|Home | About | Journals | Submit | Contact Us | Français|
Saccharomyces cerevisiae Hmo1 binds to the promoters of ~70% of ribosomal protein genes (RPGs) at high occupancy, but is observed at lower occupancy on the remaining RPG promoters. In Δhmo1 cells, the transcription start site (TSS) of the Hmo1-enriched RPS5 promoter shifted upstream, while the TSS of the Hmo1-limited RPL10 promoter did not shift. Analyses of chimeric RPS5/RPL10 promoters revealed a region between the RPS5 upstream activating sequence (UAS) and core promoter, termed the intervening region (IVR), responsible for strong Hmo1 binding and an upstream TSS shift in Δhmo1 cells. Chromatin immunoprecipitation analyses showed that the RPS5-IVR resides within a nucleosome-free region and that pre-initiation complex (PIC) assembly occurs at a site between the IVR and a nucleosome overlapping the TSS (+1 nucleosome). The PIC assembly site was shifted upstream in Δhmo1 cells on this promoter, indicating that Hmo1 normally masks the RPS5-IVR to prevent PIC assembly at inappropriate site(s). This novel mechanism ensures accurate transcriptional initiation by delineating the 5′- and 3′-boundaries of the PIC assembly zone.
In Saccharomyces cerevisiae, 138 ribosomal protein genes (RPGs) encode 79 ribosomal proteins (RPs). RPG transcription constitutes ~50% of RNA polymerase II (Pol II)-mediated transcription in rapidly growing cells (1) and consumes an enormous amount of energy and protein resources. RPs are found in equimolar amounts in ribosomes, and their production is coordinately regulated in response to certain environmental conditions, mainly at the transcriptional stage.
During the past 10 years, increasing numbers of factors and/or mechanisms that regulate RPG transcription have been identified. Rap1, the most extensively characterized RPG regulator, binds to most RPG promoters (2,3) and activates transcription by recruiting the NuA4 histone acetyltransferase (HAT) complex and/or TFIID (4–6). Rap1 regulates transcription by forming a nucleosome-free region (NFR) in its target promoters (7–9). Abf1, which binds to fewer RPG promoters, is thought to function similarly to Rap1 in forming NFRs (10), although it is unknown whether Abf1 recruits TFIID and NuA4. Fhl1 also binds to many RPG promoters and recruits the coactivator, Ifh1, or the corepressor, Crf1, in response to environmental stimuli (11–14). Sfp1 regulates RPG transcription and expression of the ribosome biogenesis (Ribi) regulon (15,16) via its translocation between nucleus and cytoplasm in response to certain environmental stresses (17); however, its exact function remains unclear.
Hmo1, a high mobility group B (HMGB) protein, plays roles in Pol I and Pol II transcription, rRNA processing, DNA repair and chromosome/plasmid stability (2,18–25). Previous studies showed that Hmo1 binds to the promoter and coding regions of the 35S rRNA gene in a Pol I-dependent manner (2,20,22,26,27). Hmol binds to ~70% of RPG promoters, compared to Rap1 (93%) and Fhl1 (90%), and promotes Fhl1 binding to a subset of RPG promoters. Given that Hmo1 commonly targets both rDNA and RPGs, which are transcribed by two different RNA polymerases (Pol I and Pol II, respectively), one can speculate that it plays a crucial and specialized role in coordinating the transcriptional regulation and synthesis of ribosomes. However, little is known about the molecular function of Hmo1 at either Pol I or Pol II loci. The deletion of HMO1 (Δhmo1) has a milder effect on RPG expression than Δfhl1 (13), or mutating the Rap1 binding site (28,29). Furthermore, Δhmo1 produces different effects among RPGs, which do not necessarily correlate with the amount of Hmo1 binding, suggesting that the primary role of Hmo1 on RPGs may not be transcriptional activation.
