For a significant number of the genes implied to be under-translated during normal growth conditions, ribosome loading increased under the appropriate stress conditions, suggesting the existence of specific regulatory mechanisms that are responsive to environmental signals. One possible mechanism for this enhanced translation is suggested by the surprising frequency of regulated alterations in transcript structure. Of the 17 poorly loaded, translationally controlled transcripts examined in detail here, 12 exhibited structural changes coincident with altered ribosome loading in response to exogenous cues. The remaining five (PRM4, UGA1, MON1, GCY1, and PGM2) are likely to be solely under translational control.

The observed structural alterations were detected exclusively at the 5' ends of the transcripts. The sequences of 5' RACE products, together with RNase protection assays, demonstrated co-linearity between transcript and genomic sequences, providing no evidence for a regulated splicing mechanism similar to that involved in regulation of HAC1 in response to endoplasmic reticulum stress 32. Excluding regulated splicing as a mechanism, the alternative forms seemingly arose either transcriptionally, through use of different promoters, or post-transcriptionally, either as normal intermediates of mRNA decay or through a new RNA cleavage mechanism. The requirement for STE12 revealed by this work points to a role for transcription in the pheromone-induced transcript changes described here, but this role could be direct or indirect. Consistent with a direct role for Ste12p-mediated promoter activation, TATA boxes and Ste12p binding sequences are found appropriately placed relative to the putative transcription starts of the pheromone-induced forms of the HO, PRM2, PRY3, and SAG1 transcripts (K.S. Bickel, unpublished observation). Previously, altered promoter usage was demonstrated directly for the nitrogen-regulated gene CAN1 and was suggested for DAL5, although the translatability of the alternative transcript forms was not assessed 33. Promoter elements implicated in regulation of CAN1 and DAL5 are also found in the promoter regions of AMD2 and DAL7, suggesting the possibility of a similar switch in promoter usage.

Considering possible post-transcriptional mechanisms, the normal process of mRNA decay in the cytosol involves removal of the 5' cap, followed by 5'-3' exonucleolytic degradation 34. A block to exonuclease action could produce some of the 5' truncated products described here. Perhaps related to this is that accumulation of 5' truncated transcripts in Arabidopsis was recently found to result from ribosome arrest mediated by nascent peptide 35. Importantly, all known nonsplicing post-transcriptional mechanisms would be predicted to generate uncapped 5' ends, in contrast to the termini generated by RNA polymerase II initiation.

Because this is the first large-scale study relating ribosome loading to transcript structure, the frequency with which these regulated changes in transcript structure occur across nature is unknown. However, the suggestion that 9-18% of mammalian transcripts may have alternative first exons 12 is provocative. Two mammalian genes, in which alternative first exons were found to modify translation, are the gene encoding TIMP (tissue inhibitor of metalloproteinases) and the oncogene mdm2. With both of these genes, the translational efficiencies of the transcripts are regulated by changes in promoter utilization, which lead to altered 5' leaders 3637.

In yeast, use of alternative promoters has been shown in some cases to produce different proteins. The SUC2 and KAR4 genes both contain multiple promoters, which generate different protein products with different biologic activities 3839. Similarly, the short forms of the CRH1, KAR5, PRM2, PRP39, PRY3, ASP3, and AQY1 mRNAs identified in this study lack the primary initiation codon, resulting in 5' truncated ORFs. These seven genes have the potential to create short protein products from internal AUG codons within the truncated mRNAs in the same ORFs as the primary products, although existence of these protein products has not been proven. With PRM2, CRH1, and PRY3, the putative amino-terminal truncated proteins lack signal sequences that target these three proteins to the endoplasmic reticulum. Therefore, if produced, the short protein products of these three genes probably differ in intracellular location, and possibly in function, from the full-length proteins. Similarly, the single transmembrane domain of the full-length Kar5 protein, which localizes it in the endoplasmic reticulum membrane, would be missing from the shorter, poorly translated form. These changes in protein targeting potentially could play roles in regulating cellular responses to pheromone.

