Oligonucleotide sequences are contained in the Supplementary Material. Oligonucleotides were designed to be complementary to known ncRNA sequences and flanking regions and were tiled at 5 nt intervals for most RNAs (intron-containing mRNAs were tiled every 20 nt and mitochondrial RNAs every 15 nt). Probe lengths were adjusted to have a melting temperature of ∼53°C. Ink-jet microarrays were manufactured by Agilent Technologies (Palo Alto, CA).

Homozygous deletion mutants (13) were obtained from Research Genetics. TetO₇-promoter alleles were constructed as described previously (14). rnt1-1^ts (15) and corresponding wild-type control strain W303-1a were kindly provided by Sherif Abou Elela; rat1-1^ts (16) and corresponding control strain FY23 were provided by Steve Buratowski. Strain numbers are as follows: TetO₇-BCD1: TH_3235; dbr1-Δ: TH_138; rat1-1^ts: TH_7097; TetO₇-RNT1: TH_3029; rnt1-1^ts: TH_6855; TetO₇-RRP46: TH_2706; TetO₇-MTR3: TH_3685; rrp6-Δ: TH_387; TetO₇-POP1: TH_2654; TetO₇-POP4: TH_5545; xrn1-Δ: TH_3339. The wild-type control for tet-promoter strains was R1158 (14); for deletion strains, BY4743 (13).

Isogenic wild-type and mutant strains were grown in parallel at 30°C in SC medium (with the exception of the temperature-sensitive alleles; see below) with shaking in baffled flasks (Bellco) to final cell concentrations matched as closely as possible to 10⁷cells/ml. TetO₇-promoter strains were exposed to 10 μg/ml doxycycline for a total of 20–24 h. Temperature-sensitive strains were grown at 23°C overnight, then shifted to 37°C for 4 h. For conditional mutants, the doubling times in Table 2 are based on growth under restrictive conditions. The cells were harvested and RNA extracted as described previously (10). Ten micrograms of DNase I-treated RNA was labeled with Alexa Fluor 546 or 647 according to the manufacturer's instructions (Molecular Probes ‘Ulysis’ kit), ethanol-precipitated and hybridized to the microarray as described previously (17). Formamide was added to a final concentration of 33%, as described previously (17). Hybridizations were carried out in a rotating incubator at 42°C for 18 h and washed as described previously (17). Microarrays were scanned on an Axon 4000B instrument.

Scanned images were quantified with GenePix 3.0 (Axon Instruments). Individual channels were spatially detrended (i.e. overall correlations between spot intensity and position on the slide removed) and normalized as described previously (11), resulting in log₂ intensities and ratios of mutant versus wild type.

Microarray oligonucleotide sequences and microarray data are found at . Spreadsheets containing the data displayed in Figure 2 are also available on the website.

For northern blot analysis, 5 μg of total RNA from each strain (isolated and treated with DNase I as described above) was separated on 10% denaturing polyacrylamide–urea gels and transferred to a Hybond-XL membrane (Amersham) in 0.5× TBE using a semi-dry transfer apparatus (Bio-Rad). Membranes were UV-crosslinked and hybridized in Church buffer using 5′-³²P-end-labeled complementary oligonucleotide probes as indicated in the figures. Images were visualized using a PhosphorImager (Bio-Rad Personal FX). Probe sequences were as follows (5′ to 3′): (Figure 3B) RUF3 5′flank, CACACGTACTAGACTTTATCTGTCTTGATTG; RUF3 body, CAATTGTTGTAGTCGCAACTACGGTAATTG; RUF1 5′ flank, GTACTCTCATTAACTAGCTCTGTTATTC; RUF1 body, AAAAAGTCGCAACCTCAATCATGCCTTTTCTC; (Figure 4C) tRNA LeuCAA 5′ flank, GGCCAAACAACCACTTATTTGTATGTTTCG; tRNA LeuCAA intron, TATTCCCACAGTTAACTGCGGTCAAGATATTG; tRNA LeuCAA exon 1, CTTGAATCAGGCGCCTTAGACCGCTC.

To increase the sensitivity of detection of directly labeled RNA transcripts, we created a version of our previously described microarray (11) with longer oligonucleotide probes (see Supplementary Material for probe sequences). This new microarray features probes tiled across the same ncRNA sequences, but has an average probe length of 25.1 nt [compared with 18.4 nt on the previous microarray (11)]. A comparison of the two microarrays using wild-type RNA revealed that the longer probes improved the consistency and detection levels of both tRNAs (Figure 1A) and rRNA (Figure 1B). Furthermore, we were able to detect a greater proportion of less abundant RNA transcripts (e.g. snoRNAs, Figure 1C). Although microarray spot intensity is a relative measurement (controlled by the photomultiplier tube voltage on the scanner), the fact that the flanking sequences (representing background) are slightly darker (i.e. there is lower signal-to-noise) in all of the short-oligonucleotide plots in Figure 1 confirms that the overall signal-to-noise ratio is superior on the long-oligonucleotide microarray.

