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

Background: Mammalian RNA polymerase II-associated protein 3 (RPAP3) recruits heat shock protein 90 (Hsp90) to assemble cellular machineries such as RNA polymerases. Results: Spaghetti encodes the Drosophila homolog of RPAP3. Spaghetti is essential for development. Spag protein binds and stimulates Hsp90 and Hsp70. Conclusion: RPAP3 function is conserved among metazoans. Significance: Our data suggest that


Drosophila Spag Is the Homolog of RNA Polymerase IIassociated Protein 3 (RPAP3) and Recruits the Heat Shock Proteins 70 and 90 (Hsp70 and Hsp90) during the Assembly of Cellular Machineries * ࡗ
ATPases that associate in hexameric rings and are implicated in various cellular functions such as chromatin remodeling and snoRNP 8 biogenesis (2). Because of sequence similarity with the bacterial RuvB proteins, they were initially thought to be DNA helicases. However, biochemical studies have shown a very modest or no unwinding activity at all, and their molecular functions remain poorly understood. RVB1/2 associate with a number of protein complexes (2), and it is currently unclear in which case this relates to the R2TP function. In contrast, Pih1 and Tah1 activity seems to be restricted to the R2TP. In mammalian cells, R2TP contains similar proteins, namely RUVBL1, RUVBL2, RPAP3 (RNA polymerase interacting protein 3, also called Spagh, homologous to Tah1), and Pih1D1 (Pih1 domain containing 1) (3). In addition, R2TP co-purifies with a set of seven prefoldin and prefoldin-like proteins, which together form the R2TP/prefoldin-like complex (4). The role of these additional proteins remains elusive.
One of the first described role of R2TP concerns the biogenesis of the L7Ae ribonucleoparticles (RNPs). This family of structurally related RNPs includes the snoRNPs, necessary for ribosomal RNA maturation, the spliceosomal snRNA U4, and mRNP encoding selenoproteins. This function is conserved from yeast to mammals and requires a dedicated adaptor, called Rsa1 in S. cerevisiae and Nufip in mammals (3,5). Later on, we showed that the R2TP is also involved in the early cytoplasmic steps of RNA polymerase II biogenesis (6). Finally, mammalian R2TP also stabilizes proteins from the PI3 kinase-like kinase family (PIKKs), including mammalian TOR and SMG-1, two regulators of protein synthesis (7). This function in PIKK stabilization is dependent on an adaptor called Tel2 (7). In all these processes, R2TP appears to stabilize newly synthesized proteins by recruiting Hsp90 and to assemble them into macromolecular complexes by yet poorly understood mechanisms (8).
These studies reveal that mammalian R2TP plays a role in the formation of cellular machineries that are necessary for cell growth and proliferation (8). Yet RPAP3 can be knocked down in cell lines without any gross effect on cell viability (6). In S. cerevisiae, R2TP function has been mostly documented for snoRNP biogenesis (5). Surprisingly, although snoRNPs are essential, the deletion of TAH1 is viable, with no clear effect on cell growth, although that of PIH1 results in thermo-sensitivity (5). Whether R2TP plays an essential or accessory role in metazoans and whether its clients would be conserved, besides snoRNP, remain open questions.
To address the role of the R2TP in a multicellular organism, we used Drosophila melanogaster as a model system to investigate the spaghetti gene (or spag) as it encodes a protein similar to the mammalian RPAP3. Homozygosity of a P-element insertion in the spag gene produces larval lethality. In mosaic flies, it gives rise to the formation of narrow strips of mutant cells in the wings, hence the spaghetti designation for the gene (9). Therefore, it is of particular interest to determine whether the function of Spag could be similar to that of the mammalian RPAP3 and, if so, whether Spag would be part of a multimeric Hsp90 co-chaperone R2TP complex.

EXPERIMENTAL PROCEDURES
Animals-All fly stocks were maintained on a standard Drosophila medium at room temperature, and the crosses were done at 25°C. The w 1118 stock was used as a control. The l(2)k12101 mutant line derives from a large P-element screen described previously (10). All second chromosomal mutations were further balanced with [CyO] or [CyO, Dfd GFP ϩ ] (kindly provided by M. Crozatier). Deficiency Px 4 was obtained from the Bloomington Drosophila Stock Center. Isolation of viable and lethal revertants was carried out as described previously (11).
