SUMMARY
Hop/Stip1/Sti1 is thought to be essential as a co-chaperone to facilitate substrate transfer between the Hsp70 and Hsp90 molecular chaperones. Despite this proposed key function for protein folding and maturation, it is not essential in a number of eukaryotes and bacteria lack an ortholog. We set out to identify and to characterize its eukaryote-specific function. Human cell lines and the budding yeast with deletions of the Hop/Sti1 gene display reduced proteasome activity due to inefficient capping of the core particle with regulatory particles. Unexpectedly, knock-out cells are more proficient at preventing protein aggregation and at promoting protein refolding. Without the restraint by Hop, a more efficient folding activity of the prokaryote-like Hsp70/Hsp90 complex, which can also be demonstrated in vitro, compensates for the proteasomal defect and ensures an alternate proteostatic equilibrium. Thus, cells may act on Hop to shift the proteostatic balance between folding and degradation.
INTRODUCTION
Homeostasis of the proteome, often referred to as proteostasis, is essential both at the cellular and organismic levels for health and longevity (Balch et al., 2008; Hipp et al., 2014; Labbadia and Morimoto, 2015). Proteotoxic stresses lead to protein misfolding and aggregation, which trigger and are counterbalanced by protein quality control mechanisms. The major cellular quality control mechanisms, all assisted by molecular chaperones, are protein refolding and the degradation of misfolded and aggregated proteins by the proteasome and by autophagy (Wong and Cuervo, 2010; Chen et al., 2011; Hipp et al., 2014; Balchin et al., 2016). Defects in any of these mechanisms can cause severe proteotoxicity, which in turn can lead to diseases such as cystic fibrosis, lysosomal storage diseases, cancer, and, most notably, neurodegenerative disorders such as Huntington’s, Parkinson’s, and Alzheimer’s diseases (Balch et al., 2008; Chen et al., 2011; Hipp et al., 2014; Schmidt and Finley, 2014; Labbadia and Morimoto, 2015).
In eukaryotes, the Hsp70 and Hsp90 molecular chaperone machines are major contributors to proteostasis by providing a platform for folding of both nascent polypeptides and misfolded, structurally labile and mutated proteins, collectively called “clients” (Mayer and Bukau, 2005; Echeverria and Picard, 2010; Kampinga and Craig, 2010; Picard, 2012; Taipale et al., 2012; Schopf et al., 2017; Radli and Rudiger, 2018; Moran Luengo et al., 2019). For folding and assembly of clients, both Hsp70 and Hsp90 undergo large conformational changes and collaborate with co-chaperones (Li et al., 2012; Mayer and Le Breton, 2015; Schopf et al., 2017). One of these co-chaperones is the Hsp70-Hsp90 organizing protein (Hop), encoded by the gene STIP1 in mammals. It is an adaptor molecule between the Hsp70 and Hsp90 molecular chaperone machines, which facilitates the folding or stabilization of clients by promoting their transfer from the Hsp70 to the Hsp90 molecular chaperone machines after the initial recognition and binding of clients by Hsp70 in collaboration with its J-domain containing co-chaperone Hsp40 (Scheufler et al., 2000; Kirschke et al., 2014; Mayer and Le Breton, 2015). Hop forms a ternary complex with Hsp70 and Hsp90 using its tetratricopeptide repeat (TPR) domains. Two of its three TPRs, TPR1 and TPR2A, specifically bind the extreme C-terminal peptide sequences EEVD and MEEVD of Hsp70 and Hsp90, respectively (Scheufler et al., 2000; Schmid et al., 2012; Bhattacharya et al., 2018).
Proteins, whose folding or refolding fails, either because they cannot fold by themselves or with the assistance of molecular chaperones, are degraded by the proteasome, a highly conserved and regulated eukaryotic protease complex. About 80% of total cellular protein turnover is through this complex (Collins and Goldberg, 2017); moreover, the proteasome works together with an Hsp40-Hsp70-Hsp110 protein disaggregase complex to eliminate intracellular aggregates (Shorter, 2011; Hjerpe et al., 2016). The proteasome is a 1.6 to 2.5 MDa complex consisting of a 20S proteolytic core particle (CP) and a 19S regulatory particle (RP); the CP can be capped by one or two RPs resulting in 26S or 30S particles, respectively (Murata et al., 2009; Gallastegui and Groll, 2010). The RP is divided into a lid and a base and has unique regulatory functions; it recognizes ubiquitinated substrates produced by the E1-E2-E3 ubiquitination system, promotes their deubiquitination and unfolding and the subsequent gate-opening of the CP, and finally the loading of the processed substrates into the proteolytic chamber (Collins and Goldberg, 2017). Dedicated chaperones for the assembly of CP and the RP base are well known, whereas RP lid assembly is still not well understood (Murata et al., 2009). Interestingly, Hsp90 has been proposed to be an assembly chaperone for the RP lid complex based on genetic interactions in the budding yeast (Imai et al., 2003) and the reconstitution of the RP lid complex in E. coli co-expressing yeast Hsp90 (Lander et al., 2012).
Prokaryotes do have Hsp70 and Hsp90 orthologs but lack a Hop-like protein. Bacterial Hsp90 and Hsp70 physically and functionally interact directly without a Hop-like protein (Genest et al., 2015; Kravats et al., 2017). In eukaryotes, Hop is not absolutely indispensable as mutant budding yeast, worms (Caenorhabditis elegans), and flies (Drosophila melanogaster) are viable with only mild phenotypes (Nicolet and Craig, 1989; Song et al., 2009; Ambegaokar and Jackson, 2011). In contrast, the deletion of STIP1 is lethal early in embryonic development in the mouse (Beraldo et al., 2013), possibly indicating that the function of Hop might be cell type-specific or dependent on the specific cellular state or requirements. In this study, we have explored why Hop is present in eukaryotes, what its critical functions are, and whether and how the eukaryotic Hsp70-Hsp90 molecular chaperone machines may function without Hop to ensure proteostasis. Our studies on the functions of Hop as a co-chaperone of the Hsp70/Hsp90 molecular chaperone machines and facilitator of protein folding and assembly led us to the discovery of alternative cellular strategies that ensure proper protein folding and proteostasis in human and yeast cells lacking this co-chaperone. These findings highlight the persistence of evolutionarily more ancient mechanisms in eukaryotic cells that may contribute to balance protein folding and degradation under certain conditions.
