C-terminal sequences outside the tetratricopeptide repeat domain of FKBP51 and FKBP52 cause differential binding to Hsp90.

Hsp90 assembles with steroid receptors and other client proteins in association with one or more Hsp90-binding cochaperones, some of which contain a common tetratricopeptide repeat (TPR) domain. Included in the TPR cochaperones are the Hsp70-Hsp90-organizing protein Hop, the FK506-binding immunophilins FKBP52 and FKBP51, the cyclosporin A-binding immunophilin CyP40, and protein phosphatase PP5. The TPR domains from these proteins have similar x-ray crystallographic structures and target cochaperone binding to the MEEVD sequence that terminates Hsp90. However, despite these similarities, the TPR cochaperones have distinctive properties for binding Hsp90 and assembling with Hsp90.steroid receptor complexes. To identify structural features that differentiate binding of FKBP51 and FKBP52 to Hsp90, we generated an assortment of truncation mutants and chimeras that were compared for coimmunoprecipitation with Hsp90. Although the core TPR domain (approximately amino acids 260-400) of FKBP51 and FKBP52 is required for Hsp90 binding, the C-terminal 60 amino acids (approximately 400-end) also influence Hsp90 binding. More specifically, we find that amino acids 400-420 play a critical role for Hsp90 binding by either FKBP. Within this 20-amino acid region, we have identified a consensus sequence motif that is also present in some other TPR cochaperones. Additionally, the final 30 amino acids of FKBP51 enhance binding to Hsp90, whereas the corresponding region of FKBP52 moderates binding to Hsp90. Taking into account the x-ray crystal structure for FKBP51, we conclude that the C-terminal regions of FKBP51 and FKBP52 outside the core TPR domains are likely to assume alternative conformations that significantly impact Hsp90 binding.

Hsp90, typically the most abundant cytoplasmic chaperone in vertebrate cells, serves a vital role in cellular signaling by regulating the folding, activity, and stability of a wide range of client proteins, as exemplified by steroid receptors and protein kinases (1). Client protein complexes contain not only Hsp90 but also one or more cochaperones that partner with Hsp90. One class of Hsp90-binding cochaperone is composed of proteins with a characteristic tetratricopeptide repeat (TPR) 1 domain that forms an Hsp90 binding site (2). Among the TPR cochaperones of Hsp90 are Hop/Sti1, an adaptor chaperone that also binds Hsp70 (3,4), protein phosphatase PP5 (5), and members of both the FK506-and cyclosporin A-binding families of immunophilins (6 -9). The TPR cochaperones compete for binding the C-terminal region of Hsp90 (10 -12), and the highly conserved MEEVD sequence that terminates eukaryotic Hsp90 is a common target for TPR interactions (13)(14)(15)(16). Mutation of the MEEVD sequence typically inhibits binding by a TPR cochaperone, although mutations in other C-terminal sequences impact binding by TPR cochaperones differentially (13,17).
The immunophilin-related TPR cochaperones contain peptidylprolyl isomerase domains that also serve as the binding site for immunosuppressive drugs, yet immunophilins can have differential effects on the function of Hsp90 client proteins. In particular, FKBP52 and FKBP51, which share ϳ70% amino acid sequence similarity, affect hormone binding by the glucocorticoid receptor (GR) in opposing manners. A study in the yeast Saccharomyces cerevisiae, which lacks endogenous counterparts to FKBP52 or FKBP51, has shown that FKBP52, specifically, can elevate GR responsiveness to hormone (18). Although FKBP51 alone does not reduce GR activity in yeast, it can effectively block the potentiation mediated by FKBP52 when coexpressed (18). Scammell and his colleagues (19 -21) have shown that cortisol insensitivity observed in New World primates is facilitated by a constitutive overexpression of FKBP51, suggesting a physiological relevance for FKBP interactions with GR⅐Hsp90 complexes. Bourgeois and colleagues (22) first noted that the gene for FKBP51 is up-regulated by glucocorticoids, and this observation has been confirmed repeatedly by recent gene expression profiles of steroid-responsive genes. The inducibility of FKBP51 by glucocorticoids could provide a mechanism for partial desensitization of cells subsequent to an initial exposure to hormone (23).
