Molecular Dissection of the Interaction between the Small G Proteins Rac1 and RhoA and Protein Kinase C-related Kinase 1 (PRK1)*

PRK1 is a serine/threonine kinase that belongs to the protein kinase C superfamily. It can be activated either by members of the Rho family of small G proteins, by proteolysis, or by interaction with lipids. Here we investigate the binding of PRK1 to RhoA and Rac1, two members of the Rho family. We demonstrate that PRK1 binds with a similar affinity to RhoA and Rac1. We present the solution structure of the second HR1 domain from the regulatory N-terminal region of PRK1, and we show that it forms an anti-parallel coiled-coil. In addition, we have used NMR to map the binding contacts of the HR1b domain with Rac1. These are compared with the contacts known to form between HR1a and RhoA. We have used mutagenesis to define the residues in Rac that are important for binding to HR1b. Surprisingly, as well as residues adjacent to Switch I, in Switch II, and in helix α5, it appears that the C-terminal stretch of basic amino acids in Rac is required for a high affinity interaction with HR1b.

interaction of the G protein (in its GTP-bound conformation) with an effector protein triggers a downstream signaling cascade. Effectors for Rho include at least 11 proteins: diacylglycerol kinase, phospholipase D, phosphatidylinositol phosphate 5-kinase, kinectin, rhotekin, rhophilin, p140 Diaphanous, myosin-binding subunit, citron, Rho kinase, and PRK 1 (or PKN). Within this group of effector proteins there are at least two Rho-binding motifs defined by sequence homology: REM (or class 1 Rho-binding motif) include the PRKs, rhophilin and rhotekin, whereas RKH (Rho effector homology 2 or class 2 Rho-binding motif) include the Rho kinases and kinectin (5).
The importance of the PRK effector family for the biological functions of the Rho family of small G proteins is indicated by the interactions that have been described for PRK1 and -2. Many proteins that interact with PRK1/2 are involved with the cytoskeletal network, e.g. ␣-actinin (6) and vimentin (7). As a major function of the Rho family proteins is known to be the control of the actin cytoskeleton, finding interactions between cytoskeletal proteins and Rho effectors was to be expected. PRKs are also implicated in the control of transcription factors (8,9), mitogenesis (10), and cell cycle regulation (11). PRK1 and -2 have also been shown to play a role in apoptosis, being activated by caspase 3 (12) and being involved in the regulation of Akt (13). Recently, PRKs have been shown to be involved in keratinocyte cell-cell adhesion with increased PRK activity promoting cell-cell adhesion (14). The activity of PRKs in so many processes intimately involved in disease progression makes dissection of their regulation of widespread interest.
PRK1 and -2 are highly related serine/threonine kinases that have a catalytic domain homologous to those of the protein kinase C family in their C termini and a unique regulatory domain in their N termini ( Fig. 1) (15,16). The N terminus of the protein was found to contain a pseudo-substrate sequence that acts as an auto-inhibitory region (17). This regulatory region also contains three leucine zipper type motifs (HR1 repeats a-c). The first of these repeats, HR1a, is now known to incorporate the inhibitory pseudo-substrate site (17). PRK1/2 kinase activity is enhanced by binding of the small GTP-binding proteins Rho and Rac in a GTP-dependent manner (18 -21), by binding of fatty acids such as arachidonic acid (22,23), and by the removal of the N-terminal regulatory domain from the catalytic domain by caspase 3 cleavage during apoptosis (12,24). Progress has been made recently in the dissection of the regulatory mechanisms employed by the Rho family effector molecules PAK (25) and Wiskott-Aldrich syndrome protein (26). It appears that many of the Rho family effectors are regulated in a similar manner, being maintained in a closed, inactive conformation via inhibitory intramolecular contacts. These inhibitory contacts can then be relieved by interactions with e.g. other proteins, fatty acids, and phosphoinositides.
The involvement of small G proteins in the control of PRK1/2 was first defined when PRK1 was found to be a target for Rho (18,21). The minimal region required for Rho binding was found to be residues 33-111 ( Fig. 1), which contains the first HR1 repeat, HR1a, and the pseudo-substrate site (27). Rho was subsequently found to bind to both isolated HR1a and HR1b domains but to HR1a in a more GTP-dependent manner, whereas no interaction was detected for HR1c (28). G proteininteracting proteins are classically defined as effectors if they bind selectively to the GTP-bound (active) form of the G protein. These results therefore suggested that the HR1a repeat was the important Rho-binding site as it provided discrimination in favor of the active form of Rho. Both PRK1 and -2 have been demonstrated recently (19,20) to be targets not only for Rho but also for Rac.
