Global Levels of histone methylation during the formation of quiescent cells
To investigate the histone methylation landscape during the development of quiescence, we grew yeast cells in glucose-containing rich medium and prepared lysates from cells in mid-log phase (growing), at the point of glucose depletion (diauxic shift), and from purified populations of quiescent (Q) and nonquiescent (NQ) cells isolated 7 and 14 days after inoculation [14]. Western blot analysis was then performed with antibodies against histones and histones modified by methylation (Fig. 1a). There was no change in the abundance of the core (H2A, H2B, H3) or variant (H2A.Z) histones between log cells and cells at the diauxic shift, and small decreases in the abundance of these proteins between diauxie and Q or NQ cells. In contrast to the global decrease in the levels of histone acetylation in both a mixed population of stationary phase cells [24] and in isolated quiescent cells (Additional file 1: Figure S1E) [16], the levels of H3K4 di-methylation (H3K4me2) and H3K36 di- and tri-methylation (H3K36me2, me3) were similar between log cells and purified Q and NQ cells (Fig. 1a; Additional file 1: Figure S1F). This is consistent with the report that these modified histones were present at high levels in unseparated stationary phase cells formed upon severe nutrient deprivation [24]. However, the levels of H3K4 tri-methylation (H3K4me3) decreased by about 50% in both Q and NQ cells and there was a significant loss of both H3K79 mono- and di-methylation (H3K79me1, me2) in Q cells compared to NQ cells (Fig. 1a, b; Additional file 1: Figure S1F). In contrast, the levels of H3K79 trimethylation (H3K79me3) increased at the diauxic shift and remained at a high level in both Q and NQ cells (Fig. 1a, b: Additional file 1: Figure S1F). Thus, while Q cells are distinguished from log cells by their loss of histone acetylation, they are similar to both log and diauxic cells in their retention of most forms of H3 methylation. Q cells are further differentiated from both log and NQ cells by their selective loss of H3K79me1 and me2.
To investigate when the levels of H3K79me1 and me2 began to decrease as Q cells developed, we probed lysates prepared from pure populations of Q and NQ cells that were isolated at 3, 5, and 7 days after culture inoculation (Fig. 1b). At day 3, cells had completed the final cell doubling that occurs after glucose depletion, and day 5 represented the midpoint in the development of the fully quiescent state (Additional file 1: Figure S1A) [14, 25]. H3K79me1 and H3K79me2 were present at very low levels in day-3 Q cells, while H3K79me3 levels did not significantly change in these cells throughout the sampling period (Fig. 1b). This phenotype appears to be a unique property of Q cells, as the levels of H3K79me3, H3K79me2, and to a lesser extent, H3K79me1, remained relatively unchanged in NQ cells isolated during the same time course (Fig. 1b). The loss of the mono- and di-methylation states in Q cells was specific to H3K79, as H3K4me1, H3K4me2, and H3K4me3 were present throughout Q cell development (Additional file 1: Figure S1B). Although the levels of Dot1, the H3K79 modifying enzyme, decreased in day-3 Q cells (Fig. 1c), the selective loss of H3K79me1 and H3K79me2 in Q cells suggests that either a Q-specific H3K79me1 and H3K79me2 demethylase was activated, or the activity of Dot1 was altered to promote only H3K79 tri-methylation. Alternatively, nucleosomes that contained H3K79me1 or me2 could have been selectively replaced with unmodified H3.
