Identification of Lhp1p-associated RNAs by microarray analysis in Saccharomyces cerevisiae reveals association with coding and noncoding RNAs

Abstract

La is a conserved eukaryotic RNA-binding protein best known for its role in the biogenesis of noncoding RNAs transcribed by RNA polymerase III. To broaden our understanding of the function of the La homologous protein (Lhp1) in Saccharomyces cerevisiae, we have taken a genomics approach. Lhp1 ribonucleoprotein complexes were immunoprecipitated and bound RNAs were examined by hybridization to whole-genome microarrays that include >6,000 ORFs, documented noncoding RNAs, and the intervening intergenic regions. Demonstrating the validity of this approach, associations with previously known Lhp1p-associated RNAs were detected and associations with additional noncoding RNAs, including multiple tRNAs and small nucleolar RNAs, were revealed. Indicating that this approach provides a robust method for discovering RNAs, the data also identify associations between Lhp1p and several intergenic regions, three of which encode the recently annotated putative snoRNAs: RUF1, RUF2, and RUF3. Unexpectedly, we find that Lhp1p is also associated with a subset of coding mRNAs. These mRNAs include many ribosomal protein transcripts as well as the mRNA encoding Hac1p, a transcription factor required during the unfolded protein stress response. In cells lacking LHP1, Hac1p levels are decreased 2- to 3-fold, whereas no changes are detected in the levels of spliced or unspliced HAC1 mRNA or in the stability of Hac1p. Finally, although LHP1 is dispensable for growth under standard conditions, we find that it is required when the unfolded protein response is induced at elevated temperatures. These results suggest that Lhp1p may play a novel role in the translation of one or more cellular mRNAs.

La is an abundant eukaryotic RNA-binding protein implicated in multiple steps of RNA metabolism, including transcription and 3′ end processing of RNA polymerase III (RNA pol III) transcripts, as well as translation of certain viral and endogenous mRNAs containing internal ribosome entry site (IRES) sequences (1). The best characterized role of La is in the biogenesis and processing of a variety of noncoding RNAs (1). La binds the 3′ terminal UUU_OH sequences of newly synthesized RNA pol III transcripts such as pre-tRNAs and pre-5S rRNA and protects them from degradation (2, 3). In yeast, La homologous protein (LHP1) is nonessential, which is surprising, given that its metazoan counterparts are involved in processing such a wide range of essential, noncoding RNAs. Nevertheless, genetic and biochemical analyses in Saccharomyces cerevisiae have confirmed that Lhp1p is involved in the processing of newly synthesized RNA pol III transcripts (4, 5), and revealed a similar role with noncoding RNAs generated by RNA pol II, spliceosomal small nuclear RNAs (snRNAs) (6), and U3 small nucleolar RNA (snoRNA) (7). Lhp1p is known to bind these noncoding RNA precursors and is thought to facilitate their maturation by stabilizing them from digestion. Because Lhp1p is not required for maintaining normal levels of mature ribonucleoproteins, it has been suggested that Lhp1p acts as an RNA chaperone, enhancing the efficiency of biogenesis (4, 6).

Additional metazoan studies (1) have indicated that La may function at the 5′ end of some RNAs to play a role in translation. By in vitro studies, La has been shown to stimulate translation and influence IRES selection of some viral RNAs (8, 9) and of some cellular RNAs containing IRES elements (10, 11). However, it remains unclear whether La plays these roles in vivo. More recently La has been proposed to influence the translation of mRNAs containing 5′ terminal oligopyrimidine (TOP) sequences, which include mRNAs encoding ribosomal proteins that are translationally regulated in a growth-dependent manner (12). However, one study examining La's effect on TOP mRNA translation by in vivo transfection experiments observed modest positive effects (13), whereas another group has observed inhibitory effects of La on TOP mRNA translation (14). Although La has recently been demonstrated to be associated with TOP mRNAs in vivo and in active polysomes (15–17), La's specific role with TOP mRNAs remains unclear.

To gain a comprehensive view of Lhp1p function, we set out to characterize the complete binding profile of Lhp1p by using a genomics approach. We have immunoprecipitated Lhp1p and identified associated RNAs by microarray analysis. We find that Lhp1p is associated not only with a large number of noncoding RNAs but with coding RNAs as well. Thus, Lhp1p may have a broader function than previously thought.