To further validate the new microarray, we analyzed RNA from a strain with the BCD1 (‘Box C/D snoRNAs’) gene under the control of the tetracycline-responsive promoter (TetO₇-BCD1). We previously demonstrated that BCD1 is required for biogenesis and/or stability of Box C/D snoRNAs (10). However, the previous classification was based on the reduction of only 11 Box C/D snoRNAs, but not 2 box H/ACA snoRNAs, in the TetO₇-BCD1 strain. In order to investigate the role of this protein on a more comprehensive scale, we analyzed TetO₇-BCD1 RNA on the new microarray, which encompasses all 84 known snoRNAs. In this experiment, RNA isolated from wild-type cells (labeled with Cy3) and RNA isolated from mutant yeast cells (labeled with Cy5) were hybridized to the microarray simultaneously, and the fluorescence in each channel was measured and compared as a ratio [(mutant RNA fluorescence)/(wild-type RNA fluorescence)] (see Materials and Methods for details). We observed the depletion of virtually all Box C/D snoRNAs, but there was essentially no effect on Box H/ACA snoRNAs (Figure 1D). This confirms the widespread involvement of Bcd1p in Box C/D snoRNA and/or snoRNP biogenesis, and also verifies that the majority of snoRNA probes are detecting the correct signal, since they specifically register a reduction in the amount of the RNA they are designed to detect.

We next analyzed strains deficient in nucleases known to be required for the processing of rRNA, tRNA, snoRNA, snRNA and mRNA introns (Table 1; growth rates of mutant strains are contained in Table 2). We expected that flanking sequences that are improperly processed in mutant strains would accumulate and result in a positive ratio, as previously seen using a smaller microarray (10).

To gain an initial overview of the results, we used clustering analysis to examine the data from oligonucleotide probes with ratios >4/1 (mutant/wild type) and corresponding to flanking regions (Figure 2, left hand side). Probes corresponding to accumulated flanking regions are colored according to the color bar shown, and the identities of the probes are indicated in the monochrome schematic on the right of Figure 2. It is evident that the majority of probes corresponding to any given type of processed fragment behave similarly (i.e. cluster together) and accumulate in mutants known to be responsible for their processing. The specificity of detection of the expected events is demonstrated by the relatively small number of apparent false-positive probes (i.e. probes that accumulate in mutants that are not thought to have a role in their processing) clustering in the unexpected regions of the figure (i.e. outside the black regions in the right hand panel). For example, Dbr1p is the enzyme responsible for debranching the lariat intron generated by mRNA splicing, a step that is necessary for the degradation of intronic RNA by the cell (18). Figure 2 shows that there is a specific accumulation of mRNA introns in the dbr1-Δ strain. We did not observe an accumulation of tRNA introns (Supplementary Material), consistent with the fact that splicing of tRNAs proceeds through the tRNA splicing endonuclease (SEN) complex rather than the spliceosome (19). However, we did observe the accumulation of intron-encoded snoRNAs, consistent with previous analysis (20). This experiment showed an impressive level of specificity, with 195 of the 200 highest ratio probes corresponding to intron sequences processed by the spliceosome (Supplementary Material; the other five probes in the top 200 correspond to unrelated tRNA and rRNA sequences, likely representing noise in the data).

In some cases, flanking regions are processed by more than one enzyme. For example, the rRNA 3′ external transcribed sequence (ETS) is cleaved from the primary transcript by Rnt1p and is then degraded by Rat1p (3). This is detected in our data: both temperature-sensitive and tetracycline-regulated alleles of RNT1, as well as a rat1-1^ts mutant, accumulated 3′ ETS sequences in this analysis. A similar phenomenon is seen for the 5′ ETS: deletion of RAT1 impaired processing of 5′ ETS sequences A₀–A₁ in the 5′ ETS and A₂–A₃ in internal transcribed sequence 1 (ITS1), while deletion of XRN1 resulted in the accumulation of both A₀–A₁ sequences as well as ITS1 sequences D–A₂ and A₂–A₃.

These data also underscore differences in phenotypes of mutants in proteins that have similar functions. Rrp46p and Mtr3p are both core components of the exosome, while Rrp6p is specific to the nuclear exosome. Mutations in all three enzymes resulted in the accumulation of sequences between the 5′ end of the transcript and A₀ in the 5′ ETS, as well as in the second ITS2 between E and C2. However, depletion of Rrp46p and Mtr3p primarily impact 35S pre-rRNA processing, while the rrp6 mutant largely accumulates snoRNA 3′ flank sequences, consistent with prior observations (3,21).