With the Rho family of small G proteins occupying such a central role in a diverse set of cellular processes, many of which are important and altered in disease progression, much effort has been expended over the last few years to elucidate the interactions of these proteins. Progress has been made in determining the structures of Rho family G proteins in complex with various effector proteins (29 -36). These structures have revealed much about the specificity of these interactions, information which is key to understanding the signaling pathways controlled by the Rho family. The x-ray structure of RhoA in complex with the HR1a repeat of PRK1 has been solved (35). This structure describes the fold of the HR1a domain as an antiparallel coiled-coil domain termed the ACC finger domain. The ACC finger domain describes a new category of G protein binding domains quite distinct from those found in other known Rho family effectors. It also differs from the ubiquitin fold Ras binding domains of Raf (37,38) and Ral GEF (39,40) and from the Ran binding domain of importin (41,42). The Rab binding domain of Rabphilin and the Rac/Arf effector Arfaptin are also helical, but the contacts that each make with their G proteins are quite different (43).
The results of Flynn et al. (28) suggesting the possibility of multiple G protein-interacting sites on PRK1 and further data identifying PRKs as targets for both Rac1 and RhoA (19,20) stimulated this work, in which we set out to determine the role of the multiple HR1 domains in PRK1/G protein interactions. We demonstrate here that the PRK1 HR1 repeats have different affinities for both Rho and Rac. We present the structure of PRK1 HR1b and a map of the contacts it makes in complex with Rac1. We also present data to show the necessity of the C terminus of Rac for high affinity binding to PRK1 and mu-tagenesis data that indicate the site of HR1b interaction on Rac.

MATERIALS AND METHODS
Expression Constructs-All proteins were expressed as GST fusion proteins from the pGEX vectors (Amersham Biosciences). The construct expressing residues 75-132 of PAK has been described previously (44) and that expressing residues 198 -439 of RhoGAP was a kind gift from Prof. Alan Hall (45). The HR1a expression construct consisted of PRK1 residues 1-106, HR1b residues 122-199, and HR1ab residues 1-199. Rac1 Q61L and all other mutants were expressed in pGEX2T from constructs encoding residues 1-191 cloned into the BamHI and EcoRI sites. The pGEX2T-Rac1 Q61L construct was used as a template to generate the C-terminal deletion constructs. Oligonucleotide primers were designed to amplify the required coding sequence, flanked directly by a 5Ј BamHI site and a 3Ј stop codon followed by an EcoRI site to facilitate subcloning into pGEX2T. The resulting PCR products were ligated into pCR2.1 (Invitrogen), and their sequences were verified using an automated DNA sequencer (Applied Biosystems Inc.) The coding fragments were then excised using BamHI and EcoRI and ligated into pGEX2T linearized with the same enzymes.
Rac1 Mutagenesis-Site-directed mutagenesis of the pGEX2T-Q61L Rac1 expression construct encoding residues 1-191 was performed using the QuikChange Site-directed Mutagenesis kit (Stratagene). The sequence of the Rac1 coding region of all mutants was verified using an automated DNA sequencer (Applied Biosystems Inc.).
Recombinant Protein Production-GST fusion proteins were expressed in Escherichia coli BL21. Stationary cultures were diluted 1 in 10, grown at 37°C to an A 600 of 0.8, and induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside for 5 h. Proteins were affinity-purified using glutathione-agarose beads (Sigma). Fusions of HR1a, -b, -ab, PAK GBD, and RhoGAP were eluted from the glutathione-agarose beads and used directly in SPAs. GST-Rac1 variants were cleaved from their GST tag while attached to the glutathione-agarose beads with thrombin (Novagen). Protein concentrations for all proteins were evaluated using their A 280 , amino acid compositions, and the extinction coefficients of tyrosine, phenylalanine, tryptophan, and the guanine nucleotide (46). The integrity of each protein was determined by mass spectrometry.
Labeled proteins for NMR spectroscopy were produced by growing the E. coli in a medium based on MOPS buffer, containing 50% labeled Celtone with 15 NH 4 Cl and/or [ 13 C]glucose (Spectra Stable Isotopes). Labeled HR1b was cleaved from its GST tag with factor Xa (Roche Applied Science), and the resulting protein was further purified on a Superdex-30 gel filtration column (Amersham Biosciences) in NMR buffer (20 mM Tris-HCl, pH 7.4, 20 mM NaCl, 0.05% NaN 3 ). Nucleotide exchange of G proteins was performed as described previously (44).