Materials and Methods

Oligos. For more information, see Table 1, which is published as supporting information on the PNAS web site.

Strains. The following yeast strains were used: LHP1-myc₃ and its isogenic wild-type (18), lhp1Δ::LEU and its isogenic wild-type (3), and lhp1Δ::KAN and its isogenic wild-type (Invitrogen).

Immunoprecipitation (IP) and Microarray Analysis of Lhp1p-Associated RNAs. Isogenic wild-type and LHP1-myc₃ genomically epitope-tagged cells were grown to saturation in rich media at 30°C overnight, diluted to OD₆₀₀ = 0.1 in 1 liter, grown to exponential phase (OD₆₀₀ = 0.75–1.0), and harvested by centrifugation for extract preparation and IPs as described (19), with minor modifications. Immunoprecipitated RNAs were reverse transcribed into cDNA as described (20), but without PCR amplification, in the presence of aminoallyl-dUTP (Sigma-Aldrich) by using random sequence nonamers as primers (0.25 mg/ml). The cDNA samples were labeled with either Cy3 or Cy5 fluorescent dyes (Amersham Biosciences, Piscataway, NJ) and purified. Fluorescently labeled cDNAs were hybridized to whole-genome microarrays as described (21). Also see http://microarrays.org for microarray printing and preparation protocols. Whole-genome primer sets and sequences are available upon request from Invitrogen. The microarrays were scanned with a GenePix 4000a scanner and analyzed with genepixpro 3.0 software (Axon Instruments, Union City, CA). Cy3 and Cy5 fluorescent signals were normalized so that the total signal ratio for all good quality features was equal to one. Therefore, the fluorescence signal intensity from the relatively low abundance untagged control is artificially inflated and the data do not necessarily reflect true enrichment values. For each feature, we calculated an average log₂-transformed ratio from two independent experiments by which we determined an overall rank order. See Table 2, which is published as supporting information on the PNAS web site for primary Lhp1p IP microarray data.

Quantitative PCR (QPCR). Cell-equivalent fractions of isolated RNA from LHP1-myc₃-tagged and -untagged samples were reverse transcribed with a sequence specific mixture of reverse primers (250 nM each) (see Oligos in Table 1). We found that QPCR results were more linear with input template when reverse transcription was done with sequence specific primers, rather than with random primers (data not shown). The cDNA was ethanol precipitated twice and used as template for QPCR. QPCRs contained 250 nM primer concentration and 0.75× SYBR green (Sigma-Aldrich) for fluorescent detection of product formation. Standard curves were generated by making a dilution series of template cDNA, which was reverse-transcribed from total RNA. Fold enrichments between tagged and untagged samples were calculated based on standard curves when the data were within the linear range. The presence of single PCR products was confirmed both by analyzing melting curve transitions and by electrophoresing completed QPCRs on 2.2% 96-well agarose gels (Amersham Biosciences). All QPCRs were carried out and analyzed with Opticon detection systems (MJ Research, South San Francisco, CA).

Unfolded Protein Response (UPR) Growth Assays. Wild-type and lhp1Δ strains were grown in rich or synthetic media. Five-fold serial dilutions were made in 96-well plates, were spotted on SD complete plates lacking inositol and were top spread with 0–1.5 μg/ml tunicamycin, and equilibrated for 5 h. The plates were incubated at room temperature, 30°C, and 37°C.

HAC1 Northern and Western Analysis. Wild-type and lhp1Δ strains were grown in synthetic media at room temperature and then shifted to 37°C with the addition of 1.0 μg/ml tunicamycin or in an sec14–3^ts strain background to induce the UPR. Aliquots were harvested at time points from 0 to 3 h. Total RNA was prepared and Northern analysis was carried out as described (22). For Western analysis, protein was prepared as described (22). Samples were separated by SDS/PAGE and transferred to nitrocellulose and immunodetected by using anti-Hac1pⁱ (a gift from P. Walter, University of California, San Francisco), anti-Npl3p mAb 1E4 (a gift from M. Swanson, University of Florida, Gainesville), ECL and ECF detection systems (Amersham Biosciences).