RESULTS
Human Hop Knock-out Cells Maintain Cellular Fitness and Proteostasis and Are Not Hypersensitive to Proteotoxic Stress
To study the functions of Hop in eukaryotic cells, we knocked out the gene STIP1, which encodes mammalian Hop, in several different human cell lines by using the CRISPR/Cas9 gene editing technique. The absence of the full-length Hop in knock-out (KO) clones was confirmed by immunoblotting using a specific antibody to Hop (Figure 1A). We did note that the HEK293T clone KO1 expresses a residual low level of a truncated form of Hop, which we characterized by mass spectrometry (MS) (Table S1); however, in subsequent experiments presented below, KO1 proved to behave essentially like the other HEK293T clone (KO33), which is devoid of any detectable trace of Hop. The frequency of obtaining KO clones with these human cell lines ranged between 33-46% indicating that Hop is not absolutely essential in human cells similarly to what had previously been found for the budding yeast Saccharomyces cerevisiae (Nicolet and Craig, 1989). Morphological examination revealed no obvious differences between wild-type (WT) and KO cells (Figure 1B). Growth rates of KO cells are only moderately reduced compared to WT cells (Figure 1C); this observation was supported by cell cycle analyses in that KO cells show less cyclic phase cells (S - G2/M) and more cells in the G0/G1 resting phase (Figure 1D, Figure S1A-B). The flow cytometric analyses of cells stained with annexin V and propidium iodide (PI) showed that none of the KO cell lines have any obvious viability issues (Figure 1E, Figure S1C-D).
We then determined how these KO cells survive in various stress conditions. We exposed WT and KO cells to thapsigargin, DTT, and A23187, which all induce the unfolded protein response of the endoplasmic reticulum, and to the oxidative stress inducer H2O2. Remarkably, KO cells are clearly not hypersensitive to these stress agents since these lead to comparable numbers of dead cells (Figure 1F) and to a rather reduced impact on cell morphology compared to WT cells (Figure S1E). Upon exposure to heat shock (HS), a stress which induces protein misfolding and aggregation, KO cells showed a similar sensitivity in the HEK293T background, but a markedly reduced sensitivity in the HCT116 and A549 backgrounds (Figure 1G). We also challenged both WT and KO cells with azetidine-2-carboxylic acid (AZC), a toxic analog of proline; incorporation of this proline analog into nascent proteins causes misfolding and subsequent protein aggregation. KO cells are more resistant to this proteotoxic stress, based on the comparison of the cellular morphology of AZC-treated WT and KO cells (Figure S1F).
A label-free MS quantitation of the whole cell proteome indicated that the vast majority of the proteome is unaltered between WT and KO cells (Figure 1H, Figure S1G). With our cut-offs for biologically significant changes of protein levels (see STAR Methods), only those of the top and bottom 4 and 12% of the identified proteins of HCT116 and HEK293T cells, respectively, are altered by the KO. These results indicated that the absence of Hop does not compromise the overall proteostatic equilibrium; instead, if anything, proteostatic buffering of KO cells becomes more robust and resilient to proteotoxic stress conditions.
The Proteasome and the Hsp70-Hsp90 Molecular Chaperone Axis Are Differentially Required in KO Cells
Cells experience intrinsic or environmental stress-induced proteotoxicity and they notably cope with these stresses by using two important protein quality control mechanisms: degradation of misfolded proteins by the proteasome and in some circumstances by autophagy, and refolding and stabilization of misfolded proteins by molecular chaperones, including the Hsp70-Hsp90 molecular chaperone machines (Figure 2A). To assess how these mechanisms are operating in KO cells, we pharmacologically blocked the functions of Hsp90, Hsp70, the proteolytic activity of the proteasome, and ubiquitination. We observed that KO cells are hypersensitive to Hsp90 inhibition with geldanamycin (GA), as demonstrated by an enhanced cell cycle arrest in the G2/M phase, and enhanced accumulation of apoptotic cells (Figure 2B, Figure S2A). This observation was further confirmed by a striking GA-induced morphological collapse (Figure S2B) and increased accumulation of dead cells with KO cells (Figure S2C). KO cells also showed a higher sensitivity to the chemically different Hsp90 inhibitor PU-H71 (Figure S2D) and the C-terminal Hsp90 inhibitor novobiocin (Figure S2E). Remarkably, these findings are reminiscent of the observation with yeast that Hsp90 inhibition and Δsti1 are synthetically lethal (Liu et al., 1999), indicating that the synthetic lethality of a STIP1/STI1 deletion and Hsp90 inhibition is evolutionarily conserved. Similarly to Hsp90 inhibitors, the Hsp70 inhibitor JG-98 blocks the growth of KO cells more efficiently than that of WT cells (Figure 2C, Figure S2F). We conclude from these experiments that both Hsp90 and Hsp70 continue to be functionally required in KO cells, which appear to be even more dependent on these molecular chaperones.
To check the other side of the coin of proteostasis, we next performed the cytotoxicity assays with proteasomal inhibitors. The treatments with both MG132 and bortezomib resulted in reduced cytotoxicity with all KO cell lines (Figure 2D, Figure S2G-H) and a less severe impact on cellular morphology than with WT cells (Figure S2I). KO cells also displayed a higher resistance to the E1 ubiquitin ligase inhibitor PYR41 (Figure 2E, Figure S2J). These results suggested that KO cells are less dependent both on ubiquitination and proteasomal degradation. Moreover, transient overexpression of full-length WT Hop in KO cells completely reversed the sensitivity to Hsp90 and proteasomal inhibitors (Figure S2K-M). These results led us to hypothesize that proteostasis in KO cells is ensured by an alternative equilibrium between protein degradation by the proteasome and protein stability/refolding supported by the Hsp70-Hsp90 molecular chaperone machines (Figure 2F). This raises two questions: (1) Is the function of the ubiquitin-proteasome system compromised in the absence of Hop? (2) How can the Hsp70-Hsp90 molecular chaperone machines function efficiently in the absence of Hop?
