Hsp90 is a molecular chaperone present both in prokaryotes and eukaryotes. Its three-dimensional structure as well as the conformational changes associated with the ATPase cycle are conserved from bacteria to man. However, there are major functional differences in that prokaryotic Hsp90 (called HtpG for high temperature protein G) seems to act on its own while eukaryotic Hsp90 has a cohort of Hsp90-specific co-chaperones. Furthermore, a large number of ‘client’ proteins in eukaryotes depend strongly on Hsp90 to reach or maintain their biologically active states. In contrast, prokaryotic Hsp90 largely lacks established client proteins, as the deletion of the htpG gene does not lead to striking phenotypes. What prokaryotic Hsp90 is good for has remained puzzling for decades. The paper by Sato et al. in this issue of Molecular Microbiology is an important step towards solving this enigma. The authors demonstrate that cyanobacterial Hsp90 stabilizes specific proteins required for the assembly of large protein complexes called phycobilisomes. Their in vitro experiments using purified components suggest that the middle domain of Hsp90 is mainly responsible for binding the client protein studied. This authentic client system will allow delineating further the chaperone mechanism of prokaryotic Hsp90.
All organisms studied so far respond to adverse environmental conditions such as unphysiologically high temperatures by co-ordinately producing a set of specific stress resistance proteins called ‘heat shock proteins’ (Hsps). The major classes of Hsps are conserved between the kingdoms of life (Lindquist and Craig, 1988). Cells in which the heat shock response is abrogated by deleting the respective transcription factor (sigma 32 in Escherichia coli) accumulate large protein aggregates like the inclusion bodies observed when certain proteins are overproduced in E. coli (Gragerov et al., 1991). This phenomenon is not only observed under stress but also, albeit to a lesser extent, under physiological conditions, implying that at least some Hsps are required to maintain protein homeostasis. The major classes of Hsps have been termed molecular chaperones, as they share the remarkable ability to recognize non-native proteins and prevent unproductive folding processes such as the unspecific aggregation of these polypeptide chains.
One of these molecular chaperones, Hsp90, is an abundant protein in the cytosol of eukaryotes and bacteria, where it is called HtpG. The E. coli htpG gene was discovered through its homology to eukaryotic Hsp90 (Bardwell and Craig, 1987). Interestingly, mitochondria of higher eukaryotes contain an HtpG homologue called Trap1 (Felts et al., 2000; Leskovar et al., 2008). The deletion of htpG in E. coli is well tolerated, causing only a slight growth defect at high temperatures (Bardwell and Craig, 1988), while deletion of Hsp90 in yeast is lethal (Borkovich et al., 1989). Under stress conditions, bacterial Hsp90 seems to be involved in supporting de novo protein folding (Thomas and Baneyx, 2000). Despite the power of bacterial genetics, progress on deciphering its physiologic function in bacteria has been slow compared with the situation in eukaryotes. On the structural level, the three-dimensional structure of Hsp90 and the conformational cycle connected to ATP hydrolysis are conserved from bacteria to man (Ali et al., 2006; Richter et al., 2006; Shiau et al., 2006; Richter et al., 2008; Graf et al., 2009; Hessling et al., 2009). Hsp90 binds ATP in the N-terminal domains (Fig. 1). These domains then associate transiently and make contacts with the middle domain. Only then, in the closed state, is ATP hydrolyzed, allowing return of Hsp90 to the open state. Like eukaryotic Hsp90, HtpG suppresses the aggregation of model substrates such as citrate synthase (Jakob et al., 1995; Krukenberg et al., 2009; Sato et al., 2010). Despite this conservation in structure and mechanism, there are notable differences between Hsp90 from prokaryotes and eukaryotes.
