Prefoldin Recognition Motifs in the Nonhomologous Proteins of the Actin and Tubulin Families*

Nascent actin and tubulin molecules undergo a series of complex interactions with chaperones and are thereby guided to their native conformation. These cytoskeletal proteins have the initial part of the pathway in common: both interact with prefoldin and with the cytosolic chaperonin containing tailless complex polypeptide 1. Little is understood with regard to how these chaperones and, in particular, prefoldin recognize the non-native forms of these target proteins. Using mutagenesis, we provide evidence that (cid:1) -actin and (cid:2) -tubu-lin each have two prefoldin interaction sites. The most amino-terminally located site of both proteins shows striking sequence similarity, although these proteins are nonhomologous. Very similar motifs are present in (cid:1) - and (cid:3) -tubulin and in the newly identified prefoldin target protein actin-related protein 1. Actin-related proteins 2 and 3 have related motifs, but these have altered charge properties. The latter two proteins do not bind prefoldin, although we identify them here as target proteins for the cytosolic chaperonin. Actin fragments containing one-step PCR with primers appropriate restriction sites. A two-step PCR used to constructs for the internal deletion mutants of (cid:2) -tubulin. First, both sequences flank- ing the intended deletion were amplified by appropriate primers using Pfu DNA polymerase and gel-purified. Equimolar amounts of both fragments were phosphorylated by T4 kinase and ligated. A second PCR was performed on one-fifteenth of the gel-purified ligation mixture with primers complementary to the 5 (cid:1) and 3 (cid:1) end of the (cid:2) -tubulin cDNA. All PCR fragments were cloned into the Nco I- Bam HI-linearized expression vector pET11d (Stratagene), with the exception of Arp1-(55–208), which was cloned into the Hin dIII- Eco RI linearized expression vector pcDNA3 (Invitrogen), and Arp2 and Arp3, which were cloned into Kpn I- Xba I linearized vector pcDNA3.1 (Invitrogen). GFP was transferred from the pEGFP-C1 vector (CLONTECH) to pET11d via the Nco I- Bam HI sites. (cid:3) -Actin-(51–203) was inserted in GFPpET11d via the Xho I- Bam HI sites. Alanine scan mutants were made with the QuikChange site-directed mutagenesis kit (Stratagene) using (cid:3) -actin-(1–350), (cid:3) -actin- (51–203), or (cid:2) -tubulin-(1–323) as template. Combined alanine scan mutants were made by exchanging fragments between the single mu- tants using Bst EII ( (cid:3) -actin) or Cla I ( (cid:2) -tubulin). Constructs were sequenced at the 5 (cid:1) and/or 3 (cid:1) end of their coding sequence, and the alanine scan mutants were sequenced site of the introduced mutations. mutants were described previously (11). Protein Methods— We purified prefoldin from rabbit reticulocyte lysate as described previously (4). We expressed PFD5 as a recombinant protein in Escherichia coli and purified it on a Superdex 200 column on a fast protein liquid chromatography system (Amersham Pharmacia Biotech) in a buffer containing 5 M urea in 10 m M Tris-HCl, pH 7.2, and 1 m M dithiothreitol. We dialyzed it against water, whereupon PFD5 precipitated as a pure protein. Rabbit polyclonal antibodies against PFD5 were raised by the Center d’Economie Rural (Belgium) and af- finity-purified using CNBr-Sepharose-coupled PFD5. In vitro transcrip-tion translation reactions in reticulocyte lysates, gel electrophoresis, and quantification of 35 S-labeled proteins bound to prefoldin were car-ried Ref. except that products were on native gels as in Ref. We chose (cid:3) -actin-(1–350) and (cid:2) -tubulin-(1–323) as 100% binding references because the full-length proteins interact very transiently to prefoldin. These stably binding mutants allow more accurate monitoring of differences in prefoldin binding capacity. were expressed as nonlabeled recombinant proteins in E. coli . They were purified as described above and dialyzed against 10 m M Tris, pH 7.2, and 1 m M dithiothreitol. For competition experiments, we added increasing amounts of the actin fragments to 15 (cid:4) l of prefoldin (0.5 (cid:4) M ). We incubated these mixtures for 20 min at room temperature. Urea-denatured 35 S-labeled target protein was then diluted 200-fold into folding buffer. Immediately after dilution, we mixed 10 (cid:4) l of this solution containing 2.5 pmol of target protein into each prefoldin-pep- tide mixture and continued the incubation for 15 min. We analyzed and quantified (11) the amount of target protein-bound prefoldin on nonde- naturing polyacrylamide gels (3).