The Hsp70-Hop-Hsp90 Ternary Complex Is Physically and Functionally Associated With the Proteasome
To determine whether Hop is responsible by itself for optimal proteasomal function or whether it requires an association with Hsp70 and Hsp90, we carried out immunoprecipitation (IP)-MS experiments with Hop mutants that are defective for Hsp70/Hsp90 binding. Briefly, we generated the TPR1 mutant K8A, which does not bind Hsp70, the TPR2A mutant K229A, which does not bind Hsp90, and the corresponding double mutant (Bhattacharya et al., 2018). We confirmed the expected interaction patterns of these HA-tagged Hop mutants with Hsp70 and Hsp90 by IP (Figure 3A). Based on this result, we defined the Hop-specific interactome by an IP-MS analysis with transfected HEK293T KO cells. After an initial quality control by SDS-PAGE (Figure S3A), samples were subjected to label-free LC/MS-MS analysis. By comparison with the proteins associated with the TPR double mutant, ~41% were identified only with WT Hop and ~57% were enriched with WT Hop (Figure 3B). Thus, the Hop interactome is largely dependent on the ability of Hop to bind Hsp70 and Hsp90. We took a closer look at the Hop interactome by focusing on the components of the Hsp70-Hsp90 molecular chaperone machines. All of these proteins are either enriched or only present with WT Hop, including Hsp70 (HSPA1), Hsc70 (HSPA8), Hsp90α (HSP90AA1), and Hsp90β (HSP90AB1) (Figure 3C). By comparison, Grp94 (HSP90B1), the endoplasmic reticulum-specific Hsp90 isoform, and not a known interactor of Hop, yields the lowest enrichment (~1.5 fold, Figure 3C) with WT Hop in this subset of proteins. When we performed a gene ontology (GO) term enrichment analysis with all preferred interactors of WT Hop using the Enricher web server (http://amp.pharm.mssm.edu/Enrichr), focusing on KEGG and WIKI pathway-annotated proteins, we observed that proteasomal and proteasome-associated ubiquitin-related proteins are overrepresented with WT Hop (Figure S3B-D). Upon searching for all the proteasomal core components and ubiquitin-related proteins, we found that all of these proteins are either enriched or only present with WT Hop (Figure 3D). Interestingly, out of the 19 identified proteasomal components, 16 are associated with the RP of the proteasome (Figure 3D). We reevaluated a published MS dataset of IPs of a panel of proteasomal components from yeast (Guerrero et al., 2008). Remarkably, three of the bait proteins jointly pulled down Sti1 (yeast Hop), Hsp82/Hsc82 (yeast Hsp90s), and Ssa1/2 (yeast Hsp70s), compatible with the conclusion that the ternary complex associates with the proteasome in yeast (Figure S3E). Thus, the Hsp70-Hop-Hsp90 ternary complex is physically associated with proteasomal components, most notably with RP proteins.
The next experiments were designed to test whether the Hsp70-Hop-Hsp90 ternary complex is functionally relevant for the proteasome. We performed an in vitro proteasomal activity assay with extracts from WT and KO cells and noticed that the total steady-state activity of the 26S/30S proteasome is reduced across all KO cell lines (Figure 3E, Figure S3F). In contrast, the rate of proteasomal activity is not significantly altered in KO cells (Figure 3F, Figure S3G). These results suggested that functional 26S/30S proteasome particles might be less abundant in KO rather than less functional on a per particle basis. We confirmed these in vitro results with an in vivo proteasomal activity assay (Dantuma et al., 2000); this involved the transient expression and flow cytometric quantitation of a degradation-prone ubiquitin-GFP fusion protein (Ub-R-GFP) in parallel to its stable counterpart (Ub-M-GFP) as a control (Figure 3G, Figure S3H). Upon transient expression of WT and TPR mutant Hop in KO cells, the proteasomal activity is rescued by WT Hop, whereas single TPR mutants are not as efficient and the TPR double mutant behaves like a completely dead mutant regarding this function (Figure 3H, Figure S3I). These results established that the Hsp70-Hop-Hsp90 ternary complex is not only physically associated with proteasomal components, but also functionally required for proteasomal activity.
Previous studies in different models had indicated that reduction or deletion of Hsp90 could negatively influence the proteasomal activity (Imai et al., 2003; Yamano et al., 2008; Nanduri et al., 2015; Choutka et al., 2017). When we deleted the genes encoding Hsp90α (HSP90AA1) or Hsp90β (HSP90AB1) in HEK293T cells, we found a reduced level of steady-state proteasomal activity (Figure S3J). Moreover, the pharmacological inhibition of Hsp70 (with JG-98) or Hsp90 (with GA) in WT cells similarly led to a reduction of proteasomal activity (Figure 3I), whereas proteasomal activity remained unaffected when GA is added to the extracts after cell lysis (Figure S3K). Interestingly, a combination of suboptimal concentrations of JG-98 and GA did not show any additive or synergistic effects in the reduction of proteasomal activity (Figure 3I, right panel). These results collectively indicate that functional Hsp70 and Hsp90, including the possibility to form a ternary Hsp70-Hop-Hsp90 complex, are essential for optimal proteasomal function and activity. Furthermore, we found that the requirement for the Hsp90-Hop-Hsp70 ternary complex for full proteasomal activity is conserved in yeast; analogously to mammalian cells, the steady-state proteasomal activity, but much less so the rate of proteasomal activity, is reduced in a Δsti1 strain (Figure 3J, Figure S3L).
Stability of Individual Proteasomal Components Is Independent of the Hsp70-Hop-Hsp90 Ternary Complex
We considered the following two possibilities to explain the impact of the STIP1(Hop) KO on the proteasome: (1) Proteasomal components are clients of the ternary complex, and in the absence of Hop, they become unstable and subsequently degraded; (2) the ternary complex is required for 26S/30S proteasome assembly and/or maintenance. To address the first possibility, we reanalyzed previously published datasets of whole-cell proteomic experiments, where HeLa and Jurkat cells had been treated with Hsp90 inhibitors (Sharma et al., 2012; Fierro-Monti et al., 2013). As expected, the levels of some well-known intracellular clients of Hsp90 are decreased and Hsp90-associated molecular chaperone and co-chaperone proteins are increased by Hsp90 inhibition (Figure 4A). In contrast to what we observed with clients, Hsp90 inhibitors do not reduce the protein levels of any of the known core proteasomal proteins (Figure 4A, right panel). We also experimentally checked the impact of GA on the levels of a few RP components in WT cells by immunoblotting with specific antibodies and found that none of them are affected (Figure 4B).
