Defining the TRiC/CCT interactome links chaperonin function to stabilization of newly-made proteins with complex topologies

Principles of TRiC substrate selection

TRiC is part of a chaperone network linked to protein synthesis ¹³ and has been shown to facilitate folding of newly translated proteins in vivo ^9,14. Previous studies established that TRiC interacts transiently with a subset of cellular proteins during biogenesis ¹⁴. We thus examined the flux of newly translated proteins through TRiC in mammalian cells using a previously established pulse-chase analyses, whereby newly-made proteins are specifically labeled by a brief pulse and folding and maturation occurs during the chase ¹⁴. Newly synthesized polypeptides interacting with TRiC were isolated by immunoprecipitation with antibodies against TRiC subunits β and ε (Supplementary Fig. 1; Fig. 1A). Two-dimensional PAGE analysis showed that soon after translation a large number of newly made proteins associated with TRiC. Following a period of chase, these proteins were dissociated, as expected for chaperone substrates which should be released upon completion of folding (Fig. 1A, left and center panel). Chaperonin complex assembly was also observed during the time course of the chase, whereby the β and ε subunits associated with the remaining TRiC subunits (Fig. 1A, center, see also ref.14). Mass spectrometry analysis of TRiC interacting proteins only identified highly abundant substrates, namely the WD-repeat containing translation initiation factor 3β and GAPDH, in addition to the known TRiC substrates actin and tubulin (Supplementary Table 1). Thus, identification of low abundance cellular substrates of TRiC required alternate genome-wide approaches.

To better define the principles that govern TRiC substrate specificity, we next examined what determines association of cellular proteins with TRiC. In principle, chaperonin-substrate interactions may be solely determined by the presence of specific TRiC-binding motifs in the substrates, such as sequence elements, that distinguish them from the rest of the proteome. A prediction of this model is that the in vitro substrate specificity of TRiC towards denatured cytosolic proteins will mirror that observed in vivo. Accordingly, we compared the subset of cellular proteins that bind TRiC upon translation in vivo (Fig. 1B, left) with those proteins binding TRiC when the same ³⁵S-labeled extract is denatured and presented to the chaperonin in vitro (Fig. 1B, center). The sets of eukaryotic proteins interacting with TRiC in vivo and in vitro were strikingly different. While actin and tubulin were prominently bound in either condition, examination of the merged gels revealed less than 10% overlap between the proteins selected by TRiC in vitro and in vivo (Fig. 1B, right). We conclude that TRiC does not select its substrates based solely on the presence of specific sequence motifs, as was proposed from in vitro studies ^15,16. Instead, TRiC-substrate selection in the cell is strongly dependent on the context of translation, where both cotranslational folding events and cooperating chaperone systems may affect the conformation of de novo folding intermediates ^11,17,18.

TRiC was proposed to be a highly specific chaperone, in contrast to the broader specificity prokaryotic chaperonin GroEL ⁷. To further define the substrate-recognition principles of TRiC, we next determined whether TRiC recognizes a more restricted range of proteins when compared to GroEL ¹⁹. To compare the cellular proteins recognized directly by TRiC and GroEL we presented denatured ³⁵S-labeled cytosolic proteins to purified GroEL or TRiC, followed by immunoprecipitation of chaperonin substrate complexes (Supplementary Fig. 2 and Fig. 1C). A large fraction of cellular proteins, including actin and tubulin, was recognized by both chaperonins. Comparison of the GroEL and TRiC-bound protein spectra revealed that the recognition specificities of these chaperonins were strikingly similar, with over 80% overlap between the protein sets recognized by either chaperonin (Fig. 1C, right). Given the well-established affinity of GroEL for hydrophobic substrate determinants, this result indicates that hydrophobicity is a strong component of TRiC substrate recognition, consistent with previous findings ^8,9,20,21. Importantly, the similar binding specificity of both chaperonins suggests that TRiC sustains the capacity to recognize a wide breadth of proteins and is not intrinsically a highly specific chaperonin. We conclude that TRiC possesses broad recognition specificities, yet in the cell it only interacts with a defined set of substrates.

