Co-translational determination of quaternary structures in chaperone factories

Philippe, Manon; Salloum, Soha; Slimani, Floric; Normanno, Davide; Barbosa, Isabelle; Chassé, Héloïse; Robert, Marie-Cécile; Urbach, Serge; Imbert, Jacques; Séveno, Martial; George, Simon; Le Hir, Hervé; Boulon, Séverine; Verheggen, Céline; Bertrand, Edouard

doi:10.1038/s41467-026-68687-8

Download PDF

Article
Open access
Published: 23 January 2026

Co-translational determination of quaternary structures in chaperone factories

Nature Communications volume 17, Article number: 1978 (2026) Cite this article

3478 Accesses
2 Citations
Metrics details

Subjects

Abstract

The HSP90/R2TP quaternary chaperone assembles key cellular machines, including the three nuclear RNA polymerases and many non-coding RNPs. Here, we characterize the RNAs associated with R2TP and find that it binds many proteins co-translationally. Its co-translational interactome reveals potential clients and distinguishes clients bound only co-translationally, only post-translationally, or both. For subunits assembling together in the same complex and bound co-translationally by R2TP, only a marginal proportion of their mRNAs is co-localized and co-translated. Instead, HSP90 and R2TP induce the formation of polysome condensates accumulating these chaperones and specific client mRNAs, thus favoring co-translational interactions between them. During the assembly process, R2TP cycles between co- and post-translational steps and this is regulated by ATP: it binds co-translationally in absence of ATP and becomes released from post-translational assembly intermediates by ATP hydrolysis. This co-translational mechanism, named co-translational chaperone channeling (cha-cha), substitutes for the rarity of co-localized/co-translated mRNAs. In turn, chaperones have the remarkable ability to spatially organize the translational apparatus in the cytoplasm.

Co-translational binding of importins to nascent proteins

Article Open access 09 June 2023

Scaffolds with optimized quaternary symmetry for de novo cryoEM structure determination of small RNAs

Article Open access 09 March 2026

An integrative structural biology approach reveals the dynamic organization of the R2SP quaternary chaperone complex

Article Open access 10 February 2026

Introduction

Non-coding RNPs are complex macromolecular assemblies and studies performed on many systems have established that they require specific factors for their assembly. For instance, a large number of comprehensive studies performed in the last 20 years have revealed that ~200 assembly factors are required to produce functional ribosomes¹. Likewise, despite their apparent simplicity, small RNPs like the spliceosomal snRNPs or the nucleolar snoRNPs also require a surprisingly large number of specialized assembly factors². These factors are evolutionarily conserved, and their number frequently exceeds the set of components of the mature RNP particle, highlighting the generality and importance of this process^2,3. Assembly machineries also occur for stable protein complexes devoid of RNAs. For instance, the formation of RNA polymerases requires an assembly machinery that includes more than 14 factors and whose mechanism of action remains poorly understood⁴. Interestingly, some of these assembly factors, like the HSP90/R2TP chaperone, are linked to the protein folding machinery. Indeed, quaternary protein folding can be seen as an extension of tertiary folding, with the main difference being that it occurs in trans and not in cis. Assembly factors have been shown to chaperone unassembled subunits, improve efficiency and specificity of the assembly process, transport subunits across the cell and perform quality controls^1,2,3.

R2TP is a conserved HSP90 co-chaperone that has a unique and essential role in the cell as a generalist assembly chaperone that helps build multi-subunit complexes^5,6,7,8. Specifically, R2TP promotes the biogenesis of: (i) small non-coding RNPs such as the box C/D and H/ACA snoRNPs^2,9,10; the telomerase RNP^11,12; U4 and U5 spliceosomal snRNPs^13,14,15; and miRNPs¹⁶; (ii) the three nuclear RNA polymerases^4,17; (iii) ciliary dyneins¹⁸; (iv) complexes containing PIKKs (mTOR, ATM/ATR, DNA-PK, TRRAP, SMG1)^19,20 ⁹; and (v) the Tuberous Sclerosis Complex (TSC), which controls mTOR activity²¹. This list of client complexes is likely incomplete, as large-scale interaction assays have revealed additional R2TP partners^22,23,24,25. In agreement with this diversity of roles, the R2TP subunit RPAP3 is essential in mice, and its targeted deletion in the adult intestine leads to rapid disruption of tissue homeostasis because of a lack of assembly of key complexes such as RNA polymerase II²⁶.

R2TP is highly conserved across evolution, and it can be tracked to the root of the eukaryotic lineage²⁷. It is composed of a heterodimer of RPAP3:PIH1D1 associated with the AAA+ ATPases RUVBL1 and RUVBL2^24,28^,. In humans, R2TP further associates with a set of prefoldins and prefoldin-like subunits to form the PAQosome²⁹. RUVBL1/RUVBL2 form hetero-hexameric and dodecameric rings and are believed to carry a chaperone activity on their own^30,31,32. Compared to other AAA+ ATPases, they have an extra domain called DII, which is located outside the ring and whose conformational changes are believed to play important roles in the assembly function of R2TP^7,30,33,34. Indeed, biochemical and structural data have shown that RUVBL1/RUVBL2 make ATP-dependent contacts with clients and cofactors^7,35. Because of their hexameric nature, it is thought that this enables them to hold various subunits in place during the assembly process, with nucleotides acting as a switch for client binding or release (³⁶ for a review). Human RPAP3 possesses two TPR domains that bind HSP70 and HSP90 and a C-terminal domain that associates with RUVBL1/RUVBL2 at the opposite side of DII^27,33,37,38. RPAP3 forms a heterodimer with PIH1D1, which is important for client recognition and the regulation of the ATPase activities of RUVBL1/RUVBL2 and HSP70/90 during their chaperone cycle^37,39,40. The recognition of R2TP clients can occur in several ways. Some clients are recruited with the help of cofactors, such as the TTT complex for PIKKs^41,42,43; ZNHIT2 and ECD for U5 snRNP^7,14,15; and NOPCHAP1, NUFIP1, ZNHIT3, and ZNHIT6 for C/D snoRNPs^9,25,44. Others are directly recruited by R2TP, and this can occur via the PIH domain of PIH1D1, which contains a basic pocket that binds acidic peptides phosphorylated on the DpSDD/E consensus motif⁴³, or via RPAP3, as in the case of the Dicer co-factor TRBP¹⁶ or clients containing Armadillo repeats²⁴.

Recent studies suggested that the assembly of multi-subunit complexes is intimately coupled with translation^45,46,47. For instance, biochemical studies in yeast and mammals have shown that subunits of the same complex can bind each other co-translationally, sometimes reciprocally, and other times in a directional manner⁴⁵. In mammals, detailed studies of the general transcription factor TFIID and the co-activators SAGA and ATAC have revealed hierarchical co-translational assembly pathways that mix co- and post-translational steps^48,49,50. For TFIID in particular, some subunits join others during their synthesis to create sub-modules, which, after their release from the ribosome, assemble with the largest TAF1 subunit during its translation⁴⁹. Importantly, the co-translational joining of one subunit with another can occur either by associating and co-translating the two polysomes together (called the co-co pathway) or by transporting one fully translated subunit to its partner polysome (the co-post pathway). Current evidence suggests that the co-co assembly pathway may be common^51,52, although single RNA imaging has so far only revealed low frequencies of mRNA co-localization^47,52, leaving the exact usage of the co-co pathway an open question. In the case of the co-post pathway, how fully translated subunits are transported to their partner polysomes for assembly is not well understood. In the case of the SMN complex, which assembles the Sm ring on snRNAs, the newly translated Sm proteins remain on the ribosome until they are picked up by the pICln assembly chaperone, which brings them to SMN⁵³. Despite the different mechanisms observed, many pathways converge to the ribosome, indicating that it may function as a hub driving assembly.

Here, we analyze co-translational events for chaperone-assisted assembly of macromolecular complexes and show that the HSP90/R2TP assembly chaperone binds many of its clients during their translation. The co-co pathway appears to be marginal, and instead, the chaperone induces the formation of dedicated translation factories that accumulate client mRNAs and thus favor chaperone-client co-translational interactions. R2TP thus channels the formation of macromolecular complexes starting from individual subunit synthesis, likely compensating for the scarcity of subunit mRNA co-localization.

Results

RUVBL1 immunoprecipitates client mRNAs in a translation-dependent manner

The HSP90/R2TP chaperone (Fig. 1A) is specialized in the assembly of macromolecular complexes, and recent evidence has linked this process to translation⁴⁷ (for a review). We thus tested whether R2TP can bind its client proteins while they are being translated. We performed a RIP-seq experiment using a heterozygous HCT116 cell line having a GFP tag fused at the C-terminus of one allele of RUVBL1. This cell line was characterized by PCR genotyping, Western blot, and immunofluorescence, confirming the proper insertion of the tag and the localization of the RUVBL1-GFP fusion in the cytoplasm and the nucleoplasm as expected (Fig. 1B and Supplementary Fig. 1A, B⁵⁴). We then immunoprecipitated RUVBL1-GFP using GFP-Trap® beads and sequenced the co-precipitated RNAs. Remarkably, we found 172 RNAs specifically enriched in the GFP immunoprecipitation (IP) over the control (FDR < 0.05), with enrichment values varying from 84 to 2 fold (Fig. 1C; Supplementary Data 1). Some mRNAs coded for known clients of R2TP, such as the PRPF8 and EFTUD2 proteins of U5 snRNP, the large subunits of the three nuclear RNA polymerases (POLR1A, POLR2A, POLR2B, POLR3A), and the PIKK SMG1. Other well-known R2TP clients, such as the snoRNP core proteins DKC1 and NOP58, did not pass the statistical threshold but were nevertheless among the top-enriched mRNAs (ranked 194 and 330, respectively). RUVBL1/RUVBL2 are part of several chromatin complexes (TIP60, SRCAP, INO80), and this is believed to be independent of their participation in the R2TP chaperone. Remarkably, the mRNAs coding for key subunits of these complexes were also highly enriched in the IP (EP400, SRCAP, BRD8, INO80). Most interestingly, many mRNAs encoding proteins not previously known to interact with RUVBL1 were also identified, and several formed clearly identifiable families of related proteins. This was the case for the AGO proteins (AGO1, AGO3, and AGO4, with AGO2 also in the top hits, although below the statistical threshold), the CREB3 family (CREBRF, CREB3, and CREB3L2), and SCAF family members that share a similar domain that interacts with the RNA polymerase II CTD (PCF11, SCAF4, SCAF8, CHERP). It was also noteworthy that RUVBL1-GFP associated with mRNAs that code for subunits assembling together in mature complexes, such as POLR2A and POLR2B in RNA polymerase II, PRPF8 and EFTUD2 in U5 snRNP, and BRD8 and EP400 in the TIP60 chromatin remodeling complex. Other mRNAs coded for seemingly unrelated proteins, although many were long proteins, larger than 1000 amino acids. Finally, we also found an enrichment for several box C/D snoRNAs (SNORD3, SNORD13, SNORD46, and SNORD31), as expected from previous studies^54,55. In order to determine if the association of RUVBL1 with mRNAs was translation-dependent, we repeated the RIP-seq experiment after a 30 min treatment with puromycin, which acts as a translation terminator and releases the nascent peptides from the ribosome. We found that most of the mRNAs previously enriched in the RUVBL1-GFP IP were lost or considerably less enriched upon puromycin treatment (Fig. 1D, S1C, E, F), suggesting that RUVBL1-GFP binds to mRNAs via the nascent proteins, while they are translated by ribosomes.