H3K4 and H3K79 methylation are regulated by H2B monoubiquitylation (H2Bub1), and H3K4 and H3K36 methylation are regulated by phosphorylation of the RNA polymerase II (RNAP II) C-terminal domain (CTD) on serine 5 (Ser 5) or serine 2 (Ser2), which mediate transcription initiation and elongation, respectively [21,27,28,29,, 26–30]. These modified forms of RNAP II recruit the lysine methyltransferases, Set1 and Set2, to chromatin and thus couple H3K4 and H3K36 methylation to on-going transcription. H2Bub1 disappeared at diauxie in response to glucose depletion, as previously reported, and was absent in both Q and NQ cells isolated after this event (Fig. 1d and data not shown) [31]. RNAP II (Rpb3-Myc), TBP, and Mediator subunit, Srb4, were present in both log and 7-day Q cells, consistent with retention of the pre-initiation complex (Additional file 1: Figure S1C, D) [16, 32]. However, these same Q cells contained only low levels of both CTD Ser5-P and CTD Ser2-P modified RNAP II compared to log cells (Additional file 1: Figure S1C). This result is similar to what has been reported in unseparated stationary phase cells [32], and supports the conclusion that both transcription initiation and elongation are generally repressed in mature Q cells. To determine when Ser5 and Ser2 CTD phosphorylation declined during Q cell development, we followed the two forms of RNAP II in purified Q cells isolated 3, 5, and 7 days after culture inoculation (Fig. 1e). Surprisingly, both RNAP II Ser5 and Ser2 phosphorylation remained at high levels in day-3 Q cells before decreasing in abundance by day 5. In contrast, Set1 was not detected in day-3 Q cells, and Set2 was present at reduced levels in Q cells from day-3 through day-7. (Fig. 1c). This suggests that H3K4me2/me3 and H3K36me2/me3 were established on chromatin by the coupled activity of RNAP II and the two histone methyltransferases shortly before or soon after diauxie, and then retained in Q cells in the absence of their usual regulatory signals.
Genome-wide localization of histone methylation and RNAP II in quiescent cells
We defined the genome-wide distributions of H3K4me3, H3K36me3, H3K79me3, and RNAP II on chromatin isolated from 7-day Q cells and compared the patterns to those found in log cells. A genome browser view of PTM occupancy across yeast chromosome III showed similar patterns in log and Q cells, with the majority of the PTM binding sites associated with gene coding regions (Additional file 1: Figure S2A). Moreover, an unbiased analysis across the genome revealed a strong correlation in all three H3 methylation profiles, as well as in the unmodified H3 profile, between the two cell types (Fig. 2b-e). This is in contrast to the weak correlation in the H3K36me3 profile between log cells and unseparated stationary phase cells formed following complete nutrient deprivation after their transfer into water [24]. Consistent with the genome-wide PTM profiles, log and Q cells contained similar numbers of gene binding sites for the three H3 modifications with each individual modification enriched on at least 35% of all identified S. cerevisiae ORFs (Additional file 1: Figure S2B). However, with the exception of H3K36me3, the average enrichment of H3K4me3 and H3K79me3 on genes in Q cells was lower than in log cells (Fig. 2f). This was not the result of decreased H3 occupancy in quiescent cells, as similar levels of H3 were present on chromatin in both log and Q cells (Fig. 2f).
In contrast to the genomic profiles of the H3 PTMs, the RNAP II genome-wide occupancy patterns in log and Q cells were different in several respects. First, as previously reported, there were significant differences in the distribution of the polymerase between the two cell types (Additional file 1: Figure S2A), and a reduced correlation in the RNAP II profile between log and Q cells (Fig. 2a) [16]. Second, Q cells contained over 900 fewer RNAP II gene binding sites than log cells (1868 vs. 914) (Additional file 1: Figure S2B). Moreover, the mean enrichment of RNAP II on genes was also reduced in Q cells, (Fig. 2f). Together, the results show that while transcription is globally repressed in quiescent cells, these cells retained some features normally associated with transcriptionally active chromatin.
Histone methylation and RNAP II occupancy profiles on genes in quiescent cells
We assigned the genomic H3 methylation and RNAP II binding sites to three subsets of genes: genes marked only in log cells (Log), only in Q cells (Q), and in both log and Q cells (Common) (Additional file 1: Figure S5A-D; ChIP-chip data processing). The majority of genes enriched with each H3 modification were common to both cell types, with a much smaller number of genes marked only in Q or only in log cells (Fig. 2g; Additional file 2: Table S2). In contrast, a larger percentage of the RNAP II marked genes were log cell specific (65%) (Fig. 2g; Additional file 2: Table S2). The log-only H3 methylation and RNAP II marked genes were predominantly associated with genes involved in cell growth and division, while the Q-only marked genes have roles in the cellular response to stress, protein catabolism, and energy production (Additional file 1: Figure S2C; Additional file 3: Table S3). These GO categories suggest that the PTMs and RNAP II were associated with genes that are actively transcribed in each cell type or have the potential to be transcribed.