It is not clear which structural features are responsible for the differential effects of FKBP52 and FKBP51. Three-dimensional crystallographic structures for FKBP51 have recently been solved (24), but there is no corresponding structure for FKBP52. Based on the 70% amino acid sequence similarity of FKBP51 and FKBP52 and the conservation of apparent domain sequences, there is no reason to suspect a dramatic structural difference. Nonetheless, previous studies from us showed that mutations in the C-terminal half of Hsp90 have different effects on binding by FKBP52 versus FKBP51 (13). Also, we observed that an exchange of sequences between the FKBP TPR domains had no effect on Hsp90 binding by FKBP52 but blocked binding by FKBP51 (25). These observations suggested that each FKBP has a distinctive interaction with Hsp90. In the current study, we have taken advantage of the x-ray crys-tallographic structure of FKBP51 to map functionally sequences of FKBP51 and FKBP52 that are important for Hsp90 binding. We find that sequences in the C-terminal region, both inside and outside the TPR domain, greatly influence FKBP binding to Hsp90.

EXPERIMENTAL PROCEDURES
Preparation of Mutant cDNAs-In vitro expression plasmids containing either human FKBP51 or FKBP52 cDNA inserted into pSPUTK (Stratagene, La Jolla, CA) were used to generate mutants used in these studies. To construct the C-terminal truncations, stop codons were introduced by site-directed mutagenesis (QuikChange kit, Stratagene). To help clarify our naming convention for many of the mutants generated for this study, it is helpful to note that FKBP52, relative to FKBP51, contains a two-amino acid insert in the loop connecting FK1 and FK2 domains; thus, the position of corresponding mutations in the C-terminal halves of either FKBP will differ by two residues. For example, the FKBP51 truncation mutant N404 contains an engineered stop codon at position 405; the equivalent FKBP52 truncation mutant, N406, contains a stop codon at position 407. The FK mutants were generated by introducing stop codons at position 258 or 260, respectively, which lies in the linker region between FK2 and the TPR domain. The TPR mutants were created by first introducing an F252M or F254M substitution and then removing all upstream coding sequences. FKBP chimeric cDNAs were constructed by two different approaches. The FKBP51-395 (positions 1-395 code for FKBP51 and positions 396end code for FKBP52) and the converse FKBP52-397 chimeric cDNA were originally cloned into yeast expression vectors (18). These cDNA were PCR amplified from yeast vectors and subcloned into pSPUTK. Other chimeras were created as follows. A gene fragment encoding the N-terminal portion of the desired chimeric protein was generated by PCR. A second gene fragment encoding the C-terminal portion of the desired chimera was separately generated by PCR. Primers for the two fragments were designed such that sequences surrounding the desired fusion site were complementary. The resulting DNA products were gel purified and used as megaprimers in a reaction with the appropriate 5Јand 3Ј-primers to generate the full-length chimeric cDNA. The final PCR product was subcloned into pSPUTK. Sequence changes in mutated cDNAs and all PCR-generated products were confirmed by automated DNA sequencing.
In Vitro Hsp90 Binding and PR Complex Assembly-The Hsp90 and PR binding abilities of wild type proteins and mutants were analyzed by a coimmunoprecipitation approach as described previously (25). Briefly, radiolabeled immunophilins were synthesized from plasmid DNA templates in an in vitro transcription/translation system (TNT lysate, Promega). Radiolabeled products were quantitated by densitometry of gel autoradiographs, and an equimolar amount of each radiolabeled product was added separately to 100 l of rabbit reticulocyte lysate (RL; Green Hectares, Oregon, WI). Each RL sample was added to a 10-l pellet of protein G-Sepharose (Amersham Biosciences) prebound with 10 g of H90-10, a specific anti-Hsp90 mouse monoclonal antibody. After incubation on ice for 1 h, unbound proteins were removed by washing three times in 1 ml of wash buffer (20 mM Tris-HCl, pH 7.4, 50 mM KCl, 1% Tween 20). Bound proteins were then extracted with SDS sample buffer and separated by SDS-PAGE. The assembly of PR complexes was analyzed in a similar manner, except 200 l of RL was supplemented with an ATP-regenerating system, samples were incubated with 1 g of recombinant PR bound to PR22-protein A-Sepharose, and incubations were at 30°C for 30 min. After electrophoresis, gels were stained with Coomassie Brilliant Blue to visualize total proteins. Gels were then dried and autoradiographed to visualize radiolabeled proteins, and bands were quantitated by densitometry using a Fluor-S Imager (Bio-Rad).
