| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Laboratory of Toxicology, University of Leuven, Belgium
1Correspondence: Laboratory of Toxicology, University of Leuven, E. Van Evenstraat 4, Leuven 3000, Belgium. E-mail: jan.tytgat{at}pharm.kuleuven.ac.be
ABSTRACT
The amino-terminal domain (N-domain) of Hsp90 represents the ATP binding site and is important for interaction with its cochaperone, p23. Whereas some evidence suggests that p23 may bind to this domain in an ATP-dependent manner and that this process requires the dimerization of two N-domains, the interaction sites between them and the molecular mechanism of coupling these two events to p23 binding remain unsolved. As a first step toward establishing the interaction mechanism, we used the evolutionary tracing (ET) method [Lichtarge, O., Bourne, H. R., and Cohen, F. E. (1996) J. Mol. Biol. 257, 342–358] to identify the putative functional surfaces of Hsp90 and p23, and combined with protein-protein docking techniques, to predict their binding interface. Both evolutionarily privileged surfaces of Hsp90 and p23 identified by ET appear on this putative interface. An analysis of the complex model produced using the ET results combined with available experimental data highlights a putative conformational pathway in the ATP binding domain of Hsp90, where a series of conformational changes transfer the ATP-induced N-domain dimerization signal for the binding of p23. In this pathway, the closure of "lid" may result in reorientation of the helix 
1 and the following loop (residues 10-27 in yeast Hsp90), which will expose more hydrophobic surface, and thus triggers the dimerization of N-domain.—Zhu, S., Tytgat, J. Evolutionary epitopes of Hsp90 and p23: implications for their interaction. FASEB J. 18, 940–947 (2004)
Key Words: molecular chaperone • evolutionary tracing • protein-protein interaction • conformational pathway
PROTEIN-PROTEIN interaction represents a common mechanism by which proteins can underlie complex biological functions in the cell (1)
. Molecular chaperones and their cochaperones comprise pairs of interaction partners, thereby offering an excellent model for studies of interaction effects on their functions. Hsp90 is a unique family of chaperone proteins responsible for folding some key molecules involved in signal transduction pathways (2
3
4
5)
. A variety of transcription factors, protein kinases, and cell cycle regulators constitute the major classes of substrates of Hsp90. The widespread distribution in nature, ranging from bacteria to plants and humans, and high conservation in sequence and structure together with distinct functions give them a central role in life and evolution. Although the bacterial homologue of Hsp90 (named HtpG) is dispensable, deletion of Hsp90 has been proved lethal to eukaryotic organisms.
Structurally, Hsp90 belongs to intracellular multidomain proteins with an N-domain of 24–28 kDa, a middle region of 38–44 kDa, and a C-domain of 11–15 kDa (6)
. A variable charged region links the N-domain and other segments of Hsp90. Accordingly, each domain fulfills its own function by interacting with different partner proteins (e.g., substrates and cochaperones) as well as ATP (2
3
4
5)
. In general, the N-domain is an ATPase and can bind and hydrolyze ATP. Moreover, it is likely involved in a direct contact with the cochaperone p23 (3
4
5
, 7
8
9
10)
. The middle region is thought to be a major site for client protein binding and also binds the cochaperone Aha1 (11)
. The C-domain functions as a major dimerization site by binding Hop/TPR proteins (12)
. However, communication among different domains is needed for maturating a substrate.