**Fig. 1: RUVBL1 associates co-translationally with mRNAs.**

The ATPase activity of RUVBL1/RUVBL2 regulates their binding to polysomes

The AAA+ ATPase activity of RUVBL1/RUVBL2 is central to their chaperone activity and the nucleotide-bound state (ATP, ADP, or none) controls which partners they interact with^30,33,35,55. To test how ATP affects their interactions with RNAs, we performed a RIP-seq experiment in the presence of CB-6644, a small molecule that inhibits their ATPase activity and blocks them in the ATP-bound state^56,57. Most of the mRNAs previously found in the RUVBL1-GFP IP were lost or less enriched after CB-6644 treatment, including the mRNAs coding for the U5 snRNP proteins, the subunits of the RNA polymerases, and the chromatin remodeling complexes (Fig. 1E, S1D). RUVBL1 may thus bind the corresponding nascent proteins in the ADP or empty state. In contrast, many box C/D snoRNAs were associated more strongly with RUVBL1-GFP upon CB-6644 treatment, suggesting that the drug prevents the release of RUVBL1/RUVBL2 from immature C/D snoRNPs. A few mRNAs also displayed stronger binding with CB-6644. This was the case for the PCF11/SCAF family (with RPRD2 as an additional member), and for WASHC2A and WAHSC2C mRNAs, which encode two almost identical proteins involved in actin dynamics. RUVBL1/RUVBL2 may thus bind more strongly to these nascent proteins when loaded with ATP. Altogether, this indicated that ATP regulates the co-translational binding of RUVBL1 to its clients.

To gain more insight into this process, we performed a quantitative label-free proteomic analysis of the RUVBL1-GFP IPs (AP-MS), with and without CB-6644 treatment. The proteomic analysis revealed the known partners of RUVBL1 as expected, including all subunits of R2TP, the PAQosome and the INO80, SRCAP and TIP60 chromatin remodeling complexes, as well as many subunits of the three nuclear RNA polymerases, the U5 snRNP and the C/D and H/ACA snoRNPs (Fig. 2A; Supplementary Data 2). RUVBL1-GFP was also associated with known R2TP co-factors, such as ZNHIT2, ZNHIT3, NUFIP1, ZNHIT6, NOPCHAP1, and the TTT complex. In the presence of CB-6644, the subunits of the chromatin remodeling complexes became less enriched, while most of the other clients, such as the H/ACA core component DKC1, the RNA polymerase II subunits (POLR2A, POLR2B, POLR2C, POLR2H), U5 snRNP proteins (PRPF8, EFTUD2) and SMG1, as well as cofactors known to be involved in their assembly (POLR2M, RPAP2 for RNA polymerase II; TELO2, TTI1, TTI2 for SMG1; SHQ1 for DKC1), became more enriched (Fig. 2B). As in the RIP-seq, binding to WASHC2A increased dramatically after CB-6644 treatment. This was also the case for the C/D core component SNU13, which directly binds snoRNAs and likely reflects the increased binding of RUVBL1-GFP to C/D snoRNAs observed in the RIP-seq with CB-6644 (Fig. 1E). In contrast, some R2TP cofactors dissociated (NOPCHAP1, DPCD, WDCP). This was consistent with the ATP-dependence observed in vitro for these proteins^44,58, suggesting that the differences observed in vitro and in vivo reflect the genuine ATP-dependent affinity of RUVBL1/RUVBL2 for these partners.

**Fig. 2: Comparison of the co- and post-translational interactomes of RUVBL1.**

Interestingly, when we plotted the fold change induced by CB-6644 in the binding of RUVBL1 to mRNAs versus proteins (Fig. 2C), it became apparent that the inhibition of the RUVBL1/RUVBL2 ATPase activity had a differential effect, with many mRNAs losing binding upon CB-6644 treatment while the corresponding protein showed increased binding (POLR2A, POLR2B, PRPF8, EFTUD2, SMG1, DKC1). This indicated that RUVBL1/RUVBL2 followed an ATPase cycle linking co- and post-translational steps: they bind co-translationally to these client mRNAs in an empty or ADP state and require ATP hydrolysis to be released from later, post-translational, assembly intermediates.

RUVBL1 association to its partners can be strictly co-translational, strictly post-translational or both co- and post-translational

Next, we wanted to determine which partners bind RUVBL1 co- or post-translationally, and we plotted the enrichments measured in the RIP-seq versus the AP-MS proteomic analysis, using the top 300 candidates in each of the two experiments made in untreated conditions (Fig. 2D). This clearly identified three classes of RUVBL1 partners (Fig. 2D, E): (i) only co-translational (41 partners), which spread along the x-axis and were enriched only in the RIP-seq experiments; (ii) only post-translational (84 partners), located along the y-axis and enriched only in the AP-MS proteomic analysis; (iii) co- and post-translational (11 partners), enriched in both experiments (Fig. 2D, E). Interestingly, the latter corresponded to key subunits of the main R2TP client complexes (U5 snRNP, the RNA polymerases), as well as subunits of the chromatin remodeling complexes. In contrast, the post-translational partners corresponded mainly to the other subunits of these chromatin remodeling complexes and to R2TP cofactors, while the strictly co-translational partners corresponded mostly to previously uncharacterized RUVBL1/RUVBL2 partners. This indicated that the co-translational interactome of the R2TP chaperone is quite unique and not easily accessible by traditional proteomic approaches. We verified by IP/Western blot that RUVBL1-GFP interacts with BRD8 (Supplementary Fig. 2A). This is in agreement with the AP-MS and RIP-seq results that categorize BRD8 as a co- and post-translational partner.

RPAP3 binds clients co-translationally and associates with a similar set of polysomes as RUVBL1

RUVBL1/RUVBL2 are not only components of the R2TP complex but also of several other complexes, such as the R2TP-like chaperones²⁷ and the INO80, SRCAP and TIP60 chromatin remodeling complexes²⁹. To determine whether the association of RUVBL1 with polysomes was taking place in the context of R2TP, we performed a RIP-seq experiment with RPAP3, another R2TP subunit. We used CRISPR/Cas9 genome editing to generate a heterozygous HCT116 cell line with a GFP tag fused at the C-terminus of one allele of RPAP3. The cell line was characterized by PCR genotyping and Western blot, which confirmed the proper integration of the tag (Supplementary Fig. 3A, B). RNA sequencing of GFP IPs showed that 153 mRNAs were associated with RPAP3-GFP, as compared to the control IP (FDR < 0.05; Supplementary Data 1; Fig. 3A, C). Comparison of the enrichment values measured in the RPAP3-GFP and RUVBL1-GFP IPs revealed a good correlation between the two RIP-seq experiments (Fig. 3C): mRNAs highly enriched in the RUVBL1-GFP IP were also highly enriched in the RPAP3-GFP IP, with, however, some preferences. For instance, RUVBL1-GFP associated more strongly with C/D snoRNAs, SMG1 and EFTUD2 mRNAs, while RPAP3-GFP bound better the mRNAs coding for SNAPC4, the SCAF4/SCAF8/CHERP family of RNA polymerase II CTD binding proteins, and the CHD family of chromatin remodelers (Fig. 3C). Interestingly, RPAP3 also bound the PCNT and NIN mRNAs, which code for centrosomal proteins and which were recently shown to be locally translated at centrosomes, possibly to enable their co-translational assembly within centrosomes⁵⁹. Finally, although the presence of RUVBL1 in the INO80, SRCAP and TIP60 chromatin remodeling complexes is thought to be unrelated to R2TP²⁹, RPAP3-GFP was strongly associated with mRNAs coding key subunits of these complexes (EP400, SRCAP, INO80, BRD8; Fig. 3A, C). These mRNAs were the same ones that were bound by RUVBL1-GFP, raising the possibility that the R2TP chaperone could be involved in the biogenesis of these chromatin remodeling complexes, beyond the participation of RUVBL1/RUVBL2 in their mature forms.

**Fig. 3: RPAP3 and RUVBL1 associate with similar RNAs in a translation-dependent manner.**

Importantly, and as previously observed for RUVBL1, most mRNAs were lost or much less associated with RPAP3-GFP after a 30 min puromycin treatment (Fig. 3B, and Supplementary Fig. 3C; Supplementary Data 1). Only 20 out of the 153 enriched RNAs were associated in puromycin-treated cells as well as in untreated cells, and these generally had low enrichment values over the control IP (<1.5; Supplementary Data 1). One interesting RNA of this class was TERC, the non-coding RNA of telomerase. This RNA carries an H/ACA motif and the R2TP chaperone is involved in telomerase assembly^9,11,12,60 suggesting that the translation-independent RNA partners of RPAP3 could also be R2TP clients. In any case, these results demonstrated that the association of RUVBL1 with polysomes reflects an engagement of the full R2TP chaperone, and not a separate function of RUVBL1.

R2TP binds its main mRNA targets via the nascent protein emerging from ribosomes

We then aimed to confirm the association of R2TP with the mRNAs identified above and to unambiguously show that it occurs co-translationally. To this end, we performed RUVBL1-GFP IPs followed by RT-qPCR and compared the association of RUVBL1-GFP with selected mRNAs after translation inhibition with various inhibitors. We observed that puromycin and harringtonine, two translation inhibitors that remove nascent protein chains from the ribosome, abolished the binding of RUVBL1-GFP to POLR2A, POLR2B, EFTUD2, PRPF8, EP400 and BRD8 mRNAs. In contrast, cycloheximide, a translation inhibitor that freezes ribosomes and keeps the nascent polypeptides attached, increases the association of RUVBL1-GFP with these mRNAs (Fig. 3D). This indicates that it is not translation inhibition per se that prevents binding of RUVBL1 to its target mRNAs, but the removal of the nascent protein chain, indicating that this is the binding determinant of RUVBL1.

Next, we performed a polysome fractionation and tested the presence of R2TP by Western blot (Supplementary Fig. 2B). Indeed, a small amount of RUVBL2 could be detected in light polysomal fractions, indicating that it interacts with the translation machinery as expected from its translation-dependent interactions with some mRNAs. Finally, we developed reporters to evaluate the role of translation independently of global translation inhibitors. Using the RUVBL1-GFP cell line, we stably expressed mCherry-POLR2A and mCherry-EFTUD2 fusion proteins, as well as variants in which a STOP codon was inserted immediately after the mCherry. We then IP’ed RUVBL1-GFP and analyzed by RT-qPCR the presence of these mRNAs. We found that mCherry-POLR2A and mCherry-EFTUD2 mRNAs associated with RUVBL1-GFP, while the mRNAs containing the internal STOP codon did not (Fig. 3E, F and Supplementary Fig. 2C, D). Altogether, these results indicated that the association of RUVBL1 with its client mRNAs depends on polysome integrity and, in particular, the translation of the corresponding protein, hence confirming that RUVBL1 is recruited co-translationally by the nascent protein chains of its client subunits.

