Control of Nutrient-Sensitive Transcription Programs by the Unconventional Prefoldin URI

See allHide authors and affiliations

Science  14 Nov 2003:
Vol. 302, Issue 5648, pp. 1208-1212
DOI: 10.1126/science.1088401


Prefoldins (PFDs) are members of a recently identified, small–molecular weight protein family able to assemble into molecular chaperone complexes. Here we describe an unusually large member of this family, termed URI, that forms complexes with other small–molecular weight PFDs and with RPB5, a shared subunit of all three RNA polymerases. Functional analysis of the yeast and human orthologs of URI revealed that both are targets of nutrient signaling and participate in gene expression controlled by the TOR kinase. Thus, URI is a component of a signaling pathway that coordinates nutrient availability with gene expression.

The evolutionarily conserved phosphatidylinositide (PI) 3-kinase–related kinase TOR (target of rapamycin) pathway occupies a central role in the integration and transduction of nutritional cues into a coherent cell-growth and proliferative response. Nutrient-rich conditions sustain TOR activity, which in turn fuels cell growth. In contrast, nutrient-depleted environments (or treatment with the immunosuppressant rapamycin) cause inhibition of TOR, which results in the activation of a response program that includes the induction of nutrient-sensitive gene expression (14). Here we describe an evolutionarily conserved member of the prefoldin (PFD) family, termed URI (for Unconventional prefoldin RPB5 Interactor), that participates in the regulation of nutrient-sensitive, TOR-dependent transcription programs.

While searching for proteins associated with the F-box protein SKP2 (for S-phase kinase–associated protein 2), which is the substrate recognition component of the cell cycle–regulatory SCFSKP2 ubiquitin ligase (5, 6), we identified a member of the PFD family of small–molecular weight (14 to 23 kD) proteins. PFD family members are composed of N- and C-terminal, α-helical, coiled-coil structures connected by either one (β-class PFDs) or two (α-class PFDs) β hairpins (7). Yeast and human PFDs 1 to 6 assemble into an α2β4 hexameric complex, referred to as the prefoldin/GimC complex, that functions as a molecular chaperone in actin and tubulin folding (810). Because the identified protein is an α-class PFD that associates with SKP2 in vivo, we termed it STAP1 (for SKP2-associated alpha PFD 1) (fig. S1).

We reasoned that STAP1, which is not part of prefoldin/GimC, could be a component of a unknown prefoldin-like complex. Mass spectrometric identification of STAP1-associated proteins from HeLa cells (Fig. 1A) revealed two β-class PFDs, PFD2 and PFD4-related (PFD4r); three proteins of unknown function; and four proteins whose functions have been linked to transcription. These are RPB5, a subunit shared by RNA polymerases (pols) I, II, and III (11); the adenosine triphosphatases TIP48 and TIP49 (12, 13); and RBP5-mediating protein (RMP), which is known to bind RPB5 (14). We refer to RMP hereafter as URI.

Fig. 1.

Isolation of a PFD-like complex. (A) Immunoprecipitates from HeLa cell extracts with control mouse immunoglobulin G (ctr. IgG) or STAP1 mouse antibody 105-128. Polypeptides that yielded unambiguous mass spectrometry spectra are indicated. MW St., molecular weight standard (in kD). (B) In vitro translated (IVT) human and yeast URI proteins were tested for binding to indicated GST fusion proteins. Input lanes display 20% of the input IVT protein.

On gel filtration, STAP1 eluted with URI as a single peak at ∼1 MD (fig. S2). RPB5, TIP48, and TIP49 were present in multiple fractions, including those that contained STAP1 and URI. A fraction of total SKP2 protein and its partner proteins CUL1 and SKP1 eluted together with STAP1. Thus, STAP1 and URI may reside in vivo in a single high–molecular weight complex that contains these components and subunits of SCFSKP2. In support of this, URI, RPB5, TIP48, and TIP49 were only detected in anti-STAP1 immunoprecipitates derived from those fractions of the gel-filtration column that also contained the majority of STAP1 (fig. S2). These proteins coimmunoprecipitated with each other from whole cell extracts (fig. S3). In addition, depletion of cellular URI by two sequential immunoprecipitations with URI antibodies codepleted STAP1 and PFD4r (fig. S3).

