Secretory Stress Boosts HAC1 mRNA Abundance
To define the basic circuitry of signal transduction in the UPR, we evaluated the HAC1 mRNA processing step in a quantitative manner. To this end, we induced the UPR with either dithiothreitol (DTT) or tunicamycin (both agents that cause protein misfolding selectively in the ER) and monitored HAC1 mRNA by Northern blot analysis (A). In agreement with previous results, we observed rapid and efficient splicing of HAC1 mRNA, as apparent from the conversion of unspliced HAC1u mRNA (u for UPR-uninduced) to spliced HAC1i mRNA (i for UPR-induced). Quantitation of the results shows that the relative abundance of HAC1 mRNA (the sum of HAC1u and HAC1i mRNAs) remained unchanged over at least 12 h (A; unpublished data). These data demonstrate that acute induction of unfolded proteins triggers a simple on/off switch that controls HAC1 mRNA splicing.
ER-Distal Secretory Stress Boosts HAC1 mRNA Abundance
In light of these observations, we were surprised to find that blocking the secretory pathway distal to the ER resulted in a pronounced increase in HAC1
mRNA abundance. As shown in B, HAC1
mRNA levels increased 3- to 4-fold in mutant strains compromised at various steps in the secretory pathway when shifted to the nonpermissive temperature (sec12–1:
ER → Golgi, lanes 5–8; sec14–1:
intra-Golgi, lanes 9–12; and sec1–1:
Golgi → plasma membrane, lanes 13–16) (Novick et al. 1980
). Splicing was also induced, albeit to a lesser degree than was observed with DTT or tunicamycin treatment. The observed splicing suggests that blockages in ER-distal compartments of the secretory pathway lead to activation of Ire1p in the ER. Temperature shift alone only transiently induced HAC1
mRNA splicing and had no effect on HAC1
mRNA abundance (B, lanes 1–4). To determine if any disruption of the secretory pathway had similar consequences, we blocked earlier stages of protein traffic. Mutations that blocked protein entry into the ER had no effect (C: sec62–101,
lanes 13–16; sec63–201,
lanes 17–20) or only a mild effect (sec61–101,
lanes 9–12) on HAC1
Thus, a surveillance pathway operates to adjust HAC1
mRNA levels in response to altered conditions in the secretory pathway. In the experiments described above, we observed HAC1
mRNA induction only in sec
mutants that block transport distal to the ER, not in those that block protein entry into the ER. One common consequence of blocking the secretory pathway at later stages is that proteins in transit will eventually back up into the ER (Rose et al. 1989
; Chang et al. 2002
). This condition results in protein folding defects, thereby activating Ire1p, as indicated by the observed HAC1
mRNA splicing. From the data discussed above (A), however, we know that an accumulation of unfolded proteins alone is insufficient to trigger an upregulation of HAC1
mRNA, suggesting that an additional inducing signal is required.
HAC1 mRNA Induction Requires a Bipartite Signal
To determine the nature of this second signal, we sought conditions that induce HAC1 mRNA when combined with ER protein misfolding drugs. Canvassing different conditions, we found two scenarios under which wild-type (WT) cells can be induced to upregulate HAC1 mRNA: (1) ER protein misfolding combined with a temperature shift from 23 °C to 37 °C (A) and (2) ER protein misfolding combined with inositol starvation (B). Intriguingly, while ER protein misfolding and inositol starvation each activated the UPR individually (as shown by the activation of HAC1 mRNA splicing; A, lanes 5–8; B, lanes 1–4 and 5–8), neither stress alone was sufficient to cause HAC1 mRNA upregulation. Similarly, the temperature shift reproducibly caused a transient UPR induction (see B, lanes 1–4; A, lanes 1–4) but by itself did not affect HAC1 mRNA levels. Only the combination of ER stress with either temperature shift (A, lanes 9–12) or inositol starvation (B, lanes 9–12) led to an increase in HAC1 mRNA abundance. Subjecting cells to both temperature shift and inositol deprivation had no additive effect, nor did treating cells with both DTT and tunicamycin (unpublished data). Thus, HAC1 mRNA induction requires a bipartite signal, consisting of one input provided by unfolded proteins in the ER (UP signal), and the other input provided by inositol starvation or temperature shift (I/T signal).