Similarly, we did not find any striking and cell line-independent differences in protein levels of proteasomal proteins between WT and KO cells in the datasets of our own whole-cell proteomic analysis (Figure 4C). Thus, it seems unlikely that proteasomal components are dependent on Hsp90 for accumulation and/or stability.
Overall Composition of the Assembled Proteasome Is Similar in the KO Cells
To evaluate the impact of the absence of the ternary complex on proteasome assembly/maintenance, we purified 26S/30S proteasome particles from both WT and KO cells. 26S/30S proteasome particles were pulled out by an affinity purification strategy targeting the RP protein S5a (human gene name PSMD4); note that this purification scheme enriches for single-(26S) and double-capped (30S) CPs and that any subsequent analysis with this material could not report on free CPs or unassembled proteasome proteins. The integrity of the purified 26S/30S proteasomal complex was analyzed by native gel electrophoresis (Figure S4A-B), and by functional assays without and with MG132 (Figure S4C-D). To characterize the composition of the proteasome purified from WT and KO cells, we performed a comparative label-free LC/MS-MS analysis. We did not find any consistent cell line-independent differences of any identified stoichiometric components of the proteasome between WT and KO cells (Figure S4E-F). Although we noticed that three substoichiometric components and chaperones of the proteasome were reduced in preparations of proteasome particles from HEK293T KO cells (Figure S4F), we did not further consider them since they were unchanged in the HCT116 background (Figure S4E). The aforementioned Hop IP-MS analysis showed that many proteasomal proteins are associated with WT Hop (see Figure 3D). Interestingly, in our own proteomic analyses of purified proteasome from WT HEK293T and HCT116 cells, all components of the Hsp70-Hop-Hsp90 ternary molecular chaperone complex could be identified, albeit at very low substoichiometric levels. This again demonstrates that the Hsp70-Hop-Hsp90 ternary complex is specifically associated with the proteasome, while the low stoichiometry suggests that the association may be regulatory and transient. We concluded from these experiments that overall the composition of fully assembled proteasome particles is similar in the absence of Hop. However, we cannot absolutely rule out the possibility that a minor fraction of purified free RP influenced the overall proteasomal composition.
Optimal Proteasomal Assembly Requires the Hsp70-Hop-Hsp90 Ternary Complex
To check the impact of the absence of Hop and of the ternary molecular chaperone complex on the structure of the proteasome, we studied the structural integrity of the purified proteasome particles by negative staining transmission electron microscopy (TEM). 2D class averaging of all visible TEM structures led to four different proteasomal structural projections. We could see side views of single-capped 26S and double-capped 30S proteasome particles (Figure 4D, Figure S4G). Moreover, we could observe the two expected top views: an open form of the proteasome with a central hole and a closed form corresponding to CP and RP face up, respectively (Figure 4D). Based on the purification scheme and the relative levels of different forms in the native-PAGE analysis, we did not expect any structures related to free CP (see Figure S4A-B). We measured the dimensions of all four proteasomal projections and confirmed that regardless of the presence of Hop, they are similar to the known values (Walz et al., 1998; Unno et al., 2002) (data not shown). The main difference is the higher abundance of the double-capped proteasome (30S) in proteasome preparations from WT cells while the single-capped proteasome (26S) is more prevalent in preparations from KO cells (Figure 4E, Figure S4H); the opposite situation applies for the relative abundance of open versus capped proteasome particles (Figure 4E). Considering these structural analyses, we hypothesized that Hop, possibly as part of an Hsp70-Hop-Hsp90 ternary complex, is important either for the process of capping of proteasome particles and/or for maintaining the stability of the assembled 26S/30S forms of the proteasome. Collectively we concluded that the individual proteasomal components are not clients of the ternary Hsp70-Hop-Hsp90 chaperone complex, but rather that the ternary complex is involved in the assembly and/or maintenance of the proteasome (Figure 4F).
We next evaluated the ensemble of proteasome particles in whole cell extracts of both WT and KO cells by native-PAGE. We detected RP, CP and 26S/30S proteasome particles using specific antibodies against the CP and RP components Psma3 and Psmc5, respectively, whose total levels are not significantly altered between WT and KO cells (see Figure 4C, Figure S4I). We found that the abundance of the 26S/30S proteasome complexes is reduced in KO cells, whereas free RP and CP are not varying strikingly (Figure 4G-H). We further illustrated this finding by running the native-PAGE with 1.5-fold more total cell extract from KO cells next to an extract from WT cells (Figure S4J). As expected, only WT Hop and not the TPR double mutant can rescue proteasome assembly (Figure S4K-L). Moreover, the reduced proteasome activity in Δsti1 yeast is mirrored by the reduced assembly and/or stability of proteasome particles (Figure 4I, Figure S4M). Thus, we concluded that the Hsp70-Hop-Hsp90 ternary complex is essential for efficient proteasomal capping and/or for optimal proteasomal stability and maintenance (Figure 4J).
KO Cells Are Less Dependent on the Proteasome Even With Proteotoxic Stress
Considering the proteasome defect in KO cells, one would expect the degradation flux through the proteasome to be impaired as well. We therefore checked the relative abundance of all identified proteins in the MS datasets of the purified proteasome preparation. This revealed a differential abundance of “proteasomal interactors” between WT and KO samples (Figure S5A). This observation suggested that the utilization and requirement of the proteasome might be different between WT and KO cells. We then filtered the MS data for all proteasomal substrates known to be degraded following either mono- or poly-ubiquitination (Braten et al., 2016). Among all identified mono-ubiquitinated substrates, about 60% are relatively more abundant in proteasome preparations derived from WT cells (Figure S5B). This result indicates that the proteasome is more utilized in WT cells, and thus, that the proteasomal degradation flux might be reduced in KO cells compared to WT cells. However, known poly-ubiquitinated proteasomal substrates are equally abundant in proteasomal preparations of both WT and KO cells (Figure S5C).