In our previous study, we found that Δhmo1 caused an upstream shift in the transcriptional start site (TSS) of Hmo1-enriched RPG promoters and rescued the growth defects of certain sua7 (TFIIB) mutants, which, themselves, caused a downstream TSS shift (23). Such suppression phenotypes for sua7, which probably depend on a TSS shift in the direction opposite to that of sua7, have been found only with mutations in four polypeptides within the pre-initiation complex (PIC): the Tfg1 and Tfg2 subunits of TFIIF (30–33), and the Rpb2 and Rpb9 subunits of Pol II (34–37). Recent studies using a cross-linking technique demonstrated that multiple interactions between TFIIF and Rpb2, which may be reinforced by Rpb9, are critical for TSS selection (38,39). Presumably, mutations that affect these interactions may impair the specific function of Pol II that is required for selecting the appropriate TSS, or for stabilizing RNA–DNA hybrids during initiation, leading to an upstream TSS shift (30,38,39). In contrast to TFIIF, a direct interaction between Pol II and Hmo1 has not been observed (our unpublished data). Furthermore, in Δhmo1 cells, a TSS shift was only observed at Hmo1-enriched RPGs, while in tfg1, tfg2, rpb2 and Δrpb9 cells, a TSS shift was observed for most class II (Pol II-driven) genes, regardless of Hmo1 binding (30,37). Therefore, we suppose that the upstream TSS shift in Δhmo1 is caused by a different mechanism than in other mutants, and reflects a defect in a specialized function(s) of Hmo1 with respect to the regulation of transcriptional initiation at the RPG promoter.
The aim of this study was to unveil such a mechanism by determining how Δhmo1 induces an upstream TSS shift in Hmo1-enriched RPG promoters. From the results of extensive chromatin immunoprecipitation (ChIP) and primer extension analyses, we identified the IVR (intervening region) between the upstream activating sequence (UAS) and the core promoter (Core) of RPS5 as the binding site of Hmo1, and found that the IVR is nucleosome depleted. In wild-type (WT) cells, the PIC assembled at a site between the IVR and a nucleosome overlapping the TSS (+1 nucleosome), while it assembled within the IVR in Δhmo1 cells. These results strongly suggested that Hmo1 and +1 nucleosome determine the 5′- and 3′-boundaries, respectively, of a zone available for PIC assembly, thereby directing PIC assembly at a biologically relevant site.
Standard techniques were used for the growth and transformation of yeast (40). The yeast strains used in this study are listed in Supplementary Table S1. Detailed information for each strain is described in the Supplementary Data. The yeast culture conditions for each experiment are described in the figure legends. The detailed protocol used to construct the plasmids in this study is described in Supplementary Data. Oligonucleotides used in this study are listed in Supplementary Table S2.
Transcription start sites were mapped by primer extension analysis as described previously (23). The primers used were TK3212 (RPS5), TK3214 (ADH1), TK9589 (RPL27B), TK9911 (RPS5-mini-CLN2) and TK10595 (ADE2-C reporter). Electrophoretic images were acquired by exposing gels to imaging plates (BAS2500, Fuji Film), and the scanning of each lane was carried out using Multi Gauge version 3.0 software (Fuji Film).
ChIP analysis was conducted according to the Hahn laboratory protocol (http://labs.fhcrc.org/hahn/Methods/mol_bio_meth/hahnlab_ChIP_method.html) with minor modifications. Briefly, DNA was fragmented by sonication to an average size of 400–500bp for standard ChIP or 100–200bp for high-resolution ChIP. Immunoprecipitation was conducted using Dynabeads Protein G (Invitrogen) and monoclonal antibodies against FLAG (Sigma-Aldrich; M2), Pk (AbD Serotec; SV5-Pk1) and Myc (Santa Cruz; 9E10); or polyclonal antibodies against histone H3 (Abcam; ab1791), Rap1 (Santa Cruz; yC-19) and Sua7 (in this study, raised against full-length recombinant Sua7 in rabbit). Real-time quantitative PCR analyses were performed using a KAPA SYBR Fast qPCR kit (KAPA) and Mx3000P (Agilent Technologies). PCR conditions were: 95°C for 40s; 40 cycles of 95°C for 10s, 52°C for 30s and 72°C for 10s. Each experiment was conducted in triplicate and the average and SD for the ratio of immunoprecipitated DNA versus input DNA (IP/input) was calculated. The positions of amplified regions are depicted in each figure. The primer pairs used for PCR are described in the Supplementary Data.