We generated three different transgenes in the CaSpeR4 P-element vector (9,12) carrying different DNA fragments from the spag locus. The transgenic P[DnaJ60] is made of a 2.7-kb segment containing the DnaJ60 gene (12); the transgenic fragment P[spag ϩ ] consists of a 4.3-kb DNA segment covering spag transcription unit located upstream from the DnaJ60 gene, and the transgenic fragment P[DnaJ60/spag ϩ ] contains a 6.5-kb DNA segment encompassing both genes ( Fig. 2A). Several transgenic lines were established for each construct. Transgenic flies expressing RNAi by the Gal4 upstream-activating sequence system were obtained from Vienna Drosophila RNAi Center and maintained at 25°C: fly strains 23896 and 103353 were used to induce RNAi against spag, and strains 21784 and 106393 were used against Nufip (13).
Protein Extracts, Immunoprecipitations, Western Blots, and Antibodies-For protein extracts, 10 snap-frozen animals (larvae or pupae) were crushed, lysed in Laemmli buffer, boiled, and centrifuged to discard cell debris and lipids. For immunoprecipitations, Schneider's S2 cells were extracted in HNTG buffer (6). Following incubation at 4°C for 10 min, extracts were centrifuged at 15,000 ϫ g at 4°C to sediment cell debris. Supernatants were collected and incubated for 1 h with agarose beads previously bound with serum or mouse monoclonal anti-Rpb1 antibody PB-7C2 (Euromedex, Souffelweyersheim, France). Bound complexes were then analyzed by Western blot.
Proteins separated by SDS-PAGE were transferred onto nylon or PVDF (small proteins) membranes, according to the size of proteins to be detected. Polyclonal antibodies against a 22-mer synthetic peptide corresponding to the C-end of Spag (CKNWPSKNPAVLDNLFKEYGVA) were raised in rabbits. Polyclonal anti-dHsp90 antibody was kindly given by Renato Paro. Proteins were detected as follows: Rpb1 detected with mouse monoclonal PB7-C2 antibody; Rpb2 with goat S20 from Santa Cruz Biotechnology; Nop58 with polyclonal antibodies generated from rabbits immunized with an KKLQEVD-SLWKEFETPEK peptide (14); p70 S6K with monoclonal antibody SC-9027 from Santa Cruz Biotechnology; phospho-Thr-398 p70 S6K with monoclonal antibody provided by Cell Signaling Technology (reference 9209); fibrillarin with monoclonal antibody 5821 from Abcam; tubulin with monoclonal 12G10 (Developmental Studies Hybridoma Bank, Iowa City, 8  IA); polyclonal anti-15.5K antibodies from Santa Cruz Biotechnology (SC-86760); polyclonal anti-P40 antibodies have been described previously (15), and anti-FLAG M2 monoclonal antibody was from Sigma. Tissue Analysis-Larvae were dissected in PBS, and tissues were fixed with 4% paraformaldehyde and mounted in Vectashield mounting medium (Vector Laboratories).
Northern Blots-RNA from animals at the indicated stage was extracted, separated on agarose gel, and transferred to nylon membrane as described (12). The Northern blots were hybridized with 32 P-labeled spag or DnaJ60 cDNA probes. 32 P-Labeled ␤-tubulin and actin cDNA probes were used as loading controls.
Isolation of Spaghetti Interactants-S2 cells were stably transformed with constructs in pMT/V5-His vector as follows: FLAG-Spag (aa 1-534), FLAG-Spag⌬TPR (lacking aa 96 -197), or full size Spag without the FLAG epitope as a control. Cells were collected by centrifugation and washed with 20 mM Tris (pH 7.5), 50 mM NaCl and lysed in a buffer containing 20 mM HEPES (pH 7.2), 50 mM KCl, 2 mM MgCl 2 , and 0.2% Nonidet P-40 supplemented with 1ϫ protease inhibitor mixture (complete, Roche Applied Science) on ice. Insoluble components were separated by sedimentation for 15 min at 10,000 ϫ g, and the supernatant was centrifuged at 4°C for 30 min at 100,000 ϫ g. The FLAG-Spag or FLAG-Spag⌬TPR complexes were then separated by affinity chromatography on anti-FLAG M2 antibody-agarose column. After washing in lysis buffer, the complexes were released with 200 g/ml FLAG peptide in low salt lysis buffer, and the proteins were separated by PAGE. The most abundant protein bands were then analyzed by high mass accuracy matrix-assisted laser desorption/ionization (MALDI) peptide mapping.