Results

To identify Lhp1p-associated RNAs, we performed IP assays followed by microarray analysis (see Materials and Methods). Lhp1p-associated RNAs were immunoprecipitated by anti-myc antibodies from LHP1-myc₃ genomically epitope-tagged cell extracts under native conditions. They were then extracted and converted to cDNA by reverse transcription without PCR amplification, which had been required in other studies (20, 23). In our hands, this step decreased data reproducibility (data not shown) and was eliminated because we found it to be unnecessary. The cDNA from Lhp1p myc-tagged and -untagged control IPs were fluorescently labeled with Cy5 and Cy3 dyes, respectively, and were analyzed by simultaneous hybridization to S. cerevisiae whole-genome microarrays (Fig. 1). These microarrays were designed to cover the entire genome according to the Saccharomyces Genome Database (SGD) annotation in May 1999 with ∼13,000 features, corresponding to >6,000 ORFs, known noncoding RNAs, and their intervening intergenic regions (24). Although SGD annotation has been constantly updated since the time of design, newly annotated genes are nonetheless represented on the microarray, because they are encoded within the PCR fragments of intergenic region features.

Fig. 1.

Schematic of Lhp1p IP microarray assay. Features enriched in the Lhp1p IP versus the untagged control are red, and others are yellow or green.

For each microarray feature, we obtained a Cy5 to Cy3 fluorescence signal ratio, representing the relative abundance of each RNA in the Lhp1p versus control IP. Because the data are normalized so that the total signal ratio is equal to one, the intensity in the relatively low abundance control IP is artificially inflated. Therefore, the normalized data do not necessarily reflect true enrichment values. Data from two independent IP experiments were linearized by log₂ transformation, were averaged, and were found to be highly correlated (Pearson correlation, r = 0.85; data not shown). The dynamic range of the average log₂ ratios spanned from +4.4 to –4.6 and defined a rank order of Lhp1p IP-enriched features (see Table 2). The data displayed a roughly Gaussian distribution centered around –1.2 (mean), with an extended tail comprising ∼100 of the most highly enriched features or the top 1% of the data set (Fig. 2A). Moreover, because most known Lhp1p-associated RNAs appeared reproducibly in the top 100 (see Table 3, which is published as supporting information on the PNAS web site), we interpreted this subset as highly likely to be associated with Lhp1p. While distinguishing between true associations and false positives becomes more difficult further down the list, we have arbitrarily classified the top 10% or 1,000 features as significantly enriched. For ease of discussion, we will present our results for three classes of Lhp1p-associated RNAs: noncoding, intergenic, and coding mRNAs (Fig. 2 A).

Fig. 2.

Histograms of RNA classes. Different RNA classes are depicted on scaled y axes. (A) The distribution of all features is in blue, the noncoding RNAs are in orange, the intergenic regions are in green, and the coding mRNAs are in black. (B) The ribosomal protein gene (RPG) transcripts (red) are more enriched in the Lhp1p IP compared with the distribution of all ORFs (black).

Class I: Noncoding RNAs. In accordance with previous results that all mammalian RNA pol III transcripts are believed to interact with La (1), we find several yeast RNA pol III transcripts are also highly enriched in Lhp1p IPs. Three known Lhp1p-associated RNA pol III transcripts are among the 10 most highly enriched RNAs. These transcripts are U6 snRNA, SCR1, the RNA component of the signal recognition particle, and RPR1, the RNA component of RNase P (see Table 3). Lhp1p has been shown to bind and stabilize newly synthesized U6 snRNA (4). Lhp1p has also been shown to associate with SCR1 (3). The feature corresponding to SCR1 itself was not included on the microarray, but the intergenic region immediately downstream was identified with high enrichment (Fig. 3A). We infer that this reflects association of Lhp1p with a pre-SCR1 species elongated at its 3′ end. Previous work (25, 26) has indicated that Lhp1p binds and may function to stabilize pre-RPR1. We observe an interaction with the adjacent intergenic region directly downstream of RPR1 and believe this to be an association with 3′ extended pre-RPR1.

Fig. 3.

Chromosomal maps of Lhp1p IP microarray enrichment. Feature boundaries and corresponding average log₂ ratios are indicated. (A) Whereas the feature corresponding to SCR1 is absent from the microarray (X), the intergenic feature iSCR1 immediately 3′ of SCR1 shows high enrichment. (B) The feature representing YNL017c shows high enrichment, most likely because tI(AAU)N2 is located on the opposite strand. (C) The intergenic region encoding a highly homologous 5S rRNA variant, iYLR159w, shows modest enrichment. (D) The snoRNA SNR18 lies within the intron of YAL003w, which shows high enrichment.