Scintillation Proximity Assays (SPA)-Affinities of Rac1 and RhoA proteins for GST-HR1 domains, GST-PAK-(75-132), and GST-RhoGAP were measured using SPAs. GST fusion protein was attached to a fluoromicrosphere via an anti-GST antibody in the presence of Q61L Rac/Q63L Rho⅐[ 3 H]GTP. Binding of the G protein to the GST fusion protein brings the labeled nucleotide close enough to the scintillant to obtain a signal. Apparent K d values for Q61L Rac⅐[ 3 H]GTP, Q63L Rho⅐[ 3 H]GTP, and proteins incorporating further mutations were measured as described previously (47,48) by varying the concentration of Rac/Rho⅐[ 3 H]GTP at a constant concentration of GST fusion protein.
These assays were performed with 30 nM GST fusion protein. By using this method the upper and lower limits of the K d values that can be measured are 1000 and 1 nM, respectively. For each affinity determination, data points were obtained for at least 10 different Rac1 concentrations. Binding curves were fitted using the appropriate binding isotherms (47,48) to obtain K d values.
NMR Spectroscopy-NMR samples were produced to a final concentration of ϳ1 mM in 500 l with 10% D 2 O. Experiments were recorded on Bruker DRX spectrometers operating at 500 or 800 MHz, and all experiments were recorded at 298 K. 15  the standard deviation from TALOS, whichever was larger. In those cases where the 3 J HN-HA could be measured, the error on the angle was set to Ϯ10°. The titration of unlabeled Rac1 into 15 N-labeled HR1b was performed in the same buffer conditions as the HR1b spectra. Titration points were measured at 1:0.2, 1:0.4, 1:0.6, 1:0.8, and 1:1, HR1b:Rac1. Chemical shift changes were calculated as a final deviation ␦ HN ϭ ͌(␦ N ) 2 ϩ (10 ϫ ␦ H ) 2 , where ␦ N ϭ change in 15 N chemical shift and ␦ H ϭ change in 1 H chemical shift.
Structure Calculations-Structures were calculated using CNS 1.0 and ARIA 1.0 (52), where the ambiguity of NOEs is reduced for each iteration by comparing the NOE tables with structures from the previous iteration. 20 structures were calculated for each iteration, and 7 were used for ARIA analysis. The total contribution of the accepted NOEs to the cross-peak volume was decreased during the 9 iterations from 1.01 (keeping all possibilities) to 0.8 (rejecting the possibilities that contribute less than 20% of the cross-peak volume). The parameters used for the calculation were essentially as described in Linge et al. (52) except that the length of the high temperature dynamics was increased to 6 ps and the cooling to a total of 39 ps (24 ps of cooling to 1000 K and 15 ps of cooling to 50 K). The HR1b/Rac1 modeling was performed with the protocol HADDOCK, interfaced to CNS 1.0 (53). For all sets of residues, the solvent accessibility was estimated using the program NACCESS (54), and those residues that were less than 50% solvent-exposed were not used. The HR1b residues whose chemical shifts had changed by more than the mean shift change were the active residues. The passive residues were defined as those that were within 5 Å of the active residues. The Rac1 active residues were defined as those that caused a greater than 2.5-fold increase in the K d . The passive residues were those that were within 5 Å of the active residues.

RESULTS
NMR Structure of PRK1 HR1b-The backbone resonances of HR1b were assigned using standard triple resonance NMR experiments. Complete assignments were obtained except for the backbone amides of residues 155-159. Side chain assignments were obtained using HCCH-TOCSY, 15 N-separated TOCSY, and H(CC)(CO)NH experiments. 2077 distance restraints were extracted from 15 N-separated and 13 C-separated NOESY spectra, of which 1443 were unambiguous at the start of the calculation and 634 were ambiguous. These were translated by ARIA into 1575 restraints (1040 unambiguous and 535 ambiguous) for the first iteration. The remaining 502 restraints were removed, either because they were non-unique or because they were ambiguous restraints with more than 50 assignment possibilities. At the end of the final iteration there were 1207 unambiguous and 400 ambiguous restraints. 52 restraints were included from the 3 J HN-HA measurements and an additional 69 loose and restraints from TALOS calculations. Table I shows the structural statistics for the 25 lowest energy structures of 100 calculated. The final structures had no NOE violations of more than 0.5 Å and no dihedral violations of more than 5°. The structures are of good quality as judged from a Ramachandran plot (87% of the residues are in the most favored region). The backbone r.m.s.d. over all non-terminal residues (124 -191) is 0.82 Å. The HR1b structure is shown in Fig. 2. It consists of two ␣-helices, which pack together into an anti-parallel coiled-coil. The packing between the helices is mediated by a regular array of Leu, Ile, and Ala side chains. The structures are well defined over both the ␣-helices. The turn between the helices, residues 152-160, is less well defined by the NMR data. The NH resonances within this loop were absent from the spectra, suggesting that they are undergoing exchange on a ms time scale.