Yeast Strains, Plasmids, and Methods-FKBP functional assays were performed in S. cerevisiae essentially as described previously (18). All experiments were performed using a GR reporter strain in the W303a background (MATa leu2-112 ura3-1 trp1-1 his3-11, 14 ade2-1 can1-100 GAL SUC2). Parental cells were transformed with a plasmid that constitutively expresses rat GR, a second plasmid that expresses ␤-galactosidase from a glucocorticoid-regulated promoter, and a third plasmid constitutively expressing one of the human immunophilins. Cells were grown in minimal medium containing 0.67% (w/v) yeast nitrogen base without amino acids, 2% (w/v) glucose, the appropriate SC supplement mixture (Q-biogene, Carlsbad, CA); for growth on plates the culture medium was supplemented with 1.6% w/v agar.
Hormone induction assays were performed as described previously (18). Briefly, yeast strains were grown in selective medium at 25°C, and the absorbance at 600 nm (A 600 ) of the culture was monitored to ensure exponential growth. Reporter gene expression was induced by adding deoxycorticosterone (25 nM final concentration) to log phase cultures. Starting 70 min later, cells were sampled at 10-min intervals over the next 40 min for reporter activity. ␤-Galactosidase activity was measured by adding 100 l of culture to an equal volume of Gal-Screen assay solution (Tropix, Bedford, MA) according to the manufacturer's instructions. Reporter expression rate was calculated as the linear slope of relative light units versus A 600 /1,000. For each strain tested, hormone-induced reporter expression rate was determined with two to four independent transformants to assure consistency of results.

RESULTS
The three-dimensional crystallographic structure for FKBP51 (Protein Data Bank code 1KT1) depicted in Fig. 1A reveals three major structural domains. Of the two FKBP12like domains, peptidylprolyl isomerase and FK506 binding activities reside in FK1 (24). The Hsp90 binding TPR domain has a structure similar to that of other described TPR domains, in particular the Hsp90 binding domains from PP5 (26), Hop (16), and CyP40 (27). The consensus TPR motifs terminate with helix 6 (H6, Fig. 1B); however, similar to structures for PP5 and CyP40, there is a seventh helix that extends beyond the core TPR domain for a minimum of 23 amino acids. How much further H7 may extend is unclear because the final 36 amino acids could not be resolved from crystal data. Although a crys- tal structure for FKBP52 has not been reported, it seems likely that it will share the same overall domain organization as FKBP51.