Artificial R2TP condensates recruit mRNAs coding for client proteins

It has previously been observed that interactions can form artifactually in extracts after cell lysis⁶¹. To validate the interaction of R2TP with its target mRNAs, we thus used a co-recruitment assay to evaluate these interactions directly in intact cells^14,17. We generated artificial cytoplasmic condensates by using a fragment of DDX6 (amino-acids 289-483, or DDX6fm) that is sufficient to target the protein to P-bodies, and fusing it to DsRed2 on one side and either RUVBL1 or RPAP3 on the other side. By fusing RUVBL1 and RPAP3 to DDX6fm, we artificially created a high accumulation of R2TP in P-bodies and could then check by smFISH whether the candidate mRNAs or proteins were recruited into these artificial condensates (Fig. 4). We tested some of the mRNAs most enriched in the previous RIP-seq experiments, coding for: the RNA polymerase subunits POLR2A, POLR2B, POLR1A, the U5 snRNP component PRPF8 and the subunits EP400 and BRD8 of the TIP60 histone acetylase complex. Remarkably, the endogenous POLR2A and PRPF8 mRNAs were efficiently recruited to the artificial P-bodies formed by DsRed2-DDX6fm-RUVBL1 or DsRed2-DDX6fm-RPAP3, revealing a specific interaction in cells with RUVBL1 and RPAP3, respectively (Fig. 4A, B). In addition, this recruitment was lost after a 30 min puromycin treatment. It also did not occur with DsRed2-DDX6fm-KPNA3, an unrelated protein used as a negative control, or with native P-bodies (Fig. 4C, D and Supplementary Fig. 6C).

**Fig. 4: The POLR2A and PRPF8 mRNAs are recruited to artificial R2TP condensates.**

To further confirm that the mRNAs recruited to these artificial P-bodies were being translated, we performed a triple color experiment, adding an immunofluorescence labeling of the POLR2A protein (Fig. 4D). Indeed, this protein also accumulated into the DsRed2-DDX6fm-RUVBL1 foci together with the POLR2A mRNA, and the protein labeling mostly disappeared after a 30 min incubation with puromycin. This confirmed that the recruited POLR2A mRNAs were translated. The other tested mRNAs were also recruited to the artificial DsRed2-DDX6fm-RUVBL1 cytoplasmic foci, although with a lower amount of mRNA per foci (Supplementary Fig. 4). Altogether, these results validated the interactions observed in the RIP-seq experiments.

Single-molecule imaging reveals R2TP factories accumulating POLR2A and PRPF8 mRNAs

Next, we wanted to quantify the interactions of R2TP with mRNAs in a more native context, and we thus tagged the endogenous RPAP3 with the SunTag. Indeed, this is a repeated epitope that enables single protein detection in native cells with a recombinant mono-chain antibody fused to GFP⁶². Genotyping and Western blots confirmed the proper heterozygous integration of the SunTag in HeLa cells and revealed that RPAP3-SunTag was much less expressed than its untagged counterpart (~50 fold less; Supplementary Fig. 5A, B). Interestingly, fluorescence microscopy showed that RPAP3-SunTag could be observed not only as single proteins in the cytoplasm but also as foci containing multiple molecules (Fig. 5A). We then performed smFISH against POLR2A and PRPF8 mRNAs. Surprisingly, both mRNAs were present as single molecules and also as foci, which colocalized with the RPAP3-SunTag foci (Fig. 5A). Counting both the mRNAs present as foci and single molecules, we measured that an average of 30% and 35% of the PRPF8 and POLR2A mRNAs colocalized with RPAP3-SunTag, respectively (Fig. 5B). After treatment with puromycin for 30 minutes, the foci containing RPAP3-SunTag, POLR2A and PRPF8 mRNAs disappeared and the colocalization of these mRNAs with RPAP3-SunTag dropped significantly (Fig. 5B). Importantly, the colocalization of POLR2A and PRPF8 mRNAs with RPAP3-SunTag occurred not only in the foci but also when these mRNAs were present as single molecules (Supplementary Fig. 5C). This colocalization appeared lower than in the foci (20% versus 75%), but it was still significant as it decreased after puromycin treatment (Supplementary Fig. 5C). The relatively high remaining colocalization in puromycin-treated cells (16% for POLR2A mRNA) likely reflected background events due to the density of the RPAP3-SunTag labeling. Indeed, similar values were obtained with α-catenin mRNA, a control unrelated to R2TP (Fig. 5B). The other mRNAs did not form foci (POLR2B, EFTUD2, BRD8, EP400). Nevertheless, we could detect a puromycin-dependent colocalization with RPAP3-SunTag for all of them, which was not the case for the control α-catenin mRNA (Fig. 5B). Their colocalization was lower than for POLR2A and PRPF8 mRNAs (15-25% vs. 30-35%), but the large excess of untagged RPAP3 (50 fold) is likely to underestimate the true colocalization values.

**Fig. 5: Colocalization of the RPAP3 protein with the PRPF8 and POLR2A mRNA in cytoplasmic foci.**

The foci observed with either RPAP3-SunTag or the POLR2A and PRPF8 mRNAs were intriguing as they suggested the existence of structures dedicated to the translation and co-translational binding of POLR2A and PRPF8 proteins to the R2TP chaperone. To explore this possibility, we used native, untagged HeLa Kyoto cells and labeled endogenous RPAP3 by immunofluorescence together with POLR2A mRNA by smFISH. Remarkably, both POLR2A mRNA and RPAP3 protein formed foci and these foci co-localized (Fig. 6A). On average, cells contained four RPAP3 foci and 75% of the POLR2A mRNA foci colocalized with RPAP3 (Supplementary Fig. 5D, E).

**Fig. 6: Translation dependency of the colocalization of POLR2A and PRPF8 mRNAs with endogenous RPAP3.**

To further characterize these naturally occurring POLR2A mRNA foci, we checked whether they contained other components by IF and smFISH. Interestingly, we observed that the ribosomal protein RPS6 and the chaperone HSP90, which have a diffuse localization in the cytoplasm, were both present, although they did not accumulate, in POLR2A mRNA foci (Supplementary Fig. 6A, B). In addition, the POLR2A protein accumulated in these foci, suggesting that the POLR2A mRNAs were indeed translated there (Supplementary Fig. 6C). In agreement, treatment with puromycin or harringtonine led to the near total disappearance of these foci (Supplementary Figs. 5E, 6E, 7A). In contrast, they were preserved upon cycloheximide treatment, a translation inhibitor that does not induce polysome dissociation, indicating that their formation is likely driven by the nascent protein chains. We then tested PRPF8 mRNA, which also formed foci in the RPAP3-SunTag cells (Fig. 5A, B). Indeed, this mRNA formed foci in native, untagged HeLa Kyoto cells, which were also puromycin sensitive and which frequently co-localized with POLR2A mRNA foci (~50%; Fig. 6B and Supplementary Figs. 5F, 7B). Several mRNAs were recently reported to form translation factories, which are small cytoplasmic foci accumulating specific mRNAs undergoing translation and sensitive to puromycin inhibition⁶³. Dual-color smFISH showed that the POLR2A mRNA foci did not colocalize with the translation factories of DYNC1H1, which is not a client of R2TP (Supplementary Fig. 6D). Likewise, the POLR2A mRNA foci were unrelated to P-bodies (Supplementary Fig. 6C). Because the POLR2A mRNA foci accumulate both the nascent POLR2A protein and RPAP3, they appear to be specialized translation sites requiring R2TP activity.

RNAs coding for pairs of subunits that assemble together rarely colocalize

Recent evidence suggests that mRNAs coding for proteins that assemble together can, in some cases, be colocalized and co-translated, a pathway termed co-co assembly^45,52. The colocalization of POLR2A and PRPF8 mRNAs in foci thus prompted us to test additional mRNAs, as our R2TP RIP-seq experiments revealed several mRNAs coding for pairs of subunits belonging to the same complex, such as POLR2A/POLR2B for RNA polymerase II, PRPF8/EFTUD2 for U5 snRNP and EP400/BDR8 for the TIP60 chromatin remodeling complex (Fig. 7A). Surprisingly, dual-color smFISH showed only a very limited colocalization between these mRNA pairs (Fig. 7B, C). The highest level was observed for BRD8 and EP400, with an average of 5% of BRD8 mRNA colocalizing with EP400 mRNA (Fig. 7C). This low level nevertheless appeared significant because it dropped to less than 1% after a 30 min puromycin treatment, reaching a level similar to the JUP control mRNA. We also observed a low level of colocalization, but still significant, between POLR2B and POLR2A mRNAs (3.5%; Fig. 7B, C), while no significant colocalization was observed between EFTUD2 and PRPF8 mRNAs. Altogether, these data are not in favor of the co-co assembly model where polysomes of subunit pairs associate together. This was further reinforced by the fact that POLR1A mRNAs colocalized with POLR2A mRNAs to the same extent as POLR2B mRNAs, even though they belong to different protein complexes (Supplementary Fig. 8A, B). This suggested that the low level of mRNA colocalization that we observed is not related to co-co assembly.

**Fig. 7: Pairs of mRNAs coding for proteins that assemble together show little colocalization.**

The HSP90 and RUVBL1/2 ATPase activity is required for localized mRNA translation

The accumulation of the RPAP3 protein in cytoplasmic foci, together with POLR2A and PRPF8 mRNAs, suggests that these structures function as R2TP factories to enhance interaction of nascent proteins with the chaperone via a proximity effect. The formation of these structures depends on translation (Supplementary Figs. 5E, 6E, 7A and B), and we wondered whether the chaperones, which interact with the nascent proteins, would themselves be involved in their formation. To address this question, we inhibited RUVBL1/RUVBL2 ATPase activity with CB-6644 and analyzed POLR2A mRNA localization (Fig. 8A, B). Interestingly, this led to a diminution of the number of POLR2A mRNA foci after 4 h and 8 h of treatment, and their complete disappearance after 16 h. The same effect was observed for PRPF8 mRNAs and, remarkably, the mRNA foci also disappeared when the ATPase activity of HSP90 was inhibited with Geldanamycin for 1 h (Fig. 8C, D, E). Thus, the HSP90 and RUVBL1/RUVBL2 chaperones play an essential role in the formation of the POLR2A and PRPF8 translation factories, indicating that they not only interact with the client proteins but also organize their localized translation.

**Fig. 8: Inhibition of the ATPase activity of HSP90 and RUVBL1/RUVBL2 leads to the disassembly of the POLR2A and PRPF8 mRNA foci.**

Interestingly, the previous RIP-seq experiments raised the possibility that the R2TP chaperone could be involved in another case of local translation, as RPAP3-GFP pulled down several mRNAs localized near the centrosomes: NIN, PCNT and AKAP9 (Fig. 3A). Moreover, other centrosomal mRNAs were among the top hits in the RPAP3 RIP-seq (Supplementary Data 1), although their enrichment was not individually statistically significant: CEP350 (ranked 321 out of ~18000; Supplementary Data 1), CCDC88C (534), BICD2 (1111), NUMA1 (1415) and ASPM (1673). In fact, out of the 8 mRNAs known to localize to human centrosomes in a translation-dependent manner⁵⁹, all but one were found in the top hits, representing a very significant enrichment for the entire mRNA family (pValue < 0.001). To test whether R2TP plays a role in the local translation of centrosomal mRNAs, we measured by smFISH the centrosomal localization of NIN mRNAs with and without CB-6644 treatment (Fig. 9A, B). We found a nearly 2-fold decrease in the amount of NIN mRNAs at centrosomes upon treatment, indicating a role of the RUVBL1/RUVBL2 ATPases in its localized translation. Altogether, these data demonstrated an important role of the R2TP chaperone in the local translation of its mRNA targets.

**Fig. 9: Inhibition of the R2TP ATPase activity prevents the centrosomal localization of NIN mRNAs.**

Discussion

Assembly of protein complexes has recently been shown to involve co-translational steps⁴⁷. One free, fully translated subunit may bind another subunit while it is translated (co-post pathway), or two mRNAs encoding partner subunits may associate to be co-translated, with the subunits being co-assembled during their translation (co-co pathway). These questions of co-translational association have generally not been investigated in the context of chaperone-mediated assembly. Here, we studied the HSP90/R2TP chaperone, and surprisingly, we found little evidence for the co-co pathway. Instead, the fate of client subunits appears to be determined co-translationally by the loading of the R2TP chaperone in dedicated factories. Remarkably, the HSP90 and R2TP chaperones both play essential roles in the formation of these factories, indicating that they previously unsuspectedly organize the translational apparatus in the cytoplasm.