Selective Binding of HR1 Repeats to Rho Family G Proteins-PRK1 was originally isolated as a Rho effector protein (18,21). However, as Drosophila PRK1 and -2 have also been described as Rac-interacting proteins (19,20), we decided to examine the binding affinities of the independent HR1 domains of PRK1 to both Rho and Rac. We used SPAs, which measure proteinprotein interactions under equilibrium conditions. These have been used previously to quantify the affinity of interaction between G proteins and their effectors; Ras and neurofibromin 1 (55), Ras and Raf (56), and Cdc42 with PAK activated Cdc42associated kinase, Wiskott-Aldrich syndrome protein, and RhoGAP (44).
In these assays GST-HR1 fusion protein was bound to protein A SPA beads via an anti-GST antibody. An SPA signal is obtained when binding of GST-HR1 to Rho (or Rac)⅐[ 3 H]GTP b ͗SA͘ c represents values for the structure that is closest to the mean. c The Lennard-Jones potential was not used at any stage in the refinement. occurs. Experiments were performed using the GTPase-deficient mutants, Q61L Rac and Q63L Rho, so that stable [ 3 H]GTP complexes could be obtained. Affinities were determined for the interactions between HR1a, HR1b, and the didomain HR1ab and both Rac and Rho. As a control, affinities were also measured for the interactions with the regulatory protein RhoGAP.
The measured K d values are summarized in Table II. HR1a binds to both Rho and Rac, with a similar high affinity. HR1b, by contrast, binds selectively to Rac with a high affinity; its affinity for Rho is greater than 1 M, which is the upper limit of detection in SPAs. The di-domain, HR1ab, behaves more like HR1a in that it binds to both G proteins with a similar affinity. However, its binding affinities for Rho and Rac are somewhat higher than the HR1a affinities for these G proteins. Binding affinities measured for Rho and Rac with RhoGAP as a control are in agreement with previous results (48) and indicate that both G proteins are functionally active in these assays.
NMR Mapping of the Rac Binding Surface on HR1b-An NMR titration was performed to map the binding sites of unlabeled Rac⌬7⅐GMPPNP on the 15 N-labeled HR1b domain. Fig.  3a shows the HSQC spectrum of uncomplexed 15 N-labeled HR1b overlaid with that of 15 N-HR1b bound to unlabeled Rac⌬7⅐GMPPNP. Resonances that experience a change in their environment when Rac is bound can be identified by a change in their chemical shift. This is accompanied by an increase in line width, because the tumbling time of the complex is slower than that of the free HR1b. Those residues that exhibit chemical shift changes of more than 0.2 ppm are labeled in Fig. 3a. Two resonances, those of Val-141 and Ala-145, disappeared during the titration, implying that these residues are also involved in the interaction. Those solvent-exposed residues whose backbone amide resonances shift or disappear when Rac1 binds are shown on the structure of HR1b in Fig. 3b.
When the x-ray-derived structure of Rho with HR1a was solved, there were two possible contact sites for the HR1a domain on RhoA (35) (Fig. 3c). In both HR1a and HR1b, the residues involved in contacting Rho family G proteins map to the ends of the helices nearest the hairpin loop.
Mutational Mapping of the HR1b Binding Surface on Rac-The crystal structure of Rho in complex with the HR1a repeat showed two possible contact sites for HR1a on Rho. To investigate whether residues in the contact regions identified in Rho were also involved in the interaction between Rac and HR1b, we mutated the homologous residues in Rac. We then examined the binding affinities of these mutants to HR1b. The measured K d values are listed in Table III,  To investigate further the interaction between HR1b and Rac1, we prepared Rac1 with a deletion of the C-terminal 7 residues (Rac1⌬7). This modification (or similar) is used extensively in work with G proteins (57)(58)(59)(60). The modified proteins are much less prone to aggregation and can be produced in sufficient quantity for biophysical studies. This C-terminal deletion usually has no effect on the binding of effector proteins (31,32,44). However, when Rac1⌬7 was used in isothermal titration calorimetry experiments, no interaction was observed between Rac1 and the HR1 domains (data not shown). To check whether the lack of binding affinity was due to the absence of the C-terminal 7 residues, or simply a lack of a heat change associated with binding, we then performed SPAs. SPAs were carried out with full-length Rac1 and Rac1⌬7 complexed with [ 3 H]GTP (Table IV). Full-length Rac1 binds to HR1b with an apparent K d of 68 nM, whereas the affinity of Rac1⌬7 for HR1b was decreased by at least 15-fold. Binding affinities for both Rac proteins were also determined for a fragment of the CRIB effector PAK. In this case both Rac proteins bound with the same affinity (Table IV), in agreement with previous results (44). Thus, the effects of the C-terminal truncation seem to be specific for HR1 domains.