Transient transfections in S2 cells were performed using Effectene (Qiagen), according to the Drosophila RNAi Screening Center (16) experimental procedures. 3ϫFLAG-Pih1D1 was obtained by recombination of the pDON clone LD15349 in the pAFW vector (both from the Drosophila Genomics Resource Center, Bloomington, IN), using Gateway technology (Invitrogen).
The Hsp70 ATPase measurements were performed as described (17,18) using His-Spag and recombinant bovine Hsc70 and human Hsp40 available from StressGen/Enzo Life Sciences. 1 g of the indicated recombinant proteins was mixed in 10 l of reaction containing ATPase buffer. 2.5 l of ATP mix, including 50 M ATP and 2.5 Ci [␣-32 P]ATP (3000 Ci/mmol, 10 mCi/ml; PerkinElmer Life Sciences), was added and incubated for 30 min at 30°C. At the end, the reaction was loaded on a Micro-Spin G-50 column (GE Healthcare) and centrifuged for 2 min. 2-l aliquots of the G-50 spin column flowthrough (containing the proteins and the bound nucleotides) were chromatographed on Polygram Cel 300 PEI/UV thin layer plate (Macherey-Nagel) using 1 M LiCl, 1 M HCOOH buffer. Unlabeled ATP and ADP were simultaneously run as standards and identified by UV light. The dried plate was exposed to X-ray film.
The Hsp90 ATPase measurements were performed with recombinant His-tagged Drosophila Hsp90 and Spag proteins using a previously described coupled enzymatic assay (19), with modified conditions (50 mM HEPES buffer (pH 7.5), KCl 50 mM, 2 mM ATP and 5 mM MgCl 2 ). ATPase activity was evaluated using 5 M Hsp90. 20 M of Spag co-chaperone was added to the reaction, and background ATPase activity was determined by the addition of 15 M of geldanamycin. All experiments were carried out at 37°C on a UVIKON spectrophotometer. Reagents were purchased from Sigma.
The aggregation of denatured rhodanese was determined according to Ref. 20. Bovine liver rhodanese (50 M, Sigma) was denatured in 6 M guanidine HCl, 30 mM MOPS-KOH (pH 7.2), and 2 mM DTT buffer for 30 min at room temperature. 2 l of the denatured rhodanese was added to 198 l of 20 mM MOPS-KOH (pH 7.2), 50 mM KCl, 2 mM MgCl 2 , 2 mM DTT, 2.5 mM ATP buffer containing the proteins (His-Spag, bovine Hsc70, human Hsp40, and bovine serum albumin; 15 g of each) in the described combinations. Aggregation of the denatured rhodanese was determined at the indicated time points by measuring the light scattering (absorbance) at 320 nm. The measured time point values were subtracted with the corresponding to 0-min value.

Characterization of the Spaghetti Gene in D. melanogaster-
The spag gene is referred to in FlyBase as Dmel/CG13570 and encodes a protein of 534 residues (NCBI accession number

Spaghetti/RPAP3 Function Is Conserved in Drosophila
. Spag protein displays one complete set of three adjacent tetratricopeptide repeats (TPR). TPR domains are responsible for Hsp binding and typically contain a triple repetition of 34 amino acids as follows: X 5 (WLF)X 2 (LIM)(GAS)-X 2 (YLF)X 8 (ASE)X 3 (FYL)X 2 (ASL)X 4 (PKE). S. cerevisiae Tah1 possess a single TPR domain, and residues responsible for its interaction with Hsp90 are conserved in Spag (21,22). Additional residues important for the binding to the C-terminal EEVD motif in most Hsp70 and Hsp90 are also present ( Fig. 1A) (23). Spag TPR domain is preceded by a structured domain at its N terminus, and followed by a second, incomplete TPR domain and a "potential Monad-binding" domain in the C terminus (InterPro 025986). Human RPAP3 (AAH5615.1) harbors two domains of triple TPR motifs and the potential Monadbinding domain at its C terminus (24). Such a potential Monadbinding domain is absent in Tah1 (Fig. 1A) (21).