The cochaperone p23 protein modulates Hsp90 activity during the last stages of the chaperoning pathway (13)
. An obvious role of p23 in the Hsp90 system is known to function as a substrate release factor by coupling the ATPase activity of Hsp90 to polypeptide dissociation (14)
. The binding to Hsp90 requires only its amino-terminal ß-sandwich domain but the unstructured carboxyl-terminal region is necessary for its chaperoning activity (15)
. Despite much biochemical and genetic data about their association, little progress has been made in understanding their molecular interactions. The difficulty most likely is explained by the lack of significant correlation between mutations and binding effects (16)
. Many mutations in Hsp90 that impair p23 binding are found to virtually affect ATPase activity and conformation changes, and consequently provide no conclusive evidence. Because p23 binding to Hsp90 is a stringent test and requires 1) that the amino-terminal nucleotide binding domain is in an ATP-bound conformational state; 2) communication between the nucleotide binding domain and a noncontiguous downstream domain; and 3) dimerization and amino-terminal domain proximity, it appears that genetic studies or point mutation strategy alone cannot be used to infer their molecular interactions (7
8
9)
. Because rapid advances in evolutionary genomics have led to better methods for extracting structurally and functionally important information from the evolutionary history of proteins based on a multiple sequence alignment, it is possible to use these new approaches to elucidate evolutionarily privileged sites (evolutionary epitopes) and correlate them to putative binding interfaces based on available experimental data and other related techniques.
By using the evolutionary tracing (ET) method (17)
, we identified several evolutionarily important clusters in the N-domain of Hsp90 and the core domain of p23, and combined molecular docking techniques (18)
to highlight a putative interface between these two molecules. A "conformational pathway" hypothesis was proposed to explain their molecular interaction mechanistically.
EVOLUTIONARY EPITOPES OF Hsp90 GENE FAMILY
The ET method developed by Lichtarge et al. is a powerful computational approach of genetic analysis that has been used to analyze the functional surfaces and interaction sites for protein families involved in diverse biological processes (17
, 19
20
21
22
23)
. This method extracts evolutionarily important information based on the phylogenetic tree of a homologous protein family, where a position is identified as class specific if its residues become invariant within one branch but vary between branches. The smallest number of branches at which one position becomes invariant within each branch defines its rank. The root branch is defined as rank 1, indicating that all residues in one position are conserved across the alignment.
To perform ET, a BLAST search of GenBank database (http://www.ncbi.nlm.nih.gov) was used to obtain raw data. Using human and yeast Hsp90 sequences as queries, we recovered 120 related sequences (Fig. 1
A),all of which were identified to contain the conserved signature motifs of Hsp90 family with easily recognized domain organization, indicating that they are homologous. This is further supported by structural prediction of N-domain sequences by comparative modeling (24)
. Based on 
40% sequence similarity, this prediction shows that the domain from five distinct subgroups (bacterial, cytosol, GRP94, chloroplast, and Hsp75/TRAP1) share a similar 3-dimensional fold (Fig. 1B
). The N-domain sequences of all 120 proteins were therefore chosen for ET analysis.
|
ET extracted 61 trace residues at ranks less than or equal to 45, of which 23 are conserved across the Hsp90 family, including 8 bulky hydrophobic residues (24Y, 84M, 96I, 104F, 117I, 120F, 130V, and 148W), 4 charged residues (32R, 33E, 40D, and 79D), and 11 small residues (36S, 37N, 81G, 83G, 99S, 100G, 118G, 121G, 123G, 170G, and 198S) (residues are numbered according to yeast Hsp90). When locating the residues on the 3-dimensional structure, we identified two distinct but contiguous clusters: A1 and A2 (Fig. 1C
). Cluster A1 lies at or near the ATP binding pocket and is composed of nearly all conserved and some class-specific residues. Because four sides comprising the nucleotide binding pocket (25)
contain an array of trace residues (Fig. 1C
), we hypothesize that most trace residues in this cluster are likely to be important for direct regulation of ATP exchange and pocket opening. A striking feature in the distribution of trace residues in the protein secondary structure elements is that three -helices (2, 5, and 6) contributes many trace residues to cluster A1. By contrast, the distal face of the pocket formed by ß-strands displays little ET signal.