HAC1 mRNA Induction Requires a Bipartite Signal and Is IRE1-Independent
The heat shock response is transiently induced by shifting cells from 23 °C to 37 °C. To determine whether the heat shock response is an important component of the I/T signal, we tested whether continued growth at 37 °C or expression of a constitutively active allele of the heat shock factor Hsf1p (Sorger 1991
; Bulman et al. 2001
) would substitute for the temperature shift described above. Constitutive expression of active Hsf1p (C, lanes 5–8) led to upregulation of SSA1,
a known target of the heat shock response (Slater and Craig 1989
), but did not substitute for the I/T signal for HAC1
upregulation. In contrast, continued growth at 37 °C (C, lanes 9–12) allowed for modest induction of HAC1
mRNA. Thus, elevated temperature elicits effects other than heat shock, which are important for HAC1
HAC1 Induction Is IRE1-Independent
The UP signal was experimentally induced by DTT or tunicamycin treatment of the cells. As Ire1p is a sensor of folding conditions within the ER lumen, we tested next whether Ire1p was required to transmit this signal. Surprisingly, it was not. HAC1 mRNA abundance was induced 2.6-fold in Δire1 cells (D, lanes 9–12), similar to the 3-fold induction observed in WT cells (A, lanes 9–12). These results show that a previously unrecognized Ire1p-independent surveillance mechanism must exist that monitors protein folding in the ER.
HAC1 mRNA Abundance Is Regulated Transcriptionally
Increase of HAC1 mRNA abundance could result from increased transcription, reduced degradation, or both. To distinguish between these possibilities, we constructed a reporter gene consisting of the HAC1 promoter driving transcription of the open reading frame encoding the green fluorescent protein (GFP) flanked by ACT1 untranslated regions (HAC1pro-GFP). The resulting heterologous GFP mRNA therefore contained no HAC1 mRNA sequences. Under conditions providing both the UP and I/T signals, the change in abundance of the GFP mRNA (A, lanes 5–8) mirrored that of the endogenous HAC1 mRNA (A, lanes 1–4), both in the kinetics and magnitude of the response. These data demonstrate that the observed increase in HAC1 mRNA abundance was caused by increased transcriptional activity of the HAC1 promoter.
Activation of the HAC1 Promoter Controls Increase in HAC1 mRNA Abundance
To further test this notion, we compared the rate of decay of HAC1 mRNA under both HAC1mRNA-inducing and noninducing conditions. To this end, we employed a strain bearing a temperature-sensitive allele of RNA polymerase II, which was subjected to either elevated temperature alone, or to both elevated temperature and DTT treatment. In both cases, polymerase II transcription ceased upon temperature shift, and mRNA decay was measured. As shown in B, the rate of decay of HAC1 mRNA was indistinguishable under the two conditions. Therefore, the increase in HAC1 mRNA abundance in response to the combination of UP and I/T signals is due solely to activation of the HAC1 promoter.
HAC1 Promoter Regulation Is Required to Survive Certain Stress Conditions
The results presented so far define a novel regulatory mechanism whereby cells adjust the amount of HAC1 mRNA. This mRNA is the substrate for the Ire1p-mediated splicing reaction, which in turn produces HAC1i mRNA that is translated to produce Hac1p transcription factor. We therefore asked whether elevated levels of HAC1 mRNA led to a proportional increase in the level of Hac1p. Quantitative Western blot analysis showed that this is indeed the case: when cells were treated with DTT and concomitantly shifted to 37 °C, the levels of Hac1p increased 3-fold (A, lanes 5–8), relative to the Hac1p levels observed in cells subjected to DTT treatment alone (A, lanes 1–4). Therefore, the transcriptional induction of HAC1 mRNA combined with Ire1p-mediated splicing results in elevated Hac1p levels, characterizing a new physiological state. Henceforth, we refer to this state as the “Super-UPR” (S-UPR).
HAC1 Promoter Regulation Is Required to Survive Stress
To assess the physiological role of the S-UPR, we sought conditions that would allow us to directly monitor the consequences of changes in HAC1 mRNA levels under otherwise identical growth conditions. To this end, we engineered a yeast strain unable to transcriptionally upregulate HAC1. In these cells, HAC1 mRNA expression was removed from the control of the HAC1 promoter and was instead driven by the heterologous ADH1 promoter (ADH1pro-HAC1), at levels closely approximating the uninduced HAC1 state (B, compare ADH1pro-HAC1, lanes 5–8, to HAC1pro-HAC1, lanes 1–4). Expression from the ADH1 promoter was constitutive, and the levels of HAC1 mRNA did not change significantly under the various inducing conditions described above. As expected, induction of the UPR in these strains led to efficient HAC1 mRNA splicing and Hac1p production. This strain therefore allowed us to fix the cellular Hac1p concentration to a level closely approximating the basal HAC1 expression state observed during the UPR.