The R2TP assembly chaperone binds many of its clients in a co-translational manner

Our RIP-seq experiments using RUVBL1-GFP and RPAP3-GFP knock-in cell lines show that the HSP90/R2TP chaperone binds its clients while they are being translated. We found that ~150 mRNAs are bound by R2TP in a translation-dependent manner, but this likely underestimates the number because some mRNAs coding for known R2TP clients ranked among the top IPs, although they did not pass the statistical threshold. Overall, this indicates that the co-translational binding of R2TP to its clients is widespread. This includes well-known clients, such as subunits of the RNA polymerases or the U5 snRNP, as well as factors previously known to associate with RUVBL1/RUVBL2 but not as R2TP clients, such as subunits of the chromatin remodeling complexes INO80, SRCAP, and TIP60. The co-translational interactions of R2TP also reveal many partners, and this is quite remarkable as RUVBL1/RUVBL2 have been heavily studied, including by a number of large-scale interactomic approaches^22,23,24,25. This indicates that co-translational interactions represent a part of the interactome that is difficult to access by more traditional methods.

It is noteworthy that RUVBL1/RUVBL2 not only associate with polysomes but also cross-link with polyA+ mRNAs⁶⁴, raising the possibility that they may be atypical RNA-binding proteins. However, the fact that translation inhibition disrupts most of their interactions with mRNAs suggests that their cross-links with polyA+ RNAs may be due to a proximity effect upon nascent protein binding, rather than a traditional sequence-specific interaction. Interestingly, some RNAs remained associated with R2TP in the absence of translation. This was the case for a few mRNAs, although their binding was generally weak, and for many non-coding RNAs such as the C/D snoRNAs and the telomerase RNA. Remarkably, inhibition of the ATPase activity of RUVBL1/RUVBL2 strengthened their binding to C/D snoRNAs and to the core C/D protein SNU13, indicating that ATP hydrolysis is required for their release from immature C/D snoRNPs. Their dissociation could trigger the release of additional assembly factors, similar to the role of R2TP in releasing SHQ1 from DKC1¹¹, and this could drive the final maturation of C/D snoRNPs.

In contrast to C/D snoRNAs, inhibition of the ATPase activity of RUVBL1/RUVBL2 dissociated them from most mRNAs, with some exceptions such as the WASHC2A/WASHC2C mRNAs and the PCF11/SCAF family. The fact that RUVBL1 associates much less with a series of client mRNAs when ATP-locked, while at the same time it binds more strongly to the corresponding proteins (POLR2A, POLR2B, PRPF8, SMG1, DKC1, NOP58; Fig. 2C), suggests a cycle in which RUVBL1/RUVBL2 are loaded co-translationally on their client in an apo state and remain associated post-translationally in an ATP state, probably as part of a partially assembled intermediate containing additional subunits and cofactors. Possibly, ATP hydrolysis may be required here to remodel these assembly intermediates and release RUVBL1/RUVBL2 and the other assembly factors. The ATP cycle of RUVBL1/RUVBL2 thus regulates cotranslational client loading and its release upon completion of assembly, thereby coordinating co- and post-translational steps.

The co- and post-translational interactomes of R2TP are distinct

A comparison of the RIP-seq of RUVBL1-GFP with the corresponding proteomic data identifies three classes of partners that bind RUVBL1: (i) only co-translationally, (ii) only post-translationally, or (iii) both co- and post-translationally. Only a minority of partners is bound both co- and post-translationally, indicating that the co-translational interactome of R2TP is distinct and quite specific, thus revealing a facet of this important chaperone. Post-translational partners include proteins that bind indirectly to RUVBL1, such as some small subunits of the RNA polymerases or the chromatin remodeling complexes, and cofactors involved in assisting assembly of key client complexes, such as ECD/TSSC4 for U5 snRNP, SHQ1 for H/ACA snoRNPs and ZNHIT3/ZNHIT6/NUFIP1/NOPCHAP1 for C/D snoRNPs. Possibly, some of these assisting factors may themselves bind co-translationally to their client proteins, chaperoning them and facilitating R2TP loading as shown for NOPCHAP1⁴⁴. Proteins that require to be fully folded to associate with RUVBL1 should also bind post-translationally. This may be the case for PIKKs because, while R2TP is involved in their biogenesis^19,20, we did not observe co-translational interactions of RPAP3 or RUVBL1 with their mRNAs, with the exception of SMG1. The PIKKs have been shown in yeast to recruit the TTT complex co-translationally, to enable folding of the C-terminal hydrophobic amino-acids back into the protein core⁶⁵. The TTT recruits R2TP via a phospho-dependent interaction of TELO2 with PIH1D1⁴³, but this may occur post-translationally after the complete folding of the PIKKs, as suggested for their other partners⁶⁵.

Proteins bound both co- and post-translationally by R2TP include key subunits of its best characterized client complexes, such as the largest subunits of the nuclear RNA polymerases and key components of the U5 snRNP. Although the INO80, SRCAP, and TIP60 chromatin remodeling complexes are not usually classified as R2TP clients, some of their key subunits also bind co-translationally to R2TP. All these proteins are large subunits of nuclear complexes that, for U5 snRNP and RNA polymerase II, require many assembly factors²⁹. These subunits are thus likely accompanied by the chaperone during a large part of their biogenesis, and in agreement, chaperone deficiencies induce their retention in the cytoplasm^17,66. Surprisingly, we found a number of R2TP partners that are only detected in the co-translational dataset. They often constitute protein families, as for instance the AGO proteins, the PCF11/SCAF family, and the CHD chromatin remodelers. This suggests a very early chaperone function of R2TP for these clients, perhaps to deposit a partner co-translationally or to help folding the protein. In this regard, it is interesting to note that the AGO proteins have hydrophobic C-terminal residues that fold back in the protein core much like the PIKKs⁶⁷. As previously proposed⁶⁵, co-translational chaperone binding may be especially important to maintain these proteins in a folding competent state until the completion of translation.

The HSP90/R2TP chaperone assembles translation factories and regulates local translation

The assembly of proteins may be favored by the physical association of their polysomes⁴⁷ and the so-called co-co pathway has recently been proposed to be widespread⁵². R2TP is well known to function in the assembly of protein complexes, and we found here that it associates co-translationally with pairs of subunits that assemble together, such as POLR2A/POLR2B, PRPF8/EFTUD2, and EP400/BRD8. Surprisingly, we found only little colocalization between these mRNAs, indicating that the co-co pathway is marginal. Instead, we found that POLR2A and PRPF8 mRNAs concentrate in cytoplasmic foci that also accumulate R2TP, which we thus named R2TP factories. These foci require ongoing translation and, remarkably, also the ATPase activity of HSP90 and R2TP. Polysomes have multiple nascent protein chains and RUVBL1/RUVBL2 form heteromeric hexamers, thus enabling multivalent interactions that could drive condensate formation. Moreover, RPAP3 can bind multiple times to RUVBL1/RUVBL2 and its N-terminal domain can also multimerize³⁷, enhancing the valency of the system. Because HSP90 directly binds RPAP3 and client proteins, the multivalency of R2TP provides a possible explanation for the role of HSP90 in the condensation of these polysomes. It is remarkable that chaperones control the spatial organization of translation.

The mRNA foci induced by HSP90 and R2TP are heterotypic, as they can contain both POLR2A and PRPF8 polysomes. POLR2A and PRPF8 are both large proteins (>210 kDa) that must be assembled into large macromolecular complexes, RNA polymerase II and U5 snRNP, respectively. Even though these two proteins are not part of the same complex, they need to be taken in charge and stabilized by the same chaperone, HSP90/R2TP, as soon as they come out of the ribosome. Indeed, we have previously shown that the cytoplasmic unassembled POLR2A becomes unstable upon HSP90 or R2TP inhibition¹⁷. Likewise, PRPF8 is rapidly degraded in the absence of its co-chaperones⁶⁸. The risk for these large proteins, if they are not properly taken in charge by their chaperones, is that they get misfolded, interact with the wrong partners, aggregate, or become unstable. Chaperone factories have a high local concentration of R2TP and may thus favor their co-translational interaction with the chaperone, likely preventing these unwanted events. It is unclear why the other mRNAs that also associate with R2TP are absent from the R2TP factories. It may be that POLR2A and PRPF8 have particularities favoring polysome condensation, such as a stronger or higher valency of binding to HSP90 and R2TP, or a controlled ribosomal pausing providing more time for binding the chaperones and condensing the polysomes.

Importantly, the R2TP factories reported here are present in native cells and they do not contain mRNA pairs coding for subunits of the same complex, ruling out that they function in the co-co assembly pathway. Rather, they appear to channel assembly by concentrating the nascent protein chains and chaperones, promoting co-translational interactions. This provides an alternative model of assembly, which we termed channeling by chaperone (cha-cha). In this model, the joining of subunits is determined co-translationally by the loading of the appropriate assembly chaperones, enabling assembly itself to occur elsewhere (Fig. 9C).

Methods

Cell culture and drug treatments

HCT116 cells CCL247 are from ATCC source, HeLa H9 cells were a gift from E. Emiliani and HeLa Kyoto a gift from I. Poser⁶⁹. HCT116 cells and HeLa cells were grown in McCoy’s 5 A medium (Gibco) or Dulbecco’s modified Eagle’s Medium DMEM (Gibco), respectively, containing 10% fetal bovine serum (FBS, Sigma-Aldrich) and 1% penicillin/streptomycin (Gibco) at 37 °C, 5% CO₂. All cells were routinely tested for the absence of mycoplasma. For cell treatments, drugs were used as follows: puromycin at 100 µg/mL for 30 min (Invivogen), CB-6644 at 500 nM for 4 to 16 h (MedChemExpress), geldanamycin at 2 µM for 1 h (Invivogen), cycloheximide at 200 µg/mL for 20 min (Thermo Scientific), and harringtonine at 3 µg/mL for 30 min (MedChemExpress).

Cloning, plasmids and transfection

Plasmids were constructed using the Gateway technology (Invitrogen) with donor vectors from the human ORFeome⁷⁰. Detailed maps are available upon request. DsRed2-DDX6fm was designed and used as described before¹⁴. Stable cell lines expressing DsRed2-DDX6fm-RUVBL1 or DsRed2-DDX6fm-RPAP3 were obtained using the Flp-In system (Thermo-Fischer Scientific) with hygromycin selection (50 mg/mL). Transient transfection was performed for the DsRed2-DDX6fm-KPNA3 construct. All transfections were performed using JetPrime according to the manufacturer's instructions: 200µL JetPrime Buffer (Polyplus), 2 µg vector DNA, and 4 µL Jetprime reage nt (Polyplus) for cells cultured in 2 mL in 6-well plates. Unless indicated otherwise, oligos were provided by Integrated DNA Technologies (IDT).

Generation of stable cell lines using PiggyBac transposon system

POLR2A and EFTUD2 cDNA WT or containing a STOP were fused at the C-terminus of a mCherry tag in the pDest-PiggyBac vector under the control of the UbC promoter. The constructs were transfected in the HCT116-RUVBL1-GFP cell line. 0.3 μg of each plasmid, alongside 1.7 μg of a construct expressing the transposase⁷¹ were transfected overnight with JetPrime (Polyplus). The day after, the transfection medium was replaced by fresh medium. Cells positively expressing mCherry were sorted using MelodyFACS. mCherry-POLR2A protein expression and localization were further verified by fluorescent microscopy and Western blots.