There are 42 unique tRNA features on the microarray, representing at least one tRNA for 38 of 41 total anticodon isoacceptor types. In addition, because of tRNA gene duplication events and the high level of homology between tRNA isoacceptors, the 42 tRNA features on the microarray represent the vast majority of the 274 total tRNA genes present in the S. cerevisiae genome. We find that 22 of the 42 tRNA features appear in the 100 most highly enriched RNAs, with 33 in the top 1,000 (see Table 4, which is published as supporting information on the PNAS web site). Significantly, tRNA^Ser(CGA), previously shown to be associated with Lhp1p (5), is in the top 100. Of the 11 other tRNAs previously suggested to be associated with Lhp1p, based on the observation that aberrant pre-tRNA processing was observed in strains deleted for LHP1 (5, 25), 10 are in the top 100, and one was undetected. Interestingly, tRNA^Leu(CAA), which was not shown to require LHP1 for processing (5), is Lhp1p-associated. In addition to the tRNA features, Lhp1p is associated with eight intergenic regions directly 5′ or 3′ of tRNAs. Whereas unannotated RNAs encoded within these intergenic regions may be responsible for the observed enrichment (see Class II: Intergenic Regions), it is equally plausible that although yeast tRNA leader and trailer sequences are generally short, these associations reflect interaction with the 5′ and/or 3′ extended pre-tRNA species. Also, because the microarray consists of double-stranded PCR products, we believe we detect three tRNAs by way of ORF features residing on the opposite strand or overlapping with the tRNAs. For example, the microarray feature for YNL017c is highly enriched, most likely because tI(AAU)N2 is located on the opposite strand (Fig. 3B). All three of these ORFs were recently annotated as Dubious ORFs by comparative genomic analysis (27–29), strongly suggesting that the associations detected are with the tRNAs. Finally, our failure to see significant enrichment of Lhp1p with the remaining nine tRNAs on the microarray may indicate low binding, low signal from low abundance, or low reverse transcription efficiency of a highly structured RNA.

The feature corresponding to 5S rRNA shows only a relatively modest enrichment compared with other RNA pol III transcripts, even though an association of Lhp1p with pre-5S rRNA has been previously observed (3, 26). We find 5S rRNA and two other intergenic regions encoding highly homologous 5S rRNA variants (30) within the 700 most highly enriched RNAs (Fig. 3C and Table 3).

In addition to RNA pol III-transcribed noncoding RNAs, we also detect association with RNA pol II-transcribed noncoding RNAs, including U2, U4, and U5 snRNAs (see Table 3). Lhp1p has been shown to bind and stabilize 3′ extended precursors of RNA pol II-transcribed spliceosomal snRNAs (U1, U2, U4, and U5) (6), as well as U3 snoRNA (7). Whereas the genes encoding U1 snRNA and U3 snoRNA were not included as unique features on our microarray due to incomplete noncoding RNA annotation, we detect enrichment with the intergenic regions encoding these RNAs. We have not excluded the possibility that other unannotated RNAs within these regions may contribute to the high enrichment ratios, but the most likely explanation is Lhp1p association with the snRNAs.

Whereas U3 was the only snoRNA previously shown to be associated with and its processing facilitated by Lhp1p (7), we find a large number of snoRNAs associated with Lhp1p. There are currently 66 total known snoRNA genes in the S. cerevisiae genome. Approximately one-third (20) are found in the 100 most enriched RNAs with more than half (38) in the top 1,000 (see Table 5, which is published as supporting information on the PNAS web site). On the microarray, the snoRNAs are represented as unique features, encoded within intergenic regions, or for seven intronically encoded snoRNAs are included within host ORF PCR products. For example, SNR18 is encoded in the intron of YAL003w, which is highly enriched (Fig. 3D). For the five of seven intronically encoded snoRNAs found in the top 1000 via their host ORF feature, we have not determined whether the enrichment is due solely to the snoRNA, the host ORF (see Class III: Coding mRNAs) or both.