FKBP Regions Required for Hsp90 and Receptor Interactions-As presented in Fig. 2, an initial set of FKBP51 and FKBP52 mutants ( Fig. 2A) was surveyed to compare interactions with Hsp90 and steroid receptor complexes. Coimmunoprecipitations of radiolabeled FKBP51 mutants (Fig. 2B) and FKBP52 mutants (Fig. 2C) were used to monitor Hsp90 binding (left panels) and assembly into PR complexes (right panels). We consistently found that incorporation of FKBP51 exceeded that of FKBP52 by 2-3-fold in Hsp90 complexes (compare lanes 1 and 2 in each data set). In line with previous observations (25), FKBP51 recovery in PR complexes exceeded the recovery of FKBP52 by 5-fold or greater (compare lanes 8 and 9 in each set). The TPR truncation mutants, which included the TPR domain plus C-terminal sequences, were sufficient for binding Hsp90 (lane 3) and assembling with PR complexes (lane 10). The TPR region was necessary for binding Hsp90 and assembling with PR because the FK domains showed no interactions (lanes 4 and 11 in each set). The importance of the TPR domain is demonstrated further by point mutation at one of the carboxylate clamp residues (K352A for FKBP51 and K354A for FKBP52) that abrogates Hsp90 binding and PR association (lanes 5 and 12). To test whether C-terminal sequences influence protein interactions, truncation mutants (N404 and N406) were generated which lacked sequences within H7 extension (see Fig. 1B) and beyond. Despite retention of the core TPR domain in these constructs, Hsp90 binding (lane 6 in both sets) and PR association (lane 13 in both sets) were largely abrogated. The final samples in this mutant series were a pair of chimeric constructs (395 for FKBP51 and 397 for FKBP52) in which the region from the H7 extension through the C terminus was exchanged between FKBPs. This exchange had little effect on FKBP52 (Fig. 2C, lanes 7 and 14) but greatly diminished interactions of FKBP51 with Hsp90 and PR (Fig. 2B,  lanes 7 and 14). In an earlier study (25) in which the C-terminal region was retained but TPR motifs were exchanged, we observed a similar phenomenon; the FKBP52-based construct functioned normally, but the FKBP51-based construct lost the ability to bind Hsp90 or assemble with PR complexes.
Taking advantage of the yeast model for FKBP52-dependent potentiation of GR signaling (18), we further tested FKBP52 constructs for function in vivo (Fig. 2D). First, note that FIG. 2. Mutant FKBP interactions with Hsp90 and steroid receptor. A, the diagrams depict a series of mutant forms generated for either FKBP51 or FKBP52, respectively. In descending order, these are the full-length wild type protein (wt), a truncation mutant containing the TPR domain through the C terminus (TPR), a truncation mutant containing the N terminus through the FK domains (FK), a point mutant in the TPR domain (K352A or K354A), a truncation mutant that terminates just beyond the TPR domain (N404 or N406), and a chimera in which the C-terminal sequences were exchanged (chimera 395 or 397). B, the ability of mutant forms of FKBP51 to bind Hsp90 (left panels) and assemble in vitro with PR complexes (right panels) is shown in comparison with human FKBP52. Radiolabeled proteins were generated by in vitro expression, and equivalent amounts of each form were added separately to RL before immunoprecipitation of Hsp90 complexes or assembly of PR complexes. The top panel in each set is a Coomassiestained gel image of total proteins recovered in each sample. Major protein bands, including antibody heavy chains (HC), are identified. The middle panel ( 35 S bound) is an autoradiograph of the stained gel which reveals bound FKBP forms. The bottom panel ( 35 S input) is an autoradiograph of a separate gel that demonstrates the relative level of radiolabeled protein added to each sample. C, same as B except the corresponding FKBP52 mutants replaced the FKBP51 mutants. D, the ability of FKBP52 mutant proteins to enhance GR signaling was measured in a yeast strain that constitutively expresses rat GR and expresses ␤-galactosidase in a hormone-inducible manner. The reporter strain was transformed with an empty vector (Vector) or vector constitutively expressing FKBP51 or the indicated FKBP52 form. For each strain the linear rate of increase in ␤-galactosidase activity was measured over a 40-min period beginning 70 min after the addition of deoxycorticosterone (25 nM final concentration). Reporter activities were normalized to that seen with yeast expressing FKBP51, and the results shown are typical for each of four independent isolates of each strain. Western immunostains confirmed that each FKBP form was present at approximately equivalent levels in total cell extracts (not shown). The results shown are typical measurements from four independent isolates for each strain.

Hsp90 Binding by FKBP Cochaperones
FKBP52 significantly enhances hormone-induced ␤-galactosidase activity compared with yeast expressing FKBP51 or lacking either FKBP (Vector). None of the FKBP52 mutants displayed potentiation of GR signaling except the C-terminal chimera 397. This is consistent with the dual requirement for Hsp90 binding and FK1 peptidylprolyl isomerase activity for FKBP52 to elevate GR hormone binding affinity (18). Because 397 retains wild type activity in this assay, it appears that specific sequences in the tail region of FKBP52 are not critical for its function in GR potentiation.
