined using initial model phases from the wild type structure. A difference Fourier map calculated between CAP91 and wild type showed the 2 alanine to threonine sequence changes in the dimer and also a change in orientation of cysteine 178 in one of the subunits. The CAP91 coordinates were refined by restrained least squares to an R factor of 0.186. Dif- ferences in the atomic positions of the wild type and mutant protein structures were analyzed by a vector averaging technique. There were small changes that included concerted motions in the small domains, in the hinge between the two domains and in an adjacent loop between ,&strands 4 and 5. The mutation at resi- due 144 apparently causes changes in the position of some protein atoms that are distal to the mutation site. The Escherichia coli catabolite gene activator protein (CAP)’ (l),

The Escherichia coli catabolite gene activator protein (CAP)' (l), also known as cAMP receptor protein (2), forms a complex with cAMP which regulates transcription from several operons, including those that encode enzymes involved in the catabolism of sugars such as lactose, maltose, and arabinose (3). In the absence of CAMP, CAP has a lower, sequence-independent affinity for DNA and does not regulate transcription. In adenylate cyclase-deficient mutants, added cAMP is normally required for CAP to function. One class of mutation in an adenylate cyclase-deficient background produces the CAP* phenotype in which transcription of the CAPdependent operons is regulated in the absence of added cAMP (4, 5). The crp genes of several CAP* mutants have been sequenced ( 6 4 , including the mutation in NCR91 that results from the base change C to T in the second position of codon 144 which substitutes threonine for the alanine of wild * Certain commercial equipment, instruments, or materials are identified in this paper in order to adequately specify the experimental procedure. Such identification does not imply recommendation or endorsement by the National Bureau of Standards. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
' The abbreviation used is: CAP, catabolite gene activator protein.
type CAP. The NCR91 mutation produces the protein which is designated CAP91.
CAP91 has been characterized in uitro; in the absence of added cyclic nucleotide, CAP91 activates the laCP1 promoter contained on a supercoiled DNA template (8). The addition of cAMP increases the apparent affinity of CAP91 for hcP1 DNA to the same level as wild type CAP with CAMP. The mutant shows related specificity for cyclic nucleotide since either cAMP or cGMP will activate CAP91 while wild type CAP is very specific for CAMP.
CAP is a dimer with subunit molecular weight of 22,500. The sequence of 209 amino acids has been determined from the DNA sequence of the cloned crp gene (9,lO). The crystal structure of the wild type CAP dimer with bound cAMP has been determined (11,12). Each subunit in the CAP dimer folds into two domains which are connected by a hinge region. The amino-terminal domain binds cAMP between a / 3 roll and a long a helix and the carboxyl-terminal domain binds to DNA. The CAP dimer structure is asymmetric: in the subunit in the more "closed" conformation the two domains are closer together than in the more "open" subunit (12,13). We report here on the structure of the CAP91 mutant crystallized with cAMP and the comparison of the CAP91 and wild type structures.

EXPERIMENTAL PROCEDURES
The gene for the mutant CAP91 of E. coli strain NCR91 has been cloned and sequenced (8). A single base change of C to T in codon 144 causes the substitution of threonine for alanine. CAP91 was purified to homogeneity from a strain of E. coli harboring a plasmid (pCRP91-37) in which the mutant crp gene was under the control of the thermoinducible PL promoter. Crystals of CAP91 were grown in the presence of cAMP under the same conditions as the wild type protein: 0.5 mM CAMP, 50 mM phosphate at pH 8.0, and room temperature (14). These crystals are isomorphous with the wild type CAP crystals with unit cell dimensions of a = 46.5, b = 96.7, and c =

A and space group P212121.
A single crystal measuring about 0.3 X 0.4 X 1.0 mm3 was used for x-ray data collection at room temperature. Diffraction data were measured on a NicoIet/Xentronics Imaging Proportional Counter which was positioned 12 cm from the crystal with the carriage angle set at 20'. In this configuration, data ranging from infinity to 2.3-A resolution were intercepted. Diffraction data were collected as a series of discrete frames or electronic images (15), each comprising a 15min oscillation. The individual frames were contiguous in that the start of each small oscillation range coincided with the end of the previous range. The raw data frames were transferred to a Digital Equipment Corp. VAX 11/780 computer for subsequent processing.
The crystal orientation and integration of reflections were performed with the XENGEN program system? A total of 61,342 observations were recorded which merged to give 19,328 unique reflections out of the 24,435 possible at 2.3-A resolution. The 12,447 reflections with differences occurring along the D helix. The highest peak in the difference map occurs near Cys-178 in the open subunit (Fig. 3a). This peak is at 7.0 u and corresponds to a movement of the side chain of Cys-178 which lies in the turn between CY helices E and F. There are no large differences around Cys-178 in the closed subunit, as shown in Fig. 3b. The negative density gave the largest peak of 5.5 u near 'the side chain of Thr-144, a peak at 5 u at CP of Cys-178 (both in the open subunit), and one at 4.8 (r near Gly-33 in the closed subunit. Smaller differences are also observed (4.5 a) at the end of the hinge in the closed subunit.
This initial difference Fourier confirms the site of the The refined coordinates of CAP3 were used for initial phasing of the CAP91 structure. The difference in x-ray amplitude between the data for CAP91 and CAP was 14.4% and includes both differences in the methods of data collection and in the crystal structures. A difference Fourier was calculated using the program, PROTEIN (la), between CAP91 and CAP (wild type) with the phases of the known structure. The difference Fourier was examined using FRODO (19) running on a PS330 computer graphics system attached to an IBM4381. The difference map clearly indicated the sequence change and also a movement of 1 Cys-178 and these changes were made to the coordinates to provide a starting model for refinement. The model coordinates were refined by restrained least squares procedures using the program, PROLSQ (20). and a fast Fourier version, PROFFT.4 This was alternated with manual adjustment using 2F,-Fc and F,-F, maps as a guide for repositioning the model coordinates. Initially, water molecules were positioned as in the original CAP structure. added at peaks in the 2F0-Fc or F,-F, density that were close to Several waters were removed during refinement and new waters were possible hydrogen bonding groups. When the refinement had converged and no further adjustments were deemed necessary the refined coordinates were analyzed and compared to the coordinates of the wild type structure, The difference in position of equivalent atoms in the two structures was computed and correlated motions were located by a vector averaging technique that minimizes random errors?