For sequential ChIP analysis, the first immunoprecipitation was performed as for standard ChIP analysis, except that 5µg of anti-FLAG antibody and cell extracts containing 5mg of protein were used. After a final wash with TE, precipitates were eluted by incubating beads with 50µl of ChIP lysis buffer containing 3xFLAG peptide (200µg/ml; Sigma-Aldrich; MDYKDHDGDYKDHDIDYKDDDDK) at 4°C for 30min. Elution was performed four times in total, and the combined eluates were diluted with ChIP lysis buffer (to a concentration of 100µg/ml 3xFLAG peptide), and were subjected to a second immunoprecipitation using an anti-Pk antibody. All steps after the second immunoprecipitation were the same as for standard ChIP analysis.
Northern blot analyses were conducted as described previously (2). For the detection of the TEF2 and ADE2-C reporter genes, DNA fragments were amplified by PCR using the primer pairs TK6965–TK6966 (TEF2) and TK10425–TK10426 (ADE2-C) and were then 32P-labelled using random priming.
5′ RLM-RACE (RNA Ligase Mediated Rapid Amplification of cDNA Ends) analysis was conducted as previously described (41), using the FirstChoiceTM RLM-RACE Kit (Ambion) with total RNA from H2450 (WT) and YTK8276 (Δhmo1) strains. The experiment was conducted according to the instruction manual of the manufacturer (http://www.ambion.com/jp/techlib/prot/fm_1700.pdf). The nested PCR was conducted using universal primers, which bind the RNA adaptor region, and gene-specific primers TK10942 (outer)/TK11567 (inner) for RPS5 or TK11350 (outer)/TK9589 (inner) for RPL27B.
In a previous study, we showed that Δhmo1 caused an upstream TSS shift in Hmo1-enriched RPGs and suppressed the temperature sensitive growth of some sua7 mutants (e.g. sua7-R78C, -E62K), which caused a downstream TSS shift in many class II genes (23). To determine whether the suppressive effect of Δhmo1 is specific to sua7 mutants, we tested for a genetic interaction between HMO1 and RPB1, which encodes the largest Pol II subunit. The mutations, rpb1-N445S (42) and rpb1-R344A (43), are in or near to the active centre of Pol II, and cause a downstream TSS shift. As previously reported (42,43), both rpb1 mutants showed significant growth defects at all temperatures tested, and no growth at 37°C (Figure 1A). In contrast, Δhmo1 cells showed less severe growth defects at high temperature than at low temperature. Importantly, Δhmo1 suppressed the growth defect of the rpb1 mutant at 37°C (Figure 1A), as observed for the sua7 mutants (23).
Next, we tested the effect of these rpb1 mutations on the TSS in the RPS5 and ADH1 promoters by primer extension analysis in the same strains. As previously reported, both rpb1 mutants caused a downstream shift of the TSS in both promoters (Figure 1B and C, lanes 1–3 and 7–9); namely, the ratios of the intensities of the major band and lower bands were altered modestly (RPS5; −36 versus −22) or significantly (ADH1; −38 versus −27). In contrast, Δhmo1 caused an upstream shift of the TSS specifically in the RPS5 promoter (Figure 1B and C), one of the most Hmo1-enriched and transcriptionally Hmo1-dependent RPGs, but not in the ADH1 promoter, which binds Hmo1 weakly and is transcriptionally independent of Hmo1 (Figures 1B and C, compare lanes 7 and 10). Briefly, compared to WT cells, in Δhmo1 cells we observed a decrease in the intensity of two minor bands (−26 and −22, Figure 1B, left panel) situated below the most intensely stained band, corresponding to the major TSS (−36; marked with an asterisk) in the RPS5 promoter. We also noticed an increase in intensity of two bands (−71, −87), and the presence of three new bands (−133, −215, −225) for TSSs above the major TSS (Figures 1B and C, compare lanes 1 and 4). Although the upper bands in lane 10 appear stronger than those in lane 7 (Figure 1B), the ratio to the band at −38 (double dagger) was nearly identical in lanes 10 and 7 (Figure 1C).
Note that the upstream TSS shift in Δhmo1 cells can be also detected by another method, 5′ RLM-RACE. The 5′ RLM-RACE is a modified 5′ RACE to amplify selectively the 5′-end of full-length mRNA that contains a base corresponding to TSS. The TAP-dependent bands, which were amplified by PCR, reflect the intact 5′-end of mRNAs of RPS5 (Supplementary Figure S1A, lane 1 and 3). The result shows that TSSs of RPS5 were shifted upstream in Δhmo1 cells (Supplementary Figure S1A, compare lane 1 and 3). A similar result was obtained with the same analysis for RPL27B (Supplementary Figure S1B, compare lane 1 and 3), which showed a more drastic TSS shift than RPS5 in Δhmo1 cells (Supplementary Figure S2A). The results indicate that the upstream TSS shift that was identified by primer extension analysis occurred in Δhmo1 cells.