Having observed that a significant number of genes were enriched for each H3 PTM in log and Q cells, we investigated whether the gene-association patterns of the modifications were different in the two cell types. We first visualized the distribution of the H3 PTMs across all genes in log or Q cells as a function of transcript length (Fig. 3a) [33, 34]. In both cell types, H3K4me3 was present around the transcription start site (TSS) and H3K36me3 and H3K79me3 were enriched in gene bodies, consistent with previously observed patterns in growing cells [35, 36]. To further assess the gene-association patterns of the PTMs, we examined the averaged H3 methylation profiles on genes in log or Q cells (Fig. 3b-d). In both log and Q cells, H3K4me3 was enriched around the promoter and 5’ coding region, while H3K36me3 and H3K79me3 occupancy was restricted to the coding region. The analysis also showed a reduced overall occupancy of both H3K4me3 and H3K79me3 on genes in Q cells, which was also observed in the unbiased genome-wide analysis (Fig. 2f). While the reduction in H3K4me3 occupancy on genes in Q cells correlated with the reduced global levels of this H3 mark, this was not the case for H3K79me3. Importantly, there was no redistribution of the H3 marks to non-canonical locations within genes, or to intergenic regions, during the development of Q cells.
In contrast to the distribution of the PTMs, the enrichment of RNAPII on genes was strikingly different between log and Q cells (Fig. 3e). In log cells, the gene-averaged Rpb3-Myc profile showed RNAP II enrichment throughout the coding region and into the 3’ UTR, with peak occupancy occurring around the transcription start site. In Q cells, RNAP II was also enriched throughout the coding region but with several distinct differences: promoter proximal RNAP II enrichment was absent, RNAP II enrichment increased at the 3’ transcription end site (TES), and the overall levels of gene-averaged RNAP II were lower in quiescent cells compared to log cells (Fig. 3e; Fig. 2f).
To investigate the gene association patterns of RNAP II in more detail, we defined the averaged distribution of RNAP II on genes in the three gene categories (log, Q, and common) in log and Q cells (Additional file 1: Figure S3A, B). In log cells, RNAP II association with Q-specific genes was significantly reduced, and similarly, there was minimal association of the polymerase with log-specific genes in quiescent cells. These results are consistent with the repression of transcription of the two groups of genes in cells where they should not be expressed. We divided the common genes into two categories: those with RNAP II enriched more highly in log cells, and those with RNAP II enriched more highly in Q cells. Despite the different levels of enrichment, both categories of common genes showed similar patterns of RNAP II distribution (Additional file 1: Figure S3A, B). In log cells, RNAP II was enriched on these genes in a pattern similar to that seen on log-specific genes (Additional file 1: Figure S3A, B left panels). However, in quiescent cells, the pattern of RNAP II association with the common genes differed in two respects from that associated with log-specific genes (Additional file 1: Figure S3A, B right panels). First, the polymerase remained associated with the common genes. Second, the pattern of RNAP II association with these genes was now similar to the pattern seen on Q-specific genes, showing a higher occupancy at the 3’ gene region. These data support the conclusion that there was a shift in the properties of RNAP II on the common genes during the development of quiescent cells, which could reflect changes in the dynamics of polymerase initiation and processivity during this period.
It was previously reported that RNAP II was present at a significant number of intergenic regions (IGRs) in unseparated stationary phase cells, and that this association might poise a subset of genes for rapid activation upon exit from quiescence [32]. We re-examined this issue in purified 7-day Q cells by measuring the distribution of Rpb3-Myc on IGRs that separate divergently transcribed genes (Additional file 4: Table S4). The data showed that 14.1% of RNAP II binding sites in Q cells were associated with IGRs, versus 7.9% of these sites in log cells. In contrast, a higher percentage of gene containing regions (GCRs) were associated with RNAP II in log cells (72.4%) than in Q cells (51.6%). Interestingly, the majority of IGRs with bound RNAP II in both log and 7-day Q cells (77, log; 74, Q) contained CUTs, SUTs, mRNAs involved in yeast transposon assembly and function, or transcripts from the long terminal repeat (LTR) of Ty elements (Additional file 1: Figure S3C; Additional file 4: Table S4). Moreover, the IGR regions occupied by RNAP II in Q cells were not present upstream of the so-called “rapid exit” category of genes [32], although they did include several genes with roles in chromatin remodeling and modification. The presence of RNAP II in the promoters of this latter group of genes might play a role in resolving the compacted genome of Q cells upon their re-entry into the cell cycle [37, 38]. Together, the data suggest that RNAP II is not poised at the majority of protein-encoding genes in Q cells and that its presence in intergenic regions is related to the production of noncoding RNAs.