While in search of homologous proteins encoded by the yeast genome, we recognized a homology between STAP1 and the uncharacterized open reading frame (ORF) YFL023w (15). Conceptual translation of YFL023w predicted a 796–amino acid protein (referred to as scUrip) with an N-terminal segment that contains all the features of an α-class PFD. Subsequent database searches uncovered orthologs of YFL023w in the genomes of Arabidopsis thaliana, Drosophila melanogaster, Caenorhabditis elegans, and Homo sapiens (fig. S4). The human ortholog of YFL023w is identical to URI, which we had identified as a STAP1-associated protein (Fig. 1A). We noticed that the published cDNA that encodes RMP (URI) was rearranged at the N terminus. The corrected cDNA sequence encodes a protein of 534 amino acids with an α-class PFD-homology domain at its N terminus. In addition, a large-scale yeast two-hybrid screen revealed scUrip as an interaction partner for the proteins scRpb5p and scPfd6p (16). Indeed, URI and scUrip efficiently bound their respective PFD and RPB5 partners in glutathione S-transferase (GST) pull-down assays (Fig. 1B) in an RPB5 and PFD domain–dependent manner (fig. S5). Therefore, we refer to this polypeptide as URI/scUrip.

To probe the function of scUrip, we deleted the corresponding gene in yeast. URI deletion resulted in viable cells. Genome-wide expression analysis identified 39 genes whose expression was consistently affected by the loss of scUrip (table S1). The products of these genes function primarily in amino acid metabolism (Fig. 2A). URI deletion also resulted in the down-regulation of genes that encode different tRNA species. Similar effects were seen in a sigma (Σ) yeast strain deleted for URI (table S2). Thus, part of the normal function of scUrip is to contribute, directly or indirectly, to the expression of both RNA pol II and pol III transcripts, specifically those involved in the nutrient response pathway. Loss of scUrip also caused cell elongation (Fig. 2B) and agar penetration (Fig. 2C), hallmarks of invasive growth that are induced by nutrient limitation (17, 18). Moreover, scUrip levels decreased when cells were grown either in amino acid–deprived medium (Fig. 2D) or in a low-quality nitrogen source (urea media) (Fig. 2D), or were treated with rapamycin (19). They also decreased when cells entered stationary phase (Fig. 2D). Down-regulation of scUrip protein levels may therefore provide a mechanism to fine-tune the expression of scUrip-dependent genes during long-term nutrient starvation.

Fig. 2.

Loss of URI affects the expression of nutrient-sensitive genes in Saccharomyces cerevisiae. (A) Functional categorization of yeast ORFs, whose expression is changed in uriΔ cells (strain yBM80) compared to wild-type cells (strain yBM79). (B) Phase-contrast images of wild-type (wt) or uriΔ cells grown to log phase in a yeast extract, peptone, and dextrose medium. The percentage of elongated cells was determined from at least 200 cells. (C) Wild-type and uriΔ cells were grown for 2 days at 30°C on rich media, and photographs were taken before and after cells were gently washed off the agar. (D) Immunoblotting of yeast extracts prepared from cells that expressed myc-scUrip, subjected to either amino acid or nitrogen starvation or to diauxic shift.

Exposure of yeast cells to rapamycin or amino acid starvation results in a rapid and rich transcriptional response (4, 20). Rapamycin affected the expression of 710 ORFs. Amino acid starvation altered the expression of 587 ORFs, out of which 331 ORFs were identical to the ones affected by rapamycin. In fact, 29 out of the 36 ORFs affected in the uriΔ mutant overlapped with the group of 331 ORFs (Fig. 3A). Thus, the majority of genes affected by loss of URI are those that are common transcriptional targets of both amino acid starvation–sensitive and rapamycin-sensitive signaling.

Fig. 3.

scUrip is required for adaptive transcriptional responses. (A) Venn diagram depicting the overlap of genes whose mRNA expression is changed in response to rapamycin or amino acid starvation in wild-type cells with mRNA genes whose expression is altered in uriΔ cells (table S1). (B) Genes affected by rapamycin or amino acid starvation in wild-type cells were clustered according to their expression profile in wild-type and uriΔ strains with the program dCHIP. Mean expression values are displayed in white. Values above the mean are in red, and values below the mean, in blue. scUrip-dependent clusters are indicated (RC1 to RC4 for rapamycin and AC1 to AC6 for amino acid starvation).