Chaperones assist folding and prevent aggregation of proteins in the crowded environment of the cell. The eukaryote cytosol contains several chaperones, of which hsp70 family members and the cytosolic chaperonin containing TCP-1 (CCT) 1 have been best studied (1). The cytoskeletal proteins, actin and tubulins, are the major target proteins for CCT, although other proteins that require this chaperonin have been identified, most notably, actin-related protein (Arp) 1 and ␥-tubulin (2). It is thought that CCT mainly acts posttranslationally (3) and that nascent actin and ␣-, ␤-, and ␥-tubulin chains are captured by another cytosolic chaperone, prefoldin (also called GimC; Refs. [3][4][5]. Eukaryotic prefoldin is composed of six different subunits (4), whereas its archaebacterial counterpart contains only two types of different subunits present in two and four copies, respectively. The crystal structure of the latter shows a jellyfish-like protein of which the tentacles are formed by coiled coils. The tips of these were proposed to interact with target proteins (6). After completion of synthesis, the prefoldinactin complex is thought to form a transient ternary complex with CCT, and non-native actin is transferred to this chaperonin (4). CCT contains eight different subunits (7,8), and some of these serve as binding sites for target proteins (9 -11). CCT assists folding of its target proteins in an ATP-dependent manner; however, no energy source appears to be required for the transfer of the translated target protein from prefoldin to CCT (4).
Experimental evidence indicates that CCT (a class II chaperonin) recognizes its target proteins in a different manner than class I chaperonins (11,12). In particular, class I chaperonins appear to recognize general hydrophobic properties of target proteins (13), whereas CCT interacts with target proteins through discrete binding determinants (11, 14 -18) that may have common features (11,15). Similarly, we hypothesized that prefoldin interacts with its target proteins, actin and tubulins, via a common mechanism. Therefore, we sought to identify similar features in the two nonhomologous proteins ␤-actin and ␣-tubulin enabling them to interact specifically with prefoldin.
Using truncation analysis, we identified in both target proteins two regions necessary for prefoldin interaction. Mutations of particular residue stretches within these regions to alanine reduce or abolish the binding of ␤-actin or ␣-tubulin to prefoldin and reveal a common signature sequence in the most amino-terminal site. The presence of a strikingly similar prefoldin interaction motif in these nonrelated target proteins suggests that they are recognized by similar prefoldin subunits. This argument is strengthened by our observation that fragments of actin containing these two regions compete efficiently with ␤-actin and with ␣and ␤-tubulin for prefoldin binding. Similar motifs are present in ␥-tubulin and in the newly identified prefoldin target protein Arp1.
¶ Recipient of fellowships from the Flemish Institute for Promotion of Scientific-Technological Research in Industry (IWT). 1 The abbreviations used are: CCT, cytosolic chaperonin containing TCP-1; Arp, actin-related protein; GFP, green fluorescent protein; PCR, polymerase chain reaction; PFD, prefoldin.
were generated by a one-step PCR with primers containing appropriate restriction sites. A two-step PCR was used to obtain the constructs for the internal deletion mutants of ␣-tubulin. First, both sequences flanking the intended deletion were amplified by appropriate primers using Pfu DNA polymerase and gel-purified. Equimolar amounts of both fragments were phosphorylated by T4 kinase and ligated. A second PCR was performed on one-fifteenth of the gel-purified ligation mixture with primers complementary to the 5Ј and 3Ј end of the ␣-tubulin cDNA. All PCR fragments were cloned into the NcoI-BamHI-linearized expression vector pET11d (Stratagene), with the exception of Arp1-(55-208), which was cloned into the HindIII-EcoRI linearized expression vector pcDNA3 (Invitrogen), and Arp2 and Arp3, which were cloned into KpnI-XbaI linearized vector pcDNA3.1 (Invitrogen). GFP was transferred from the pEGFP-C1 vector (CLONTECH) to pET11d via the NcoI-BamHI sites. ␤-Actin-(51-203) was inserted in GFPpET11d via the XhoI-BamHI sites. Alanine scan mutants were made with the QuikChange sitedirected mutagenesis kit (Stratagene) using ␤-actin-(1-350), ␤-actin-(51-203), or ␣-tubulin-(1-323) as template. Combined alanine scan mutants were made by exchanging fragments between the single mutants using BstEII (␤-actin) or ClaI (␣-tubulin). Constructs were sequenced at the 5Ј and/or 3Ј end of their coding sequence, and the alanine scan mutants were sequenced at the site of the introduced mutations. The ␤-actin truncation mutants were described previously (11).