Rac1 contains a stretch of six positively charged residues at C terminus, which is followed by the CAAX box (where C indicates cysteine, AA indicates aliphatic amino acid, and X is any amino acid) (Fig. 5a). In vivo, the protein undergoes a series of modifications in this region. A geranyl geranyl group is added to the CAAX box cysteine, the AAX residues (LLL in Rac1) are then removed by proteolysis, and finally the cysteine is methylated. These modifications leave the stretch of basic residues intact in the functional protein in vivo. The C-terminal three basic residues were removed in the Rac1⌬7 variant (residues 186 -188). To investigate further the role of these Cterminal charged residues of Rac1 in the binding of HR1 domains, we mutated each residue individually to serine and also designed a new C-terminally deleted protein in which none of these six basic residues were removed (Rac1⌬4). We then measured the apparent K d values of the relevant proteins for HR1a, HR1b, and HR1ab, together with PAK-(75-132) as a control, by SPA (Table IV). The affinities of all these Rac variants for PAK were similar to that determined previously for Q61LRac⅐GTP (see Ref. 46 and Table IV), indicating that all the new variant proteins were functional. The measured K d values of Rac1⌬4 for HR1a, HR1b, and HR1ab were similar to that of the fulllength protein. However, Rac1⌬7 and all the individual mutants showed a dramatic decrease in affinity. Thus it seems that removing any of the basic residues in the C terminus of Rac is deleterious to binding of HR1 domains. No single residue is more crucial than any other, but the effect is specific for HR1 domain binding because there is no effect on PAK binding. R185S has a smaller effect on HR1a, HR1b and HR1ab binding, suggesting that this residue is not as important as the other residues. It is notable that this is also the only residue that is not conserved between Rac1 and RhoA (Fig. 5a).
Modeling of the Rac/HR1b Interaction-The NMR chemical shift mapping data and the mutagenesis data were used to generate ambiguous interactive restraints as input for the program HADDOCK to dock HR1b and Rac1. The residues on HR1b that change chemical shift when Rac binds and are also solvent-exposed in the three-dimensional structure were defined as the following active residues: Leu-139, Lys-140, Gln-143, Gly-144, Asn-147, Leu-162, Gln-168, Gln-171, Asp-172, Thr-175, Asp-178, and Ile-179. The residues that are close on the surface to these residues were defined as the following passive residues: Gln-134, Ala-136, Ile-137, Val-141, Gln-150, Thr-151, Ser-153, Asn-154, Gly-155, Ser-156, Thr-157, Lys- 158, Asp-159, Arg-160, Lys-161, Leu-164, Thr-165, Gln-167, Lys-174, Arg-181, Met-182, and Gln-183. The residues in Rac which, when mutated, caused a significant decrease in affinity (greater than 2.5-fold), were defined as the following active residues: Ala-27, Asn-43, Asp-63, Leu-67, Gln-162, Lys-166, and Thr-167. The residues close to the active residues on the surface were defined as the following passive residues: Thr-25, Asn-26, Asn-52, Ser-41, Glu-62, Arg-66, Leu-70, Leu-160, Thr-161, Gly-164, and Arg-163. The lowest energy structure from the lowest energy cluster of models is shown in Fig. 6a. In all of the models produced by HADDOCK, the HR1b binds to Rac1 at a site on the G protein that corresponds to Contact I RhoA . The   FIG. 3. a, 15 N-HSQC spectrum of HR1b free (black) and complexed with unlabeled Rac1⌬7 (red). Residues whose resonances changed significantly are labeled. b, ribbon representation of HR1b on which the residues whose resonances shift when Rac1 binds are shown. These residues map to both faces of the coiled-coil. c, ribbon representation of HR1a on which residues in close contact with RhoA are shown. Contacts were derived from the crystal structure of HR1a/RhoA described by Maesaki et al. (35). Site I HR1a is shaded in dark blue and Site II HR1a is shaded in pale blue. These figures were produced using Molscript (65).