Insertion of GFP and SunTag cassettes by CRISPR/Cas9 knock-in

HCT116 cells were CRISPR/Cas9 edited to insert an mAID-GFP-IRES-Neomycin cassette in fusion to the C-termini of RPAP3 and RUVBL1. The recombination cassettes contained 500 bases of homology arms flanking a cassette composed of mAID-GFP, a stop codon, an IRES containing a start codon fused to a Neomycin resistance gene, and a stop codon. HeLa Flp-In cells stably expressing the SunTag scFv-sfGFP were CRISPR/Cas9 edited to insert a SunTag*32 cassette at the C-terminus of RPAP3. Cells were selected on 800 µg/mL G418 neomycin for a few weeks. Individual clones were then picked and analyzed for GFP expression by fluorescence microscopy, Western blotting and PCR genotyping. The sequences targeted by the guide RNAs were (PAM sequences are lowercase): TCAGCAGTAAGAGACTCCCCagg for RUVBL1 guide 1, ACTTCATGTACTTATCCTGCtgg for RUVBL1 guide 2; GAGCTGAAAAAGAGGTATGGtgg for the RPAP3 guide. PCR genotyping was conducted using Phusion polymerase (New England Biolabs) and genomic DNA prepared with Wizard® Genomic DNA Purification Kit (Promega). The sequences of the oligonucleotides were as follows (5’ to 3’): Fw1 in the gene before the homology arms: GGTGAAGTCCCCATGTCCTG for RUVBL1 and TGTCGACCGATCAGTGGTAGTAA for RPAP3; Rev1 in the mAID-GFP cassette: CGGTGCTCCGTCCATTGATA; Rev2 in the SunTag*32 cassette: ACTAGTGACCGGTCCTCTAGA; Fw2 in the repair cassettes: GTTCATAAACGCGGGGTTCG; Rev2 in the gene after the homology arm: CGCAGCTCTCATCCTACTGG for RUVBL1 and ACGCTACACTCACATCCAACA for RPAP3.

Western blots

Proteins were resuspended in Laemmli 1X, boiled for 5 min, separated by 10% SDS-PAGE, and transferred to a nitrocellulose membrane (Protran). Membranes were blocked with 5% non-fat milk (weight/volume) in TBST (TBS with 0.5% Tween-20) and incubated in 1% non-fat milk in TBST with the primary antibodies overnight at 4 °C followed by three times 5 min washes in TBST at RT, and incubation with secondary antibodies conjugated with horseradish peroxidase (HRP) for 1 h at RT, and finally washed three times 10 min in TBST at RT. Enzymatic activity was detected using the SuperSignal West Pico/Femto Chemiluminescent Substrate (Genetex) or Immobilon (Roche). The primary antibodies used were: anti-RPAP3 (Sigma SAB1411438-100UG, 1/1000 dilution), anti-RUVBL1 (Proteintech 10210-2-AP, 1/500 dilution), anti-GAPDH (Abcam 8245, 1/20000 dilution), anti-GFP (Roche 11814460001, 1/1000 dilution), anti-mCherry (abcam ab213511, 1/1000), anti-BRD8 (Proteintech 10476-1-AP,1/1000 dilution). Secondary antibodies were anti-rabbit (Abcam ab6721, dilution 1/10000) and anti-mouse (Abcam ab6728, dilution 1/5000) coupled to HRP.

Immunofluorescence

2.10⁵ to 3.10⁵ cells were grown on Ø12mm glass coverslips, treated as indicated and after 48 h washed with PBS and fixed with 4% paraformaldehyde in PBS for 20 min at RT then washed with PBS. Permeabilization was done in 0.1% Triton X-100 (Thermo-scientific) in PBS for 10 min at RT. After 3 PBS washes, coverslips were incubated in PBS/BSA 1% or PBS/FBS 1% with primary antibodies for 1h at RT, washed three times in PBS, incubated with secondary antibody for 1h at RT, and washed three times again in PBS. Primary antibodies used were anti-POLR2A (PB-7C2, Euromedex, 1/100 dilution), anti-RPAP3 (Sigma SAB1411438-100UG, 1/100 dilution), anti-HUR (Santa Cruz Biotechnology sc-5261, 1/100 dilution) and anti-HSP90α/β (Santa Cruz Biotechnology sc-13119, 1/100 dilution) and anti-p70 S6 Kinase (Santa Cruz Biotechnology, sc-8418, 1/100 dilution). Secondary antibodies were: anti-mouse Cy3 diluted 1/1000 (#115-165-003); anti-rabbit Cy3 diluted 1/1000 (#711-166-152); anti-mouse FITC diluted 1/200 (#115-546-006); anti-rabbit FITC diluted 1/200 (#111-096-047) anti-mouse Cy5 diluted 1/200 (#115-176-003) and anti-rabbit Cy5 diluted 1/200 (#111-175-144). Coverslips were mounted in Vectashield medium containing DAPI (Vector Laboratories).

Single molecule fluorescent in situ hybridization (smFISH) with DNA probes

Cells were grown on glass coverslips or glass-bottom 96-well plates (Greiner 655892), fixed with 4% paraformaldehyde in PBS for 20 min at RT and permeabilized with 70% ethanol overnight at 4 °C. To detect a specific mRNA, 24–48 unlabeled primary DNA probes complementary to the RNA of interest were used (Supplementary Data 3). In addition to hybridizing to their targets, these probes contained a FLAP sequence that was pre-hybridized to a complementary secondary probe coupled to a fluorophore, as described previously⁷¹. Pre-hybridization was done in NEBuffer 3 (NEB, 10 mM NaCl, 5 mM Tris-HCl, 1 mM MgCl₂, pH 7.9) according to the ratios indicated in Supplementary Data 3 and using a thermocycler with the following program: 85 °C for 3 min, 65 °C for 3 min, and 25 °C for 5 min. Cells on coverslips were first incubated in 1x SSC, 5 M urea (Sigma-Aldrich) for 15 min at RT, and then incubated with the hybridization mixture containing the probe duplexes (2 µl of pre-hybridization mix per 100 µl of final volume), with 1X SSC, 0.34 mg/mL tRNA, 5 M urea, 2 mM VRC, 0.2 mg/mL RNAse-free BSA, 10% Dextran sulfate, overnight in a wet chamber at 37 °C. The day after, the samples were washed three times for 30 min in the wash buffer (1xSSC, 5 M urea) at 37 °C and rinsed twice in PBS at RT. The 96-plates were incubated in the dark with 5 µg/mL DAPI for 40 min at RT, then mounted in Vectashield medium without DAPI (Vector Laboratories, Inc.).

SmFISH in 96-well plates with RNA probes

Cells were grown in a 96-well plate with glass bottom (Greiner 655892), fixed with 4% paraformaldehyde in PBS at RT for 20 min and permeabilized with 70% ethanol overnight at 4 °C. Probe sets of DNA oligonucleotides (GenScript) targeting the mRNAs of interest were generated based on the Oligostan script⁷². Each oligo is composed of a specific hybridization sequence flanked by 2 readout sequences and a barcode shared among the gene probeset (Supplementary Data 3). Gene-specific probes were amplified by PCR with the gene-specific barcodes before being transcribed in vitro to obtain transcript-specific primary RNA probesets, as described previously⁷³. For each mRNA of interest, 50 ng of primary RNA probes and 25 ng of each fluorescent secondary DNA probes (TYE-650-labeled LNA/DNA oligonucleotides, Qiagen)⁷³ targeting the readout sequences were pre-hybridized in 100 µl of a solution containing 7.5 M urea (Sigma-Aldrich), 0.34 mg/ml tRNA (Sigma-Aldrich) and 10% Dextran sulfate. Pre-hybridization was performed in a thermocycler, with the following program: 90 °C for 3 min, 53 °C for 15 min. Cells in 96-well plates were washed with PBS and incubated for 15min at RT in hybridization buffer (1XSSC, 7.5 M urea) before adding 100 µL of the hybridization mixture to each well. Hybridization was performed at 48 °C overnight. The next day, the plate was washed 8 times for 20 min each in 1XSSC, and 7.5 M urea at 48 °C. Finally, cells were washed twice with PBS at RT for 10 min each, stained with 5 µg/mL DAPI, and mounted in 90% glycerol (VWR), 1 mg/mL p-Phenylenediamine (Sigma-Aldrich), in PBS pH 8.

Image acquisition

Coverslip images were acquired using a 63x oil immersion objective (1.4 NA, PL Apo, oil) on a Leica DM6B microscope. DAPI, GFP, CY3, and CY5 channels were imaged using the following filter combinations: 405/60, 455, 470/ 40; 470/40, 495, 525/50; 545/30, 570, 610/75; 620/60, 660, 700/75. Alternatively, coverslips were imaged using a 63X objective (Plan Apochromat, 1.4 NA, oil) on a Zeiss Axioimager Z2 widefield microscope. DAPI, GFP, CY3, and CY5 channels were then imaged using the following filter combinations: <365, FT 395, 445/50; 470/40, FT 495, 525/50; 545/25, 570, 605/70; 610/30, 660, 640/50. Images were taken as Z-stacks with one plane every 0.24 µm (Zeiss Z2) or every 0.376 µm (Leica Thunder). 96-well plates were imaged on an Opera Phenix High-Content Screening System (PerkinElmer), with a 63× water-immersion objective (NA 1.15). Three-dimensional images were acquired, consisting of 18 slices with a spacing of 0.6 µm. RPS6 and HSP90 staining were acquired using a Zeiss LSM980 Airyscan confocal microscope equipped with a 63X Plan Apo oil 1.4NA objective.

Image analysis

Image quantifications were based on an image analysis pipeline that included cell segmentation, threshold-based spot detection, dense region deconvolution, and a Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. Firstly, nuclei and cytoplasm segmentation were inferred in 2D on image stack mean projection using Cellpose neural network⁷⁴. Secondly, single molecules were detected as 3D spots using the Big-FISH package⁷⁵, which relies on an intensity Laplacian of Gaussian (LoG) filter followed by a local maxima filter and a user-adjusted threshold. During analysis, we ensured that the same thresholds were applied to images made for the same RNA, for all cell treatments. As local maxima detection can lead to a misestimation of single molecule number in bright regions where individual molecules cannot be distinguished, a dense region deconvolution step was added. It consists of fitting the median detected spot with a 3D-gaussian to construct a reference spot and then reconstructing the bright region signal using multiple reference spots, thus providing the number of single RNAs in bright regions. Thirdly, a DBSCAN was used to assign spots to clusters (foci). At this point, detected spots were divided in clustered spots (spots assigned to a cluster) or free spots. Finally, a centroid calculation was performed to assign coordinates to each cluster. This single-molecule quantification process was repeated on all smFISH or SunTag channels, enabling quantification of RNA colocalization. During analysis, filtering was used to discard aberrations: (i) cells without nucleus or with more than one nuclei; (ii) any cell whose mask laid on the border of the field of view and that was thus considered incomplete; (iii) any spot detected outside of segmented cells; (iv) any regions containing an abnormally high average of 2 spots per pixel in a sphere of 1400 nm. Distances between spots were computed using the exact Euclidian distance. Finally, spots were considered colocalizing when found closer together than a cutoff distance of 310 nm.