So far, we had compared the Hop requirements for the proteasome under normal conditions. We now turned to investigate the effects of proteotoxic stresses. When cells are treated with the Hsp90 inhibitor GA, more proteins form aggregates (Figure S5D), and during a HS, more insoluble and ubiquitinated material accumulates in KO cells (Figure S5E); this demonstrates that the reduction of proteasomal activity and cellular dependency on the proteasome of KO cells is not a symptom of compromised ubiquitination. Upon inhibiting the proteasome in the recovery phase after a 6-hour treatment with GA, we observed that WT cells are dying significantly more than KO cells (Figure 4K, Figure S5F-G). Similarly, inhibiting the proteasome after HS kills WT cells strikingly more than KO cells (Figure 4L, Figure S5H).
To evaluate the proteasomal degradation flux more directly, we blocked the proteasomal activity with MG132 for 24 hrs and looked at the accumulation of ubiquitinated proteins. Significantly less ubiquitinated substrates accumulated in KO cells (Figure 4M), confirming a reduced degradation flux of the proteasome in KO cells since overall ubiquitination is not compromised in the KO cells (see above and Figure S5E). We conclude that WT cells not only have a higher abundance of assembled proteasome particles, but that they are also more dependent on proteasomal function, and even more so in stressed conditions.
Hsp70 and Hsp90 Functionally Collaborate Without Hop In Vivo
Even though the KO cells display a compromised proteasomal assembly and function, overall they nevertheless seem to maintain proteostasis (see Figure 1H, Figure S1G). To provide an additional general assessment of this conjecture, we biochemically fractionated soluble and insoluble proteins from WT and KO cells. Irrespective of genotype, we found similar amounts of ubiquitinated proteins and protein aggregates (Figure 5A; see also untreated samples in Figure S5D-E). Therefore, it seemed conceivable that more efficient chaperoning functions are compensating for the proteasomal defects of KO cells. We started to explore this with in vivo luciferase refolding assays; we found a higher rate of refolding of heat-inactivated luciferase in KO cells, consistent with a better chaperoning activity in these cells (Figure 5B, Figure S6A). This higher rate of luciferase refolding in KO cells could be completely reverted to that of WT cells by pharmacological inhibition of Hsp90 or Hsp70 (Figure 5C, Figure S6B). Thus, Hop appears to restrain the refolding activity of Hsp70/Hsp90. Reconstitution experiments with overexpression of WT and TPR mutants of Hop in KO cells showed that only WT Hop inhibits luciferase refolding whereas all TPR mutants do not affect it (Figure 5D).
From the literature, Δsti1 yeast cells are known to be HS-sensitive (Nicolet and Craig, 1989), unlike what we have found for human KO cells (see Figure 1G). In order to check whether the Hop/Sti1-independent chaperoning mechanism is evolutionarily conserved in yeast, we performed an in vivo luciferase refolding experiment with the HS-sensitive Δsti1 yeast strain in the W303 background (Figure S6C). Despite its heat sensitivity, this strain displayed a faster rate of luciferase refolding during the early recovery phase (Figure S6D). To mirror the experiments with our HS-resistant human KO cells, we also performed the experiment with a Δsti1 mutant yeast strain in the BY4741 background, in which the mutant is as heat-resistant as the WT (Figure S5E). In this genetic background, the enhanced luciferase refolding in the absence of Hop/Sti1 is even more striking (Figure 5E) and there is a much higher residual luciferase activity after a milder HS (Figure 5F). Hence, in the absence of Hop/Sti1, Hsp70 and Hsp90 are functional and responsible for enhanced folding of a model substrate.
Another aspect of chaperoning is keeping aggregation-prone misfolded proteins in a soluble state. Accordingly, we checked the solubility of the aggregation-prone polyglutamine model protein Q74-EGFP (Narain et al., 1999) in WT and KO cells along with a non-aggregating control (Q23-EGFP). We observed a reduction of the aggregated form of Q74-EGFP in KO cells in the HEK293T background (Figure 5G, Figure S6F). This improved solubility seems to be Hsp90-dependent as it could be substantially suppressed by GA (Figure 5H). In HCT116 cells, we did not see any difference in insolubility between the two genotypes (Figure S6G), possibly indicating that cell lines may differ with respect to their intrinsic anti-aggregation activities.
Considering that cell viability and proliferation are relatively unperturbed in the absence of Hop and if indeed Hsp70/Hsp90 function even better for some activities in the absence of Hop, we expected only a minimal impact on Hsp90 clients and co-chaperones. To test this hypothesis, we filtered our whole cell MS dataset for proteins of the Hsp90 interactome (https://www.picard.ch/downloads/Hsp90interactors.pdf) (Echeverria et al., 2011). We observed that the levels of the vast majority of the identifiable Hsp90 interactors are unaffected (Figure 5I-J, Figure S6H). Interestingly, the heat inducible isoform of Hsp70 (encoded by HSPA1) and Hsp110 (encoded by HSPH1), a nucleotide exchange factor (NEF) of Hsp70, are moderately upregulated in the absence of Hop (Figure 5I, Figure S6H), whereas the expression of other co-chaperones of the Hsp70-Hsp90 chaperone systems, including several J-proteins, remains unaltered (Figure 5I, Figure S6H).
To exclude the formal possibility that Hsp90 clients accumulate to normal levels, but do so in an inactive conformation, we checked the activities of different classes of Hsp90 clients. As there are many tyrosine and serine/threonine kinases amongst the clients of Hsp90, we compared total levels of serine- and tyrosine-phosphorylated proteins between WT and KO cells; we could not see any significant global differences between these two genotypes (Figure 5K). We also did not observe any strong differences between WT and KO cells in the phosphorylation of specific sites indicative of activation of c-Src, an Hsp90 client (Bijlmakers and Marsh, 2000), and of its downstream MAP kinases Erk1/2 (Figure S6I). Using luciferase reporter assays, we checked the transcriptional activities of several transcription factors, which are either Hsp90 clients by themselves or the downstream factor of Hsp90 clients. We found that either the activity is enhanced (in HEK293T cells) or not strikingly compromised (in HCT116 cells) in the absence of Hop (Figure 5L). We conclude that Hsp90 clients not only maintain their steady-state protein levels in KO cells, but that by and large they also maintain their activity compared to WT cells.