We next assessed whether we could identify physiological conditions under which elevated HAC1 mRNA levels were required for cell growth. Therefore, we subjected WT cells and the engineered strain described above to the combinations of stresses described in . Cells expressing HAC1 from the endogenous or from the ADH1 promoter grew equally well on plates lacking inositol (C, left, first and third rows). This condition induces the UPR and requires the expression of at least a minimal amount of HAC1 mRNA, as Δhac1 cells fail to grow (C, left, second row). In contrast, only WT cells, which are able to upregulate HAC1 mRNA production, grew on plates lacking inositol and also containing tunicamycin. Cells expressing HAC1 mRNA only at the basal levels from the ADH1 promoter were nonviable on these plates (C, right, third row). As shown previously in B, this combination of stresses induces the S-UPR. The data therefore reveal that regulation provided by the HAC1 promoter is necessary for cells to survive certain stress conditions that otherwise are lethal.
Differential UPR Target Gene Induction by Elevated Hac1p Levels
To begin to characterize the cause for increased viability, we next determined differences in UPR target gene expression resulting from either UPR or S-UPR induction.
To this end, we used DNA microarray chip analysis to determine the complete mRNA profile of cells grown under UPR and S-UPR conditions. The results of this analysis are shown in A. Each spot represents the fold induction of a UPR target under UPR conditions (x-axis) or S-UPR conditions (y-axis) (see Materials and Methods
for definition of the UPR target set used in this analysis). UPR target genes for which the S-UPR has no additional effect should undergo equal induction under both conditions, and are expected to scatter around the diagonal, indicated by the dashed line. This was the case for many UPR targets. However, induction of a substantial number of genes was skewed to the top of the graph, indicating stronger induction under S-UPR conditions than under UPR conditions. These same data are displayed in B to highlight and categorize these differences. In the histogram, the x-axis represents the ratio of the induction of a target gene during S-UPR and UPR conditions, and the y-axis shows the number of genes with a given ratio. We have operationally divided UPR target genes into three classes, based on their fold induction during the S-UPR compared to their fold induction during the UPR. (1) Class 1 targets (, red bars) exhibit little if any difference in induction during the UPR and S-UPR (S-UPR induction / UPR induction < 2). Thus, the increased Hac1p during the S-UPR does not lead to enhanced transcription, indicating that for these genes the response is already saturated at UPR Hac1p levels. Class 1 targets include many of the known genes encoding ER lumenal chaperones (including KAR2, SCJ1, LHS1,
) and redox proteins (including PDI1, EUG1,
). (2) Class 2 targets (, blue bars) are induced to a 2- to 4-fold greater extent during S-UPR than during the UPR. Transcription of these genes is therefore roughly proportional to the Hac1p levels in the cell. Class 2 targets include YIP3,
involved in ER-to-Golgi transport, OPI3,
encoding a phospholipid methyltransferase, and the hexose transporters HXT12, HXT15, HXT16,
. (3) Class 3 targets (, green bars) are induced by the S-UPR greater than 4-fold more than by the UPR. Class 3 contains the UPR targets DER1,
involved in ER-associated degradation (Knop et al. 1996
; Ng et al. 2000
; Travers et al. 2000
), and INO1,
critical for membrane biogenesis (Hirsch and Henry 1986
Differential UPR Target Gene Induction by Elevated Hac1p Levels
Role for a Putative UPR Modulatory Factor
The increased transcriptional output under S-UPR conditions could occur for two reasons. It could be due to increased Hac1p concentrations in the cell, or it could result because an additional S-UPR-specific transcription factor is produced or activated (perhaps the same that regulates HAC1 transcription). It could also be due to a combination of these two scenarios. To distinguish among these possibilities, we determined the target gene induction profile in cells in which the HAC1 mRNA concentration was artificially elevated to a similar level as that found after S-UPR induction. We took advantage of a specific 15-bp deletion in the HAC1 promoter (HAC1proHI), which increases basal expression by about 3-fold, as compared to the endogenous promoter (C). In cells bearing a HAC1proHI-HAC1 gene (“HAC1proHI cells”), splicing of HAC1 mRNA was somewhat reduced upon UPR induction (47%, compared to 67% for WT); however, even with this reduction, HAC1proHI cells produced approximately 2.5-fold more spliced HAC1i mRNA than WT cells (C, compare lanes 3 and 4 to lanes 1 and 2). The increased levels of HAC1i mRNA led to a corresponding increase in Hac1p (D, compare lanes 3 and 4 to lanes 1 and 2). The amount of Hac1p produced by DTT induction of HAC1proHI cells is approximately the same as the amount of Hac1p produced during the S-UPR (compare D, lanes 2 and 4 with A, lanes 4 and 8).