Of 12 point mutations that affected ATP binding and/or ATPase activity and/or p23 recruitment (7
8
9
10
, 26)
, 10 (33E, 37N, 41AS, 79D, 81G, 101ST, 118G, 121G, 123G, and 170G, numbered according to yeast Hsp90) were identified as trace residues, of which some are likely involved in direct protein interactions to ATP (such as conserved residues 37, 41, 79, and 121). Mutations of these residues (N37A, A41V, D79A, and G121V) normally lead to low ATP affinity but do not affect ATPase activity. Some mutations impair p23 binding (such as N37A, D79A, G81D, G118V, G121V, G123V, and G170D), which is due to uncoupling ATP binding or results from a changed conformation that cannot be recognized by p23. The 33E, conserved in the GHKL superfamily (27)
, represents the catalytic residue required for ATP hydrolysis. Its mutation is accompanied with a sharp loss in ATP hydrolysis activity. Both mutations (D79N and T101I) can reduce ATPase activity, but the former also inhibits ATP binding.
Outside the pocket, ET detected a small exposed patch, named cluster A1', composed of five trace residues (198S, 199DENQSY, 200FH, 201IN, and 204DP) and located in a loop linking helix 9 and strand ß8) (Fig. 1C
). Evolutionary importance together with a large exposure onto molecular surface suggests that they form a binding site for a downstream region of Hsp90 or a cochaperone.
Cluster A2 emerges when rotating the molecule relative to cluster A 90° about x axis, then 120° about y axis. It consists of 13 trace residues [one conserved residue (24Y) plus 12 class-specific residues (15ILMV, 17DHKNQS, 18IL, 19IM, 20AIV, 21EHKNR, 25KRST, 26DEHKN, 28ED, 114AEILMNQSTV, 152ADGNQST, and 178AEKPRSTV)], most located in the region of the first 30 amino acids and primarily contributed by the helix 1 and the following loop. These residues form a continuous stretch with trace residues on the nucleotide binding pocket. In this stretch, residues 24, 25, 26, and 28 are located on a loop linking two helices (1 and 2) and are partially exposed onto cluster A1 (Fig. 1C
). These unique features make them a candidate region for functional connection between two clusters. Cluster A2 is a probable contact surface for the dimerization of N-domains. Two lines of evidence support this hypothesis. First, this cluster contains four bulky hydrophobic residues (15, 18, 19, and 20) (all contributed by 1) surrounded by several polar or charged residues (17, 21, 152, and 178), which form a typical contact interface (28)
. A second line of evidence comes from experimental work performed by Richter et al. who demonstrated that deletion of the first 24 amino acids of yeast Hsp90 abolished the trans-activation between the two N-domains (29)
. Cluster A2 functioning as the dimerization surface also accounts for the inability of Hsp75/TRAP1 in binding to p23 (30)
. An examination of residue type on the trace positions reveals a clear difference in the substitution pattern in two positions (20 and 21). In Hsp75/TRAP1, these positions are occupied by 20A and 21EHKR. In the cytosolic Hsp90s, which are known to have the ability to bind p23, they are 20IV and 21N. The substitution from Ile or Val to Ala (IV20A) would significantly decrease the hydrophobic intermolecular interaction and thus destabilize N-domain dimerization, whereas N21EHKR would produce an electric charge expelling effect between two surfaces. Thus, in both cases the N-domain dimerization is disfavored and leads to the loss of p23 binding ability in Hsp75/TRAP1.