These data are posted in their entirety at and are available as a Supplemental Data file accessible via the journal's website. While the data primarily reflect known roles of these RNA processing enzymes, there are numerous examples of processing events that to our knowledge have not been previously characterized. Two selected instances of particular interest that we have confirmed by directed analysis are described here.

We observed novel processing events involving Rnt1p (RNase III), an endonuclease with many known targets in yeast, including rRNA, snRNAs and snoRNAs [reviewed in (3)]. We analyzed RNA from two strains containing different mutant alleles of RNT1 by microarray (TetO₇-RNT1 and rnt1-1^ts; Figures 2 and 3A; Supplementary Material). In addition to the accumulation of known targets (e.g. 35S 3′ ETS and Box C/D snoRNA 5′ flank sequences), we observed the accumulation of 5′ flank sequences of both RUF1 and RUF3 RNAs (Figure 3A). Initially identified as ncRNA transcripts by a comparative genomic strategy, these RNAs of Unknown Function were classified as Box H/ACA snoRNAs (22,23). To confirm that these snoRNAs are indeed processed by Rnt1p, we performed northern blot analysis of RUF3 RNA isolated from strains depleted for Rnt1p (Figure 3B, top). Strains containing either conditional allele of RNT1 accumulated an extended RUF3 species in addition to the mature RNA. Reprobing the same blot (after stripping the first probe) with a 5′ leader-specific probe confirmed that the larger RNA in both strains is a 5′-extended version. Identical results were obtained from northern analysis of RUF1 (Figure 3B, bottom). Similar to other Rnt1p targets, the 5′ flank sequences of both RUF1 and RUF3 are predicted to contain a consensus AGNN tetraloop upstream of the cleavage site (Figure 3C) (24,25). Together, these data reconfirm RUF1 and RUF3 as bona fide RNAs and establish that they are processed at the 5′ end by Rnt1p.

tRNAs in yeast are extensively processed: they are subject to both 5′ and 3′ end processing (6,26), many covalent modifications (5), and intron-containing tRNAs are spliced by the SEN complex (19). Pop1p and Pop4p are protein components of RNase P (the ribonucleoprotein complex responsible for removing the 5′ leader sequence from tRNAs) and RNase MRP (which processes sequences between A3 and B1 in the 35S pre-rRNA transcript). We analyzed RNA from strains with tetracycline-regulated alleles of POP1 and POP4 by microarray and observed the accumulation of both tRNA 5′ leader and A3-B1 sequences (Figure 2). Figure 4A shows schematic diagrams of selected tRNAs with the relative fluorescence from the TetO₇-POP1 and TetO₇-POP4 experiments as described above. The concentration of high-ratio (i.e. red) probes in the region immediately 5′ of the mature tRNA (highlighted in red on the schematic) indicates that the leader sequences are accumulating in both mutants.

In addition to the accumulation of 5′ leader sequences, we observed the accumulation of intron sequences in several intron-containing tRNAs (Figure 4A, LeuCAA and LysCTT). This was surprising because it was thought that there is no relationship between tRNA 5′ end processing and tRNA splicing (27). To confirm that this effect was caused by the depletion of the RNase P complex and not a general response to slow growth or depletion of an essential gene product, we compared the relative fluorescence of LeuCAA tRNA probes in nine different experiments (Figure 4B). Accumulation of tRNA intron probes was exclusive to the two experiments affecting the RNase P complex. We next performed northern blot analysis of LeuCAA tRNA from TetO₇-POP1 and TetO₇-POP4 strains. Figure 4C confirms that the TetO₇-POP1 and TetO₇-POP4 strains accumulated unspliced, 5′ leader-containing RNA. These strains also contained a roughly equivalent amount of spliced leader-containing RNA (Figure 4C). As a control, we analyzed RNA from a strain with a tetracycline-regulated allele of SEN34, the catalytic subunit of the tRNA SEN. As expected, this strain accumulated the intron-containing (but not 5′ extended) tRNA (Figure 4C, center blot). Together, these data indicate that the presence of the 5′ leader does not prevent, but may delay splicing as ∼50% of the 5′ leader-containing RNA is unspliced (compared with <5% of tRNA with a mature 5′ end in the wild-type samples). Conversely, the presence of the intron had no effect on the removal of the 5′ leader, as shown by the complete removal of the leader in the TetO₇-SEN34 experiment. The northern blot in Figure 4C (left) also appears as though it may contain 3′ unprocessed forms, consistent with previous demonstrations that 5′ processing generally precedes 3′ processing (26,28). Because there is no mutant known to be defective specifically in tRNA 3′ end processing, we did not examine further the relative ordering of 5′ and 3′ steps, or the ability of the arrays to measure such a defect.