Hsp90 Binding by C-terminal Truncation Mutants-The defect in Hsp90 interactions apparent with FKBP51-N404 and FKBP52-N406 was unexpected and raised the question of which sequences downstream from the core TPR domain are minimally required for Hsp90 binding. To begin addressing this question, a series of FKBP truncation mutants was generated which focused on this region, and these were tested for coimmunoprecipitation with Hsp90. Densitometric measurements were taken from gel autoradiographs similar to those in Fig. 2, and these data were plotted as shown in Fig.  3. Significant recovery of FKBP51 mutants in Hsp90 complexes (solid circles) began with constructs that included residues 415-420, near the downstream end of H7. Full Hsp90 binding was only observed with constructs containing sequences beyond 430.
Truncations of FKBP52 (open circles) differed from FKBP51 mutants in two ways. First, note that there is a leftward shift in the Hsp90 binding curve for the FKBP52 series compared with the FKBP51 series. Thus minimally sized FKBP52 truncations that retain measurable levels of Hsp90 binding are 5-8 amino acids shorter than the smallest FKBP51 constructs that retain Hsp90 binding. The second difference pertains to the influence of sequences beyond position 430. Whereas the Cterminal region heightened Hsp90 binding by FKBP51, there appears to be a modest inhibition of Hsp90 binding by the corresponding region of FKBP52. Mutants terminating at approximately position 430 bind Hsp90 equally. However, the inclusion of sequences beyond this point moderated binding of FKBP52 to Hsp90 but enhanced binding by FKBP51.
Hsp90 Binding by FKBP51 Chimeras Containing C-terminal Portions of FKBP52-In marked contrast to the corresponding FKBP52 chimeras, swapping either the TPR domain (25) or the adjacent C-terminal tail (chimera 395 in Fig. 2B) abrogated FKBP51 binding to Hsp90. Swapping both regions together had no effect on Hsp90 binding. These observations suggest important intramolecular interactions between the C-terminal tail region and the TPR domain which are distinctive in FKBP51 and FKBP52. As shown in Fig. 4, additional FKBP51 chimeras focusing on the TPR domain boundary from H5 through H7 were constructed to probe more thoroughly for FIG. 3. Binding of FKBP truncation mutants to Hsp90. Truncation mutants of FKBP51 (solid circles) and FKBP52 (open circles) terminating at the indicated amino acid were compared for interactions with Hsp90. Radiolabeled protein products were added to RL before immunoprecipitation of Hsp90 complexes. Binding levels were quantitated by autoradiography of gel-separated samples and plotted relative to full-length FKBP51 (defined as 100% binding). The diagram above the plot illustrates the corresponding positions for H7 and the C-terminal sequences of FKBP51.

FIG. 4. Binding of FKBP chimeric proteins to Hsp90.
A, the H5-H7 regions (amino acids 348 -421, which is the final resolved residue in the crystal structure) of FKBP51 and FKBP52 are illustrated with coloring to denote hydrophobicity (blue indicates hydrophobic side chains; red, hydrophilic). The FKBP52 structure is modeled on the solved structure for FKBP51. The numbers on FKBP51 indicate six of the eight positions at which C-terminal sequences from FKBP52 were placed to generate chimeric proteins. Downstream chimeras started at 422 and 438. B, FKBP51 chimeras containing the tail sequences from FKBP52 were compared with wild type FKBPs for coimmunoprecipitation with Hsp90. Radiolabeled products were added to RL before immunoprecipitation of Hsp90 complexes. The top panel is an image from a Coomassie-stained gel separation of each immunoprecipitate. The numbers above the lanes indicate the final FKBP51 residues prior to the switch to FKBP52 sequences. The middle panel (bound) is an autoradiograph of the stained gel used to detect bound FKBP forms. The bottom panel (input) is a separate autoradiograph that demonstrates the relative levels of radiolabeled protein added to each sample. C, the Hsp90 binding data were plotted for FKBP51 chimeras terminating with FKBP52 sequences (solid circles) and for FKBP52 