In vitro translation of spag ORF produced a single peptide with an apparent molecular mass of ϳ68 kDa corresponding to the predicted molecular mass of Spag (data not shown). Polyclonal antibodies raised against a 22-mer synthetic peptide corresponding to the C-terminal end of Spag reacted on a Western blot with a protein of ϳ70 kDa (see below for the specificity of the antibodies). In embryo extracts, Spag migrated with a slightly higher mass than expected. Treatment with potato acidic phosphatase (PAP) indicated that Spag is phosphorylated in embryos but not in larvae nor in S2 Schneider's cells (data not shown). Spag expression is at its highest level during early embryogenesis and then drops to a low level during the second half of embryogenesis to become barely detectable in successive developmental stages and in adult flies. This developmental profile partly reflects mRNA expression and suggests that spag may play a critical role during development (Fig. 1B).
The spag I(2)k12101 mutation (hereafter termed spag k12101 ) was previously identified in a screen for P-element insertion causing tissue overgrowth during larval development (10). In this line, the P-lacW-transposon is inserted 52 nucleotides downstream from the spag ORF initiation codon ( Fig. 2A-a). As a result, expression of both spag and its upstream gene DnaJ60 mRNAs is reduced, whereas the expression of the neighboring genes Dat and CG3328 is not (Fig. 2B, and data not shown) (12). Accordingly, we did not detect Spag mRNA nor protein in spag k12101 animals, also demonstrating the specificity of the antibodies (Fig. 2, B and C).

JOURNAL OF BIOLOGICAL CHEMISTRY 6239
concerns more particularly the distal and the secondary lobes, which in wild-type larvae are usually the smallest. All other tissues such as imaginal discs (Fig. 2D, panels c and f-h) and salivary glands (Fig. 2D, panel d versus panel e) are reduced in size.
Following P-element excision, we recovered numerous white flies producing homozygous viable progeny, as well as a series of 20 nonviable w Ϫ revertants with larval lethality. Molecular studies using Southern blotting and PCR analysis revealed that the majority of these revertants contained a partially excised P-element, whereas the lethal revertant line R7 displayed an interstitial chromosomal deficiency ( Fig.  2A-b). These findings indicate that the P-insert is responsible of the lethality.  The Northern blots were hybridized with 32 P-labeled spag or DnaJ60 cDNA probes. The 32 P-labeled ␤-tubulin mRNA was used as loading control. C, Western blot analysis shows a signal of ϳ70 kDa in extracts from wild type but not in homozygous null mutants spag k12101 nor in spag k12101 /spagh R7 animals. D, lymph gland (lg), but not the brain (br), is overgrown in the spagh k12101 larvae (panel b) as compared with the wild-type w 1118 (panel a). Arrow points toward a melanocytic tumor. In contrast, there is atrophy of the imaginal discs for the legs and halteres (panel c), the salivary glands (panels d and e), and the imaginal discs of the wings (panels f-h) in the spag k12101 null mutants (panels e, g, and h). This effect is more pronounced at 25°C (panel h) than 18°C (panel g). In addition, transgenes encompassing the spag locus (transformation fragments P(spag ϩ ] and P(DnaJ60/spag ϩ ] from Fig.  2A-c) rescued the lethality of animals with either P-element insertion (homozygous spag k12101 ) or deletion in the spag locus (heterozygous spag k12101 /spag R7 , spag k12101 /Px 4 , or spag R7 /Px 4 flies; see Fig. 2A-b for a representation of these deletions). In contrast, a DnaJ60 transgene alone (P(DnaJ60] from Fig. 2A-c) was unable to rescue the development of the mutant animals. These data show that DnaJ60 is dispensable, although spag is essential for the survival of the fly (12).

Spaghetti/RPAP3 Function Is Conserved in Drosophila
We then took advantage of two different strains carrying transgenes with inducible RNAi against spag (P(GD8058) and P(KK100112), hereafter referred to as spag RNAi1 and spag RNAi2 , respectively) (13). Induction of gene silencing was performed with the upstream activating sequence Gal4 system. RNAi expression driven by Gal4 under either an act5c or a tubulin promoter resulted in pupal lethality shortly after metamorphosis. These animals die later than homozygous spag k12101 , probably because gene silencing is a lengthy process, as compared with a genetic null mutation. This result further demonstrates that spag is necessary for fly development.