Next, we used hierarchical cluster analysis to compare transcriptional responses between wild-type and uriΔ cells after rapamycin treatment or amino acid starvation (Fig. 3B). Although loss of scUrip had little impact on the ability of rapamycin to repress genes (n = 325 genes), it did affect the induction of rapamycin-sensitive genes (Fig. 3B and table S3). In response to amino acid starvation, loss of scUrip affected the extent and kinetics of both gene activation and repression (Fig. 3B and table S3). Thus, the function of scUrip is critical for adaptive transcriptional responses that originate from distinct nutrient signals.

To identify underlying transcriptional pathways engaged upon scUrip loss, we analyzed the 5′ promoter regions of scUrip-dependent genes. A large fraction of genes activated in uriΔ cells contained a consensus-binding site for Gcn4p (Fig. 4A), a transcription factor known to activate amino acid biosynthesis genes in response to amino acid deprivation (21). A role of GCN4 in gene activation in uriΔ cells is evidenced by the fact that these genes failed to be induced in a URI/GCN4 double mutant (fig. S7). However, not all scUrip-dependent genes are also GCN4-dependent, including certain tRNA genes and GAP1, an established rapamycin-induced gene.

Fig. 4.

scUrip represses GCN4-dependent transcription. (A) Conserved motifs within the promoter region of scUrip-dependent genes. The consensus site for Gcn4p is underlined. Numbers in parentheses refer to the number of allowed mismatches. P values correspond to the false-positive probabilities for each motif. (B) Extracts from indicated strains were analyzed by Western blotting with the 9E10 antibody to myc. Actin was used as a loading control. (C) Indicated strains harboring the reporter plasmid GCN4μORF-lacZ or the mutant GCN4μORF4only-lacZ were analyzed for lacZ activity. (D) Standardized mRNA expression values of GAP1, GCN4, GLN3, and GAT1 in the indicated strains relative to the wild type were obtained from genome-wide transcription analysis. (E) Wild-type and uriΔ mutant cells were analyzed for Gcn4p-myc expression after 100-nM rapamycin treatment. (F) Wild-type and uriΔ mutant cells were analyzed for Gcn4p-myc expression after amino acid starvation. eIF2α was used as a loading control.

Consistent with the above, Gcn4p levels were up-regulated under conditions of amino acid sufficiency in uriΔ cells (Fig. 4B). This up-regulation of Gcn4p was at least in part Gcn2p-independent. A lacZ reporter controlled by the wild-type, but not the mutated, GCN4 5′ leader sequence was markedly activated in these cells (Fig. 4C), suggesting that the increase in Gcn4p in uriΔ cells involves a relief of translational repression. This activation was again, in part, GCN2-independent (Fig. 4C). Similar results were obtained in the prototrophic yeast strain Σ (fig. S8). No changes in Gcn4p protein stability were detected in uriΔ cells (fig. S8). Hence, scUrip may contribute, directly or indirectly, to eIF2-GTP-tRNAiMet ternary complex formation and consequently to the suppression of GCN4 protein translation and thus of GCN4-dependent transcription programs under nutrient-rich conditions.

Only those Gcn4p target genes are induced in uriΔ cells that are also repressed by the TOR pathway. The latter is known to repress the translation of GCN4 and the activation of the transcription factors Gln3p and Gat1p. Because Gcn4p can facilitate the induction of certain Gln3p target genes, this negative effect of TOR on GCN4 translation is critical to prevent the expression of genes under the dual control of GLN3 or GAT1 and GCN4 in rich media (22, 23). GLN3 and GAT1 are indeed induced in uriΔ cells in a Gcn4p-dependent manner (Fig. 4D), providing a potential explanation for the activation of TOR-controlled genes in uriΔ cells. Consistent with this view, Gcn4p is expressed in uriΔ cells at levels similar to those of rapamycin-treated wild-type cells (Fig. 4E). Moreover, rapamycin did not affect Gcn4p levels in uriΔ cells (Fig. 4E). On amino acid starvation, however, Gcn4p can be induced in uriΔ cells as in wild-type cells (Fig. 4F). Combined, these results suggest a model in which loss of scUrip triggers the activation of genes under the dual control of Gcn4p and Gln3p or Gat1p, at least in part through the translational derepression of Gcn4p.