and the FKBP52-397 chimera (open circles). Also plotted are binding data for truncation mutants of the FKBP51-395 chimera (solid triangles). The level of Hsp90 binding to full-length FKBP51 was defined as 100%. The diagram above the plot illustrates the corresponding positions of H5, H6, and H7 in the FKBP51 crystal structure. potential interactions. Comparing hydrophobicity patterns in this region from FKBP51 and FKBP52, the two proteins are similar (Fig. 4A), although there is only 50% amino acid sequence identity. Another series of FKBP51 chimeras was generated in which an exchange occurred at one of eight positions along H6, H7, or just beyond H7 (seven of these sites are indicated in Fig. 4A). Radiolabeled products were synthesized and compared for coimmunoprecipitation with Hsp90 (Fig. 4B). As observed consistently, recovery of wild type FKBP51 in Hsp90 complexes exceeded that of wild type FKBP52 (compare first two lanes). Chimeras containing the H6 region and beyond from FKBP52 (365, 374, and 378) retained Hsp90 binding. When the exchange occurred within H7 (395, 404, and 413), there was a complete loss of Hsp90 binding. Hsp90 binding was again observed when the exchange was restricted to sequences beyond 422. To reiterate, we observed no defect in Hsp90 binding by FKBP52 chimeras that contained similar tail regions from FKBP51 (for example, for results with FKBP52 chimera 397, see Fig. 2C). Collectively, these observations suggest that there is an interaction within the region from H6 through H7 extended which is uniquely required by FKBP51 for Hsp90 binding.
An additional analysis was undertaken in which a series of C-terminal truncation mutants was generated from the defective chimera FKBP51-395. Similar to previous binding assays, truncations of 395 were examined for coimmunoprecipitation with Hsp90 on gel autoradiographs. The quantitated data were plotted in Fig. 4C (solid triangles). Also plotted in this figure are the quantitated data for FKBP51 chimeras (solid circles) that were examined in Fig. 4B and for FKBP52 and the FKBP52-397 chimera (open circles). Consistent with the behavior of FKBP52 truncation mutants (Fig. 3), the FKBP52-397 chimera displayed greater binding than wild type FKBP52 (compare open circles) presumably because of the loss of inhibitory sequences beyond position 430 of FKBP52. As seen from the gel data in Fig. 4B, the FKBP51 chimeras (solid circles) had a pattern of loss then recovery of Hsp90 binding as the fusion point in chimeras progressed from H6 through H7. Interestingly, truncations of FKBP51-395 (triangles) showed that Hsp90 binding could be partially restored in constructs that terminate between 410 and 430. The sharp boundary at ϳ410 corresponds well with the Hsp90 binding boundary observed with FKBP52 truncation. Likewise, the trailing boundary near 430 corresponds to the inhibition boundary deduced from FKBP52 truncation mutants.
Functional Analysis of FKBP52 Sequences in the 400 -420 Region of H7-Experimental results with FKBP52 and FKBP51 mutants point to the extended portion of H7 (amino acids 400 -420) as being important for Hsp90 binding and FKBP function. This region of FKBP52 was analyzed further, as shown in Fig. 5. Alignment of FKBP sequences in this region highlights the conservation of amino acids 406 -415 (Fig. 5A,  FKBP52 numbering), a 10-amino acid stretch consisting of a highly charged segment followed by YANMF. A series of FKBP52 truncation mutants that target this region was generated and tested for Hsp90 binding and assembly with PR complexes (Fig. 5B). As seen previously (Fig. 2), binding to Hsp90 and assembly into receptor complexes were greatly reduced with N406, but both interactions were restored when the C terminus was extended to 414 and beyond. Yeast GR reporter strains were generated to correlate protein-protein interactions in vitro with FKBP52 function in vivo (Fig. 5C). Corresponding with weak Hsp90 and PR interactions, N406 and N410 lacked the ability to potentiate GR signaling; however, potentiation was significantly boosted with N414 and larger mutants in parallel with Hsp90 and PR interaction patterns.