First, prokaryotic Hsp90 lacks the long, charged linker between the N-terminal and the middle domains that is characteristic of all eukaryotic versions, and second, the Hsp90 system of the eukaryotic cytosol comprises a large number of co-chaperones, none of which is present in prokaryotes. Finally, and most importantly, we know of hundreds of client proteins for eukaryotic Hsp90 (Picard, 2002), but hardly any prokaryotic client proteins are known.
The paper by Sato and colleagues in this edition of Molecular Microbiology is an important step towards unveiling the function of prokaryotic Hsp90 (Sato et al., 2010). The Nakamoto lab had shown previously that in cyanobacteria Hsp90 is important for general stress management (Tanaka and Nakamoto, 1999). Here they demonstrate that cyanobacterial Hsp90 is an important player in the assembly and stability of phycobilisomes (Fig. 2), large protein complexes involved in light harvesting (Grossman et al., 1993), whose complex architecture requires a sophisticated assembly pathway (Anderson and Toole, 1998; Arteni et al., 2009). In a strain lacking htpG, the phycobilisomes are devoid of specific proteins, in particular the so-called linker protein LR30, which is important for the assembly and architecture of the phycobilisomes. As can be seen in Fig. 2, LR30 sits at a specific position in the rod part of the phycobilisome. It integrates into the adjacent disc and is important both for the formation of the correct rod structure and for the light harvesting properties of the complex (Liu et al., 2005). Consequently, phycobilisome complexes isolated from the htpG deletion strain differed in composition from the wild-type form (Sato et al., 2010). This is an exciting result, as it shows that HtpG is involved in the assembly of a macromolecular complex by specifically stabilizing a protein that has an important scaffolding role. There is also evidence that in E. coli, Hsp90 binds ribosomal proteins (F. Motojima, pers. comm.). These findings are reminiscent of an earlier observation from the Walsh lab on the involvement of HtpG in protein complex assembly in E. coli (Li et al., 1996). Here, HtpG was purified as a stoichiometic component of a peptide antibiotic synthase complex. However, this complex also assembled in the htpG deletion strain. Thus, in this context, this role of HtpG seems less important than its role in the assembly of the phycobilisome structure. Sato and co-workers go on to demonstrate with purified components that the phycobilisome assembly proteins are indeed recognized and bound by HtpG directly. Furthermore, they show that these proteins are labile; they precipitate at elevated temperatures and form inclusion bodies when produced in E. coli. One could argue that E. coli Hsp90 should keep these proteins soluble. However, as the authors report, unlike cyanobacterial Hsp90, purified E. coli Hsp90 cannot prevent the aggregation of LR30 upon dissociation of phycobilisomes. Thus, although the prokaryotic Hsp90s are quite homologous, they seem to exhibit functional specialization.
Despite decades of research on the mechanism of Hsp90 action, even a simple, fundamental question such as the localization of the binding site for the client protein remained unanswered. The jury is still out on the position of the client protein binding site in Hsp90, with reports claiming evidence for the interaction of model client proteins with each of the Hsp90 domains. If one takes the scattered pieces of evidence together, the N-terminal and the middle domain have been found repeatedly to interact with clients in vitro (Scheibel et al., 1988; Young and Hartl, 1997; Nemoto et al., 2001). However, there is also evidence for the involvement of the C-terminal domain (Yamada et al., 2003; Harris et al., 2004), and the linker region between the N-terminal and the middle domain (Scheibel et al., 1989).