Hop Determines a Unique Spectrum of Hsp90 Client Proteins
Since the vast majority of Hsp90 clients were unaltered in KO cells, we wondered whether challenging their proteostasis by overexpression of Hsp90 clients could reveal Hsp90 vulnerabilities. We overexpressed several steroid receptors, which need Hsp90 for proper folding, stability, and transcriptional activity (Picard et al., 1990; Bohen and Yamamoto, 1993; Nathan and Lindquist, 1995). We observed that the accumulation and transcriptional activity of the glucocorticoid receptor is compromised in KO cells compared to WT cells (Figure 5M, Figure S6J-K). Other steroid receptors, the estrogen receptor α and the progesterone receptor, are only moderately affected in KO cells both in terms of accumulation and transcriptional activity (Figure 5M, Figure S6J, L). The overexpression of v-Src, a viral tyrosine kinase client of Hsp90 (Whitesell et al., 1994), yielded lower protein levels and also strongly reduced kinase activity in KO cells (Figure S6M-N). We overexpressed several other Hsp90 clients (Hif-1α, Hif-2α, and androgen receptor (AR)) and observed that they are not at all compromised in KO cells (Figure S6O). We wondered whether the attenuation of steroid receptor accumulation and activity in KO cells is limited to the case of their exogenous overexpression. At least for GR, this is not the case as we found that both expression and transcriptional activity of endogenous GR are remarkably reduced in A549 KO cells (Figure 5N). Overall, these results suggested that affected steroid receptors, kinases such as v-Src, and the proteasome are rather the exceptions amongst Hsp90 clients with regards to their pronounced Hop-dependence; anecdotally, GR and v-Src are also the very same Hsp90 clients that were the first ones to be discovered and that have been studied for a long time (Picard et al., 1990; Whitesell et al., 1994; Whitesell and Cook, 1996). Moreover, in agreement with a previous publication showing the plasticity of the Hsp90 co-chaperone network for folding of exogenous Hsp90 clients in yeast (Sahasrabudhe et al., 2017), we hypothesized that Hop might determine unique criteria of client selectivity for Hsp90 even in higher eukaryotes.
An Alternative Prokaryote-like Hsp70-Hsp90 Binary Complex Maintains the Hsp90 Interactome
KO cells maintain proteostasis even though the proteasome function is inefficient; the latter may be counterbalanced by more efficient Hsp70/Hsp90-dependent chaperoning (Figure 6A). Since bacterial Hsp90 and Hsp70 orthologs HtpG and DnaK, respectively, can physically and functionally interact without a Hop-like protein (Genest et al., 2015; Kravats et al., 2017), we investigated the possibility that an alternative Hop-independent Hsp70-Hsp90 molecular chaperone complex might form in human cells. We performed an Hsp90-IP-MS analysis and found Hsp70/Hsc70, encoded by HSPA1 and HSPA8, among the top hits even in KO cells, despite a 4-10-fold reduction (Figure 6B, Figure S7A). We confirmed this Hop-independent interaction of Hsp70 and Hsp90 by a targeted co-IP experiment (Figure 6C, Figure S7B).
In search of a mechanism promoting a Hop-independent interaction of Hsp70 and Hsp90, we checked whether Hop activity might be functionally redundant in human cells. We revisited our own Hsp90-IP-MS datasets and extracted the data for proteins with a Hop-like architecture, that is with multiple TPR domains. Note that there are only very few of these proteins, which are enriched with Hsp90 in the absence of Hop, and they are very moderately so (Figure S7C). Most importantly, all of these "enriched" TPR proteins are present at very low levels compared to the abundance of Hsp90 and Hsc70/Hsp70 in total cellular extracts from either WT or KO cells, and to Hop in WT cells (Figure S7D). Thus, it is very unlikely that any of these proteins could substitute for Hop for the formation of another Hsp70-(multiple TPR protein)-Hsp90 ternary complex.
These results led us to speculate that human Hsp70 and Hsp90 could directly interact as they do in bacteria (Genest et al., 2015; Kravats et al., 2017) and in Δsti1 yeast (Kravats et al., 2018). We tested this notion using an in vitro association assay with recombinant proteins. To minimize client-type interactions of Hsp70, we used the substrate-binding mutant V435F of Hsp70. The co-IP experiment revealed a direct interaction between Hsp90 and Hsp70 both in the presence and absence of ATP (Figure 6D). We then checked the involvement of evolutionarily conserved surface residues that had been shown to be important for the direct interaction of Hsp70 and Hsp90 orthologs in bacteria (Genest et al., 2015) (Figure 6E). As in bacteria, these residues are well surface-accessible in both human Hsp90 isoforms (Figure 6F, Figure S7E), and modifying these residues in human Hsp90α strongly suppresses the interaction with Hsc70/Hsp70 in KO cells (Figure 6G, Figure S7F).
Our whole-cell MS analysis had already indicated that the binary Hsp90-Hsp70 complex may be sufficient to support the Hsp90 interactome (see Figure 5I, Figure S6H). To address this issue more directly, we compared the levels of all proteins in the Hsp90-IP-MS datasets between WT and KO cells. The majority of the interactions do not change and there are similar proportions of proteins that are enriched or depleted in the absence of Hop (Figures 6B and 7A, Figure S7A and S7G). Thus, client-Hsp90 and co-chaperone-Hsp90 interactions are largely maintained in KO cells, which supports the conclusion that the prokaryote-like Hsp70-Hsp90 binary complex in human cells is functional.