Sequence data bases were accessed to determine whether other TPR-dependent Hsp90 cochaperones share a motif similar to the 406 -415 segment. For each cochaperone we selected sequences that lie in the same relative downstream juxtaposition to TPR motifs. These juxta-TPR sequences are aligned in Fig. 6. Included in the comparison are the Hsp90-binding FKBP family members FKBP52, FKBP51, FKBP36 (28), and Xap2/AIP (29,30). CyP40 and PP5 sequences are present, as well as sequences from Hop (31) and CHIP (32). As shown above the alignment, an 11-amino acid motif, what we term the charge-Y motif, was identified with the consensus organization ϪϩϪϩX⌽YXXMF, where Ϫ represents Glu or Asp, ϩ represents Lys or Arg, ⌽ represents a hydrophobic amino acid, and X represents any amino acid. There is also a negatively charged amino acid 5 positions further downstream which may relate to FIG. 5. Minimal FKBP52 sequences required for protein interactions and function. A, FKBP sequences that correspond to the extended portion of H7 are aligned. Arrows denote the positions for a series of FKBP52 truncation mutants that terminate in this region. B, FKBP52 wild type (wt) and H7 truncation mutants were compared for coimmunoprecipitation with Hsp90 and assembly with PR complexes. Shown are images from Coomassie Blue-stained gels (top panels), the corresponding autoradiographs (middle panels), and separate autoradiographs of synthesis products (bottom panels). C, the yeast GR reporter strain was separately transformed with empty vector (Vector) or plasmids expressing each form of FKBP52, as indicated. Hormoneinduced reporter gene activity was measured in each strain, and results were normalized to the activity observed in the vector strain. The results are typical for multiple independent strain isolates. the consensus. Six of the nine Hsp90 cochaperones match the consensus in at least four of nine positions (indicated by left arrowhead).

DISCUSSION
FKBP52 and FKBP51 are closely related Hsp90-binding cochaperones, yet they have distinctive patterns of interaction with Hsp90 and Hsp90 client proteins such as steroid receptors. In an effort to understand better the structural basis for distinctive interactions with Hsp90, we performed a mutagenic analysis of sequences in FKBP51 and FKBP52 which impact Hsp90 binding. Both immunophilins have a core TPR domain that is necessary for binding to Hsp90, but this core domain is not sufficient for full binding to Hsp90 (Fig. 2). We have identified a conserved region, the charge-Y motif, which lies immediately downstream from the TPR domain and is required for Hsp90 binding (Figs. 3 and 5). A similar sequence is found in some other Hsp90-binding TPR proteins (Fig. 6). Further downstream in either FKBP, unique sequences within the final 30 amino acids appear to distinguish the relatively higher Hsp90 binding affinity of FKBP51 compared with FBP52 ( Figs.  3 and 4). The FKBPs are distinguished further by an intramolecular interaction peculiar to FKBP51 which involves H7, the final helix in the core TPR domain, and the adjacent charge-Y motif (Fig. 4). Thus, the Hsp90 binding properties of these two TPR cochaperones result from a combination of the core TPR domain and the influence of C-terminal sequences outside this domain. Consistent with the notion of alternative modes of interaction with Hsp90, point mutations in the C-terminal region of Hsp90 have distinct effects on the binding of individual TPR cochaperones (13,17). This suggests that cochaperones might interface with distinct structural features of Hsp90 in addition to a common interaction with the C-terminal MEEVD.
Jackknife Model for Charge-Y Motif Participation in Hsp90 Binding-According to the crystal structure for FKBP51 (Fig.  1), the charge-Y motif lies in a portion of H7 that extends beyond the core TPR domain. As depicted in Fig. 7, we hypothesize two alternative conformational states of H7 sequences that could account for our mutagenic data. In the first state (Fig. 7A), H7 exists in the extended conformation that is consistent with the FKBP51 crystal structure. Here, charge-Y could directly contact Hsp90 and complement or facilitate binding through the core TPR domain. However, it is difficult to reconcile this conformational state with data from FKBP51 chimeras which argue for matching of sequences between the core and extended portions of H7 (Figs. 2 and 4). Yet based on the FKBP51 crystal structure, there is unlikely to be direct contact between these regions. We propose the alternative possibility (Fig. 7B) that H7 may be disrupted in some circumstances, forming an eighth helix that continues the anti-parallel pattern of interactions observed with H1 through H7. In this conformational state, the putative H8 may contribute to core TPR domain interactions that enhance TPR affinity for Hsp90.