A pipeline named SmallFish was also developed to quantify single mRNAs and clusters, and evaluate the percentage of colocalization between clusters. It provides a ready-to-use graphical interface to combine the Python packages used for cell analysis. In the pipeline, cell segmentation (2D) was performed using Cellpose (https://github.com/MouseLand/cellpose), while spot detection was carried out using BigFish (https://github.com/fish-quant/big-fish). Clusters were considered colocalizing when found closer together than a cutoff distance of 310 nm. Boxplots graphs were constructed on Python using Seaborn package⁷⁶ and Statannotations package⁷⁷.

RNA immunoprecipitation for RIP-seq

HCT116 cells with GFP-tagged RUVBL1 or RPAP3 alleles were seeded in 15 cm plates (1 plate/condition). At 70% confluency, cells were left untreated or treated as follows: with puromycin for 30 min at 100 μg/mL; with CB-6644 for 16 h at 500 nM; with both CB-6644 for 16 h and puromycin for 30 min. All treatments were done at 37 °C. After treatments, cells were put on ice and washed twice with ice-cold PBS and then lysed with 0.5mL HNTG lysis buffer (20mM Hepes pH 7.9, 150mM NaCl, 1% Triton X-100, 10% glycerol, 1mM MgCl2, and 1mM EGTA) supplemented with a cocktail of protease inhibitors (Roche). Cells were incubated for 30 min on a tube roller at 4 °C before being centrifuged for 10 min at 16,000 × g at 4 °C. The supernatant was subjected to immunoprecipitation for 1.5 h at 4 °C using GFP-Trap agarose beads (Chromotek-gta-20). Agarose control beads (Chromotek-bab-20) were used for non-specific control IP. All experiments were done in duplicates. Beads were washed 5 times with ice-cold HNTG. RNeasy mini kit (Qiagen-74106) was used to purify RNAs. DNA contamination was eliminated by a TURBO-DNA Free kit, as recommended by the manufacturer (Invitrogen-AM1907).

RNA-sequencing and analysis

SMART-Seq Stranded Kit (Takara Bio) was used to generate cDNA libraries for high-throughput sequencing using NovaSeq 6000 Illumina. RNA and DNA libraries' quality was subjected to quality control checks periodically (or after each step) using a fragment analyzer (Kit high sensitivity NGS) and qPCR (Roche Light Cycler 480). Ribosomal cDNAs were cleaved with scZapR enzyme in the presence of scR-Probes designed against mammalian ribosomal RNAs. Libraries were then sequenced with NovaSeq Reagent Kits on a flow cell SP (single read, −100 nucleotides) using the 2-channel sequence-by-synthesis method. Before demultiplexing, sequences were mixed with PhiX sequences as an Illumina control (not indexed). Demultiplexing was realized using Illumina bcl2fastq software (v2.20.0.422), and all PhiX sequences were removed.

FastaQC software (v0.11.9) provided a tool to check read quality after high-throughput sequencing, and the MultiQC program (V1.12) served to group data that emerges from all samples. FastQ screen software served to detect contaminating genomes. Sequences were aligned on 15 species genomes, mycoplasma, and ribosomal RNAs using Bowtie2 software. No contamination was detected.

TopHat2 software (v2.1.1) was used to align the obtained reads on the human genome (hg38 version) and to annotate them according to NBCI database. FeatureCounts software was utilized to count reads per gene. All reads that overlap on two genes or show alignment at different positions on the genome were discarded. Counts were normalized in each sample to eliminate differences due to manipulation. This was done by a normalization method known as relative log expression (RLE). Differential expression genes were identified using EdgeR (v3.34.1). Tests are based on a generalized linear model and allow count modeling using a negative binomial distribution. To calculate the reads per kilobase per million mapped reads (RPKM), the formula below was used: (read counts aligned on a specific gene/ (total aligned counts x gene length) x109.

Label-free quantitative proteomics

HCT116 cells with RUVBL1 or RPAP3 endogenously tagged with GFP were seeded in 15 cm plates (1 plate/condition). At 70% confluency, cells were treated or not with CB-6644 for 16 h at 500 nM. Cells were washed with ice-cold PBS, then lysed with 0.5 mL HNTG lysis buffer supplemented with proteinase inhibitor cocktail (Roche; see above for the composition of HNTG). Cells were scraped and incubated for 15 min at 4 °C before being centrifuged for 10 min at 16,000 × g at 4 °C. The supernatant was subjected to immunoprecipitation using GFP-Trap agarose beads (Chromotek-gta-20). Agarose control beads (Chromotek-bab-20) were used for non-specific control IP. All experiments were done in triplicates. Extracts were incubated with beads for 1.5 h at 4 °C and washed 5 times with ice-cold HNTG. Proteins were then eluted in Laemmli buffer and heated for 10 min at 95 °C.

For sample digestion, proteins were loaded onto SDS-PAGE (10% polyacrylamide, Mini-PROTEAN® TGX™ Precast Gels, Bio-Rad) and stained with Protein Staining Solution (Euromedex). Gel lanes were cut into one single band and destained with 50 mM TriEthylAmmonium BiCarbonate (TEABC, Merck), followed by three washes in 100% acetonitrile. After reduction (with 10 mM dithiothreitol in 50 mM TEABC at 60 °C for 30 min) and alkylation (with 55 mM iodoacetamide in 50 mM TEABC at room temperature for 30 min), proteins were digested in-gel using trypsin (1 g/band, Gold, Promega). Digest products were vacuum-dried before LC-MS/MS analysis. For mass spectrometry, a Q Exactive HF mass spectrometer (Thermo Fisher Scientific) coupled with an Ultimate 3000 nano-HPLC (Thermo Fisher Scientific) was used. Peptide samples were solubilized in 0.05% trifluoroacetic acid (TFA) / 2% ACN, and were injected for desalting and pre-concentration on a PepMap®100 C18 precolumn (0.3 mm x 5 mm; ThermoFisher Scientific). Peptides separation were done on a 50 cm analytical reversed-phase column (75 mm inner diameter, Acclaim Pepmap 100® C18, Thermo Fisher Scientific) using a 120 min gradient of 2 to 40% of buffer B (80% ACN, 0.1% formic acid) and a flow rate of 300 nL/min. MS/MS analyses were performed in a data-dependent mode. Full scans (375-1500 m/z) were acquired on the Orbitrap mass analyzer with a 60,000 resolution at 200 m/z. For the full scans, 3e⁶ ions were accumulated within a maximum injection time of 60 ms and detected in the Orbitrap analyzer. The twelve most intense ions with charge states ≥ 2 were sequentially isolated to a target value of 1e⁵ with a maximum injection time of 100 ms and fragmented by HCD (Higher-energy collisional dissociation) in the collision cell (normalized collision energy of 28%) and detected in the Orbitrap analyzer at 30,000 resolution. First mass was set to 100 m/z.

Raw spectra were processed using the MaxQuant environment (v.2.0.3.0), and Andromeda for database search with label-free quantification (LFQ), match between runs, and the iBAQ algorithm (CIT)⁷⁸. The MS/MS spectra were matched against the UniProt Reference proteome of Homo sapiens (Proteome ID UP000005640; release 2022_03; https://www.uniprot.org/) and a homemade contaminant database. Different release versions were used, depending on the date of analysis. Enzyme specificity was set to trypsin/P, and the search included cysteine carbamidomethylation as a fixed modification and oxidation of methionine, and acetylation (protein N-term) as variable modifications. Up to two missed cleavages were allowed for protease digestion. FDR was set at 0.01 for peptides and proteins and the minimal peptide length at 7. A representative protein ID in each protein group was automatically selected using an in-house developed bioinformatics tool (Leading tool 3.4) as the best annotated protein in UniProtKB (reviewed entries rather than automatic ones⁷⁹. Analysis of quantification data, statistical analyses and graphical representations were performed using Perseus (v1.6.15). Briefly, label-free quantification (LFQ) intensity-based values were log 2 transformed. Imputation of missing values was performed using Perseus with standard parameters (normal distribution, width = 0.3, shift = 1.8). Normalized LFQ intensities were compared between replicates, using a 1% permutation-based false discovery rate (FDR) in a two-tailed Student’s t test (250 randomizations)⁸⁰.

RT-qPCR

RNA immunoprecipitation was performed as described before except that DNAse Treatment was neglected. The reverse transcription reaction (RT) was carried out by mixing 2 μL of RNA with final 2.5 μM of random primers (Thermofisher Scientific) and 0.5 mM of dNTP mix (Thermofisher Scientific). The RNA-primers mix was incubated for 5 min at 65 °C then for 1 min on ice. 100 U of SuperScript IV enzyme was added to the mix in addition to a final 1X SuperScript Buffer, 5 mM DTT (Thermofisher Scientific) and 20 U of the RNase Out recombinant RNAse inhibitor (Invitrogen). The final mix (Vf = 20 μL) was incubated for 10 min at 23 °C, 10 min at 50 °C followed by 10 min at 80 °C to inactivate the enzyme. The RT reaction was then diluted 4x by adding 60 μL of RNAse Free H2O.

The qPCR was performed using the LightCycler 480 instrument II (Roche). 2 μL of the RT reaction were mixed with a final 1X of the LightCycler 480 SYBR Green I Master (Roche) and 1 μM of RNA-specific primers (Integrated DNA Technologies or Eurofins Genomics), Vf =10 μl. Samples were loaded in a LightCycler 480 white multiwell plate 384 (Roche) and sealed well with the LightCycler 480 Sealing foil (Roche). PCR amplicon sizes range between 100 and 120 bp. The sequences of designed primers are listed in the Table 1.

Table 1 Primers used for RT-qPCR

Full size table

Sucrose gradient and Western Blot

All buffer solutions were treated with diethyl pyrocarbonate. HEK293 Trex cells were washed in chilled HKM200 buffer (10 mM Hepes pH7.6, 5 mM MgCl2, 200 mM KCl) and lysed in the same buffer supplemented with 1% NP-40, 1 μl/mL protease inhibitor cocktail (Calbiochem #535140), 40 units/ml rRNAsin (Recombinant RNasin Ribonuclease Inhibitor, Promega) and 0.5 μM DTT. Cell lysates were clarified by centrifugation 10 min at 4 °C at 16,000× g, and loaded on 11–54% sucrose gradients in 20 mM Hepes pH7.6, 100 mM KCl, 5 mM MgCl2. Ultracentrifugations were performed at 38,000 RPM during 2 h at 4 °C in a Beckmann SW41 rotor. One ml fractions were collected manually and proteins were precipitated by TCA and resuspended in SDS loading dye. For Western blots, samples were separated by electrophoresis in 4–12% Tris-glycine SDS/PAGE (Life Technologies) and were transferred onto 0.2 µ nitrocellulose membranes (Protan-BA83; GE Healthcare) using Thermofisher Transblot systems. Membranes were blocked in PBS with 10% (w/v) milk and 0.05% Tween-20 (Euromedex) before incubation with primary antibodies diluted 1:1000 in PBS 0.05% Tween overnight at 4 °C. After washing with 1XPBS/0.05% Tween-20, membranes were incubated with primary antibodies anti-RUVBL2 (Proteintech 10195-1-AP, 1/1000 dilution) or anti-RPS3 (Proteintech 66046-1-Ig, 1/1000 dilution) and secondary antibodies coupled to HRP (1/5000 dilution) and visualized using SuperSignal West Femto or Pico (Thermo Scientific) with LAS 4000 mini (GE Healthcare).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD061298. The PRIDE database resources in 2022: A Hub for mass spectrometry-based proteomics evidences⁸¹. The sequencing data have been uploaded to GEO repository: GSE291802 (Co-translational determination of quaternary structures in chaperone factories). The RIP-seq data and the proteomic data are provided as Supplementary Data. Source data are provided with this paper. Source for data analysis are provided here: https://github.com/Flo3333/Co-translational-determination-of-quaternary-structures-in-chaperone-factories Source data are provided with this paper.