Protein Methods-We purified prefoldin from rabbit reticulocyte lysate as described previously (4). We expressed PFD5 as a recombinant protein in Escherichia coli and purified it on a Superdex 200 column on a fast protein liquid chromatography system (Amersham Pharmacia Biotech) in a buffer containing 5 M urea in 10 mM Tris-HCl, pH 7.2, and 1 mM dithiothreitol. We dialyzed it against water, whereupon PFD5 precipitated as a pure protein. Rabbit polyclonal antibodies against PFD5 were raised by the Center d'Economie Rural (Belgium) and affinity-purified using CNBr-Sepharose-coupled PFD5. In vitro transcription translation reactions in reticulocyte lysates, gel electrophoresis, and quantification of 35 S-labeled proteins bound to prefoldin were carried out as described in Ref. 11, except that products were analyzed on native gels as described in Ref. 3. We chose ␤-actin-(1-350) and ␣-tubulin-(1-323) as 100% binding references because the full-length proteins interact very transiently to prefoldin. These stably binding mutants allow more accurate monitoring of differences in prefoldin binding capacity.

RESULTS
Discrete Binding Sites for Prefoldin in ␤-Actin and ␣-Tubulin-We previously identified the CCT recognition determinants in ␤-actin using truncated actin variants by monitoring their CCT binding on native gels. In these types of assays, we observed for some of the mutants an additional band of nonnative, radioactively labeled products migrating between CCTbound complexes and native actin (see Fig. 2 in Ref. 11). We reasoned that this would be non-native actin associated with prefoldin (4) and confirmed this using anti-prefoldin 5 antibodies (data not shown). Moreover, this band for actin associated with prefoldin was more intense after the addition of purified prefoldin to translating lysates (see Fig. 1A). Similarly, a complex between ␣or ␤-tubulin and prefoldin can be formed (Fig.  1A), which was also confirmed using the anti-prefoldin 5 antibodies (Fig. 1B). We delineated the regions of ␤-actin and ␣-tubulin required for prefoldin binding using truncation analysis ( Fig. 1, CϪF). Using this assay, one must keep in mind that the fragments produced in lysates can bind to both CCT and prefoldin and that the binding of a given fragment will be driven to the chaperone for which it has the most binding information (see below). Thus, removing parts of CCT binding information will direct binding to prefoldin, whereas deleting prefoldin binding information will drive the binding to CCT (for instance, see Fig. 2 and Table II in Ref. 11). This bias is due to recognition; thus, having the two chaperones in the lysate increases the specificity toward a given chaperone but obviously complicates quantitative analysis. Binding of truncated forms of ␤-actin and ␣-tubulin was analyzed on native gels as described in Ref. 3, allowing better quantification of prefoldin complexes. Upon removal of certain portions of the target proteins, prefoldin binding capacity suddenly decreased. We observed the most pronounced reduction upon shortening ␤-actin fragments 1-203 to residue 179 (Fig. 1C) and 51-375 to residue 76 (Fig. 1D) and upon shortening ␣-tubulin fragments 1-249 to residue 219 (Fig. 1E) and 9 -451 to residue 47 (Fig. 1F), suggesting that these regions contain critical information for prefoldin interaction. Additional decreases are noticed for removal of regions 151-179 in ␤-actin and 199 -219 in ␣-tubulin. From the truncation analysis, we conclude that the major binding information for prefoldin in ␤-actin and ␣-tubulin in both cases is present in two regions, i.e. between amino acids 51-75 and 151-203 in ␤-actin and between amino acids 9 -46 and 199 -249 in ␣-tubulin. Alanine scans in these regions confirmed that they contain binding information and additionally allowed us to define discrete sites (see below).