Identification of Spag Partners-The yeast Tah1 and mammalian RPAP3 act in the context of a multimolecular co-chaperone to assist client stabilization by Hsp90. To investigate the partners of Drosophila Spag, we isolated soluble, cytoplasmic complexes from Schneider's S2 cells expressing either FLAG-Spag (full length) or FLAG-Spag⌬TPR, in which the three TPR motifs were excised from amino acid residues 96 -197 (Fig. 1A). Following binding to an anti-FLAG M2 antibody-agarose column, complexes were eluted with the FLAG peptide, to reduce contaminants. High mass accuracy matrix-assisted laser desorption/ionization (MALDI) peptide mapping identified numerous heat shock proteins, including Hsp90 and Hsp70 isoforms (Table 1). In the case of Hsc70-3 and Hsc70-5 (both devoid of a C-terminal EEVD), in contrast to the other Hsps, more peptides were detected with Spag⌬TPR than with full-length Spag. In addition, we detected peptides for Pontin and Reptin in full-length Spag immunopurified complexes (Table 1).
We also performed a yeast two-hybrid screen using Spag protein fused to the DNA-binding domain of Gal4 as bait. We identified interacting clones corresponding to the genes CG1242 (Hsp90), CG5792, in accordance with another twohybrid screen (26), and CG13849 (Nop56) ( Table 1). Alternative splicing of CG5792 gives rise to two transcripts, one of which (CG5796-RA) produces a polypeptide of 334 amino acids, which contains a Pih1 domain (CG5792-PA or NP-609590.1). It is thereafter described as D.mel/Pih1D1, as its likely ortholog. Interaction of Spag with Hsp90 parallels the proteomic data described above, as in both cases it required the presence of the TPR domain. This was also the case for Nop56, a core protein from the box C/D snoRNPs. S. cerevisiae Nop56 also interact with Tah1 in two-hybrid assays (3). In contrast, Spag interaction with dPih1D1 was independent from the TPR domain.
Co-immunoprecipitations in S2 cells, using a 3ϫFLAGtagged dPih1D1 confirmed that it binds to endogenous Spag, Pontin, and Reptin (Fig. 3). Altogether, these data indicate that the R2TP core complex is conserved in Drosophila.
Spag Is a Co-chaperone of Hsp90-To test whether the interaction between Spag and Hsps was direct, we used GST pulldown assays. Full-length in vitro translated Spag protein, albeit not Spag⌬TPR (lacking aa 96 -197), could bind to immobilized C-terminal domain of Hsp90 (aa 535-717, encompassing the EEVD-binding motif for co-chaperones) (Fig. 4A). This finding is in accordance with the results from the proteomic and yeast

Identification of full-length Spag or Spag⌬TPR (lacking aa 96 -197) interactants by proteomic analysis or yeast two-hybrid
The number of prototypic peptides in the proteomic is indicated. ND ϭ not detected. The presence of an EEVD motif at the C terminus of the Hsps is indicated.  FEBRUARY 28, 2014 • VOLUME 289 • NUMBER 9

JOURNAL OF BIOLOGICAL CHEMISTRY 6241
two-hybrid assays. In Tah1, the TPR domain is also responsible for binding to Hsp90 (21,27).
To monitor the effect of Spag on Hsp90 function, we followed its effect on Hsp90 ATPase activity. We observed a weak stimulation in the presence of a 4-fold excess of Spag (Fig. 4B). The level of stimulation of Hsp90 ATPase by Spag appears similar to the effect reported for Tah1 on yeast Hsp82 (22,27,28).
Spag Is a Co-chaperone of Hsp70-In vitro GST pulldown assay showed that binding to Hsc70-4 was also direct and depended on the TPR domain (Fig. 5A). This is consistent with the fact that Hsc70-4 harbors a C-terminal EEVD motif, known to bind to TPR domains of co-chaperones (5,23,27). Yet, in S. cerevisiae, Tah1 does not bind directly to Ssa-1, a cytoplasmic Hsp70 isoform (28). Conversely, direct binding of Spag to Hsc70-3, which lacks a C-terminal EEVD motif, was not dependent on the Spag TPR domain, further supporting the fact that deletion of the TPR did not induce a gross disorganization of the protein (Fig. 5B). Indeed, Hsc70-3 was unable to bind Spag⌬2Ct, which lacks aa 231-490. Altogether, our results indicated that Spag contains distinct domains able to bind to distinct chaperones.