Importantly, Δhmo1 partly reversed the TSS shift in the RPS5 promoter (Figure 1B and C, compare lanes 1, 4, 5 and 6) but not in the ADH1 promoter (Figure 1B and C, compare lanes 7, 10, 11 and 12) in both rpb1 mutants. This effect was stronger in the Δhmo1 rpb1-R344A mutant (Figure 1B and C, compare lanes 1 and 5, 6), consistent with the observation that Δhmo1 suppressed the growth defect of rpb1-R344A more strongly than for rpb1-N445S (Figure 1A). These results suggested that Δhmo1 suppresses the growth defects of rpb1 mutants by reversing the TSS shift in certain Hmo1-enriched and transcriptionally Hmo1-dependent genes, such as RPS5.
While all mutant genes that are known to cause the upstream TSS shift encode subunits of Pol II or limited components of PIC (30–37), only Hmo1 is not a component of PIC. This suggests that the upstream TSS shift in Δhmo1 is caused by a different mechanism than in other mutants. Therefore, its mechanism may involve a specialized function(s) of Hmo1 with respect to the regulation of transcriptional initiation at the RPG promoter. To understand the mechanism behind the upstream TSS shift in Δhmo1 and other mutants, transcriptional phenotypes for Δhmo1, Pol II (Δrpb9) and TFIIF (tfg1-E346A) mutants were compared.
First, we compared the TSS profiles of the RPS5 and ADH1 promoters in three single mutants (Δhmo1, tfg1-E346A and Δrpb9) and three double mutants (Δhmo1 tfg1-E346A, Δhmo1 Δrpb9 and tfg1-E346A Δrpb9) with those in WT cells. As previously reported (31), the tfg1-E346A Δrpb9 double mutant exhibited more severe growth defects than either of the single mutants (tfg1-E346A, Δrpb9) at 25°C, 30°C and 35°C (Supplementary Figure 3A). Primer extension analyses revealed that the tfg1-E346A and Δrpb9 mutations caused upstream TSS shifts in the RPS5 and ADH1 promoters, while the shift in each promoter was enhanced in the double mutant (Supplementary Figure S3B and C). Consistent with a previous study (30), although the degree of TSS shift was slightly different between tfg1-E346A and Δrpb9, the positions of the upstream TSS in both promoters were nearly the same in both mutants (Supplementary Figure S3B and C, compare lanes 2, 6 and 3, 7, respectively), suggesting that the mechanism for the TSS shift may be similar in these mutants.
In contrast, the feature of TSS shift was quite different in Δhmo1 and tfg1-E346A mutants (Figure 2). Primarily, tfg1-E346A caused an upstream TSS shift in all promoters tested (RPS5, ADH1, SPT15, HTB1, GAL1, GAL10, HIS3, HIS4, SNR7, SNR14, SNR19 and SNR20) [in this study, and (30)], while Δhmo1 shifted the TSS specifically in the Hmo1-enriched RPG promoters, e.g., RPS5, RPL32 (23), and RPL27B (Supplementary Figure S2A). Furthermore, it is noteworthy that transcription from −51A in the RPS5 promoter was markedly enhanced by tfg1-E346A, but not by Δhmo1 (Figure 2B, lanes 1–4). Conversely, TSSs around −220 (−215 and −225) were induced uniquely by Δhmo1 (Figure 2B, lanes 1–4). Similar but weaker effects were observed in Δhmo1, Δrpb9 and Δhmo1 Δrpb9 cells (Figure 3, lanes 1–4). These results suggested that the mechanism(s) underlying the TSS shift in the Δhmo1 cells might be different from that in the TFIIF/Pol II mutants. However, direct evidence will be required to confirm this possibility.