Both regions in each of the target proteins appear to be necessary for efficient and stable complex formation, e.g. actin-(76 -228) (Fig. 1G), which contains only one of the regions, does not bind to prefoldin. In contrast, actin fragment 51-203, which contains both regions, interacts strongly with prefoldin and, when fused to GFP, can even direct GFP to prefoldin (Fig. 1G). ␣-Tubulin fragment 9 -249, which contains the two regions, also binds strongly to prefoldin, whereas fragments 47-249 and 1-198 show little complex formation (Fig. 1, G and E). In addition, ␣-tubulin mutants with internal deletion of one or both regions show little or no binding to prefoldin, respectively (Fig. 1H).
To further investigate the importance of these sequences, we performed alanine scans of regions 50 -79 and 150 -204 in ␤-actin, changing 5 consecutive amino acids each time. Because the amount of prefoldin-bound full-length actin is quite low (most likely resulting from rapid transfer of actin to CCT; Ref. 4), we used ␤-actin-(1-350), a truncated form of actin that binds more strongly to prefoldin, to perform the alanine scan (Fig. 2B). This allows more accurate quantification of the prefoldin binding capacity. The mutants covering sequence 60 -79 of actin in the first region show strongly impaired prefoldin interaction (we define this sequence as site I; see Fig. 2A). Similarly, some alanine mutants of the second region show a significant reduction in binding, i.e. those covering sequences 170 -183 and 194 -198 ( Fig. 2A), whereas replacement with alanines in region 150 -169 does not decrease prefoldin binding. Thus, we define the sequence 170 -198 as site II. Note that the effects of mutations in site II are less dramatic than those in site I.
Comparing these sites with the sequences of the regions of ␣-tubulin, identified above as being required for prefoldin interaction readily suggests that a similar motif for site I is present (underlined in Fig. 2A). Indeed, a tetrapeptide (EHGI) preceded by hydrophobic residues at particular positions can be observed. An alanine scan across this similar region in ␣-tubulin-(1-323) (we used a truncated form of ␣-tubulin for the same reasons explained above for actin) confirms that these residues are implicated in prefoldin binding. A potential analogy between actin and ␣-tubulin is less obvious for site II. In view of the fact that ␣-, ␤-, and ␥-tubulin are prefoldin targets (Refs. 4 and 5; see below), we focused on those sequences in this region of 50 amino acids that are most similar between these tubulin family members (i.e. 202-208 and 218 -249). Mutants covering amino acids 221-231 and 240 -244 show significantly reduced binding ( Fig. 2A), whereas those covering 202-208 do not show altered binding (data not shown). It is noteworthy, however, that similar to the case for actin, the binding contribution of site II residues is smaller than that of site I residues. Despite this delineation of site II, based on the reduction in binding percentages, it is still difficult to discern a common pattern between actin and tubulin.
As mentioned previously, target protein binding is driven to the chaperone for which it has the most binding information. Actin-A55-59, for example, shows increased prefoldin binding; this may be due to the fact that binding information for CCT has been disturbed. This is in agreement with Ref. 17, in which actin peptide 56 -70 was found to bind to CCT.
Combining an alanine mutant of site I with one of site II, each with moderately reduced binding, results in strongly impaired interaction with prefoldin for both target proteins (Fig.  2C). We conclude that site I and II are indeed important for prefoldin binding and that amino acids in the motif of site I appear to be more important for binding than those in site II.