RESULTS AND DISCUSSION
Initial Difference Fourier Map-The structure of the CAP subunit is shown in Fig. 1 and the site of mutation from Ala-144 to Thr is indicated. A difference Fourier map was calculated for CAPSl-CAP(wild type) data with phases from the wild type structure. Difference density appeared as expected at the site of the mutation, residue 144, in the D helix: a peak at 5. The refined coordinates of the mutant, CAP91, were compared with the refined coordinates of the wild type CAP (R = 20.5%). The overall change between the two structures was evaluated by calculating the root mean square differences in atomic positions. The terminal residues were omitted from the comparison which is summarized in Table I. The root mean square distances calculated for CY carbon atoms are slightly smaller than for the side chains which tend to have more torsional freedom. Note that the two CAMP molecules are displaced less than main chain or a carbon atoms. This can be compared to results from other refined protein structures at high resolution. The root mean square displacements listed in Table I are  for all atoms. The two CAP subunits have been refined independently and have different crystallographic contacts as well as different conformations, whereas the mutant CAP91 was refined from the same starting model as CAP. Although the differences between the CAP91 and CAP structures are small, they are not distributed randomly as is described in the next section.
The major known change in the refined structure of CAP91 is the substitution of Thr for Ala-144.

Concerted Shifts in Local
Regions-The differences between the CAP91 and CAP structures were analyzed in more detail by various methods. There was no significant change in the radial distribution of atoms around the site of mutation or around CAMP. When the coordinates of CAP91 and CAP were displayed alternately on the computer graphics screen, correlated movements of atoms were seen, especially in the small domains. The movements were analyzed by calculating a vector average of the differences in atomic position for several consecutive residues. This method averages out random changes of direction and identifies regions with concerted motions? The average was taken over 11 consecutive residues to reduce refinement artifacts arising from the correlated motion of neighboring atoms which were coupled by stereochemical restraints.
The magnitude of the vector difference between CAP91 and CAP is plotted against sequence number in Fig. 5. The terminal residues with high thermal factors, residues 1-9 and These small motions are more readily examined in three dimensions: the vector between a carbon atoms of CAP and CAP91 is shown in Fig. 6a and the average vector difference is displayed in Fig. 6b The vector between equivalent a carbon atoms has been scaled by a factor of 3.0. b, the vector average is plotted on a stereo a carbon representation of the CAP dimer. The local vector averaged over all atoms of 11 residues is shown extending from the central (Y carbon atom. The vector is scaled by a factor of 5.

I L
Experiments in solution show that CAP91 in the absence of cAMP appears to have a similar conformation as the wild type CAP in the presence of cAMP (8). This has also been shown for other CAP* mutants (17). However, CAP91 exhibits half-maximal activation of transcription at a concentration of CAMP that is approximately 40-fold lower than required by wild type (8) (7); these have different phenotypes since the mutation with Phe-62 is not activated by cGMP unlike the mutation with His-53. The structure provides an explanation since residue 53 is on the protein surface and lies close to the E helix whereas residue 62 is normally serine and lies next to the adenine ring of the bound CAMP. Therefore changes at position 62 are likely to disrupt the binding of CAMP, especially a change to the large Phe side chain which might sterically interfere with the binding of cyclic nucleotides.
Two other mutations are located in the CAMP-binding pocket (6,8). One is in residue 127 which in CAP is Thr and forms a hydrogen bond with the 6-amino group of CAMP. Another results in the substitution of Ala for Glu-72 as part of a double mutation which would eliminate the hydrogen bond between Glu-72 and the ribose 2'-OH of CAMP. Therefore a CAP* phenotype may involve alteration or elimination of interactions between CAP and CAMP.
Conclusions-The refined structures of wild type CAP and the CAP91 mutant protein have been compared, revealing that the two structures with bound cAMP are very similar overall. However, small concerted movements occur in the residues of the hinge between the two domains of the closed subunit, in the adjacent loop of p4 to -5 in the open subunit, and in the turn between the DNA-binding E and F helices in both subunits. These changes are distal to the site of mutation at residue 144, unlike the changes observed in the structures of subtilisin mutants which are seen only in close proximity to the mutated residue. These small changes are seen even though CAP91 and CAP are both crystallized'in a complex with CAMP, however, in solution, CAP and CAP91 show different affinities for cAMP and other cyclic nucleotides. The structures of other mutants or of CAP crystallized in the absence of cAMP will probably be required to understand the small conformation changes due to the Ala-144 to Thr mutation described here.