To investigate the functional consequences of Spag binding to Hsp70, we first measured the in vitro ATPase activity of Spag in the presence of commercially available Hsc70 (HspA8, homolog of Hsc70-4). Neither Spag nor Hsp40 was able to hydrolyze ATP to ADP (Fig. 5C). In contrast, Spag could significantly enhance the intrinsic ATPase activity of Hsc70, to a level comparable with that of the Hsc70-Hsp40 complex (Fig. 5C). Finally, we investigated the effect of Spag on rhodanese disaggregation. In this assay, rhodanese was first denatured to induce aggregate formation. Addition of Spag, Hsp40, or the control protein BSA alone exerted no effect on those aggregates, which were moderately dissociated by Hsc70. In contrast, combination of Hsc70 with either Hsp40 or Spag strongly enhanced rhodanese disaggregation by as much as ϳ75% (Fig. 5D). Hence, similarly to Hsp40, Spag can act as an Hsp70 co-chaperone. All these data show that Spag is also a bona fide Hsp70 co-chaperone.
Conservation of the R2TP Clients-Several substrates have been identified for R2TP. These include the box C/D snoRNPs (3,5). Experiments in mammalian cells showed that accumulation of newly synthesized 15.5K and Nop58 depend on Hsp90 activity (3). We assessed the stability of Nop58 (also called Nop5 in Drosophila) (29) and 15.5K (Hoip encoded by the hoipolloi gene or hoip (9)) in spag null mutant larvae (spag k12101 ), as well as in pupae in which spag was knocked down by the induction of RNAi under the control of Gal4 act5c driver. Western blot analysis showed a decrease in Nop58 and 15.5K content, in spag k12101 larvae and act5CϾspag RNAi pupae. Different protein turnovers in larvae and pupae could explain the differences observed for Nop58 and 15.5K destabilization in spag k12101 or act5CϾspag RNAi animals. In contrast, fibrillarin, another core protein of the C/D snoRNPs, was not affected (Fig.  6, A and B). This is coherent with results from mammals.

JOURNAL OF BIOLOGICAL CHEMISTRY
Indeed, fibrillarin is incorporated at a latter step during snoRNP assembly, and it does not seem to be a client of R2TP (3,30). Interaction between R2TP and 15.5K requires Nufip as an adaptor (3). CG4076 encodes a protein with 18% identity to mammalian Nufip. Induction of two different transgenic lines expressing RNAi against CG4076/Nufip driven by Gal4 act5c produced early pupal lethality. In these animals, as in the case of spag, there is a decrease in endogenous Nop58 and 15.5K, as compared with controls (Fig. 6C). In conclusion, Spag and Nufip are required for the stabilization of the C/D snoRNP core proteins.
A second class of clients described for mammalian R2TP are the PIKKs. Mammalian R2TP is recruited to these kinases via the adaptor Tel2, with a predominant effect on mammalian TOR and SMG1 stability (7). In Drosophila, proteomic data indicate that TOR similarly interacts with Pontin, Reptin, and Lqfr, a distant ortholog of Tel2 (31). To test whether these interactions are functionally relevant, we measured TOR activity in wild-type and Spag-deficient flies (as null mutant or transgenic RNAi). Phosphorylation of S6K is classically used as a surrogate marker of TOR activity, to circumvent the lack of antibodies to detect TOR in Drosophila. In animals where spag was inactive (as null mutant or transgenic RNAi), we detected a down-regulation of S6K phosphorylation (Fig. 7,   A and B). In contrast, phosphorylation of S6K was not affected upon Nufip inactivation (Fig. 7C). These data are consistent with a role for spag, but not Nufip, in the activity of TOR and are in accordance with a recent report in the Drosophila S2 cell line (4,31).