The conservation in structure between yeast and human URI proteins led us to question whether human URI might also participate in mammalian TOR (mTOR) signaling. URI is phosphorylated in vivo (Fig. 5A). Serum-starvation of cells from the HEK293 line caused dephosphorylation of URI (Fig. 5B). Insulin reverted this effect and produced hyperphosphorylated URI (Fig. 5B), which could be blocked by rapamycin (Fig. 5B) and the PI 3-kinase inhibitor wortmannin (19, 24). Thus, the phosphorylation state of URI is dependent on signals that affect mTOR activity. Consistent with this view, transfection of a rapamycin-resistant allele of mTOR (23) recovered URI phosphorylation in the presence of rapamycin, whereas wild-type mTOR did not (Fig. 5C). These results imply a role for URI in mTOR signaling in mammalian cells.

Fig. 5.

URI participates in mTOR signaling. (A) HEK293 cells were serum-starved, treated with insulin, processed for lambda phosphatase treatment, and analyzed by Western blotting for URI. Ppase, phosphatase; URI-P, phosphorylated URI. (B) Serum-starved HEK293 cells were induced for the indicated time with insulin in the absence or presence of rapamycin and were analyzed by Western blotting for URI. (C) Untransfected HEK293 cells or cells transfected with hemagglutinin (HA)–mTOR (wt) or rapamycin-resistant mutant HA-mTOR (RR) plasmids were serum-starved, treated with insulin and rapamycin, and analyzed by immunoblotting. (D) Cell extracts of HeLa cells stably transfected with either the plasmid pSuper (25) or pSuper-si-URI were analyzed by immunoblotting for URI. ctr, control; si, small interfering. (E) Gene expression profile of pSuper-HeLa cells and pSuper-si-URI cell lines after rapamycin treatment. Silencing of URI affected the rapamycin response of 28 genes shown in the cluster analysis.

Next, we created HeLa cells that had been stably silenced for URI expression by small interfering RNA (Fig. 5D) and compared their transcriptional profiles in response to rapamycin (Fig. 5E). After treatment of control cells for 2 hours with rapamycin, the expression of 194 genes was changed. Of these 194, 28 genes, whose products are primarily involved in the regulation of cell growth and maintenance and cell communication, had transcription patterns that were significantly altered in URI-silenced HeLa cells. This suggests a functional link between the presence of URI and the implementation of a proper rapamycin-sensitive transcriptional response in human cells.

These results describe an unconventional member of the PFD family, termed URI, that is part of an ∼1 MD multiprotein complex in human cells. URI appears to occupy a central role therein as a scaffolding protein able to assemble, through its PFD-homology and RPB5-binding domains, a prefoldin-like complex that contains other PFDs and proteins with roles in transcription and ubiquitination. Functional analysis of the yeast and human orthologs of URI established a critical role for URI in TOR-controlled transcription pathways.

TOR coordinates nutrient availability with cell growth and proliferation at least in part by controlling the transcription of distinct sets of nutrient metabolism genes. scUrip appears to play a critical role in this process, in that it contributes to TOR-mediated suppression of a select class of genes under the dual control of the transcriptional activators GCN4 and GLN3 or GAT1 under nutrient-rich conditions. In this regard, scUrip protein levels are down-regulated in response to various nutritional signals known to inhibit TOR activity, such as nitrogen starvation or rapamycin. Thus, scUrip is a component of a TOR-controlled downstream effector pathway that modulates nutrient-sensitive gene expression.

In mammalian cells, URI is a phosphorylation target of the mTOR pathway and, as in yeast, contributes to rapamycin-sensitive transcription. This implies an evolutionarily conserved role for URI proteins in TOR signaling. Further investigations are clearly warranted into the functions served by URI and other prefoldin family members in human diseases in which rapamycin-sensitive mTOR signaling has been implicated.

Supporting Online Material

Materials and Methods

Figs. S1 to S8

Tables S1 to S3

References and Notes

Stay Connected to Science

Navigate This Article