Genomic screen for TRiC substrates during biosynthesis

Our finding that TRiC substrate selection in the cell is determined in the context of protein biosynthesis raises the question of what features distinguish proteins recognized by TRiC during biogenesis from the wide spectrum of potential interactors. Accordingly, we adapted a genome-wide approach that allowed identification of physiologically relevant TRiC substrates by detecting chaperone interactions in the context of translation.

We employed a screening approach that monitored which proteins interact with TRiC during translation of cDNA expression pools in a cell-free mammalian translation system ²². This approach, termed small pool expression cloning, or SPEC, allowed detection of TRiC-substrate interactions in a physiologically relevant context (Fig. 2A). Importantly, SPEC presents several unique advantages for the identification of physiological chaperone substrates. First, these translation lysates contain the full complement of chaperones and translation components required to fold most cytosolic proteins to an active state ²³. TRiC-substrate interactions are thus examined in the context of protein biosynthesis, whereby cotranslational folding can occur in the presence of physiological levels of upstream chaperones and folding cofactors. Second, proteins are translated at very low levels, in the picomolar range, and thus interactions with the chaperonin are observed without overexpression of the substrates ²⁴. Furthermore, the lack of bias toward detection of abundant proteins in the SPEC approach permits the identification of chaperone interactions for rare, low-abundance proteins, unlike large-scale immunoprecipitation reactions.

A high-throughput screen was performed using single mouse cDNAs arrayed in a 96-well plate format. Orthogonal cDNA pools were made along rows and columns for each plate and analyzed by in vitro translation in rabbit reticulocyte lysate in the presence of ³⁵S-methionine (Fig. 2A). The presence of TRiC substrates encoded by cDNAs in the pools was detected by co-immunoprecipitation with the chaperonin followed by SDS-PAGE and phosphorimaging analysis (Fig. 2B). cDNAs encoding candidate substrates expressed in two overlapping pools were isolated and their interaction with TRiC confirmed before their identification by DNA sequencing (Fig. 2C). To simplify the screen, abundant housekeeping mRNAs were subtracted from the cDNA library, thus increasing the proportion of lower abundance transcripts surveyed ²⁵. Of note, the identification of alpha-tubulin among the TRiC interacting substrates demonstrated that housekeeping proteins were still represented in the library, thus providing a good representation of transcripts in the cell (Supplementary Table 1). In all, 2600 clones were surveyed and more than one hundred TRiC-interacting proteins were isolated. The screen was not conducted to saturation but was concluded upon the repeated identification of several clones. From these data, 6–7% of all cytosolic proteins are anticipated to interact with TRiC. Importantly, this analysis is in agreement with previous estimates from pulse-chase analysis ¹⁴ and proteomics studies ²⁶. Because only high affinity interactions are detected by immunoprecipitation, the predicted number of TRiC-interacting proteins might be underestimated. Assuming that the eukaryotic cytosol contains approximately 4500 proteins, based on estimates from the yeast Saccharomyces cerevisiae, our analysis predicts that on the order of 300 cytosolic proteins interact with TRiC in vivo.

Distinct in vivo kinetics of substrate flux through TRiC

Our screen revealed that a wide range of proteins interact with TRiC upon translation, most of which are conserved across eukaryotes (Supplementary Table 1). Accordingly, we used S. cerevisiae to validate the TRiC-interactions in vivo taking advantage of the extensive annotation of the yeast genome. Initially, candidates carrying an N-terminal GST-tag ²⁷ were affinity isolated, and association with TRiC assessed by immunoblot analysis. Remarkably, all of the yeast homologues of SPEC-derived candidates surveyed bound to TRiC in vivo but neither GST alone nor GST-tagged to Ypt6, a control GTP-binding protein, associated with the chaperonin (Fig. 3A). This supports the idea that the screen identified bona fide TRiC substrates and that the interaction with TRiC is evolutionarily conserved across eukaryotic organisms.