HR1b, however, is rotated by 180°relative to the HR1a in site I of the Rho/HR1a crystal structure (Fig. 6b), so that the N and C termini are flipped. The coiled-coil is also in a slightly different orientation with respect to the G protein.

DISCUSSION
The structure that we describe here of the HR1b repeat of PRK1 shows that it is a coiled-coil domain, consistent with predictions from its sequence homology to HR1a. There are some differences however in the lengths of the helices in the two HR1 domains. The sequence alignment (Fig. 5b) shows that the first ␣-helix is longer in HR1a than in HR1b, whereas the second ␣-helix is longer in HR1b than HR1a. This difference in the length of the helices may have some relevance to their relative positions in our model of the HR1b/Rac structure. When HR1a binds to RhoA in Contact I, the longer N-terminal helix contacts the ␤2-␤3 loop and helix ␣5 in RhoA, whereas the shorter C-terminal helix contacts the ␤2 and ␤3 strands and residues at the N-terminal end of Switch I. In the HR1b/Rac model, the longer helix also contacts the ␤2-␤3 loop and helix ␣5 in Rac, but now the longer helix is the C-terminal helix. Similarly, the shorter N-terminal helix contacts the ␤2 and ␤3 strands and residues at the N-terminal end of Switch I in Rac1. The residues that are used to make contact with the G proteins are not conserved within these long helices and are a mixture of hydrophobic and polar/charged residues (Fig. 5b).
The crystal structure of the complex between HR1a and RhoA describes two possible contacts between RhoA and HR1a (Figs. 3c, 5, and 6). Contact 1 RhoA has a total buried surface area of 2080 Å and involves residues in the ␤2/␤3 sheet and the hairpin between them, part of helix ␣5, and the ends of Switch I. Contact 2 RhoA buries a total surface area of 1640 Å and involves residues in Switch I (Val-38 to Asn-41), strand ␤3, and Switch II (Trp-58 and Asp-65 to Asp-76). The size of the interface in Contact 2 means that it is unlikely to be an artifact of crystal packing. It is also noteworthy that it involves residues in the switch regions of RhoA, which are often involved in effector binding. The presence of two contact sites for PRK1 on Rho has also been suggested by in vitro experiments. Flynn et al. (28) performed binding assays using Rho/Rac chimeras, the results of which indicated that HR1a bound primarily to residues 73-143 of Rho, whereas HR1b bound to residues 1-73.
These two contacts also utilize different residues on HR1a. The contact sites for Rac on HR1b that we have determined by chemical shift mapping can be compared with the contact sites for Rho on HR1a in the Rho/HR1a x-ray structure. Fig. 5b shows that the HR1a residues involved in site I HR1a and site II HR1a are quite distinct and indeed only overlap at two residues. The total number of HR1a residues involved in site I HR1a is 16 and in site II HR1a is 15. A total of 10 resonances from solvent-exposed residues in HR1b have chemical shifts that change when Rac binds and 2 disappear; of these 12, 6 correspond to residues within site I HR1a , 4 correspond to residues within site II HR1a (2 of which are in both site I HR1a and site II HR1a ), and 4 are in neither site 1 HR1a nor site II HR1a . The two residues that are in both sites I HR1a and II HR1a are also involved in the HR1b/Rac interaction. Thus it appears that the   interaction surface on HR1b for Rac involves contact sites that only partially overlap those between HR1a and RhoA. This is clearly evident from a comparison of the contact sites on the structures (Fig. 3, b and c), which shows that in HR1a, RhoA binds to two sites that are on opposite faces of the coiled-coil. The contact sites for Rac on HR1b map to a similar region of the coiled-coil, although the number of residues involved in the contact site in HR1b is considerably less. Our chemical shift mapping of HR1b shows that the Rac/HR1b interaction must be significantly different from that between Rho and HR1a. First, our results do not point to the exclusive use of one of the two contact sites on the HR1b side. One conclusion from this would be that two Rac molecules are binding to a single HR1b domain; however, this does not appear to be the case. Chemical shift mapping by NMR generally overestimates the number of residues in a contact site, although in this case we have attempted to limit this by removing residues that are not solvent-exposed and whose shift changes are likely to be due to secondary effects. Thus, if Rac bound at 2 sites to the HR1b domain, we should see chemical shift changes over all the residues within FIG. 5. a, alignment of RhoA and Rac1. The secondary structure of RhoA is shown; ␤-strands are represented by arrows and ␣-helices by cylinders. Residues involved in the two contacts are boxed and labeled with the contact number. Contact I RhoA residues that are conserved between RhoA and Rac1 are colored red. Contact II RhoA residues that are conserved are colored blue. The C-terminal basic region is colored green. * marks the site of geranylgeranylation, proteolysis, and methylation on processed Rac1. b, alignment of HR1 domains from PRK1, rhophilin, and rhotekin. The secondary structures of HR1a and HR1b are shown above the relevant sequence as yellow cylinders. HR1a residues are colored red in site I HR1a , blue in site II HR1a , and pink if in both sites. HR1b residues that shift or disappear when Rac1 binds are shown in green. In gray are residues whose backbone amides are missing from the HR1b spectra and those that are too overlapped to track chemical shift changes. the 2 sites, and this is not the case. Also, when we performed the NMR titration, the chemical shifts stopped changing at a ratio of ϳ1:1 Rac:HR1b, whereas if there were two HR1b domain binding to Rac, the changes should stop at 0.5:1. In an effort to resolve the stoichiometry of the Rac/HR1b interaction, we attempted isothermal calorimetry experiments. Unfortunately, the heat changes associated with this interaction were too small to enable accurate data to be obtained.