With several other transcripts identified here - HO, SAG1, AMD2, DAL5, and DAL7 - the altered 5' leaders did not modify the protein encoding regions but profoundly altered the loading of ribosomes on the resulting transcripts. Although one can posit functional explanations for the truncated protein products produced from alternate transcripts, the biologic significance of 5' leaders with repressed translational activity is less obvious. The HO gene represents an extreme example in which a 5' leader as long as 2 kilobases is produced in response to pheromone treatment and the long transcripts are located primarily in untranslated mRNP particles. Likewise, poor translation of the SAG1 transcript in growing cells is mediated by an inhibitory 826-nucleotide 5' UTR. A similar situation seems to occur with the nitrogen-regulated AMD2, DAL5, and DAL7 transcripts. In addition to the changes in ribosome loading observed here, the levels of all five of these transcripts are regulated at the transcriptional level. One outcome of these parallel changes in transcript level and translation efficiency is to amplify the biologic consequence of transcriptional control by accentuating the upregulation or downregulation of protein production. This is surely one mechanism for the 'homodirectional' changes in transcript level and ribosome loading that have been noted by others on a global level in yeast 2. However, this rationalization neglects the conundrum of why the cell does not simply enhance an expression response by switching transcription completely off.

Why should the cell produce a transcript that is either poorly translated or not translated at all? One speculative role for the continued synthesis of translationally inactive transcripts, under conditions in which the protein product is not needed, could involve regulation of accessibility to the promoter regions of these genes. One suggested role for the 'intergenic' transcription, which has been found widely in eukaryotes 4041, is to assist in maintaining an open chromatin state required for facile transcriptional activation. Intergenic transcription has been found in the locus control regions of the mammalian β-globin and MHC (major histocompatibility complex) class II loci 4243, in the promoter regions of the interleukin-4 and interleukin-13 genes 44, in the V(D)J region of the mouse immunoglobulin heavy chain locus 45 and within the Drosophila bithorax complex 46. RNA polymerase II is found upstream of many apparently inactive genes in stationary phase S. cerevisiae 47. It is noteworthy that the 5' leader of the long HO transcript extends 2,000 nucleotides upstream of the coding region through a region that is devoid of genes (for example, intergenic) and which contains a multitude of transcription factor binding sites that mediate the complex transcriptional control of the HO gene (discussed by Krebs and coworkers 48). Maintenance of this extended region in an open state through continued low level transcription of the translationally inactive transcript species could allow rapid reactivation of HO transcription upon removal of pheromone.

In addition to keeping chromatin in an active state, transcription from intergenic regions can also be involved in repressing transcription from promoters. This has been found to occur either by local competition between promoters 49 or through interference by elongating polymerases coming from an upstream promoter 50. The competition model applies equally to promoters located upstream or downstream of the primary promoter, which is consistent with the occurrence of both longer and shorter 5' UTRs in this study.

Very little is known of the mechanisms that prevent inappropriate protein production from intergenic transcripts. Some 'cryptic' RNA polymerase II products are removed within the nucleus through a highly conserved process utilizing a unique poly(A) polymerase and the nuclear exosome 51. Nonsense-mediated decay 5253, another highly conserved process 5455, removes those transcripts that are recognized as having premature translation termination codons. This paper describes a third process, translational silencing, wherein continuing synthesis of transcripts with inhibitory 5' leaders contributes to an open chromatin structure while protecting the cell from inappropriate protein production. At this time, we have no evidence defining the inhibitory elements in the 5' leaders of the silenced transcripts. As discussed in the Background section (above), possible inhibitory features could be secondary structure, protein binding sites, or ATG codons upstream of the coding region. With regard to the latter mechanism, we have noted ATG sequences in all of the long, inhibitory 5' leaders. For example, the long forms of the DAL5 and AMD2 5' leaders contain five and two ATG codons, respectively, whereas neither short form contains an ATG upstream of the start codon. Further experimental work will be required to establish the inhibitory elements in the translationally silenced transcripts.

All experiments used strain BY2125 (MATa ade2-1 his3-11,15 leu2-3,112 ura3-1 can1-100 ssd1-d : W303 background). Strains VM1906 (Δfus3::LEU2 Δkss1::TRP1), VM1718 (Δste12::TRP1) and LL1 (Δgcn2::TRP1) were derived from BY2125 by gene disruption 56.

Cells were grown at 30°C in rich glucose medium, YPD (1% yeast extract, 2% peptone and 2% glucose) 57, to mid-log phase (approximately 1 × 10⁷ cells/ml) before harvesting. Preparations of cell lysates and polysome fractionation were described previously, as was pheromone treatment of yeast cultures 3. For nitrogen starvation, cultures were grown at 30°C in minimal glucose medium 57 with necessary supplements to mid-log phase, washed once with 10 mmol/l potassium phosphate (pH 7.0), and then incubated for 30 minutes at 30°C in pre-warmed nitrogen starvation medium (0.2% yeast nitrogen base [without amino acids or ammonium sulfate], 3% glucose, 20 mmol/l potassium phosphate [pH 7.0], and adenine and uracil added at 40 and 20 µg/ml, respectively) 58. For osmotic stress, exponential phase YPD cultures (approximately 8 × 10⁶ cells/ml) were diluted into an equal volume of pre-warmed YPD or YPD + 2 mol/l sorbitol. Incubation at 30°C was continued for 30 minutes before the cultures were harvested.