One hallmark of chaperones that facilitate de novo folding is that they associate transiently with newly made proteins and dissociate during the course of polypeptide folding and maturation. As such, we examined the in vivo flux of SPEC-derived TRiC substrates through the chaperonin in yeast cells. Pulse-chase analysis followed by anti-TRiC immunoprecipitation indicated that, as in mammalian cells, a range of yeast proteins transiently associate with TRiC early in their biogenesis, and dissociate during the chase, consistent with release upon folding (Fig. 3B, 15 min). To observe the flux of single substrates through the chaperonin, TRiC immunoprecipitations were subjected to a second round of immunoprecipitation against a unique tag in the substrates (Fig. 3C, 3D). Transient association kinetics were observed for SPEC-identified substrates, indicating these proteins transit through the chaperonin during their biogenesis (Fig. 3D). Notably, two types of association kinetics were observed among TRiC substrates. One class of proteins, including actin and the small purine nucleoside phosphorylase (PNP1), bound early to TRiC and rapidly dissociated (Fig. 3D, 3E). Elongation factor 2 (EFT2) also bound rapidly to TRiC, but dissociated very slowly from the chaperonin (Fig. 3D, 3E). Another set of substrates exhibited much slower kinetics of association with TRiC. For instance, the mRNA export factor MEX67, the homotetrameric citrate synthase CIT2, the multidomain protein arginyl tRNA synthetase RRS1 (or ArgRS), as well as the WD40 G_β-protein STE4 did not bind to TRiC at the earliest chase times but instead their binding seemed to peak later in the time-course (Fig 3D, 3E). Notably, the TRiC association kinetics was independent of substrate size, since the larger EFT2 associated rapidly with TRiC while the smaller STE4 did not. It is tempting to speculate that the slower TRiC-binding kinetics arises from the action of cooperating upstream chaperones delaying binding to TRiC during translation ^13,18. Intriguingly, all the slower binding proteins share the common property of belonging to larger protein complexes, raising the possibility that these substrates accumulate on TRiC, before their release into oligomeric assemblies. In support of this idea, several WD40-containing F-box proteins translated in a cell-free system accumulated on TRiC in the absence but not in the presence of their oligomeric partner protein Skp1 (Supplementary Fig 3). We conclude that a genome-wide screen for the TRiC interactome identified bona fide TRiC substrates that flux through the chaperonin during biogenesis in vivo.

Physical and structural properties of the TRiC Interactome

We next sought to define what distinguishes the relatively small fraction of TRiC-interacting proteins from the cytosolic proteome at large. Having identified a large set of proteins that interact with TRiC in a physiologically relevant context (herein the TRiC Interactome, Suppl. Table 1) we applied a bioinformatics approach to identify common features among TRiC substrates that may underlie their physiological association with TRiC. Proteins from the Interactome spanned a breadth of cellular processes, most of which are conserved across eukaryotic organisms. Interestingly, 40% of the TRiC Interactome comprises essential genes, more than twice the proportion of essential genes in the yeast genome ²⁸. Of note, the frequency of TRiC substrates belonging to oligomeric protein complexes was also enriched in the TRiC Interactome to 90% compared to less than 50% of all yeast proteins in the MIPS complex database²⁹. These two observations may be linked since proteins that belong to oligomeric complexes are more highly connected in the protein network and, thus, disproportionately essential ³⁰. We hypothesize that the eukaryotic chaperonin assists in folding proteins belonging to oligomeric assemblies and may serve as a reservoir to stabilize them against aggregation or degradation prior to complex formation.

Previous biochemical experiments using actin and tubulin, led to the proposal that TRiC selects its substrates by specific recognition of a set of polar sequence elements ^15,16. Accordingly, we analyzed the TRiC Interactome to search for the presence of specific sequence motifs that may confer TRiC binding. Sequence analysis of the TRiC Interactome failed to reveal statistically significant consensus sequences, or an enrichment in previously proposed TRiC-binding sequence elements ^15,16. This is consistent with the finding that TRiC is not a highly sequence-specific chaperonin (Fig. 1C).