References

Broeck, A. V. & Klinge, S. Eukaryotic ribosome assembly. Annu. Rev. Biochem. 93, 189–210 (2024).
Article Google Scholar
Massenet, S., Bertrand, E. & Verheggen, C. Assembly and trafficking of box C/D and H/ACA snoRNPs. RNA Biol. 14, 680–692 (2017).
Article PubMed Google Scholar
Lanfranco, M., Vassallo, N. & Cauchi, R. J. Spinal muscular atrophy: from defective chaperoning of snrnp assembly to neuromuscular dysfunction. Front Mol. Biosci. 4, 41 (2017).
Article PubMed PubMed Central Google Scholar
Garrido-Godino, A. I., Gutiérrez-Santiago, F. & Navarro, F. Biogenesis of RNA Polymerases in Yeast. Front. Mol. Biosci. 8, 669300 (2021).
Article CAS PubMed PubMed Central Google Scholar
von Morgen, P., Hořejší, Z. & Macurek, L. Substrate recognition and function of the R2TP complex in response to cellular stress. Front. Genet. 6, 69 (2015).
Google Scholar
Pal, M. et al. Structure of the TELO2_TTI1_TTI2 complex and its function in TOR recruitment to the R2TP chaperone. Cell Rep. 36, 109317 (2021).
Article CAS PubMed PubMed Central Google Scholar
Serna, M. et al. CryoEM of RUVBL1-RUVBL2-ZNHIT2, a complex that interacts with pre-mRNA-processing-splicing factor 8. Nucleic Acids Res 50, 1128–1146 (2022).
Article CAS PubMed PubMed Central Google Scholar
Lynham, J. & Houry, W. A. The Role of Hsp90-R2TP in Macromolecular Complex Assembly and Stabilization. Biomolecules 12, 1045 (2022).
Article CAS PubMed PubMed Central Google Scholar
Boulon, S. et al. The Hsp90 chaperone controls the biogenesis of L7Ae RNPs through conserved machinery. J. Cell Biol. 180, 579–595 (2008).
Article CAS PubMed PubMed Central Google Scholar
Zhao, R. et al. Molecular chaperone Hsp90 stabilizes Pih1/Nop17 to maintain R2TP complex activity that regulates snoRNA accumulation. J. Cell Biol. 180, 563–578 (2008).
Article CAS PubMed PubMed Central Google Scholar
Machado-Pinilla, R., Liger, D., Leulliot, N. & Meier, U. T. Mechanism of the AAA+ ATPases pontin and reptin in the biogenesis of H/ACA RNPs. RNA 18, 1833–1845 (2012).
Article CAS PubMed PubMed Central Google Scholar
Venteicher, A. S., Meng, Z., Mason, P. J., Veenstra, T. D. & Artandi, S. E. Identification of ATPases pontin and reptin as telomerase components essential for holoenzyme assembly. Cell 132, 945–957 (2008).
Article CAS PubMed PubMed Central Google Scholar
Bizarro, J. et al. NUFIP and the HSP90/R2TP chaperone bind the SMN complex and facilitate assembly of U4-specific proteins. Nucleic Acids Res 43, 8973–8989 (2015).
Article CAS PubMed PubMed Central Google Scholar
Malinová, A. et al. Assembly of the U5 snRNP component PRPF8 is controlled by the HSP90/R2TP chaperones. J. Cell Biol. 216, 1579 (2017).
Article PubMed PubMed Central Google Scholar
Cloutier, P. et al. R2TP/Prefoldin-like component RUVBL1/RUVBL2 directly interacts with ZNHIT2 to regulate assembly of U5 small nuclear ribonucleoprotein. Nat. Commun. 8, 1–14 (2017).
Article Google Scholar
Abel, Y. et al. The interaction between RPAP3 and TRBP reveals a possible involvement of the HSP90/R2TP chaperone complex in the regulation of miRNA activity. Nucleic Acids Res 50, 2172–2189 (2022).
Article CAS PubMed PubMed Central Google Scholar
Boulon, S. et al. HSP90 and its R2TP/Prefoldin-like cochaperone are involved in the cytoplasmic assembly of RNA polymerase II. Mol. Cell 39, 912–924 (2010).
Article CAS PubMed Google Scholar
Zur Lage, P. et al. Ciliary dynein motor preassembly is regulated by Wdr92 in association with HSP90 co-chaperone, R2TP. J. Cell Biol. 217, 2583–2598 (2018).
Article PubMed PubMed Central Google Scholar
Horejsí, Z. et al. CK2 phospho-dependent binding of R2TP complex to TEL2 is essential for mTOR and SMG1 stability. Mol. Cell 39, 839–850 (2010).
Article PubMed Google Scholar
Takai, H., Xie, Y., de Lange, T. & Pavletich, N. P. Tel2 structure and function in the Hsp90-dependent maturation of mTOR and ATR complexes. Genes Dev. 24, 2019–2030 (2010).
Article CAS PubMed PubMed Central Google Scholar
Abéza, C. et al. The HSP90/R2TP quaternary chaperone scaffolds assembly of the TSC complex. J. Mol. Biol. 36, 168840 (2024).
Article Google Scholar
Sardiu, M. E. et al. Probabilistic assembly of human protein interaction networks from label-free quantitative proteomics. Proc. Natl. Acad. Sci. USA 105, 1454–1459 (2008).
Article ADS CAS PubMed PubMed Central Google Scholar
Cloutier, P. et al. High-resolution mapping of the protein interaction network for the human transcription machinery and affinity purification of RNA polymerase II-associated complexes. Methods 48, 381–386 (2009).
Article CAS PubMed PubMed Central Google Scholar
Seraphim, T. V. et al. Assembly principles of the human R2TP chaperone complex reveal the presence of R2T and R2P complexes. Structure 30, 156–171.e12 (2022).
Article CAS PubMed Google Scholar
Bizarro, J. et al. Proteomic and 3D structure analyses highlight the C/D box snoRNP assembly mechanism and its control. J. Cell Biol. 207, 463–480 (2014).
Article CAS PubMed PubMed Central Google Scholar
Maurizy, C. et al. The HSP90/R2TP assembly chaperone promotes cell proliferation in the intestinal epithelium. Nat. Commun. 12, 4810 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Maurizy, C. et al. The RPAP3-Cterminal domain identifies R2TP-like quaternary chaperones. Nat. Commun. 9, 1–16 (2018).
Article ADS CAS Google Scholar
Zhao, R. et al. Navigating the chaperone network: an integrative map of physical and genetic interactions mediated by the Hsp90 chaperone. Cell 120, 715–727 (2005).
Article ADS CAS PubMed Google Scholar
Houry, W. A., Bertrand, E. & Coulombe, B. The PAQosome, an R2TP-based chaperone for quaternary structure formation. Trends Biochem. Sci. 43, 4–9 (2018).
Article CAS PubMed Google Scholar
Muñoz-Hernández, H. et al. Structural mechanism for regulation of the AAA-ATPases RUVBL1-RUVBL2 in the R2TP co-chaperone revealed by cryo-EM. Sci. Adv. 5, eaaw1616 (2019).
Article ADS PubMed PubMed Central Google Scholar
Gorynia, S. et al. Structural and functional insights into a dodecameric molecular machine - the RuvBL1/RuvBL2 complex. J. Struct. Biol. 176, 279–291 (2011).
Article CAS PubMed Google Scholar
Zhou, C. Y. et al. Regulation of Rvb1/Rvb2 by a domain within the INO80 chromatin remodeling complex implicates the yeast rvbs as protein assembly chaperones. Cell Rep. 19, 2033–2044 (2017).
Article CAS PubMed PubMed Central Google Scholar
Martino, F. et al. RPAP3 provides a flexible scaffold for coupling HSP90 to the human R2TP co-chaperone complex. Nat. Commun. 9, 1–13 (2018).
ADS Google Scholar
Silva, S. T. N. et al. X-ray structure of full-length human RuvB-Like 2 - mechanistic insights into coupling between ATP binding and mechanical action. Sci. Rep. 8, 13726 (2018).
Article ADS PubMed PubMed Central Google Scholar
McKeegan, K. S., Debieux, C. M., Boulon, S., Bertrand, E. & Watkins, N. J. A dynamic scaffold of pre-snoRNP factors facilitates human box C/D snoRNP assembly. Mol. Cell. Biol. 27, 6782–6793 (2007).
Article CAS PubMed PubMed Central Google Scholar
Dauden, M. I., López-Perrote, A. & Llorca, O. RUVBL1–RUVBL2 AAA-ATPase: a versatile scaffold for multiple complexes and functions. Curr. Opin. Struct. Biol. 67, 78–85 (2021).
Article CAS PubMed Google Scholar
Henri, J. et al. Deep structural analysis of RPAP3 and PIH1D1, two components of the HSP90 Co-chaperone R2TP complex. Structure 26, 1196–1209.e8 (2018).
Article CAS PubMed Google Scholar
Benbahouche, N. E. H. et al. Drosophila Spag is the homolog of RNA polymerase II-associated protein 3 (RPAP3) and recruits the heat shock proteins 70 and 90 (Hsp70 and Hsp90) during the assembly of cellular machineries. J. Biol. Chem. 289, 6236–6247 (2014).
Article PubMed PubMed Central Google Scholar
Kakihara, Y., Makhnevych, T., Zhao, L., Tang, W. & Houry, W. A. Nutritional status modulates box C/D snoRNP biogenesis by regulated subcellular relocalization of the R2TP complex. Genome Biol. 15, 404 (2014).
Article PubMed PubMed Central Google Scholar
Yu, G. et al. Yeast R2TP interacts with extended termini of client protein Nop58p. Sci. Rep. 9, 20228 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Takai, H., Wang, R. C., Takai, K. K., Yang, H. & de Lange, T. Tel2 regulates the stability of PI3K-related protein kinases. Cell 131, 1248–1259 (2007).
Article CAS PubMed Google Scholar
Goto, G. H. et al. Two separate pathways regulate protein stability of ATM/ATR-related protein kinases Mec1 and Tel1 in budding yeast. PLOS Genet. 13, e1006873 (2017).
Article PubMed PubMed Central Google Scholar
Hořejší, Z. et al. Phosphorylation-dependent PIH1D1 interactions define substrate specificity of the R2TP cochaperone complex. Cell Rep. 7, 19–26 (2014).
Article PubMed PubMed Central Google Scholar
Abel, Y. et al. NOPCHAP1 is a PAQosome cofactor that helps loading NOP58 on RUVBL1/2 during box C/D snoRNP biogenesis. Nucleic Acids Res 49, 1094–1113 (2021).
Article CAS PubMed PubMed Central Google Scholar
Shiber, A. et al. Cotranslational assembly of protein complexes in eukaryotes revealed by ribosome profiling. Nature 561, 268–272 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Halbach, A. et al. Cotranslational assembly of the yeast SET1C histone methyltransferase complex. EMBO J. 28, 2959–2970 (2009).
Article CAS PubMed PubMed Central Google Scholar
Bernardini, A. & Tora, L. Co-translational assembly pathways of nuclear multiprotein complexes involved in the regulation of gene transcription. J. Mol. Biol. 436, 168382 (2024).
Article CAS PubMed Google Scholar
Kamenova, I. et al. Co-translational assembly of mammalian nuclear multisubunit complexes. Nat. Commun. 10, 1740 (2019).
Article ADS PubMed PubMed Central Google Scholar
Bernardini, A. et al. Hierarchical TAF1-dependent co-translational assembly of the basal transcription factor TFIID. Nat. Struct. Mol. Biol. 30, 1141–1152 (2023).
Article CAS PubMed PubMed Central Google Scholar
Yayli, G. et al. ATAC and SAGA co-activator complexes utilize co-translational assembly, but their cellular localization properties and functions are distinct. Cell Rep. 42, 113099 (2023).
Article CAS PubMed PubMed Central Google Scholar
Bertolini, M. et al. Interactions between nascent proteins translated by adjacent ribosomes drive homomer assembly. Science 371, 57–64 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Mallik, S. et al. Structural determinants of co-translational protein complex assembly. Cell 188, 764–777.e22 (2025).
Article CAS PubMed Google Scholar
Paknia, E., Chari, A., Stark, H. & Fischer, U. The Ribosome Cooperates with the Assembly Chaperone pICln to Initiate Formation of snRNPs. Cell Rep. 16, 3103–3112 (2016).
Article CAS PubMed Google Scholar
King, T. H., Decatur, W. A., Bertrand, E., Maxwell, E. S. & Fournier, M. J. A well-connected and conserved nucleoplasmic helicase is required for production of box C/D and H/ACA snoRNAs and localization of snoRNP proteins. Mol. Cell. Biol. 21, 7731–7746 (2001).
Article CAS PubMed PubMed Central Google Scholar
Watkins, N. J. et al. Assembly and maturation of the U3 snoRNP in the nucleoplasm in a large dynamic multiprotein complex. Mol. Cell 16, 789–798 (2004).
Article CAS PubMed Google Scholar
Assimon, V. A. et al. CB-6644 is a selective inhibitor of the RUVBL1/2 complex with anticancer activity. ACS Chem. Biol. 14, 236–244 (2019).
Article CAS PubMed Google Scholar
Yenerall, P. et al. RUVBL1/RUVBL2 ATPase activity drives paqosome maturation, DNA replication, and radioresistance in lung cancer. Cell Chem. Biol. 27, 105–121.e14 (2020).
Dos Santos Morais, R. et al. Deciphering cellular and molecular determinants of human DPCD protein in complex with RUVBL1/RUVBL2 AAA-ATPases. J. Mol. Biol. 434, 167760 (2022).
Article PubMed Google Scholar
Safieddine, A. et al. A choreography of centrosomal mRNAs reveals a conserved localization mechanism involving active polysome transport. Nat. Commun. 12, 1352 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Jády, B. E., Bertrand, E. & Kiss, T. Human telomerase RNA and box H/ACA scaRNAs share a common Cajal body–specific localization signal. J. Cell Biol. 164, 647–652 (2004).
Article PubMed PubMed Central Google Scholar
Kittur, N., Darzacq, X., Roy, S., Singer, R. H. & Meier, U. T. Dynamic association and localization of human H/ACA RNP proteins. RNA 12, 2057–2062 (2006).
Article CAS PubMed PubMed Central Google Scholar
Tanenbaum, M. E., Gilbert, L. A., Qi, L. S., Weissman, J. S. & Vale, R. D. A protein-tagging system for signal amplification in gene expression and fluorescence imaging. Cell 159, 635–646 (2014).
Article ADS CAS PubMed Google Scholar
Pichon, X. et al. Visualization of single endogenous polysomes reveals the dynamics of translation in live human cells. J. Cell Biol. 214, 769–781 (2016).
Article CAS PubMed PubMed Central Google Scholar
Caudron-Herger, M., Jansen, R. E., Wassmer, E. & Diederichs, S. RBP2GO: a comprehensive pan-species database on RNA-binding proteins, their interactions and functions. Nucleic Acids Res 49, D425–D436 (2021).
Article CAS PubMed PubMed Central Google Scholar
Toullec, D. et al. The Hsp90 cochaperone TTT promotes cotranslational maturation of PIKKs prior to complex assembly. Cell Rep. 37, 109867 (2021).
Article CAS PubMed Google Scholar
He, X. D. et al. Overlapping peri-implantation phenotypes of ZNHIT1 and ZNHIT2 despite distinct functions during early mouse development. Biol. Reprod. 111, 1017–1029 (2024).
Article CAS PubMed PubMed Central Google Scholar
Schirle, N. T. & MacRae, I. J. The crystal structure of human Argonaute2. Science 336, 1037–1040 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
Erkelenz, S. et al. Ecd promotes U5 snRNP maturation and Prp8 stability. Nucleic Acids Res 49, 1688–1707 (2021).
Article CAS PubMed PubMed Central Google Scholar
Poser, I. et al. BAC TransgeneOmics: a high throughput method for exploration of protein funtion in mammals. Nat. Methods 5, 409–415 (2008).
Article ADS CAS PubMed PubMed Central Google Scholar
Rual, J.-F. et al. Human ORFeome version 1.1: a platform for reverse proteomics. Genome Res 14, 2128–2135 (2004).
Article CAS PubMed PubMed Central Google Scholar
Wilson, M. H., Coates, C. J. & George, A. L. PiggyBac transposon-mediated gene transfer in human cells. Mol. Ther. 15, 139–145 (2007).
Article CAS PubMed Google Scholar
Tsanov, N. et al. smiFISH and FISH-quant – a flexible single RNA detection approach with super-resolution capability. Nucleic Acids Res 44, e165–e165 (2016).
Article PubMed PubMed Central Google Scholar
Safieddine, A. et al. HT-smFISH: a cost effective and flexible workflow for high-throughput single molecule RNA imaging. Nat. Protoc. 18, 157 (2023).
Article CAS PubMed Google Scholar
Pachitariu, M. & Stringer, C. Cellpose 2.0: how to train your own model. Nat. Methods 19, 1634–1641 (2022).
Article CAS PubMed PubMed Central Google Scholar
Imbert, A. et al. FISH-quant v2: a scalable and modular tool for smFISH image analysis. RNA 28, 786–795 (2022).
Article CAS PubMed PubMed Central Google Scholar
Waskom, M. L. seaborn: statistical data visualization. J. Open Source Softw. 6, 3021 (2021).
Article ADS Google Scholar
Charlier, F. et al. trevismd/statannotations: v0 Zenodo. https://doi.org/10.5281/zenodo.8396665 (2023).
Cox, J. et al. MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat. Biotechno. l 26, 1367–1372 (2008).
Article CAS Google Scholar
Raynaud, F. et al. SNAP23-Kif5 complex controls mGlu1 receptor trafficking. J. Mol. Cell B Biol. 10, 423–436 (2018).
Article CAS Google Scholar
Tyanova, S. et al. The Perseus computational platform for comprehensive analysis of (prote)omics data. Nat. Methods 13, 731–740 (2016).
Article CAS PubMed Google Scholar
Perez-Riverol, Y. et al. The PRIDE database resources in 2022: a hub for mass spectrometry-based proteomics evidences. Nucleic Acids Res 50, D543–D552 (2022).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank Drs. D. Helmlinger and B. Pradet-Balade for stimulating discussions and advice, and Dr. B. Delaval for the gift of the Hela centrin-GFP cell line. We acknowledge (i) the sequencing facility MGX, which obtained financial support from France Génomique National infrastructure, funded as part of the “Investissement d’Avenir” program (contract ANR-10-INBS-09); (ii) the microscopy facility MRI, member of the national infrastructure France-BioImaging supported by the French National Research Agency (ANR-10-INBS-04); and (iii) Montpellier Proteomics Platform (PPM, BioCampus Montpellier), a member of the national Proteomics French Infrastructure (ProFI UAR 2048) supported by the French National Research Agency (ANR-24-INBS-0015, Investments for the future F2030). The work was supported by grants from the ANR (ANR-23-CE12-0022), the INCa (PLBIO23), and the ‘Ligue Nationale contre le Cancer’ (équipe labellisée LIGUE 2023), all allocated to EB. SS and FS were supported by MSD Avenir.