In this study with LR30 in vitro, the N-terminal domain was only weakly active in preventing protein aggregation, a 10- to 20-fold excess over the client protein being required, while Hsp90 itself acted at stoichiometric ratios (Sato et al., 2010). The C-terminal domain was inactive. Thus, the most likely site of interaction is the middle domain. As it was not possible to produce the isolated middle domain, the evidence for it containing the chaperone-binding site is somewhat circumstantial. While the deletion analyses clearly implicate this domain, other domains, especially the C-terminal domain, seem to modulate its activity. However, one has to bear in mind that bacterial Hsp90 and the eukaryotic versions are dimers (Spence and Georgopoulos, 1989). Thus, constructs containing the C-terminal domain are dimeric, while those lacking the C-terminal domain are monomeric (see also Fig. 1). This means that by creating these domain deletion constructs, one changes the domain composition and the quaternary structure. It is not known whether the Hsp90 monomer binds client proteins as efficiently as the native dimer. A cryo-EM reconstruction of a eucaryotic Hsp90-kinase complex (in the presence of the kinase-specific co-chaperone Cdc37) suggests that binding occurs to one subunit of the dimer via the N-terminal and middle domains (Vaughan et al., 2006). Asymmetric binding of co-chaperones might be used generally to functionalize the two subunits in eukaryotic Hsp90 in a different manner, as shown for the co-chaperone Aha1 (Retzlaff et al., 2010). However, as mentioned above, prokaryotes seem to lack Hsp90 co-chaperones. Therefore, in the HtpG dimer, two chaperone sites might be used simultaneously. The results reported by Sato and colleagues (2010) for LR30 suggest, however, that one client is bound per dimer.
Finally, this study addresses another aspect of Hsp90 that has puzzled the field for two decades, namely the effect of ATP on the interaction with client proteins in vitro. Interactions of the chaperones GroE and Hsp70 with clients are clearly ATP-dependent. In simple terms, ATP binding and/or hydrolysis shifts the chaperone from high affinity to low affinity, thus regulating client binding and release. Published studies, including this one, with Hsp90 from different species show hardly any ATP dependence (Wiech et al., 1992;Krukenberg et al., 2009; Sato et al., 2010): the presence of ATP did not affect the ability of Hsp90 to suppress client aggregation. This is hard to reconcile with the finding that the ATPase activity is essential for the in vivo function of yeast Hsp90 (Obermann et al., 1998; Panaretou et al., 1998). It might be explained by the fact that the in vitro studies used model proteins. This is an important point, as Hsp90 differs from promiscuous chaperones such as the bacterial GroE system, which handles a large number of different proteins, seemingly only governed by their folding characteristics (Sparrer et al., 1996). Hsp90 seems to be more selective in this respect as, at least for the eukaryotic forms, current evidence suggests that it binds a specific set of client proteins in partially folded or even native-like conformations (Picard, 2002). Sato and colleagues used an authentic in vivo client protein for their in vitro studies. Nevertheless, ATP did not have much effect and therefore the influence of ATP binding and hydrolysis on the interaction with client proteins remains mysterious. Surprisingly, however, in the presence of ADP, HtpG was unable to suppress aggregation. Thus, prokaryotic Hsp90 might adopt a conformation upon ADP binding that is incompatible with client binding. There is structural evidence that ADP induces a compact conformation in bacterial Hsp90 (Shiau et al., 2006). One could also argue that the assay used does not reflect the in vivo situation, although it did recapitulate the binding of the authentic client to Hsp90. Incorporation of the Hsp90-stabilized LR30 into rod structures in vitro could be an important next step to expand the assay. In this context, one might also want to invoke Hsp70 (DnaK in bacteria) in the scheme of events. In the chaperone cycle models for eukaryotic Hsp90, Hsp70 is required to bring at least certain clients into the Hsp90 system (Smith, 1993), and it had been shown in vitro that Hsp70, together with its J-protein co-chaperone, can reactivate Hsp90-bound model clients (Freeman and Morimoto, 1996). It will be exciting to see what, if any, relationship exists between the chaperones Hsp70 and Hsp90 in prokaryotes in general, and in the folding and assembly of cyanobacterial phycobilisomes in particular.
The study by Sato and co-workers presents an important step forward in the quest for Hsp90 clients in prokaryotes. The experimental system described here will allow one to address more specific questions concerning mechanism and physiological importance. With more client proteins to be identified, we will see whether the role for bacterial Hsp90 defined here, the assembly of oligomeric protein complexes, will be a re-occuring theme.