In principle, physical and/or structural properties could influence the folding characteristics and contribute to chaperonin binding. A comparison of the size distribution of TRiC-interacting proteins to the naturally occurring distribution of cytosolic proteins in the S. cerevisiae proteome (Fig. 4A) or the murine proteome (Supplementary Fig. 4) demonstrated a statistically significant enrichment for proteins ranging between 40–75 kDa in size, in good agreement with results obtained using pulse-chase approaches, ¹⁴ (Fig. 1 & 4A). Of note, the chaperonin chamber is predicted to accommodate polypeptides in this size range ³¹. Since a single folding domain is approximately 25–30 kDa ¹, this finding suggests that the TRiC-interactome consists mostly of multidomain proteins. Interestingly, a number of larger substrates were also identified in our screen, raising the possibility that some substrates are not completely encapsulated during folding. In support of this idea, the 100 kDa protein myosin is an known to be an obligate substrate of TRiC ³². Furthermore, the bacterial chaperonin GroEL similarly supports the folding of several large proteins ³³.

The proteins in the TRiC Interactome exhibited an enrichment in hydrophobic sequences (Fig. 4B). This finding could be explained by substrate size, as larger proteins retain correspondingly larger hydrophobic cores. Alternatively, hydrophobic inter-domain contacts within substrate proteins could contribute to the higher proportion of hydrophobic sequences. Inappropriate interactions among these hydrophobic surfaces, for instance in domain swapping, could complicate the folding of multidomain proteins ³⁴. Previous observations that TRiC can recognize hydrophobic determinants ^8,9,20,21 (see also Fig. 1C) suggests the chaperonin may prevent these inappropriate interactions.

Analysis of the structural propensities of the TRiC Interactome revealed the most striking shared property among TRiC substrates. While no specific protein fold was enriched in the TRiC interactome, we observed that proteins of high beta sheet propensity and/or low alpha helical content were highly enriched among TRiC substrates (Fig. 4C & 4D). Additional scanning window analysis demonstrated that a high proportion of TRiC-interacting proteins contain long continuous stretches of beta sheet propensity, with a particular bias toward stretches of 35–45 amino acids (Fig. 4E). Thus, analysis of secondary structure propensity identifies a clear commonality among proteins in the TRiC interactome. Importantly, the enrichment in beta-rich proteins suggests that TRiC substrates have complex topologies that are predicted to be slow-folding and aggregation-prone.

Cellular Principles of TRiC substrate selection

Our studies resolve a long-standing controversy over TRiC substrate recognition and specificity. The data presented do not support the idea that TRiC selects its substrates in the cell based solely on intrinsic determinants, such as specific sequence elements. Instead, TRiC recognition is strongly influenced by the context of protein biogenesis and may involve structural and physical features of the biosynthetic intermediates as well as cooperating chaperone systems (Fig. 5).

In vitro experiments comparing TRiC with the broad spectrum chaperonin GroEL indicate that TRiC is not a highly specific chaperone. Since these in vitro substrates differ substantially from the substrates binding to TRiC in vivo, this result implies that TRiC substrate interaction is dictated by the folding properties of the polypeptide in the context of translation. The substantial differences between the spectrum of cellular proteins that bind to TRiC during biosynthesis and those that TRiC selects when the same labeled proteins are presented in a denatured form highlights the importance of examining chaperone substrates in a physiologically relevant context. Furthermore, the similarities between the proteins bound by TRiC and GroEL in vitro suggest that the specificity of the eukaryotic chaperonin stems from the particular combination of folding kinetics and exposed recognition motifs in the co- or post-translationally-generated intermediates of its cellular substrates. The lack of sequence specificity in the TRiC substrate interaction is reinforced by the absence of sequence motifs in the in vivo subset that might confer TRiC binding propensity. Future studies should uncover the molecular determinants and the possible role of upstream cofactors in TRiC recognition.