EVOLUTIONARY EPITOPES OF p23 GENE FAMILY
A search of GenBank database (http://www.ncbi.nlm.nih.gov) using yeast and human p23 sequences for BLAST queries retrieved 25 related sequences, which can be divided into 3 subgroups based on their domain organization (Fig. 2
A-C): 1) human p23-like homologues with an amino-terminal antiparallel ß-sandwich core domain and a carboxyl-terminal unstructured acidic region (31)
; 2) yeast p23-like homologues with a GM/A-rich (G: glycine; M: methionine; A: alanine) domain situated within the acidic region (26)
. This additional domain appears to be gained by an exon shuffling event after gene duplication during evolution (Fig. 2D
). Its distribution restricted to fungi, protists, and plants, suggesting these two events occurred independent of metazoans in early evolution; and 3) B-ind1-like (butyrate-induced 1) homologues have a carboxyl-terminal PTPLA (a putative tyrosine phosphatase) domain fused with p23 (32)
. Despite diversification in domain organization, the core region interacting with Hsp90 appears to be structurally conserved among these subgroups, which has been confirmed by comparative modeling (Fig. 2B
), suggesting that these new members of p23 family might have ability to bind Hsp90. Given that the carboxyl terminus is responsible for the chaperone activity of p23, the diversity in this region likely reflects a need for substrate specificity in different evolutionary lineages. In support of this speculation, human B-ind1 protein has been suggested to function as a chaperone in the Rac1 (a member of Rho family of small GTPase) signaling.
|
In p23, ET identifies two clusters (B1 and B2, at ranks less than or equal to 11) located on one face of p23. In contrast, no ET signal was detected on the opposite face (Fig. 2E
). Cluster B1 is found to be near a previously recognized cavity (31
, 33)
and is composed of five conserved (8W, 79K, 86W, 89L, and 106W) and four class-specific residues (100AKRS, 102DN, 103FW, and 109EW) (residues are numbered according to human p23). Cluster B2 comprises 10 class-specific residues (2HILQRV, 4AP, 18DENRT, 21CILV, 23DES, 39ADFKLPST, 40ACGKS, 43GS, 71PQR, and 72CDHNQSVY). Overall, B1 and B2 display different side chain features: B1 can be considered as a hydrophobic patch where two aromatic residues (103F/W and 106W) are completely exposed to the molecular surface, whereas B2 is primarily composed of charged and polar amino acids. These two clusters are most likely involved in interaction with Hsp90. Support for this idea comes from mutagenesis experiments of Oxelmark et al. who showed that the surface containing these two clusters is important in estrogen receptor (ER) signaling and p23 binding to Hsp90 (33)
. Of four mutations that diminish p23 activity, three are found to be located within these two clusters (4, 96, and 99 in human p23, corresponding to 13, 114, and 117 of yeast p23) (Fig. 2E
). On the contrary, six mutations that do not perturb activity were not identified as trace residues. The lack of ET signal in the interface of the crystal structure dimer of human p23 supports the conclusion that this dimer is really an artifact of crystallization (31)
.
MOLECULAR COMPLEX MODEL OF Hsp90 AND p23
Though it is clear that the binding site for Hsp90 is contained in the folded amino-terminal ß-sheet domain of p23 (residues 1–110 in human p23) (15)
, it has remained uncertain for the corresponding site in Hsp90. Mutations in the N-domain affect the binding of p23, suggesting that this domain is essential for the interaction with p23. However, the isolated N-domain does not bind p23, and two deletion mutations near the carboxyl terminus and two point mutations at positions 485 and 525 of Hsp90 have been shown to abolish p23 binding (16)
. By using a series of Hsp90 mutants with different truncation in its carboxyl-terminal region, Chadli et al. showed that the amino-terminal half of Hsp90 (not including the charged region) is crucial for p23 binding (16)
. Shorter truncation mutants (NC450, NC431, NC220 of yeast Hsp90), which bind ATP but do not dimerize in its presence and lack of ATPase activity, abolished p23 binding (10)
, suggesting that the involvement of a downstream domain is important for dimerization of Hsp90, a prerequisite for p23 binding. It appears that p23 is most likely in a direct contact with the N-domain but requires Hsp90 in its dimeric state.