Although an anti-parallel H8 is not observed in the static FKBP51 crystal structure, crystal packing may not have favored a conformation of FKBP51 that exists in solution or that is induced by binding to Hsp90. Furthermore, there are precedents for alternative conformational states in TPR domains such as proposed in Fig. 7. Walkinshaw and colleagues (27) obtained two crystal forms for CyP40 from which two TPR conformations were resolved. One structure contained a TPR domain similar to PP5, Hop-TPR2a, and FKBP51. In the alternate structure, the loops separating H2-H3 and H3-H4 have been reorganized into ␣-helical forms, thus resulting in a single, greatly extended H2. A similar phenomenon was observed in the crystal structures for peroxisomal TPR protein PEX5. The structure for human PEX5 has the canonical anti-parallel helix arrangement (33). In contrast to this structure, Kumar et al. (34) obtained a structure for trypanosome PEX5 in which H5 and H6 were fused into a single extended helix. They proposed the possibility that certain TPR domains may naturally assume alternative conformations through extension versus folding back of helices, what they termed the "jackknife model" for TPR motif rearrangement (34). The alternate FKBP structures proposed in Fig. 7  Chimeric data suggest that FKBP51 may assume the closed conformation (Fig. 7B), whereas FKBP52 binds Hsp90 in an open conformation (Fig. 7A). One can visualize in the closed conformation how the charge-Y motif and adjacent amino acids may interact with amino acids in the core portion of H7 or even with side chains extending from H6. The C-terminal 30 -35 amino acids that are unresolved in the FKBP51 crystal structure are not illustrated in Fig. 7. However, this region also impacts Hsp90 binding (Figs. 3 and 4), enhancing binding by FKBP51 and moderating binding by FKBP52. In the jackknife model, this tail region would have to undergo a dramatic positional swing, but such a difference perhaps contributes to influences of the tail on Hsp90 binding.
Further structural studies are needed to test the jackknife model, and alternative explanations for our observations remain viable. For example, the charge-Y motif lies within a region that is purported to be a calmodulin binding site in FKBP52 (35). We considered whether Ca 2ϩ /calmodulin interactions could either influence the conformational state of this region or provide an alternative mechanism for distinguishing FKBP51 and FKBP52 interactions with Hsp90. However, we think a role for calmodulin is unlikely for several reasons. First, the putative calmodulin binding motif scores very weakly for FKBP52 and FKBP51 when analyzed in the Calmodulin Target Data base (calcium.uhnres.utoronto.ca/ctdb/flash.htm). Second, we have never observed calmodulin in FKBP complexes with Hsp90 or steroid receptors. Finally, neither Ca 2ϩ nor the Ca 2ϩ chelators EGTA or EDTA alter FKBP binding to Hsp90 (results not shown).
General Significance of TPR Cochaperone Interactions with Hsp90 and Client Proteins-Hsp90 serves a large number of client proteins that regulate cellular pathways, and in every case examined one or more cochaperones accompanied Hsp90 in client protein complexes. Individual clients displayed distinct preferences for certain Hsp90 cochaperones. For example, among the TPR cochaperones, Xap2/AIP is found in arylhydrocarbon receptor complexes but not steroid receptor or kinase complexes (29,30). FKBP51, FKBP52, PP5, and CyP40 associate differentially with progesterone, estrogen, and glucocorticoid receptor complexes in a receptor-specific manner (25,36,37). Preferential assembly with a client may be caused by direct interactions between client and the Hsp90 cochaperone (24), but the cochaperone may also influence how Hsp90 interfaces with client, either stabilizing or destabilizing client interactions relative to other Hsp90/cochaperone pairs. Thus, the TPR cochaperones that competitively interact with the MEEVD terminus of Hsp90 might, through unique contacts with Hsp90, induce distinct structural/functional changes in Hsp90 which elaborate the chaperoning of client proteins.