Finally, in mammals, RNA polymerase II is also a substrate of R2TP (4, 6, 32). Indeed, human Spag binds to neo-synthesized Rpb1, the largest subunit of RNA polymerase II, to stabilize it before its association with the other polymerase II subunits. In flies devoid of Spag (spag k12101 ), we found a moderate decrease in total Rpb1 (RPII215). However no effect was seen in spag RNAi flies (Fig. 8A). Immunoprecipitation of soluble Rpb1 in S2 cells shows a tight interaction with Spag and Hsp90 (Fig. 8B). This suggests a role of Spag in RNA polymerase II assembly rather than Rpb1 stabilization.

DISCUSSION
Conservation of the R2TP Complex-A multimeric Hsp90 cochaperone known as R2TP was first identified in S. cerevisiae (1) and later in human cells (3,4). In R2TP, Tah1/RPAP3 plays a central role, binding directly to Pih1/Pih1D1 and Hsp90 (3,5,27). In this study, we characterize spaghetti, a gene essential for Drosophila development. Spag interacts with Pontin, Reptin, and CG5792-PA, the Drosophila orthologs of the mammalian  35 S-labeled Hsc70 -3 is specifically retained on GST-Spag or GST-Spag⌬TPR but not GST-Spag⌬2Ct. C, Spag stimulates ATP hydrolysis by Hsc70. Chromatography on PEI-cellulose thin layer plate shows that [ 32 P]ATP hydrolysis into ADP by Hsc70 alone is weak and undetectable in the presence of Spag or Hsp40 alone. ATP hydrolysis by Hsc70 is strongly enhanced by the addition of recombinant Spag, Hsp40, or both. D, combination of Hsc70 with either Hsp40 or Spag strongly reduces rhodanese aggregation by ϳ75%, as measured by light scattering at 320 nm. In addition, addition of Spag, Hsp40, or the control protein BSA alone exerts no effect on rhodanese aggregation, which is moderately suppressed when Hsc70 is alone. This experiment is representative of three.
Pih1D1, in a complex very similar to the mammalian and S. cerevisiae R2TP.
In mammals, R2TP associates with seven prefoldin-like proteins (4,8). So far, only the association of Spag, Pontin, and Reptin with the product of the CG14353 gene has been documented (31). The CG14353-encoded protein contains a domain with 56% sequence identity with the human prefoldin Monad, associated with the R2TP (24). Besides, an ortholog of Uri, also part of R2TP/prefoldin-like system (4), was identified as a specific interactor of protein phosphatase 1␣ (33). dUri is cytoplasmic and is strongly expressed in embryo, pupae, and adult gonads. Interestingly, dUri co-localizes with RNA polymerase II on polytene chromosomes, yet its link with R2TP remains elusive. Conservation of the prefoldin-like association with R2TP in D. melanogaster thus remains open to question.
Conservation of R2TP Clients across Metazoans-In mammals, three types of clients have been found for the Hsp90/ R2TP system as follows: PIKKs, snoRNPs, and RNA polymerases. In Drosophila, we detect a decreased activity for the PIKK TOR when spag is inactivated. Proteomic experiments in Schneider S2 cells showed that dTOR interacts with Pontin, Reptin, Spag, and liquid facet-related, a distant homolog of Tel2 (31,34). In mammalian cells, recruitment of the whole R2TP by Tel2 is necessary to stabilize mammalian TOR (7). Taken together, these data support the conservation of R2TP in the stabilization and assembly of PIKK complexes.
We found that the amounts of two box C/D snoRNP core proteins are diminished in animals devoid of Spag or Nufip. This is consistent with the role described for R2TP and Rsa1/ Nufip in the biogenesis of snoRNPs in S. cerevisiae and mammals (3,5). In Arabidopsis thaliana, atNufip hypomorph mutants are viable but show severe developmental defects, with a subclass of snoRNPs being affected (35). Yet, in this organism, no gene encoding products similar to Pih1D1 has been so far identified (36).
Mammalian RPAP3 stabilizes nascent Rpb1 prior to its association with other RNA polymerase II subunits (6). In Drosophila, Spag depletion only showed a mild destabilization of Rpb1 in the null mutant. Yet the strong binding of Spag to Rpb1 FIGURE 6. Spag and Nufip are required for box C/D sno core protein stabilization. A, Western blot analysis of protein extracts from third instar larvae spag k12101 , as compared with wild-type w 1118 (Ct), showed a significant diminution in the content of Nop58 (Nop5) and 15.5K (Hoip) but not fibrillarin. Tubulin was used as a loading control. B, this phenomenon was also observed in pupae extracts from animals in which RNAi was induced by Gal4 act5C to knock down spag (Gal4 act5c /spag RNAi as compared with Gal4 act5c /ϩ). C, similar results were observed in pupae extracts from animals in which RNAi was induced against Nufip (C, Gal4 act5c /nufip RNAi compared with w 1118 ).