The Prokaryote-like Human Hsp70-Hsp90 Complex Is Also More Efficient In Vitro
Bacterial Hsp70 collaborates with the bacterial J-domain protein Hsp40 and a NEF for the initial substrate recognition, and further substrate folding can be obtained by the collaboration of bacterial Hsp70 with Hsp90 (Mayer and Le Breton, 2015; Moran Luengo et al., 2018). To evaluate the chaperoning activity of the binary human Hsp70-Hsp90 complex, we performed in vitro luciferase refolding experiments with purified components. Human Hsp70 was complemented with a J protein, the NEF Apg2, and human Hsp90α. To determine the impact of Hop on this molecular chaperone system, we measured luciferase refolding as a function of increasing concentrations of Hop. To our surprise, but in agreement with our in vivo experiments (see Figure 5B, Figure S6A), we discovered that the Hsp70-Hsp90 molecular chaperone systems refold heat-denatured luciferase most efficiently in the absence of Hop (Figure 7B). Increasing concentrations of Hop gradually decrease the final yield of luciferase refolding achieved by the Hsp70-Hsp90 system (Figure 7B-C). In conclusion, human/eukaryotic Hsp70 and Hsp90 can form an evolutionarily conserved chaperone complex that is fully functional for protein folding and maintains the proteostatic equilibrium despite proteasomal assembly defects in the absence of Hop/Sti1.
DISCUSSION
Depending on cellular states and stresses, the proteostatic equilibrium can be shifted to alternate mechanisms. For example, HS or inhibition of Hsp90 strongly induce the Hsp70/Hsp40 chaperone system and small Hsps to meet the new cellular requirements (Sittler et al., 2001; Richter et al., 2010; Neckers and Workman, 2012). Here we establish cellular models with human cell lines and yeast, which, in the absence of Hop/Sti1, adopt the more ancient and more efficient mechanism of chaperoning of bacteria and thereby compensate for proteasomal defects, reestablishing an alternate proteostatic equilibrium. Thus, depending on cell-specific requirements or external inputs, eukaryotic cells may still be able to shift to a more prokaryote-like mode of operation for the Hsp70/Hsp90 systems. We speculate that the "invention" of Hop during the evolution from prokaryotes to eukaryotes may have promoted a shift from a proteostatic system centered on refolding to a more extensive use of proteasomal degradation.
Impaired proteasomal function could potentially be compensated by increased autophagy. Proteotoxic stress induced by inhibition of the proteasome can activate autophagy by phosphorylating the autophagy receptor p62 (Lim et al., 2015). The genetic suppression of Hsp90 function in Drosophila melanogaster shifts the protein degradation balance from proteasome-mediated degradation to autophagy (Choutka et al., 2017). We cannot formally rule out that increased autophagy might contribute to the new proteostatic equilibrium in Hop KO cells, but we have not seen any upregulation of lysosomal or autophagy-related proteins in our whole cell proteomic data.
Hsp90 itself had already been linked to proteasomal integrity and activity (Imai et al., 2003; Lander et al., 2012). Our data complement this early evidence by demonstrating that it is the Hsp70-Hop-Hsp90 ternary complex that is required for optimal 26S/30S proteasomal assembly and/or for maintaining the proper abundance. According to current models, the RP ATPase-ring docks to the CP α-ring to form the functionally active proteasomal holoenzyme (Murata et al., 2009; Gallastegui and Groll, 2010). We speculate that the ternary molecular chaperone complex primarily binds to the RP, facilitates the docking with the CP and stabilizes the assembled 26S/30S proteasome. The substoichiometric proteasome component Ecm29, which itself is not dependent on Hop (see Figure S4E-F), has been shown to be a tethering factor for RP and CP (Leggett et al., 2002). We propose that the ternary molecular chaperone complex plays a similar role; further studies are required to define more clearly how the ternary complex promotes RP-CP docking and what the interplay with Ecm29 might be. Our MS analysis showed that all components of the Hsp70-Hop-Hsp90 ternary complex are associated with 26S/30S proteasome particles at substoichiometric levels; hence, it is conceivable that the ternary complex acts like other bona fide proteasome-dedicated chaperones, which dissociate from the functionally mature proteasomal holoenzyme (Roelofs et al., 2009). Although proteasomal activity and assembly is known to be regulated by posttranslational modifications of its components (Rousseau and Bertolotti, 2018), it is not likely that this can help to explain the proteasomal defects of KO cells; the vast majority of proteins in general and several of the known posttranslational regulators of proteasomal components such as MAPKs, mTORC1-related proteins, and the catalytic subunit of protein kinase A, in particular, are unaltered in KO cells. Thus, we conclude that Hop, as part of the Hsp70-Hop-Hsp90 complex, promotes the assembly of RP with CP and/or contributes to maintaining their assembled state.
Our most surprising finding is that cells lacking Hop/Sti1 compensate the proteasomal defect by improved protein folding. In view of the well-established role of Hop for substrate transfer between the Hsp70 and Hsp90 molecular chaperone machines, and as allosteric regulator of the ATPase activity of Hsp90 (Richter et al., 2003; Alvira et al., 2014; Kirschke et al., 2014; Rohl et al., 2015), this is the last thing one would have expected. Under normal conditions, one can imagine that Hop does promote substrate transfer from Hsp70 to Hsp90, but then ends up stalling the Hsp90 complex, at least temporarily, thereby slowing down further substrate folding. This may be beneficial or even essential to allow the folding or assembly of some substrates such as GR, v-Src, the proteasome, and possibly of some particularly labile clients such as the ΔF508 CFTR mutant (Bagdany et al., 2017). For these, Hop must be able to form the ternary complex. TPR mutants of Hop, which neither bind to Hsp70 nor to Hsp90 nor both are unable to rescue any phenotype in KO cells. This is further supported by the finding that the fortuitously incomplete KO of the HEK293T cell clone KO1 has the exact same phenotype as other clones with a complete KO. The residual ~36 kDa fragment of KO1 only contains the Hsp90-binding domain TPR2A and does not suffice to rescue the KO defects. In contrast, for most Hsp90 clients, the temporary stalling or slow-down of the Hsp90 molecular chaperone machine by Hop may be counterproductive. What defines the Hop requirement of Hsp90 clients remains to be deciphered, but it is clear from our data that the vast majority of Hsp90 clients can be efficiently processed by the prokaryote-like Hsp70-Hsp90 binary complex. We assume that this core molecular chaperone complex still requires the support of other co-chaperones like Hsp40, Cdc37, Aha1 and p23 to select, to process and to release Hsp90 clients (Li et al., 2012; Dunn et al., 2015; Keramisanou et al., 2016; Schopf et al., 2017; Bachman et al., 2018). Compromising the function of these co-chaperones could be particularly detrimental for Hop KO cells running on a somewhat hyperactive Hsp90 machine; indeed, the deletion of SBA1 (encoding the yeast p23 ortholog) in a Δsti1 yeast strain severely reduces cellular fitness (Fang et al., 1998).