The ability to set HAC1 mRNA levels to S-UPR levels allowed us to compare directly UPR target gene induction with the cellular Hac1p concentration being the only variable. We induced the UPR in both WT and HAC1proHI cells with DTT and determined the mRNA expression profiles. For each class of UPR target defined above, the expression analysis of UPR-induced WT and HAC1proHI cells is shown in E. In the histograms, the x-axis shows the ratio of target gene induction during the UPR driven by a high level of Hac1p from HAC1proHI cells compared to induction during the UPR in WT cells. The y-axis shows the number of genes at any given ratio. As expected, Class 1 targets (E, top panel) did not further respond to the higher levels of Hac1p produced in HAC1proHI cells. The majority of Class 2 and Class 3 targets (E, middle and bottom panels) also did not respond to higher levels of Hac1p (ratio less than 2), indicating that only raising the Hac1p concentration in cells is not sufficient to account for their full increased induction during the S-UPR. By contrast, ten of the 32 Class 2 and Class 3 targets were significantly induced (ratio greater than 2) in cells expressing high levels of Hac1p. For the Class 3 target DER1, high levels of Hac1p were sufficient to elevate expression to S-UPR levels (compare 8-fold induction in DTT-treated HAC1proHI cells to 9-fold induction in WT cells during the S-UPR). Otherwise, however, high levels of Hac1p did not fully reconstitute the induction seen during the S-UPR. For example, while the Class 3 gene INO1 was induced 7.5-fold more in the S-UPR than in the UPR, it was induced only 3-fold more by high levels of Hac1p, compared to normal levels. We conclude that elevated Hac1p levels are sufficient to selectively increase the induction of a few UPR targets, but that the full transcriptional program of the S-UPR predicts the production or activation of an additional transcriptional activator, which we term UPR modulatory factor (UMF).
To dissect further the UMF contribution during the S-UPR, we sought conditions under which UMF activity was the only variable. To this end, we induced the S-UPR in ADH1pro-HAC1 cells, which are prevented from achieving high level Hac1p expression, and compared the mRNA expression profile against the UPR in WT cells. In this analysis, Hac1p levels were approximately equivalent in the two conditions, so variations from the normal UPR transcriptional program reflect the activity of UMF. The results are shown in F, with the data displayed similarly to E: the x-axis shows the ratio of target gene induction during the S-UPR in ADH1pro-HAC1 cells, compared to induction during the UPR in WT cells, and the y-axis shows the number of genes at any given ratio. Not surprisingly, the induction of Class 1 targets (F, top panel) was unaffected: these are targets that are fully induced by even low levels of Hac1p and are not more induced during the S-UPR. Two Class 3 targets, YOR289W and YHR087W (both of unknown function) reach near WT S-UPR induction levels, without elevated levels of Hac1p; for these targets, UMF likely plays a leading role in their induction, with Hac1p having less influence. Most Class 2 and Class 3 targets (F, middle and bottom panels), however, do not reach full S-UPR induction levels in the absence of elevated Hac1p levels. For example, the Class 3 target INO1 is induced roughly 25-fold in ADH1pro-HAC1 cells during S-UPR conditions; while this is roughly twice the induction observed during the UPR, it falls far short of the 75-fold S-UPR induction in WT cells.
These results reinforce the in vivo requirement for high levels of Hac1p to survive S-UPR stress, demonstrated in C. Taken together with the data shown in E, we conclude that the full S-UPR transcriptional program results from a collaboration between elevated Hac1p levels and UMF, with the relative contribution from each varying among different target genes.