Properties of TRiC substrates

We find that approximately 7% of the proteins we screened by genomic approaches associated with TRiC upon translation. This is consistent with our pulse-chase analysis, which indicates that approximately 5–10% of cytoplasmic proteins flux through the chaperonin. Analysis of the shared features of the Interactome identified some common properties in these substrates that may illuminate the principles of substrate selection by TRiC. In particular, the enrichment of beta-rich proteins among TRiC substrates provides a link between chaperonin interaction and one of the major challenges faced by the biosynthetic machinery that must facilitate cellular folding. Indeed, regions with high beta-strand propensity are inherently aggregation-prone and difficult to fold, such that evolutionary pressures have introduced elements of natural design to protect exposed beta-edges in native protein structures ³⁵. Importantly, these design elements would not be in place until completion of folding. Thus, folding intermediates with regions of high beta-sheet propensity are especially vulnerable to misfolding and aggregation. It is tempting to speculate that the chaperonin TRiC functions to stabilize these exposed beta edges against aggregation until loops or helices in the folded protein architecture emerge to protect the beta strand in the native state.

It is also possible that the enrichment in beta-sheet propensity emerges from a more overarching characteristic of TRiC substrates. The observation that beta-sheet propensity correlates with slow-folding kinetics provides another intriguing link between our analysis of the TRiC interactome and folding in the cell. Emerging studies find that complex protein topology, such as that found in beta-rich proteins, is a major contributor to slow folding rates in vitro ^36–38. Of note, the TRiC Interactome also contains all-helical proteins such as citrate synthase, raising the possibility that the enrichment in beta sheet content is diagnostic of a more general property recognized by TRiC, such as complex topology or slow-folding kinetics. These features can lead to accumulation of toxic aggregates which have been linked to a number of amyloid diseases ⁴. The idea that TRiC may play a role in preventing aggregation of amyloidogenic folding intermediates resonates with the recent identification of this chaperonin as a potent suppressor of huntingtin amyloid formation, a polyglutamine protein that is highly aggregation-prone ^39–41.

Contribution of TRiC to the Cellular Folding Network

Our data provides a clue on how the cell solves the problem of folding proteins with complex topologies, such as those with beta-rich domains. During synthesis, these proteins are in a precarious position as they need to bring together many discontinuous regions to reach the native state, and thus are susceptible to aggregation with neighboring proteins. Binding to the ring-shaped chaperonin TRiC would solve several problems posed by this process: first, the polyvalent binding of multiple substrate regions (e.g. hydrophobic beta-strands) to different subunits in a ring would sequester several aggregation-prone sequences during translation; secondly, the allosteric communication between subunits in one ring ⁴² would provide a mechanism for concerted release of all folding elements to the central chaperonin chamber; and finally, this chamber would provide a sequestered environment to protect slow-folding proteins until they reach the native state. Taken together, the combination of proteomic and bioinformatics analyses reveal TRiC substrates have complex topologies that are slow-folding and aggregation-prone, and provide a compelling rationale for the function of this oligomeric ring shaped chaperonin in assisting cellular folding.

The additional finding that TRiC substrates are enriched in proteins belonging to oligomeric assemblies suggests that TRiC also functions to facilitate complex assembly. We envision two possible mechanisms for this role. In principle, the processes of TRiC-mediated folding and assembly could be directly coupled. Alternatively, TRiC could fold monomeric subunits and maintain them in an assembly-competent state until association with oligomeric partners. Thus, TRiC may contribute to the regulation of complex cellular processes such as signaling or cell cycle regulation by maintaining a pool of assembly-competent but inactive protein molecules. Indeed, other chaperones such as Hsp90 have also been implicated in the regulation of signaling cascades by a similar mechanism ⁴³.