Although the Δhmo1 tfg1-E346A double mutant had more severe synthetic effects on the TSS compared to either of the single mutants, we found no obvious synthetic growth defect for this double mutant when compared to the single mutants. A similar result was obtained for the Δhmo1 Δrpb9 double mutant (data not shown). Therefore, it seems unlikely that the upstream TSS shift itself is the major determinant for the growth defects in these mutants (single and double).
Because Δhmo1 decreased the binding of Fhl1 to the RPG promoter (2,22), it is possible that the upstream TSS shift in Δhmo1 cells is the result of the dissociation of Fhl1 from the RPS5 promoter. This was tested by analysing the TSS of RPS5 in Δfhl1 cells by primer extension analysis. Although Δfhl1 shifts the TSS of RPS5 modestly as compared with WT, the extent of the TSS shift was less severe in Δfhl1 cells than in Δhmo1 cells (Supplementary Figure S2B; TSSs upstream of −87 were not observed in Δfhl1 cells). Given that Δfhl1 decreases the Hmo1 binding to a subset of RPG promoters (22), the TSS shift in Δfhl1 cells may be related to a decrease in Hmo1 binding rather than the direct effect of the loss of Fhl1 function. However, Δhmo1 caused a less pronounced upstream TSS shift in an Hmo1-enriched RPG, RPS18A, whose binding to Fhl1 was not affected in Δhmo1 cells (2), than in the RPS5 promoter (Supplementary Figure S2C). Therefore, it cannot be excluded that Fhl1 has also a direct role in the selection of the correct TSS.
Besides RPG promoters, Hmo1 also binds abundantly to a subset of non-RPG promoters (2,22). Therefore, we tested whether Δhmo1 caused upstream TSS shifts in these promoters. Remarkably, the TSS of FET3, a non-RPG with strong Hmo1 binding activity, was not affected in Δhmo1 cells (Supplementary Figure S2D), suggesting that the function of Hmo1 at non-RPG promoters may be different from that at RPG promoters.
In S. cerevisiae, a ‘scanning model’ has been proposed to explain TSS selection by Pol II (44–46). In this model, following PIC assembly at the core promoter, Pol II starts scanning downstream and initiates transcription when it encounters an appropriate TSS. According to this model, two different mechanisms for an upstream TSS shift seem possible. The first mechanism involves normal PIC assembly, followed by a defect in the post-PIC assembly function of Pol II, while the second mechanism proposes an upstream shift of the PIC assembly site, reflecting a pre-PIC assembly defect but a normal post-PIC assembly function of Pol II.
The tfg1, tfg2, rpb2 and Δrpb9 mutants are thought to cause upstream TSS shifts via the post-PIC assembly defect, while Δhmo1 causes the TSS shift by a different mechanism, presumably the pre-PIC assembly defect. If a TSS shift is due to relocation of the PIC assembly site upstream in Δhmo1 cells, we predicted that insertion of a TATA element to induce upstream PIC assembly artificially would not alter the TSS significantly in Δhmo1 cells, but would cause a significant TSS shift in WT cells. As a result, the TSS pattern in WT and Δhmo1 cells would be similar. In contrast, if the TSS shift is caused by a post-PIC assembly defect, such as a change in Pol II activity, we predicted that insertion of an upstream TATA element would generate novel TSS patterns in both WT and Δhmo1 cells, and that these patterns would be different. To test this hypothesis, an ectopic TATA element was engineered at two upstream positions, −125 (TATA1) or −165 (TATA2), in the RPS5 promoter (Figure 3A). The original or modified RPS5 promoters were fused to a mini-CLN2 reporter gene, a non-functional CLN2 lacking part of the ORF (47), and inserted into a low-copy number plasmid. Importantly, the RPS5 promoter on the plasmid had almost identical properties to the chromosomal RPS5 promoter, including abundant Hmo1 binding, Hmo1-dependent Fhl1 binding, and TSS profiles in WT, Δhmo1 and/or Δrpb9 cells (Supplementary Figure S4A and B, and data not shown). Primer extension analyses were conducted using WT, Δhmo1, Δrpb9 and Δhmo1 Δrpb9 cells, each containing a mini-CLN2 reporter plasmid carrying the RPS5, RPS5-TATA1 or RPS5-TATA2 promoters (Figure 3B and C). In all strains tested, RPS5-TATA1 had no significant effect on the TSS, possibly because the endogenous PIC assembly site is close to this site (Figure 3B and C, compare lanes 1–4 with 5–8). On the contrary, RPS5-TATA2 caused a more modest upstream TSS shift in Δhmo1 cells than in WT cells (Figure 3B and C, compare lanes 1, 2 and 9, 10, respectively). As the result, the TSS patterns became similar in WT and Δhmo1 cells, at least in the region downstream of this insertion (Figure 3B, compare lanes 9 and 10). In contrast, RPS5-TATA2 drastically altered the TSS pattern in Δrpb9 and Δrpb9 Δhmo1 strains (Figure 3B and C, compare lanes 3, 4 and 11, 12). As the result, the TSS patterns of the RPS5-TATA2 promoter in Δrpb9 and Δrpb9 Δhmo1 became significantly different from those in WT and Δhmo1 (Figure 3B and C, compare lanes 9, 10 and 11, 12). These results suggested that Δhmo1 shifts the TSS by relocating the PIC assembly site upstream, while Δrpb9 shifts the TSS by causing a defect(s) in Pol II activity at a post-PIC assembly step.