Class II: Intergenic Regions. An advantage of the whole-genome microarray is the potential to discover new, uncharacterized RNAs within intergenic regions. Indeed, 11 intergenic regions in the top 100 do not contain any known RNAs. Notably, the immediately surrounding features are not significantly enriched, suggesting that the signal in the intergenic region is not due to hybridization from the 5′ or 3′ UTR of a neighboring feature. Remarkably, the three most highly enriched intergenic regions encode genes annotated only after our analysis was completed. They are RUF1, RUF2, and RUF3, which correspond to three putative H/ACA snoRNAs identified computationally by phylogenetic analysis and confirmed by Northern analysis (31). The convergence of discovering these new genes within Lhp1p-enriched intergenic regions strongly suggests that the other intergenic regions may also contain novel RNAs (see Table 6, which is published as supporting information on the PNAS web site).

Class III: Coding mRNAs. Unexpectedly, 26 coding mRNAs appear in the 100 most Lhp1p-enriched RNAs with up to 300 in the top 500 (see Table 7, which is published as supporting information on the PNAS web site). Previous descriptions of La interactions with mRNAs have been restricted to virally encoded RNAs or endogenous mRNAs with IRES sequences or mRNAs with TOP sequences, which have not been documented in S. cerevisiae (32, 33).

We used QPCR as an independent method for assessing RNA association, by reverse transcribing cell-equivalent fractions of immunoprecipitated RNA from tagged and untagged cells (see Materials and Methods). We then determined fold enrichments for 18 RNAs from the microarray data list (see Table 8, which is published as supporting information on the PNAS web site). The values determined by QPCR generally correlated well with the ratios generated from the microarray results (Pearson correlation, r = 0.8; Fig. 4). As expected, as a result of data normalization (see Materials and Methods), the enrichment ratios obtained from the microarray data generally underestimated fold enrichments determined by QPCR. Nevertheless, although not as highly enriched as many of the noncoding RNAs, the QPCR data confirmed significant association between Lhp1p and coding mRNAs.

Fig. 4.

Lhp1p IP microarray data and QPCR data are highly correlated (Pearson correlation, r = 0.8). A subset of 18 genes was chosen (see Table 8).

RPGs. The predominant highly enriched class of coding mRNAs in the microarray data set is the RPGs (Fig. 2B). Of the 130 RPGs, 14 are in the 100 most highly enriched RNAs and 100 are in the top 1,000. Although RPGs encode some of the most abundant transcripts, we did not observe a general bias for Lhp1p association with highly expressed genes (data not shown; refs. 34 and 35). Moreover, there does not seem to be any bias for Lhp1p association with the most abundant RPGs, the small versus large subunit or intron-containing versus nonintron-containing RPGs (data not shown).

HAC1 mRNA. One of the most intriguing mRNAs we found to be associated with Lhp1p was HAC1 RNA, which encodes a transcription factor required for induction of the UPR. We sought to demonstrate a functional link between Lhp1p and HAC1 mRNA by examining the cellular response to unfolded proteins in the absence of Lhp1p. We plated serial dilutions of wild-type and lhp1Δ yeast cells on yeast media supplemented with tunicamycin (0–1.5 μg/ml), which impairs protein folding by inhibiting glycosylation (36), to induce the UPR. Whereas lhp1Δ cells grew like wild-type at 30°C, LHP1 is essential for growth upon induction of the UPR at 37°C (Fig. 5). Although we cannot rule out indirect effects, these results suggest that Lhp1p is required for proper induction of the UPR at elevated temperatures.

Fig. 5.

LHP1 is required for growth during the UPR at 37°C. Serial dilutions of wild-type (WT) and lhp1Δ::KAN strains (or lhp1Δ::LEU; data not shown) were grown at 30°C and 37°C on yeast media plates containing 0.0, 0.75, and 1.0 μg/ml tunicamycin (TM) to induce the UPR.

Interestingly, HAC1 mRNA is spliced only upon induction of the UPR by a mechanism that does not use the spliceosome (37). Rather, the HAC1 transcript is cleaved by an endonuclease and ligated by tRNA ligase (38). Because the mechanism of HAC1 mRNA splicing partially resembles that of pre-tRNA splicing (39), this finding raised the possibility that HAC1 RNA processing could likewise be facilitated by Lhp1p. To determine whether Lhp1p is involved in the splicing of HAC1 mRNA, we examined HAC1 RNA levels by Northern analysis in an lhp1Δ strain at time points from 0 to 3 h after induction of the UPR. Two methods of induction were used: addition of 1 μg/ml tunicamycin (data not shown), and shifting to 37°Cina sec14–3^ts mutant background, which arrests the secretory pathway (22, 40). We observe no LHP1-dependent change in either the HAC1 total mRNA levels or the ratio of spliced to unspliced HAC1 RNA (Fig. 6A). Thus, Lhp1p has no detectable effect on the processing of HAC1 RNA.