Spaghetti/RPAP3 Function Is Conserved in Drosophila
supports a role in RNA polymerase II assembly. Quantitative analysis of Hsp90 interaction with its clients demonstrates that some, but not all, are degraded upon Hsp90 inhibition (37). Hsp90 could act as a molecular tweezers, by maintaining Rpb1 via R2TP in a state amenable for interaction with Rpb2, as it maintains glucocorticoid receptors in a state compatible with ligand binding (38).
The phenotype observed in spag null mutant larvae, in which most organs are atrophied, could thus be explained by alterations in the functions of TOR, snoRNP, and/or RNA polymerase II, as all these machineries are required for cell growth (39,40).
Spaghetti Is a Co-chaperone for Both Hsp70 and Hsp90-D. melanogaster Spag is able to bind directly through its TPR domain, with Hsp90 and Hsp70 isoforms containing a C-terminal EEVD. In addition, the conserved C-terminal domain is implicated in recruiting Hsc70 -3, an Hsp70 isoform devoid of EEVD motif. Finally, Spag is able to stimulate Hsc70 activity, suggesting a role for Hsp70 in R2TP function.
Some co-chaperones such as CHIP, display a single TPR domain, which binds alternatively Hsp70 or Hsp90 (15,17). In contrast, HOP harbors three TPR domains, which enable it to transfer clients from Hsp70 to Hsp90 (23). In invertebrates genomes, genes encoding proteins displaying both TPR and Monad-binding motifs can be identified, which are likely Spag homologs (41). Most insect Spag, including D. melanogaster, show a single TPR domain, although other insect Spag contain two such regions, as in vertebrates. Spag homologs with two TPR domains could bind simultaneously to Hsp70 and Hsp90 and transfer the client between these chaperones. In species with a single TPR domain (such as D. melanogaster), another co-chaperone or Hsp (such as Hsp22, -23, or -26) could assist the transfer of R2TP/clients from Hsp70 to Hsp90.
Intriguingly, the Ssa-1 and Ssb-1 proteins of S. cerevisiae, which are orthologs of cytoplasmic Hsp70, interact with Pih1 as well as the box C/D sno core proteins (42). This supports a role for Hsp70 in box C/D snoRNP biogenesis, and R2TPs function in general. Based on all these data, we propose a model whereby nascent R2TP substrates would first be pre-folded by Hsp70 and then transferred to Hsp90 to maintain them in a conformation amenable for integration into molecular complexes (Fig. 9).
Conclusion-In S. cerevisiae, single deletions of some R2TP components (namely Tah1 and Pih1) or adaptors (Rsa1) are not lethal unless cells are subjected to heat or nutrient stress (5). In particular, deletion of TAH1, the subunit that contacts Hsp90, generates only very mild phenotypes, thus suggesting that the chaperone plays an auxiliary role in the assembly of R2TP clients. In contrast, spaghetti is essential during fly development. This suggests that within the R2TP context, the role of Hsp90 might have evolved from a conditional machinery operating under stress conditions in yeast to a constitutive assembly fac- FIGURE 7. Phosphorylation of p70 S6K, a readout of TOR activity, is decreased upon spag inactivation but not Nufip. Western blot analysis showed a diminution of phospho-p70-S6K (Ph-S6K) but not total p70-S6K, as compared with tubulin. This was observed upon spag inactivation, either in spag k12101 null mutants as compared with w 1118 (A) or in pupae with spag knockdown (B, Gal4 act5c /spag RNAi ). This was not the case, however, when Nufip was silenced (C, Gal4 act5c /Nufip RNAi as compared with w 1118 ) (C).  FEBRUARY 28, 2014 • VOLUME 289 • NUMBER 9 JOURNAL OF BIOLOGICAL CHEMISTRY 6245 tor in metazoans. In addition, our data support a role for Hsp70 in R2TP-assisted complex assembly.