Author information

These authors contributed equally: Manon Philippe, Soha Salloum.

Authors and Affiliations

IGH, Université de Montpellier, CNRS, Montpellier, France
Manon Philippe, Soha Salloum, Floric Slimani, Davide Normanno, Héloïse Chassé, Marie-Cécile Robert, Séverine Boulon, Céline Verheggen & Edouard Bertrand
Equipe labellisée Ligue Nationale Contre le Cancer, 2023, Montpellier, France
Manon Philippe, Soha Salloum, Floric Slimani, Davide Normanno, Héloïse Chassé, Marie-Cécile Robert, Séverine Boulon, Céline Verheggen & Edouard Bertrand
Institut de Biologie de l’Ecole Normale Supérieure (IBENS), Ecole Normale Supérieure, CNRS, INSERM, PSL Research University, Paris, France
Isabelle Barbosa & Hervé Le Hir
IGF, Université de Montpellier, CNRS, Montpellier, France
Serge Urbach & Martial Séveno
BCM, Université de Montpellier, CNRS, INSERM, Montpellier, France
Jacques Imbert & Simon George

Authors

Manon Philippe
View author publications
Search author on:PubMed Google Scholar
Soha Salloum
View author publications
Search author on:PubMed Google Scholar
Floric Slimani
View author publications
Search author on:PubMed Google Scholar
Davide Normanno
View author publications
Search author on:PubMed Google Scholar
Isabelle Barbosa
View author publications
Search author on:PubMed Google Scholar
Héloïse Chassé
View author publications
Search author on:PubMed Google Scholar
Marie-Cécile Robert
View author publications
Search author on:PubMed Google Scholar
Serge Urbach
View author publications
Search author on:PubMed Google Scholar
Jacques Imbert
View author publications
Search author on:PubMed Google Scholar
Martial Séveno
View author publications
Search author on:PubMed Google Scholar
Simon George
View author publications
Search author on:PubMed Google Scholar
Hervé Le Hir
View author publications
Search author on:PubMed Google Scholar
Séverine Boulon
View author publications
Search author on:PubMed Google Scholar
Céline Verheggen
View author publications
Search author on:PubMed Google Scholar
Edouard Bertrand
View author publications
Search author on:PubMed Google Scholar

Contributions

E.B. conceived the study, and E.B., C.V., and S.B. supervised the work. S.S. optimized the RIP-seq and performed the initial experiment of Fig. 1. MP, with the help of S.S., H.C., M.C.R., D.N., S.B., and C.V., performed the other experiments and interpreted the data. F.S. developed small-FISH and quantified the microscopy images. J.I. and S.G. performed the RNA sequencing and analyzed the data. S.U. and M.S. did the mass spectrometry analysis and processed the data. I.B. and H.L.H. performed the polysome fractionation experiment. E.B., C.V., S.B., and M.P. wrote the manuscript, which was corrected by all the authors.

Corresponding authors

Correspondence to Séverine Boulon, Céline Verheggen or Edouard Bertrand.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Ayala Shiber and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information (download PDF )

Supplementary Data 1 (download XLSX )

Supplementary Data 2 (download XLSX )

Supplementary Data 3 (download XLSX )

Reporting Summary (download PDF )

Transparent Peer Review file (download PDF )

Source data

Source data (download XLSX )

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Philippe, M., Salloum, S., Slimani, F. et al. Co-translational determination of quaternary structures in chaperone factories. Nat Commun 17, 1978 (2026). https://doi.org/10.1038/s41467-026-68687-8

Download citation

Received: 11 March 2025
Accepted: 14 January 2026
Published: 23 January 2026
Version of record: 24 February 2026
DOI: https://doi.org/10.1038/s41467-026-68687-8