With the ET results in hand, we can now tentatively test whether the evolutionarily privileged clusters of Hsp90 and p23 are putative sites for their interaction. Since the cluster A2 represents a surface for the dimerization of the N-domains of Hsp90, this leaves cluster A1 (possibly including A1') as the only binding site for p23. To test this hypothesis, the docking program BIGGER was used to generate and evaluate a family of plausible binding modes between Hsp90 and p23 (18
, 34)
. Two docking tests are performed by using human p23 (1EJF) as a probe and either yeast ATP-bound Hsp90 (1AM1) or human ADP-bound Hsp90 (1BYQ) as a target. In both tests, the 1000 best solutions generated were evaluated and ranked according to a combination of additional interaction criteria (including electrostatic energy of interaction, relative salvation energy, and the relative propensity of adjacent side chains to interact). Strikingly, 20 complex models ranked as most favorable display significant similarity and two tests produced nearly identical results (data not shown). Two tests produced nearly identical results. As expected, when mapping onto the complex structure (rank=1) we found that the evolutionarily privileged residues in cluster A1 of Hsp90 appear in the putative interface, which comes close to the binding face of p23 (Fig. 3A, B
). In this complex model, a binding site is defined between cluster B1 of p23 and the helix 6 belonging to a part of cluster A1 of Hsp90, all identified as trace residues. Two exposed aromatic residues in positions 103 and 106 directly contact the nucleotide-bound pocket and 103 is close to ATP. Cluster B2 of p23 is close to the end of the long helix 2 of Hsp90 and thus is defined as the second binding site. The regions lacking ET signal in Hsp90 and p23 are all far from the interface (Fig. 3
).
|
Both noncontiguous binding sites on the interface display consistent mutation behavior—e.g., conserved to conserved (site I) and class specificity to class specificity (site II) (Fig. 3)
—which likely represents an evolutionary coupling (coevolution) between molecular chaperone and cochaperone, as seen in the system of receptors and ligands (35)
.
A PUTATIVE CONFORMATIONAL PATHWAY IN THE N-DOMAIN OF Hsp90: CRUCIAL FOR p23 BINDING?
An emerging trend in the study of the structural basis of Hsp90 function is the observation of ATP-induced conformational change associated with its biological process (2
, 10)
. Based on our complex model and ET results combined with available experimental data, a putative conformational pathway for p23 binding can be elucidated in the N-domain of Hsp90, where four helices (1, 2, 5, and 6) highlighted by ET could function as conformational switches.
Although no experimental structure has been reported for an Hsp90 N-domain with the lid in a closed conformation, different conformations of the ATP lid have been observed in multiple crystal structures of the ATP binding domain of Hsp90. This is consistent with the result of a normal mode analysis for conformational flexible prediction (36)
that indicates that two helices (5 and 6) of Hsp90 N-domain move the most and probably in concert (data not shown). According to the "molecular clamp" mechanism proposed by Prodromou et al. (10)
, ATP-bound inducing a "lid" closure that leads to transient dimerization of N-domains and p23 can recognize this conformation and bind to Hsp90. Our results strongly support the existence of such a mechanism. Because the lid segment of Hsp90 comprising helices 5/6 and their connecting loop is structurally packed against the helix 1 and the following loop formed by residues 10–27 (25)
, it is reasonable to infer that closure of the lid will trigger a remarkable conformational change in the helix 1 and the following loop that may result in an increased side chain exposure in the four hydrophobic residues for N-domain dimerization. This is consistent with the notion that these three helices (1, 5, and 6) form an integral structural unit and may have the potential to act as a conformational switch (37)
. This dimeric signal may be transferred to the end of the long helix (2) through conformational transitions (structural reorganization) mediated by the connecting loop of 1/2 and 2 itself and produce a "fit" contact site recognized by the cluster B2 of p23 (interaction site II) (Fig. 3D
). A similar mechanism has been proposed to explain how the transmembrane AT-1 angiotensin receptor can bind G-protein of cytoplasmic surface in response to an extracellular ligand (38)
.