Fig. 6.

(A) HAC1 RNA levels and rate of splicing are unaffected in an lhp1Δ::LEU strain upon induction of the UPR by shifting to 37°Cina sec14–3^ts mutant strain background. HAC1 precursor (HAC1^p) and HAC1 mature (HAC1^m) were detected, and SCR1 served as a loading control. (B) Hac1p levels are consistently reduced by at least 2-fold. Npl3p served as a loading control. Similar results were seen with an lhp1Δ::KAN strain and with 1 μg/ml tunicamycin induction of the UPR (data not shown).

To determine whether Lhp1p is involved in postprocessing steps of HAC1 gene expression, we examined Hac1p levels by Western analysis. In an lhp1Δ strain, upon induction of the UPR by 1 μg/ml tunicamycin (data not shown) or shifting to 37°C in a sec14–3^ts mutant background, we observe a reproducible 2- to 3-fold reduction in Hac1p levels (Fig. 6B and data not shown). This observation is consistent with a role for Lhp1p either in Hac1p translation or posttranslational stability of Hac1p. By Western analysis, we examined the decay of Hac1p levels (41) in an lhp1Δ strain after addition of cycloheximide to inhibit protein synthesis and found no LHP1-dependent change in Hac1p stability (data not shown). Taken together, these data suggest a role for Lhp1p in translation of HAC1 mRNA.

Discussion

We have used a genomics approach to identify Lhp1p-associated RNAs. By exploiting the whole-genome microarray, which includes intergenic regions and noncoding RNAs in addition to ORFs (24), we were able to survey Lhp1p association with the entire transcribed yeast genome in a single experiment. The wealth of available genetic and biochemical evidence for Lhp1p association with many noncoding RNAs provided robust positive controls for our study. Indeed, we detected association with many known Lhp1p substrates, including tRNA^Ser(CGA), U6 snRNA, SCR1, RPR1, and 5S rRNA, as well as the RNA pol II-transcribed spliceosomal snRNAs and U3 snoRNA.

It is important to note that some RNAs that are Lhp1p-associated may not be identified by our assay as a consequence of insufficient signal due to low abundance, inadequate hybridization from a short RNA, or poor reverse transcription efficiency of a structured RNA. Moreover, only associations stable during the course of the IP assay will be detectable. Therefore, the absence of significant enrichment by our assay does not necessarily indicate lack of an association. For example, at least one RNA predicted to be associated with Lhp1p, NME1 (25), the RNA component of ribonuclease MRP, was not significantly enriched in our IPs. Furthermore, some genes, such as TLC1, which encodes the RNA component of telomerase, were omitted either as unique features or within intergenic regions on the microarray. With these caveats, we were able to survey nearly the whole genome for Lhp1p-association to gain a broader view of its binding profile.

Discovery of Noncoding RNA-Binding Partners. Whereas previous work had demonstrated that 3′ end processing of multiple tRNAs is facilitated by Lhp1p (5, 25), it remained an open question whether the interaction with U3 snoRNA (7) was unique or indicative of class-wide recognition for snoRNAs. Our detection of Lhp1p interaction with the majority of tRNAs and snoRNAs by the IP microarray assay suggests a class-wide role for Lhp1p in the processing of these stable, noncoding RNAs. Interestingly, although no changes in the pre-tRNA patterns were previously detected for tRNA^Leu(CAA) in lhp1Δ strains (5), we nonetheless detected association with this tRNA by the IP microarray assay. Whereas 3′ processing defects may simply not be rate-limiting under the conditions assayed, another possibility is that Lhp1p plays a role in step(s) further downstream. One such possible role is the facilitation of proper pre-tRNA folding for aminoacylation (42). Whereas our work has expanded the number of RNAs associated with Lhp1p, it remains to be determined whether Lhp1p plays similar or different roles for all tRNAs and snoRNAs.