Actin Fragments Containing Prefoldin Binding Sites Compete with Actin and Tubulins for Prefoldin Binding-The analogy of site I in both target proteins suggests that at least these parts of actin and tubulins contact similar prefoldin subunits. This predicts that actin sites can compete with ␣-tubulin for binding to prefoldin. To test this, we chemically synthesized actin peptides 50 -75 and 179 -203 containing major binding information of sites I and II and preincubated them with purified prefoldin to which radiolabeled actin was added. When added separately, the peptides only weakly inhibit actin binding. Simultaneous addition significantly increases their competitive behavior to a level that is higher than the sum of the contributions of the separate sites (data not shown). We reasoned that fragments with the two sites linked would compete better. Therefore, we produced nonlabeled actin-(51-203), and native gel analysis of 35 S-labeled ␤-actin (C and D) or ␣-tubulin mutants (E and F) truncated from either the carboxyl (C and E) or amino terminus (D and F) (the end points of the fragments are indicated between the gels) produced in rabbit reticulocyte lysates (only the relevant portions containing the prefoldin fragment complex are shown). Bottom panels are denaturing gels monitoring the expression levels and size of the translation products. To calculate the relative percentage of prefoldin binding of the fragments, we divided the amount of prefoldin-bound target protein (phosphorimager quantified from the native gels) by the total amount of expressed protein (quantified from the denaturing gels) relative to a 100% binding reference (1-350 for ␤-actin and 1-323 for ␣-tubulin). The percentages shown above the gels are averages of the values obtained from at least three independent experiments. Ϯ indicates that the amount of prefoldinfragment complex was too low to be accurately quantified. it did indeed prevent in vitro binding of denatured, radiolabeled actin to prefoldin in a concentration-dependent manner. Importantly, competition of this actin fragment with ␣and ␤-tubulin was as efficient as compared with actin (Fig. 3A). To determine whether inhibition with this nonlabeled actin fragment also occurs in actively translating lysates, we added it to reactions producing ␤-actin-(51-375) or ␣-tubulin-(1-323) (see Fig. 3A, inset, lanes b and c). In both cases, we observed a dramatic loss of the radioactivity associated with the prefoldin complex. As a control for specificity, we used actin-(76 -179), which contains the sequence in between the sites and part of site II. This fragment does not bind to prefoldin when produced in reticulocyte lysates (Fig. 1G), and the recombinant produced and purified fragment does not influence the interaction of ␤-actin-(51-375) and ␣-tubulin-(1-323) with prefoldin in lysates at the highest concentration used (see Fig. 3A, inset, lanes d). In addition, it hardly competes with actin for binding to prefoldin in the competition assay, whereas actin-(51-203) does (Fig. 3B). The slight inhibition observed is probably due to residual binding information present in residues 171-179. These results suggest that amino acids in between site I and II do not contain major prefoldin binding information.
The inhibiting capacity of actin-(51-203) was strongly reduced by mutating residues 70 -74 to alanine and to a lesser extent by changing residues 179 -183. This once again indicates that these residues contain crucial prefoldin binding information and that site I appears to be more important than site II. Given the competitive behavior for prefoldin binding and the similarity of site I in actin and ␣-tubulin, we conclude that actin and tubulin dock (at least in part) to similar acceptor sites on prefoldin and that the binding of the two sites may be cooperative.

Arp1 Is a Target Protein for Prefoldin and Has Similar Recognition Sites as Actin and ␣-, ␤-, and ␥-Tubulin; Arp2 and Arp3 Do Not Interact with Prefoldin, Although They Are CCT
Target Proteins-␥-Tubulin was reported to be a prefoldin and CCT target protein (2,5). Because Arp1 is also a CCT target (2), we investigated whether this protein and the other actin homologues Arp2 and Arp3 are target proteins for prefoldin. We first checked whether Arp2 and Arp3 do interact with CCT in reticulocyte lysates. This is indeed the case (Fig. 4). At present, it is not known whether the band at the bottom of the gel represents folded Arp3 protein. Intuitively, one may expect that these proteins require their partner proteins (20) to remain stable. Inspection of the gels readily identifies Arp1 as a prefoldin target. In contrast, Arp2 and Arp3, despite their similarity to actin and Arp1, do not bind to prefoldin (Fig. 4) even when they are carboxyl-terminally truncated (data not shown). Upon inspection of their sequences, we noted that ␥-tubulin and Arp1 contain the site I consensus motif (see Fig.  2A), whereas this motif is less conserved in Arp2 and Arp3. As we did for actin, we constructed ␣-, ␤-, and ␥-tubulin fragments containing the expected prefoldin binding sites for Arp1. After translation, these products bind to prefoldin (Fig. 4), confirming that these fragments contain major prefoldin binding information.