While our data highlights the contribution of TRiC to cellular folding, it will be interesting to examine how TRiC cooperates with other components of the cellular folding machinery. While the few TRiC substrates characterized to date, such as actin, tubulin, Cdc20 and Cdh1 ^9,44,45, exhibit an obligate requirement for the chaperonin, it is possible that other TRiC-interacting proteins can be folded by alternative chaperone pathways. Indeed, an in vitro analysis of GroEL interacting proteins revealed that only a small proportion have an absolute requirement for GroEL to fold, while most GroEL substrates can fold spontaneously or use alternate chaperones ⁴⁶. It is likely that TRiC substrates fall into similar categories. Future studies should determine whether some TRiC substrates can fold via alternate chaperone systems. However, even for the well-studied chaperonin GroEL, comparison of in vitro and in vivo analyses illustrates the complexity of chaperone pathways in the cell. Thus, several GroEL-interacting proteins that can fold spontaneously in vitro nonetheless aggregate in vivo upon GroEL impairment ^33,46. Clearly, much remains to be learned about the function of chaperonins and how they cooperate with other chaperone machinery to facilitate folding in the cell.

Tissue Culture and Pulse-Chase Analysis of Mammalian Proteins

Human fibroblast TSA-201 cells used for pulse-chase analysis were cultured in complete medium at 37°C (DMEM with 10% (v/v) FBS, 100 U ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin, and 2 mM L-glutamine). Cells were starved for 15 minutes in starvation media (DMEM - methionine and cysteine, supplemented with 5% (v/v) dialyzed FBS to remove amino acids: Sigma, 10 kDa MWCO), before they were labeled with 0.8 mCi ml^{−1 35}S-methionine for 5 minute, and then chased for 30 minutes in chase media (complete medium supplemented with 0.4% (w/v) cysteine and methionine). Then cells were harvested in ATP depletion buffer (ice-cold PBS with 1 mM azide, 5 mM EDTA, and 2 mM deoxyglucose), lysed by dounce homogenization in buffer A (20 mM HEPES pH 7.4, 100 mM NaCl, 5 mM EDTA, 5% (v/v) glycerol), and clarified by centrifugation. Equivalent amounts of protein were then immunoprecipitated with 2 µl anti-TCP1 β antibodies as described ¹⁴. Briefly, lysates were incubated with antibodies for 40 minutes, rotated with 10 µl Protein G Sepharose for another 40 minute, and immunoprecipitates were washed in TBS + Tween20 buffer as described ¹⁴.

2D Gel Analysis and Mass Spectrometry

TSA-201 cells were briefly labeled with ³⁵S-methionine for 5 minutes, followed by a 0 or 30 minute chase in chase media as described above. Cells lysates were immunoprecipitated with anti-TCP1 β antibodies, and immunoprecipitates were resuspended in 250 µl IEF rehydration buffer (8M urea, 2% (w/v) CHAPS, 2% (v/v) IPG buffer pH 4–7, 2.8 mg ml⁻¹ DTT, trace bromophenol blue) with 400 µg unlabelled lysate (~15 mg ml⁻¹), before being separated on 13 cm Immobiline DryStrips pH 4–7 (Pharmacia Biotech), according to manufacturer’s instructions. The proteins were separated in the second dimension by 10% SDS-PAGE, and gels were silver-stained and analyzed by phosphorimaging. Superposition of silver-stained and radiolabelled gel images identified proteins to be further analyzed by mass spectrometry. Protein spots were excised from a keratin-free gel run in a parallel sample and digested with 12.5 ng µl⁻¹ trypsin at 37°C for 4 hr - overnight, and identified by LC/MS as described ⁴⁷. Data analysis was performed with the algorithm Protein Prospector.