RNA was isolated using Qiagen RNeasy mini-columns (Qiagen Corp., Valencia, CA, USA). An equal proportion of the RNA isolated from each sucrose gradient fraction was used directly in reverse transcription reactions using anchored oligo(dT)₂₅ primers. When comparing changes in total RNA isolated from different culture conditions or treatments, equal quantities of total RNA were used for reverse transcription reactions. QPCR was performed as described previously 3.

Northern blot analysis followed a procedure described previously 59, as did the RNase protection assays 60. The SAG1 RNase protection assay probe was 631 bases long and contained 484 nucleotides 5' of the coding region and 55 nucleotides into the coding region. 5' RACE was carried out as described by Frohman 61 using gene specific primers for the reverse transcription reaction. The Thermoscript RT-PCR system (Invitrogen, Carlsbad, CA, USA) was used allowing for the reverse transcription reaction to be done at 55°C to minimize reverse transcriptase stops due to secondary structure in the RNA. To estimate the HO 5' leader in cells treated with α-factor for 30 minutes, reverse transcription reactions were done as described above and the products were used as DNA template in a series of PCR reactions. Two reverse primers, located -500 and -1493 nucleotides relative to the initiation codon of the HO ORF and 10 different forward primers, spaced roughly 200 nucleotides apart starting at -700, were used.

Using QPCR, relative levels of mRNA across polysome gradients were determined for the indicated genes in growing cells or treated cells. The treatment was either pheromone treatment for 30 minutes, nitrogen starvation for 30 minutes or osmotic stress for 30 minutes. Using the Abs_{260 nm} traces from the polysome gradients, the number of ribosomes associated with a specific mRNA in each fraction was determined. The ribosome loading ratio was calculated by dividing the average number of ribosomes associated with a transcript from a polysome gradient from treated cells divided by the average number of ribosomes associated with a transcript from a polysome gradient from growing cells. A number greater than 1 indicates an increase in ribosome loading with treatment and conversely a number less than 1 indicates a decrease in ribosome loading with treatment.

A HIS3-HA reporter plasmid (pVW12) was constructed by insertion of the HIS3-HA sequence from pVW06 3 between the Bam HI and Eco RI sites of the multiple cloning sequence of plasmid pRS416ADH1p 62, so that HIS3-HA transcription is from the constitutive ADH1 promoter. The ADH1 5' leader in pVW12 (nucleotides -48 to -1) was replaced with either the SAG1 short 5' leader (nucleotides -48 to -1, plasmid pVW13) or the SAG1 long 5' leader (nucleotides -836 to -1, plasmid pVW14) using plasmid gap repair 63. Specifically, pVW12 was cleaved in the 5' leader with Spe I and Xba I and transformed into strain BY2125 with PCR fragments bearing the short or long SAG1 5' leader flanked with 5' and 3' sequences homologous to the ADH1 promoter (-91 to -49) and HIS3-HA (+1 to +48). Ura⁺ yeast transformants were screened by PCR to identify plasmids repaired with the SAG1 fragments and confirmed by DNA sequencing. S1 nuclease protection assays were carried out as described 6465, using gel purified oligonucleotides (Qiagen Corp.), on RNA isolated from pVW13 or pVW14 to confirm the 5' ends of each transcript.

Yeast transformed with pRS416ADH1, pVW13, or pVW14 were grown in selective medium (synthetic complete medium with casamino acids and lacking uracil) to mid-exponential phase, harvested, and lysed as described previously 3. Protein samples (5 µg for the pVW13 lysate and 50 µg each for the pVW14 and pRS416ADH1p lysates) were separated by electrophoresis on a 10% polyacrylamide gel and transferred electrophoretically to PVDF membrane. The membrane was incubated with anti-HA mouse monoclonal antibody HA.11 (Covance Research Products, Berkeley, CA, USA) and sheep anti-mouse immunoglobulin conjugated with horseradish peroxidase (Amersham Biosciences, Piscataway, NJ, USA), then developed with ECL Plus Western Blotting Detection System (Amersham Biosciences). His-HA protein was quantitated using a Storm 840 phosphorimager (Amersham Biosciences).