DISCUSSION
In ␤-actin and ␣-tubulin, we identified regions important for binding to prefoldin. Either target protein contains two such regions, of which the amino-terminal ones (site I) are strikingly similar (underlined and boxed in Fig. 2A). This is remarkable because actin and ␣-tubulin are nonhomologous. We looked for other candidate prefoldin target proteins containing the site I motif, but our data base searches were hampered by the shortness of the sequence. When restricted to one organism (e.g. man or Drosophila), we found either no hits or actins or tubulins (depending on the length and sequence of the input). However, the site I signature sequence ( Fig. 2A) is present in ␤-tubulin (Fig. 1A), in ␥-tubulin, and in the newly identified prefoldin target Arp1. Intriguingly, the EHGI sequence preceded by hydrophobic residues at the Ϫ3 and Ϫ7 position in site I is extremely well conserved in all actins, in Arp1, and in all ␣-, ␤-, and ␥-tubulin isoforms. We note that in native ␤-actin, His-73 is methylated. The role of this methylation is unclear, although it may be tempting to speculate that this modification could play a role in the recognition of non-native actin by prefoldin or in its transfer from prefoldin to CCT. The following observations argue against such a scenario. In yeast, prefoldin is required for normal kinetics of actin production (21), but His-73 in yeast actin is not methylated (22), and mutants in His-73 do not show a phenotype in vivo (23). The potential importance of His-73 methylation in these processes is difficult to verify because of the low levels of incorporation of 14 C-methyl in ␤-actin produced in lysates (data not shown) and because prefoldin can be bypassed in this system (11).
The latter is consistent with our observation that Arp2 and Arp3 do not bind prefoldin despite being target proteins for CCT. This also indicates that not all proteins interacting with the cytosolic chaperonin are also prefoldin targets. We note that the sequences of site I and site II of Arp2 and Arp3 are not absolutely conserved with respect to actin and even contain charge alterations, perhaps suggesting that charged residues are important in the interaction of eukaryotic prefoldin with target proteins.
Site I appears to be more important for prefoldin recognition because the binding behavior of alanine scan mutants in site I was more affected than that of the analogous site II mutants. Additionally, our various competition experiments confirmed the relative importance of site I. Obviously, further research is needed to completely dissect the individual importance of each residue in the motif, but based on its sequence, we put forward that polar interactions are important for the interaction between prefoldin and its target proteins. This is consistent with structural data available for prefoldin from archaebacteria. For Methanobacterium thermoautotrophicum prefoldin, it was proposed that this interaction occurs via exposed hydrophobic surface patches at the distal regions of the locally untwisted coiled coils (6). However, in eukaryotic prefoldin subunits, the tips appear less hydrophobic, supporting our view that target protein interaction occurs partly by hydrophilic interactions.
In the respective crystal structures of actin and tubulin (24,25), the site I amino acids are located in loops, whereas those of region II are in rather similar loop-␣-helix structures. These regions of the protein do not appear to be essential for nucleotide binding, with the possible exception of the site II ␣-helix in tubulin (residues Tyr-224 and Asn-228 may contact the nucleotide base; Ref. 25). In actin, the EGHI tetrapeptide sequence is close to a polymer contact (in the refined Holmes model; Ref. 26), but it is not directly involved in contacting neighboring protomers, nor has it been described as an actin-binding protein interaction site. Similarly, in ␣and ␤-tubulin, this tetrapeptide is in a loop that is not involved in tubule formation and is probably not involved in contacting microtubule-associated proteins because it is on the inside of the protofilament (27). In tubulins, residues more carboxyl-terminal of this loop are much less conserved, and in ␣-tubulin, even insertions are accommodated. Thus, this extreme conservation of the site I signature sequence suggests another important function, which may well be the interaction with prefoldin during translation. Site II seems to be generally less well conserved (for example, Arp1 and ␤and ␥-tubulin), and this may explain why these target proteins bind less well to prefoldin than actin and ␣-tubulin (see Fig. 4; the fragments of Arp1 and ␤and ␥-tu-bulin also show binding to CCT, whereas actin and ␣-tubulin do not).
If we consider the modeled structure of actin in the electron microscopic reconstruction of actin-CCT (10), site I and II are both solvent-exposed, facing out of the CCT cylinder but spatially close to CCT subunits. This is suggestive of a mechanism in which prefoldin with bound actin docks to CCT by specific interactions between prefoldin and CCT subunits in such a way that the CCT recognition sites in non-native actin are properly presented to CCT subunits. Indeed, the two prefoldin interaction sites in ␤-actin are located at either side of the aminoterminal CCT interaction site (11), and the same applies for ␣-tubulin. 2 Because prefoldin is thought to transfer non-native target proteins to CCT after the formation of a transient ternary complex with CCT (4), some prefoldin subunits may serve to specifically contact certain CCT subunits, and this may be the reason why, in eukaryotes, both these chaperones consist of multiple subunits, i.e. some are required for target protein binding, and/or some are required for proper docking onto subunits of the partner chaperone.