Specificity of in vivo vs. in vitro binding

TSA-201 cells were labeled overnight with 10 µCi ml^{−1 35}S-methionine in starvation medium supplemented with 1% (v/v) complete medium. To generate a cytosolic protein extract, the cells were harvested, lysed, and then denatured with 6 M guanidinium chloride. In vitro binding to the chaperonin was assessed by diluting the denatured extract (~5 mg ml⁻¹) by 100 times into 200 µl buffer B (25 mM HEPES pH 7.4, 100 mM KCl, 5 mM EDTA, 1 mM 2-deoxyglucose, 1mM NaN₃) containing either the chaperonin GroEL (0.1 µM) or TRiC (0.25 µM). Samples were incubated for 20 minutes at 30°C, followed by 20 minutes on ice, and then clarified by ultracentrifugation for 25 minutes at 20,000xg. Bound proteins were then coimmunoprecipitated with anti-chaperonin antibodies (5 µl rabbit anti-GroEL or 2 µl rabbit anti-TCP1 β antibodies), separated simultaneously by 2D gel, and then visualized by autoradiography.

To compare proteins bound to GroEL and TRiC, gel images were differentially colorized in Adobe Photoshop and overlayed. Minor rescaling of images was performed using actin and tubulin spots as points of reference.

Small Pool Expression Cloning

TRiC-interacting proteins were identified by screening for TRiC coimmunoprecipitation of expressed cDNAs from a substracted murine library. The cDNA library was generated from mRNAs differentially expressed in G1/S and G2/M C2C12 cells as described ²⁵. cDNAs were inserted into the pCS+ vector behind the SP6 promoter and transformed into bacteria. Bacterial cDNA clones were grown in a 16×12 array and then pooled by rows and columns from which plasmid DNA were purified (Qiagen miniprep). cDNA pools (~250 ng) were expressed in vitro in a rabbit reticulocyte lysate coupled transcription-translation system (8 µl, Promega TNT Quick Coupled Transcription/Translation Systems) in the presence of ³⁵S-methionine (0.75 µCi µl⁻¹) at 30°C for 45 minutes. Translations were stopped by incubation with 10 mM EDTA, 50 µg ml⁻¹ RNase A, and 5 mg ml⁻¹ methionine for 10 minutes on ice before they were immmunoprecipitated with anti-TCP1 β antibodies. TRiC-substrate complexes were separated by 12% SDS-PAGE and analyzed by phosphorimaging. Proteins of similar size present in two orthogonal pools were selected as candidate clones. Then the interactions of individual expressed cDNAs were confirmed by in vitro expression, followed by anti-TRiC immunoprecipitation. Confirmed clones were finally identified by SP6-primed sequencing and BLAST analysis. In total, 2600 clones were screened, from which approximately 1500 clones were estimated to have an insert generating a detectable translation product.

In vivo Validation of Substrates

Bioinformatic Characterization of the TRiC Interactome

A comprehensive list of TRiC-interacting proteins was compiled, including proteins identified in this study as well as other reports ^{9,11,12,19,32,44,45,50–54}. TRiC-interacting proteins identified in largescale proteomics studies were also included if they bound to three or more TRiC subunits ²⁶. Trends found among the TRiC-Interacting proteins were compared with the distribution of cytosolic proteins occurring in the yeast proteome (Figure 4) or the murine proteome (Suppl. Fig. 3). The cytosolic protein class was estimated by removing proteins with one or more transmembrane helices from the proteome, using the prediction server TMHMM ⁵⁵, resulting in a reduction of the proteome by approximately 25%, in agreement with predictions by Wolf and colleagues ⁵⁶.

Physical and structural properties of proteins were predicted by computer algorithms: the Kyte-Doolittle scale was used to measure hydrophobicity ⁵⁷ and PSIPRED was used to determine secondary structure ⁵⁸. Scanning window analysis determined the maximal number of hydrophobic (or beta strand) residues within a 60 residue window for each protein. P-values were calculated by chi-squared analysis. P-values calculated using the unequal variance t-test ⁵⁹ also confirmed significant differences in the means of the distributions. Functional properties of proteins such as essentiality and complex formation were also determined. Data on protein essentiality was downloaded from the Saccharomyces Genome Database (SGD) ⁶⁰. A list of yeast complexes containing at least two non-TRiC protein members was obtained from the MIPS database ²⁹.