The transcriptome of quiescent cells
While there is a large reduction in active transcription in mature quiescent cells, these cells have also been reported to contain a significant cohort of stored or sequestered RNAs important for their survival [17]. To examine this issue in more detail, we defined the transcriptomes of both 7-day Q cells and log cells using RNA-seq. As previously reported, global RNA levels were significantly lower in quiescent cells than in log cells, and both the overall number and abundance of specific transcripts were also reduced in Q cells (Fig. 4a, b; Additional file 1: Figure S4A) [16]. However, many RNAs could still be detected in Q cells (5105), although these cells contained 1100 fewer transcripts than log cells (6205) (Additional file 1: Figure S4A). A majority of the RNAs in both cell types were produced by RNAP II and fell into four main categories: coding region (ORF) RNAs, cryptic unstable transcripts (CUTs), stable unannotated transcripts (SUTs), and small nuclear RNAs (snRNAs) (Additional file 1: Figure S4A; Additional file 5: Table S5) [39]. Most of the RNAs were associated with ORFs (Additional file 1: Figure S4A) and represented full-length transcripts, suggesting that they arose from a complete round of transcription. The noncoding CUT RNAs predominated in log cells and SUT RNAs were more abundant in Q cells, while snRNAs were present in both cell types without bias (Additional file 1: Figure S4A). Venn analysis showed that the majority of the ORF RNAs were found in both cell types (3930), with <10% (417) of the Q cell RNAs specific to quiescent cells, and 27% (1472) of the RNAs in log cells specific to this cell type (Fig. 4c; Additional file 6: Table S6). As seen for all RNAs, the median level of ORF RNAs common to both cell types was lower in Q cells than in log cells, while a more profound difference in RNA levels was seen for transcripts unique to either log or Q cells in quiescent or growing cells, respectively (Fig. 4b).
Approximately 40% of the Q cell ORF transcripts (1723) corresponded to the RNAs that have been reported to be sequestered in protein-RNA complexes in quiescent cells [17], and these same RNAs were also present in the log cell ORF population (1755) (Fig. 4d, e). We asked if there was a correlation between genes marked with RNAP II and the presence of transcripts in the two cell types. There was a stronger correlation between RNAP II and transcripts in log cells than in Q cells. In log cells, 33% (1788) of ORF RNAs were associated with genes that had bound RNAP II, while in Q cells 19% (818) of ORF RNAs showed a similar association (Fig. 4f, g). Together, the results support the view that a significant fraction of the RNA population in Q cells was transcribed from genes in log cells and stored in Q cells, while a smaller fraction of the Q cell RNAs resulted from active transcription during the development of Q cells. In support of active transcription in Q cells, Q cell-specific RNAs were associated with genes involved with cell survival, including the response to stress, protein catabolism, and autophagy (Additional file 1: Figure S4B; Additional file 7: Table S7). In contrast, ORF RNAs found only in log cells encoded proteins dedicated to cell growth and division (Additional file 1: Figure S4B; Additional file 7: Table S7).
We queried genes associated with RNA only in log cells or only in Q cells for the occupancy of RNAP II and the three H3 PTMs by deriving average gene profiles (Additional file 1: Figure S3D-G). The PTM profiles were very similar to the profiles derived for genes in the absence of their transcription status (Additional file 1: Figure S3D-E; Fig. 3b-d). However, the RNAP II profile on Q genes did not show the canonical 5’ peak characteristic of active genes in log cells, although high levels of polymerase were associated with the coding region of these Q genes (Additional file 1: Figure S3G; Fig. 3e). This suggests that these Q genes underwent an earlier period of high transcriptional activity during their development, and that after the initiation of transcription was repressed, RNAP II was retained across the coding region in mature Q cells.
