The heat shock response in yeast is mediated by the transcription factors (TF) Hsf1 and Msn2/Msn4, and involves the upregulation of a number of genes for heat-shock proteins (Hsps) that participate in protein folding, trafficking and maturation, and also the genes for the protein degradation machinery. This is paralleled by a downregulation of the genes for ribosome biogenesis [8]. Some Hsp genes are specific for one of the above transcription factors, while others are governed by both Hsf1 and Msn2/4 through the respective HSE and STRE promoter motifs. Heat shock factor Hsf1 is an essential protein in S. cerevisiae, which participates in both the heat shock response and the adaptation to oxidative stress and glucose starvation [9], [10]. The partially redundant zinc-finger transcription factors Msn2 and Msn4 mediate the response to a variety of cellular and environmental stresses, including heat shock (known as the Environmental Stress Response, reviewed in [11]). During adaptation to heat shock, both factors seem to play differential roles: Hsf1 is essential for recovery from a brief exposure to extreme temperature, while Msn2/4 is required for long-term survival at high temperatures [12]. Partial overlapping has been described between the heat shock response and the unfolded protein response, which is induced by misfolded proteins at the endoplasmic reticulum and mediated by the Hac1 TF [13]. Both types of stress share a number of transcription targets.

A general view of the 16 clusters shown in Fig. 2 indicates that half of the S. cerevisiae genes for which data have been obtained are grouped in two main clusters (11 and 12). The concentration of genes in fewer clusters than in the previous GRO study on the oxidative stress response [4], demonstrates that S. cerevisiae cells display a more homogeneous response for heat stress than for oxidative stress. In the oxidative stress response, an opposed effect of mRNA stability versus RA trend has been shown [17]. This apparently paradoxical behavior has been interpreted as a way to modulate the speed of the response, as previously predicted [3]. In the same study, however, the response to DNA damage stress was seen to follow homodirectional changes for RA and mRNA stability [17]. In our current results the accelerating effect of mRNA destabilization can be seen in clusters 3–5. In these clusters, genes related with translation reduce their mRNA level rapidly (Fig. S3), perhaps to prevent unnecessary expenditures in this energetically-demanding process [30]. A homodirectional response has also been found in osmotic stress [18]. In this case the authors argue that the change in mRNA stability precedes the change in RA. Using our GRO protocol in this work, we can further explore these behaviors because we obtain separate data for TR (the transcriptional response itself) and RA (the consequence of the transcriptional response modulated by changes in mRNA stability). We see here that this stress induces mainly homodirectional changes because all the upregulated genes show mRNA stabilization profiles (clusters 11–16), and most of the downregulated genes (clusters 2–6 and 10) present mRNA destabilization. Thus, we confirm that most of the homodirectional changes (as suspected) between RA and mRNA stability are due to the homodirectional behaviors of TR and mRNA stability. However, in about one quarter of the genes, clusters 7–10, mRNA degradation counteracts the TR trend, provoking even an opposite profile between TR and RA (cluster 10). In relation to the anticipation of the mRNA stability response seen by Molin et al [18] in osmotic stress, our current data show that the destabilization peak in most clusters of the upper branch takes place between 15 and 20 min (Fig. S1), whereas the minimum in RA appears later (Fig. 2). In the genes of the lower part of the tree (Fig. 2), mRNA stabilization peaks at min 5–10, clearly before the RA (which peaks at the end of the time course, or even later). This agrees with the proposal of Molin et al [18]. However, it should be considered that an advance of the TR peak over the RA one may naturally result from kinetic laws and does not necessarily mean an advanced cell response.

Among the genes which accumulate mRNA as a result of heat shock (clusters 11–16), different TR kinetics are observed. Thus, the small number of genes included in clusters 15 and 16 displays stable or decreased TR, indicating that RA upregulation results exclusively from mRNA stabilization effects. This situation would have remained unnoticed in standard DNA microarray analyses. Remarkable stabilization is also predicted for the genes of cluster 14, for which RA upregulation (about 4- or 5-fold) is much higher than that predicted from its TR profile (Fig. 4). Other previously undetected behaviors correspond to those genes with a stable RA profile whose transcription, however, has clearly altered, as in clusters 1–2 and 7–9. They represent apparent futile behaviors in which changes in mRNA stability are devoted to compensate changes in TR (as mentioned above). This behavior was also observed in our previous work on the oxidative stress response [4]. To date, we can provide no explanation for it, but it is interesting to point that another kind of apparent futile pathway has been observed by Warringer et al [31], who found that some transcriptionally induced mRNAs are not translated at all during the osmotic stress response. These results suggest the existence of still unknown gene regulatory strategies which are not just the straightforward ones [32]. In any case, and as previously suggested [3], the increases in both the transcription and degradation rates for mRNA can be a way to speed the response. On the other hand, the simultaneous decrease in both rates can be used to save energy dispenses. Therefore, we conclude that the heat shock response seems to be contributed both by mRNA transcription and stabilization effects. This is particularly relevant in the case of genes involved in protein refolding (Fig. 7).

It has been proposed that the existence of genes commonly regulated at the mRNA stability level (post-transcriptional regulons, [33]] is due to the existence of RBPs that affect the stability of their mRNA targets [21], [22]. Those proteins have been experimentally studied by performing immunoprecipitation and microarray analysis of the RNA bound fraction [21]–[23]. Their RNA sequence binding motifs have been determined either experimentally in vitro or in vivo [21]–[23] or by computer searches using newly developed algorithms [21]–[24]. Here we have performed, for the first time, a search for overrepresented RBPs motifs in clusters of genes that are characterized by having a common mRNA stability profile during a time course. We have found that some of the gene clusters include statistically overrepresented Pub1, Npl3, Puf3 and Puf4 targets (Fig. 5). This is an additional proof of the existence of post-transcriptional regulons with an overlapping effect on gene regulation with classical transcription regulons. Our analysis has found some putative new binding motifs in other possible post-transcriptional regulons and thereby opens new paths for further research.

However, not all the genes are modulated at an important level by mRNA stability. Clusters 2 and 11 (with 35% of the analyzed genes) show almost identical experimental and theoretical RA profiles (Fig. 4). The number of genes with a high positive correlation between the theoretical RA (assuming no stability changes, see M & M) and the experimental RA (Fig. S2) is large, supporting that TR changes are the main cause of RA changes upon heat stress.

Figure S1

Figure S2

Figure S3

Table S1

Table S2

Table S3