Our data demonstrate that the Hsp70-Hsp90 binary system not only works, but somehow manages to be more efficient. Our quantitative Hsp90-IP-MS experiments show that the steady-state association of Hsp70 and Hsp90 is less prominent. This is compatible with previously reported Kd values for the interactions that are relevant in this context; compared to the Kd for the direct interaction of yeast Hsp70 and Hsp90 of 14 μM (Kravats et al., 2018), the values for the yeast or human pairs Hop-Hsp70 and Hop-Hsp90 are considerably lower (Mayr et al., 2000; Brinker et al., 2002; Wegele et al., 2003). We propose that the binary Hsp70-Hsp90 complex, albeit less stable, is more dynamic than the ternary molecular chaperone complex. Moreover, similarly to what has been reported for the bacterial DnaK-HtpG system (Nakamoto et al., 2014; Genest et al., 2015), the stimulation of the ATPase of one partner by the other might contribute to the improved chaperoning activity in the human system.
Whether Hsp70 and Hsp90 can form the binary molecular chaperone complex in the presence of Hop in WT cells and whether this alternative molecular chaperone complex has any specialized functions in normal or stressed conditions remain open questions. Our data support this hypothesis since we observe a colocalization of Hsp70, Hsp90, Hsp40, and Hsp110, but not Hop, in HS-induced protein aggregates (Figure S7H). However, the specific detection of the Hsp70-Hsp90 binary complex and its characterization in the presence of Hop is technically extremely challenging. Further methodological developments will be necessary to explore the functions of this prokaryote-like complex across different cell types and cellular conditions.
The highly proliferative state of embryonic stem cells is associated with rapid protein turnover and high proteasomal activity (Vilchez et al., 2012). Deletion of some proteasomal subunits has been demonstrated to cause lethality in the mouse, flies and in plants (Sakao et al., 2000; Szlanka et al., 2003; Brukhin et al., 2005; Hamazaki et al., 2007; Beraldo et al., 2013). It is therefore conceivable that the embryonic lethality of the Hop KO in the mouse (Beraldo et al., 2013) could be explained by the failure to meet this particular requirement in embryonic stem cells or in other rapidly proliferating cells during early development. Moreover, the absence of Hop might also compromise the folding/assembly of one or a few specific Hsp90 clients other than the proteasome, which are essential during early embryonic development. What the mouse model clearly demonstrates is that every eukaryotic cell may not be able to shift its proteostatic equilibrium to the overall more efficient chaperoning by the prokaryote-like Hsp70-Hsp90 binary complex in the absence of Hop. A systematic effort will be needed to determine whether Hop KO mouse embryos die because of a proteostasic collapse or because of a more subtle defect relating to some Hsp90 clients. It should be emphasized that the presence of a single STIP1 allele is sufficient for embryonic development, even though the reduced Hop levels of heterozygote mice does lead to increased cellular stress and vulnerability to ischemia (Beraldo et al., 2013). In this study, we established that the absence of Hop does not affect proteostasis in cellular models. However, cellular or organismic fitness could nevertheless be negatively affected in the long run. Indeed, even though Hop KO worms are viable, their lifespan is reduced (Song et al., 2009). Furthermore, Hop levels drop in the aging brain, and aging is directly correlated with reduced proteasomal and chaperoning activity (Brehme et al., 2014; Rousseau and Bertolotti, 2018). Since the absence of Hop generates a mixed outcome in different experimental models, tissue- and cell type-specific functions of Hop for the maintenance of proteostasis should be further studied.
Our discoveries may also have translational potential. If Hop could be specifically inhibited, this might promote superior chaperoning by the prokaryote-like Hsp70-Hsp90 binary complex. This could be a useful strategy to reequilibrate the proteostatic balance in diseases or altered physiological states with proteasomal dysfunctions. Neurodegenerative disorders such as Huntington’s and Parkinson’s diseases, which are associated with inefficient chaperoning (Labbadia and Morimoto, 2015; Brehme et al., 2014), could potentially benefit from improved chaperoning induced by Hop inhibitors. Some compounds have been reported to inhibit specifically the interaction between Hsp90 and the TPR2A domain of Hop (Yi and Regan, 2008; Pimienta et al., 2011) and may be promising leads for such a therapeutic strategy.
AUTHOR CONTRIBUTIONS
K.B. conceived the study, designed and performed experiments, analyzed the data, prepared figures, and wrote the manuscript. L.W. and M.Q. conducted the proteomic analyses. T.M.L and S.G.D.R. designed and performed in vitro luciferase refolding assays. P.C.E. performed bioinformatics analyses of proteomic datasets. M.V. generated α- and Hsp90β-KO cells. L.B. contributed to experiments with recombinant proteins. D.W. helped with all yeast experiments. C.B. contributed to TEM experiments and analyses. D.P. conceived the study, contributed to designing the experiments and analyzing the data, supervised the work, and wrote and edited the manuscript. All authors provided critical analysis of the data and contributed to the editing of the manuscript.
DECLARATION OF INTERESTS
The authors declare no competing interests.
ACKNOWLEDGMENTS
We are grateful to Alfred L. Goldberg, Matthias P. Mayer, Ueli Schibler, Marcello Maggiolini, Donald McDonnell, Yoshihiko Miyata and Adrienne Edkins for gifts of plasmids. We thank David O. Toft for gifts of antibodies, Jason Gestwicki for the gift of JG-98, and Stacey Mattison for critically reading the methods section of this manuscript. We are also indebted to various previous members of the Picard laboratory for miscellaneous reagents. This work has been supported by the Swiss National Science Foundation and the Canton de Genève.