On the other hand, provided their close proximity in the complex model (Fig. 3)
, lid closure may offer a more suitable conformational surface for the binding of the cluster B1 of p23 (interaction site I). In support of this hypothesis, three point mutations (G118V, G121V, and G123V in yeast Hsp90) at the lid segment impair p23 binding (8)
. Because valine replacement of glycines is believed to be associated with reduced backbone freedom, these mutations may seriously constrain the conformational change of the lid and consequently disfavor p23 binding. The lid segment has recently been characterized to be involved in a direct interaction with an "early" cochaperone p50cdc37, where it adopts a more open conformation, thereby preventing trans-activating interaction between two N-domains (39)
. However, p23 obviously cannot recognize this conformation. The fact that p50cdc37 and p23 associates with Hsp90 complexes in different stages of the chaperoning pathway suggests that this evolutionarily important segment might be able to interact with two distinct cochaperones by adjusting its conformation in the chaperoning process.
Our model thus provides a reasonable explanation as to how the ATP-induced lid closure and N-domain dimerization can affect p23 binding. Although ATP binding is the first step of this conformational pathway (Fig. 3D
), this event alone is insufficient to ensure p23 binding because it also requires that two N-domains be in a dimeric state (10)
. Considering deletion of downstream dimerization-associated elements leading to no sufficient support for N-domain proximity, which in fact abolish the ability of two N-domain dimerization even in the existence of ATP (10)
, it appears that the dimeric N-domain is a real initial signal for p23 binding (Fig. 3D
). This is consistent with some experimental observations (10
, 16)
and agrees with the role of dimerization in other biological systems (40)
. For example, in some cell signaling pathways, dimerization often functions as a common initial signal for conformational transitions.
Communication between the nucleotide binding domain and a noncontiguous downstream domain of Hsp90 is also needed for p23 binding. Given that cluster A1 of Hsp90 contacts directly with clusters B1 and B2 of p23 (Fig. 3C
) and that cluster A2 could function as a dimerization surface, cluster A1' remains the only site for communication with the downstream domain and likely mediates the association among the N-domain, downstream region, and p23. This is supported by its unique location in the complex model and the reconstructed dimer model proposed by Meyer et al. (11)
. The latter was based on a comparative structural analysis, where the N-domain and middle segment (residues 2–525 in yeast Hsp90) were included. In our complex model, cluster A1' is in a close proximity to the binding face of p23 (Fig. 3C
); in the reconstructed dimer model this cluster forms an extensive contact with the downstream domain of Hsp90. An attractive scenario is that, by direct interaction with the downstream domain, cluster A1' has the ability to mediate p23 binding to Hsp90.
CONCLUDING REMARKS
The conformational change of proteins represents a major event involved in signal transduction of many biological processes by which an initial signal can be transferred and activate a new reaction. Such a mechanism likely exists in Hsp90-p23 interaction. N-domain dimerization of Hsp90 induced by ATP binding, as an initial signal, appears to be crucial for binding of p23 to Hsp90. Although the present results do not provide direct evidence for this process, the existence of a putative conformational pathway involved in chaperone and cochaperone interaction opens a door to further investigation of this interaction from a mechanistic perspective. Successful tracing of the evolutionary epitopes of Hsp90 and p23 will undoubtedly provide a systematic and rational guide for engineering their mutations.
ACKNOWLEDGMENTS
This work was supported in part by a K. U. Leuven postdoctoral fellowship to S.Z. We gratefully thank Vadim Alexandrov for fruitful discussions on the conformational change of Hsp90 and Olivier Litchtarge for help in the ET method.
Received for publication January 27, 2004. Accepted for publication February 12, 2004.
REFERENCES
This article has been cited by other articles:
|
|
 |
|
  C. Weber, G. Guigon, C. Bouchier, L. Frangeul, S. Moreira, O. Sismeiro, C. Gouyette, D. Mirelman, J. Y. Coppee, and N. Guillen Stress by Heat Shock Induces Massive Down Regulation of Genes and Allows Differential Allelic Expression of the Gal/GalNAc Lectin in Entamoeba histolytica Eukaryot. Cell, May 1, 2006; 5(5): 871 - 875. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