Correlation between RNAP II and H3 methylation in quiescent cells
The finding that many genes in Q cells retained RNAP II, H3K4me3, H3K36me3, or H3K79me3 raised the question of whether there was a correlation between the presence of RNAP II and the presence of the three H3 PTMs on genes in these cells. An unbiased analysis of these factors across the genome in log cells showed a strong correlation between the presence of RNAP II and H3K4me3, but little to no correlation between RNAP II and H3K36me3 or H3K79me3 (Fig. 5a). In Q cells, there was a much weaker correlation between the presence of RNAP II and H3K4me3, and similar to log cells, there was no correlation between RNAP II and H3K36me3 or H3K79me3 (Fig. 5a). Another significant difference between log and Q cells was seen in the correlation between H3K36me3, H3K79me3 and histone H3, with Q cells showing a stronger correlation between the PTMs and H3 than log cells (Fig. 5a).
A similar scenario was seen when genes were specifically queried for the co-enrichment of RNAP II and the three species of H3 methylation (Fig. 5b). Significantly, more genes in log cells were co-enriched for RNAP II and combinations of H3K4me3, H3K36me3, and H3K79me3 than in quiescent cells, while more genes in Q cells were co-enriched for the PTMs in the absence of RNAP II. Together, the data support the view that a large fraction of the H3 marks on genes in mature Q cells were initially established in log phase cells or during the development of quiescence, and then retained.
Gene expression and histone methylation patterns on individual genes during the development of quiescent cells
Because pure populations of quiescent cells can be isolated shortly after diauxie [14], this allowed us to investigate gene expression and histone methylation patterns in Q cells during their development. We isolated Q cells at two-day intervals after diauxie (3, 5, and 7 days after culture inoculation), and examined a subset of genes for the levels of transcripts they produced, and the occupancy of RNAP II and the three H3 PTMs. We used three groups of genes for analysis: genes associated with transcripts present only in log cells, only in Q cells, or in both log and Q cells. Distinct patterns emerged for the different classes of genes.
The log-only expressed genes, PMA1 and CLN3, were associated with high levels of RNAP II, H3K4me3, H3K36me3, H3K79me3, and RNA in log cells, consistent with their active transcription (Fig. 6a; Additional file 1: Figure S6A and S7D). However, in Q cells isolated 3 days after inoculation (D3 Q), the levels of RNA and RNAP II associated with these genes had decreased to background levels. The levels of the RNAP II Ser5- and Ser2-phosphorylated CTD species (CTD-Ser5P and CTD-Ser2-P) also decreased at the log-only genes in the developing Q cells during the same period, reflecting the shutdown of their transcription (Additional file 1: Figure S7A). Interestingly, all three H3 PTMs persisted at these log-only expressed genes during Q cell development, albeit to varying extents: H3K4me3 occupancy decreased, H3K36me3 levels remained unchanged, and H3K79me3 levels increased, while no changes were noted for H3 occupancy (Fig. 6a; Additional file 1: Figure S6A, S7B,D). These data suggest that the marks were initially established on these genes during a period of active transcription in log cells and then retained in Q cells following the cessation of transcription.
The Q-only expressed genes, XBP1 and SNZ1, were associated with very low levels of RNAP II, H3K4me3, H3K36me3, H3K79me3, and RNA in log cells, correlating with their lack of transcription in growing cells (Fig. 6b; Additional file 1: Figure S6B, S7D). Three days after culture inoculation (D3 Q), there was a dramatic and simultaneous rise in the levels of RNA, RNAP II, and the H3 PTMs associated with the two genes in the developing Q cells. RNAP II occupancy then decreased over time, while H3K4me3, H3K36me3 and H3K79me3 occupancies remained high. Interestingly, the levels of RNAP II CTD-Ser5P and CTD-Ser2-P associated with the two Q cell genes also peaked in day-3 and day-5 Q cells before dropping to lower levels in day-7 Q cells (Additional file 1: Figure S7A). This suggests that there was an extended period of time after diauxie in which these genes had high transcriptional activity, which was then followed by a gradual shut down of transcription. The data further support the view that the histone methylation marks were deposited on the Q genes during this period of active transcription and retained at high levels despite the decrease in transcription.