Our approach provides the opportunity to explore uncharacterized regions in the S. cerevisiae genome in search of RNAs on a genomic scale. The discovery of novel RNAs by probing unusually large intergenic regions (43), underscored the fact that our current annotations are incomplete. Subsequently, others have used computational approaches to look for members of specific classes of RNAs by consensus sequences (44) or to find noncoding RNAs within structurally conserved intergenic sequences among related species (31). In contrast, the whole-genome microarray approach is unbiased for sequence information. Recent studies (45, 46) have productively used a combination of computational and biochemical methods using similar microarrays in prokaryotes. To our knowledge, however, ours is the first search for new RNAs using a complete eukaryotic genome array. Because the SGD noncoding RNA annotation was deficient at the time of design of the whole-genome microarray, a number of known snRNAs and snoRNAs were omitted as specific features. Nonetheless, intergenic regions encoding these RNAs were detected as Lhp1p-associated by the microarray assay. As further validation of the utility of this approach for gene discovery, we identified three intergenic regions, which were independently found to encode novel snoRNA genes (31). We are currently mapping the other Lhp1p-associated intergenic regions to look for novel RNAs. Interestingly, we find some of the intergenic regions are highly conserved among fungal species (data not shown).

Lhp1p Is Associated with Coding mRNAs. Our data indicate that Lhp1p is associated with coding mRNAs in addition to noncoding RNAs. In particular, we detect an interaction with HAC1 RNA, which encodes an activating transcription factor for the UPR. Importantly, we demonstrate that Lhp1p is required for growth under UPR conditions at high temperature (Fig. 6). Most simply, Lhp1p could be involved in facilitating processing of HAC1 mRNA, as it is with many noncoding RNAs. However, we did not observe an effect on HAC1 RNA levels or HAC1 mRNA splicing due to deletion of LHP1 (Fig. 5A). Moreover, HAC1 mRNA is polyadenylated (47), and although HAC1 mRNA splicing is carried out in part by tRNA ligase (38), HAC1 RNA is not believed to be acted upon by other 3′ end tRNA-processing enzymes. As the steady-state levels of many RNAs whose processing are known to be facilitated by Lhp1p remain unaffected in an lhp1Δ strain (4, 6), it is still possible that Lhp1p plays a role in HAC1 RNA processing, but its effects are undetected by our assays. However, we consistently observe a 2- to 3-fold decrease in Hac1p levels in the absence of LHP1 (Fig. 5B and data not shown), suggesting a role for Lhp1p in Hac1p production. Because La has been implicated in enhancing the translation rates of certain mRNAs in mammals (10, 13), it is intriguing to speculate that Lhp1p may be involved in formation of a structure required for efficient Hac1p translation. Notably, translation of HAC1 is attenuated by way of a base-paired interaction between the 5′UTR and the intron (22). Thus, Lhp1p could influence translation by destabilization of an inhibitory interaction such as between the 5′UTR and intron or by stabilization of a competing alternative structure. The latter mechanism would be consistent with Lhp1p's proposed role in stabilizing the anticodon stem of tRNA^Ser(CGA) (5). Similarly, loss of Lhp1p may destabilize such a structure in HAC1 RNA and compromise Hac1p translation.

As the vast majority of mRNAs in S. cerevisiae are not believed to include IRES elements (32), and yeast RPGs lack 5′ TOP sequences (33), we were surprised to identify Lhp1p association with multiple mRNAs, particularly RPGs. Although yeast RPGs lack 5′ TOP sequences, they are nonetheless regulated in a growth-dependent manner (48). However, unlike mammalian RPGs that are primarily translationally regulated through polysome association (12), yeast RPGs are primarily regulated at the level of transcription (49). Whereas future experiments are required to determine how association with yeast RPGs may affect their regulation, our data suggest an intriguing connection among Lhp1p-interacting RNAs. These RNAs are not only involved in processing of the translational apparatus (5S rRNA and tRNAs) and in processing the processing components of the translational apparatus (snoRNAs and RNase P RNA), but include transcripts encoding the protein components of translational machinery (RPGs). Thus, Lhp1p may contribute to the coordination of both RNA and protein components of translation for general cell metabolism. Intriguingly, LHP1 mRNA is expressed during logarithmic growth, but is repressed during diauxic shift and stationary phase (21, 50), suggesting that Lhp1p function may only be used during times of rapid growth, which is particularly suggestive in the case of RPGs. Moreover, Lhp1p function may only become critical during challenging growth situations as we have uncovered in the case of the UPR. Further experiments are required to assess the functional relationship between Lhp1p and coding mRNAs.