A similar conclusion can be drawn from our data for Rac mutants binding to the HR1b domain. Residues are found in Rac that are equivalent to Rho residues in both Contact I and II, which disrupt binding to the HR1b domain. Ala-27, Asn-43, Gln-162, Lys-166, and Thr-167 are equivalent to Rho residues in Contact I, and all show a significant decrease in affinity for HR1b, whereas Asp-63 and Leu-67, equivalent to residues in Contact II RhoA , show similar effects. Thus, the HR1b-binding site on Rac1 also partially overlaps both the RhoA contact sites. Predicting the effects of mutants is not as straightforward as the NMR mapping experiments, so it is not possible to say, for example, that if two HR1 domains were binding to a single Rac all of the residues in both contact sites would have an effect when they are mutated. In this case, however, only 7 mutants of the total of 20 residues in the two contact sites had an effect on the binding of HR1b.
A careful examination of the regions of RhoA that are involved in Contacts I RhoA and II RhoA shows that the residues in Contact II RhoA are completely conserved in Rac (Fig. 5a), whereas those in Contact I RhoA are almost totally non-conserved. HR1b alone binds weakly to RhoA (Table II), whereas HR1a alone binds equally well to RhoA and Rac1. Thus, it is possible that when HR1a binds to Rac, it binds preferentially to Contact II, accounting for its similar affinity for both G proteins (Table II). The residues in Rac that, when mutated, caused a significant drop in affinity for HR1b are not conserved between Rac and Rho, except Asp-63 and Leu-67 in Switch II. This again could explain why HR1b is not able to bind with a high affinity to Rho (Table II). Thus the interaction of fulllength PRK1 with Rho would involve contacts with HR1a only, whereas the interaction with Rac could utilize both HR1a and HR1b. As both G proteins contact HR1a, both would activate the kinase by alleviating the inhibitory interaction between the pseudo-substrate sequence and the catalytic domain.
In an attempt to model the structure of the Rac-HR1b complex, we used the chemical shift mapping and mutagenesis data to dock HR1b and Rac1 together. The model predicts that HR1b binds to a site on Rac that is equivalent to Contact I RhoA . This mode of binding is also consistent with a scenario whereby HR1a and HR1b could interact simultaneously with Rac1, the HR1a via Contact I Rac1 and the HR1b via Contact II Rac1 . This model of the HR1b/Rac interaction does not include the Cterminal basic region that we have shown to be important for the interaction. This is because there is little structural information at present for this region in the Rho family G proteins, as it is usually removed for structural studies. The model is not entirely consistent with our mutagenesis data for Rac1, because two residues in Switch II disrupt the interaction with HR1b, but neither of them is in contact with the coiled-coil in the model. This is because the essential Rac residues identified by our mutagenesis studies fall into two regions of the protein, which cannot interact simultaneously with a single HR1b molecule. There are several potential explanations for this. For example, it could be that these two residues are involved in an intramolecular interaction with the C-terminal basic region that we know is necessary for the high affinity interaction between Rac1 and HR1b. Alternatively, mutating these residues has caused long range changes in Rac that are transmitted to the HR1b contact site. Although few data are available relating to the structure of the C-terminal basic residues of small G proteins, these amino acids are visible in a Cdc42/ RhoGAP structure (61). Cdc42 is highly homologous to Rac, and their tertiary structures are very similar. In Cdc42, one of the C-terminal basic residues, Arg-187, is observed to make a hydrogen bond with Asp-76 in Switch II. This interaction, if present in Rac, could explain the effects we see with the two Switch II mutations, Asp-63 and Leu-67, with mutations in Switch II affecting the C-terminal basic region which would then affect HR1b binding.