The common genes represented the largest subset of genes associated with transcripts in both log and Q cells (Fig. 4c, 3930 genes). A comparison of the enrichment of RNAP II and the H3 PTMs on this class of genes between log and day 7-Q cells revealed that the majority of these genes showed a decrease in RNAP II, H3K4me3, and H3K79me3 occupancy in Q cells, with a less dramatic difference in H3K36me3 occupancy occurring between the two cell types (Additional file 1: Figure S7C). Two common genes, BAP2 and OLE1, were examined in more detail during Q cell development. Both genes were associated with approximately similar levels of transcripts in log and day-7 Q cells, but showed different patterns of RNAP II occupancy during Q cell development (Fig. 6c; Additional file 1: Figure S6C). At BAP2, RNAP II occupancy gradually decreased throughout the time course, similar to, but not as dramatic as the decrease in RNAP II occupancy at log-only transcribed genes (compare Fig. 6c and a; Rpb3-Myc; Additional file 1: Figure S7D). In contrast, the pattern of RNAP II occupancy at OLE1 was more similar to the pattern seen at Q-only transcribed genes, with RNAP II occupancy rising over log levels in day-3 Q cells before falling off in day-5 and day-7 Q cells (Additional file 1: Figure S6C; 6b). The occupancy profiles of RNAP II CTD-Ser5P and CTD-Ser2-P on the two common genes were also somewhat heterogeneous. However, the general picture was that their levels increased at the two common genes in day-3 to day-5 Q cells before decreasing over the remainder of Q cell development (Additional file 1: Figure S7A).
The levels of H3K36me3 and H3K79me3 at both BAP2 and OLE1 during Q cell development remained similar to the levels seen in log cells, while the H3K4me3 occupancy profile at these genes was more heterogeneous (Additional file 1: Figure S6A, S6C compared to Fig. 6a and c). H3K4me3 occupancy at BAP2 remained at log levels throughout the time-course, but H3K4me3 occupancy at OLE1 dropped by ~2-fold between day-5 and day-7 Q cells, similar to the global decrease in the levels of this PTM observed at the majority of common genes (Additional file 1: Figure S6C, S7C, D). Together, the data suggest that the three H3 PTMs were deposited at the common genes during an earlier period of high transcriptional activity in both log and developing Q cells and then differentially retained in Q cells as transcription was gradually shut down.
Development and maintenance of quiescence in mutants deficient for histone methylation
To assess the biological significance of the histone methylation marks to cellular quiescence, we examined the reproductive capacity of yeast mutants deficient for H3K4, H3K36, or H3K79 methylation during growth into stationary phase (Fig. 7a). We also examined the reproductive capacity of a strain deficient for H2B ubiquitylation, as this modification regulates H3K4 and H3K79 methylation. The ability of stationary phase cells to re-enter the cell cycle after prolonged glucose deprivation is a measure of chronological aging - cells with a shortened chronological lifespan (CLS) have reduced reproductive capacity, while cells with an increased chronological lifespan have enhanced reproductive capacity. CLS is related to the proportion of Q cells in stationary phase cultures because NQ cells have a reduced ability to re-enter the cell cycle [14, 40]. A wild type strain, containing an approximately equal mix of both Q and NQ cells, showed an ~50% reduction in reproductive capacity 4 days after diauxie. An H2B ubiquitylation mutant showed a 90% decrease in reproductive capacity during the same period, correlating with a dramatic reduction in the number of quiescent cells formed in this strain [41]. An H3K4 methylation mutant also lost reproductive capacity [40], although the ability of this strain to form colonies decreased more slowly over time, so that by day-18 only 20% of the cells retained reproductive capacity. This decreased survival also correlated with an increased proportion of non-quiescent cells in the stationary phase culture (data not shown). In contrast, the absence of H3K36 methylation had no effect on the ability of stationary phase cells to re-enter the cell cycle. Surprisingly, an H3K79 methylation mutant displayed a 50% increase in colony formation after diauxie, which was then followed by a gradual drop in reproductive capacity over prolonged incubation. This striking phenotype prompted us to examine this mutant in more detail. In contrast to wild type cells, a higher proportion of quiescent cells were formed in the hht-K79A mutant (Fig. 7b, c). Moreover, when purified wild type and hht-K79A quiescent cells were assessed for their long-term ability to survive in water, the mutant Q cells showed an enhanced ability to re-enter the cell cycle (Fig. 7d). Together, the data suggest that different histone methylation marks have distinct roles in the development and maintenance of the quiescent cell population, with H3K4me and H3K79me having contrasting roles in the regulation of chronological lifespan.