Previously published data, from overlay assays, investigating the interaction between Rac and the HR1 domains have yielded different conclusions. Flynn et al. (28) observed a weak interaction between Rac and HR1a with no detectable interaction with HR1b. Work with Drosophila PRK1 and -2, however, demonstrated clear interactions between the whole proteins and Rac (19,20). Differences in observations between these groups could be due to the Rac used. We have found that full-length Rac1 is highly susceptible to proteolysis at its C terminus even when it is expressed in E. coli BL21, and we check each purified protein by mass spectrometry before use. We see a high affinity interaction between Rac1 and isolated HR1b (68 nM, see Table II) along with similar high affinity binding to HR1a (ϳ170 nM). In fact the apparent K d values that we measure between Rac1 and the HR1 domains are slightly higher than those for RhoA. These data would imply a role for Rac in the regulation of PRK. Evidence to support this view comes from the observation by Lu 6. a, model of the HR1b/Rac structure generated using the HAD-DOCK protocols (53). The HR1b binds at a site that corresponds to Contact I in the RhoA/HR1a crystal structure. b, the crystal structure of HR1a with RhoA showed that HR1a can potentially bind to two sites on the G protein, which are designated Contact I and Contact II.
Our data have also highlighted a role for the C terminus of Rac in the interaction with HR1b, indicating that the C-terminal basic residues (183-188) in Rac1 are required for high affinity complex formation. Unfortunately, we encountered technical difficulties in mapping the contacts between fulllength Rac and HR1b by NMR, but at the high concentrations required for NMR studies, Rac1⌬7 will bind to HR1b. We were thus able to map the residues in HR1b that contact the other regions of Rac1 but were unable to predict the HR1b residues that are involved in the interaction with the C-terminal basic residues. Furthermore, these residues are missing from the RhoA construct used to solve the RhoA-HR1a complex (35), so we cannot predict the importance of these residues in the HR1a/RhoA interaction.
For most of the G protein-effector protein complexes studied to date, these C-terminal residues play no part in the interaction. For example, their removal has no effect on binding of Rac1 to the PAK GBD (Table IV). The involvement of these residues may not, however, be unique to the Rac/PRK1 interaction. There is circumstantial evidence for the involvement of Rac residues 175-191 in the activation of p67 phox (62). The crystal structure of the complex between the TPR domain of p67 phox and Rac unfortunately also included a Rac construct that lacks the C-terminal eight residues, and so the role of these residues in this interaction is not clear (30). In Cdc42, another member of the Rho family, mutation of Lys-183/184 has been shown to disrupt binding to the ␥-subunit of the coatomer complex (63). The C-terminal residues have also been shown recently (64) to be involved in the oligomerization of Rac1. Rac1 was found to exist in a monomer/oligomer equilibrium. Removal or mutation of these C-terminal basic residues resulted in a shift to the monomeric state. Concomitant with the shift to the monomeric state was a decrease in intrinsic GTPase activity and a decrease in PAK activation, although the mechanism of this impairment is not clear, given that the removal of these C-terminal residues does not affect PAK binding (Table III). Our data suggest that loss of the C-terminal basic residues would also lead to a decrease in Rac activation of PRK1, as seen by an impairment of binding ability. The proximity of these basic residues to the site of isoprenyl (geranyl geranyl) modification in Rac1 (Fig. 5a) poses interesting steric questions of how these residues might be able to interact with PRK in vivo. However, juxtaposition with the membrane and/or lipid groups does not necessarily preclude interaction with effector proteins.
In conclusion, the data presented here further clarify the interaction between PRK1 and the Rho family small G proteins. The HR1b domain of PRK1 is shown to have the same overall fold as that reported previously for the HR1a domain, but it binds with a far higher affinity to Rac1 than to RhoA unlike HR1a, which binds with similar affinity to both small G proteins. We have shown that the interface of the HR1b-Rac complex utilizes residues in both proteins that are in both contact sites found previously in the HR1a-Rho complex. The number of residues involved in the interaction suggests, however, that there is only one contact site between HR1b and Rac and that the interaction is thus likely to be quite different from that between HR1a and RhoA. We have presented the unexpected finding that the six basic C-terminal residues of Rac1 are critical for the high affinity binding of PRK1. The details of their role in the interaction await the determination of the structure of the complex between HR1b and full-length Rac.