Δhmo1 caused an upstream TSS shift in the Hmo1-enriched RPS5 promoter, but not in the Hmo1-limited RPL10 promoter (23). Therefore, to identify the RPS5 promoter region responsible for this shift in Δhmo1 cells, we constructed a series of chimeric promoters in which the UAS, Core or IVR of RPS5 and RPL10 were mutually exchanged (Figure 4A). These modified promoters were integrated into the ADE2 chromosomal locus in WT or Δhmo1 cells. Primer extension analysis for these strains clearly revealed that the IVR of RPS5 was required for an upstream TSS shift in Δhmo1 cells (Figure 4B) while the UAS and Core of RPS5 can be exchanged for those of RPL10 without blocking the TSS shift.
RPL27B showed almost the same properties as RPS5 with respect to Hmo1 binding, Hmo1-dependent transcription and Fhl1 binding. As expected, Δhmo1 also induced an upstream TSS shift in an additional chimeric promoter, constructed from the IVR of RPL27B and the UAS and Core of RPL10 (Figure 4A and B, lanes 17 and 18), implying that the IVR of Hmo1-enriched RPGs is a critical determinant for the TSS shift in Δhmo1 cells.
These results suggested that the IVRs in the RPS5 and RPL27B promoters are also required for abundant Hmo1 binding to these promoters. In fact, ChIP analyses revealed that Hmo1 binds abundantly to promoters containing the RPS5- or RPL27B-IVR, but not to those containing the RPL10-IVR (Figure 4C, panels 1 and 2). The roles of the IVR, UAS and Core in supporting abundant Hmo1 binding were examined by ChIP analyses of RPS5 promoters lacking each of these segments. The results clearly revealed that the IVR is essential, but the UAS and Core are dispensable, for Hmo1 binding (Figure 4C, panels 3 and 4). Previously, Hall et al. (22) reported that Hmo1 binding to the RPS11B promoter at the HIS3 chromosomal locus was dependent on Rap1 binding sequences. Therefore, we used ChIP analysis to test whether our deleted UAS (ΔUAS) construct contained a cryptic Rap1 binding site. The result confirmed that Rap1 was absent from this construct (Supplementary Figure S5A and B). Furthermore, the ΔUAS construct showed much weaker transcription than WT. These results excluded the possibility that a cryptic Rap1 (or Abf1) binding site would decrease Rap1-dependency of Hmo1 binding in our ΔUAS construct. It is possible that Rap1-dependency of Hmo1 binding differs at each genomic locus.
Further mapping analysis, using a series of promoter constructs, in which 40-bp segments in the RPS5 promoter were deleted systematically, found no segments that were indispensable for Hmo1 binding (data not shown). However, ChIP analysis using promoter constructs, in which the RPS5-IVR between the UAS and Core of RPL10 was deleted serially from upstream or downstream, revealed at least two non-overlapping sequences that supported abundant Hmo1 binding in the RPS5-IVR (−439 to −260bp and −259 to −127bp; Supplementary Figure S6A and B). In addition, the −319 to −199bp region (120bp), which overlaps these two regions, also supported full Hmo1 binding (data not shown). Thus, abundant Hmo1 binding to the RPS5-IVR occurs by the independent or cooperative functioning of multiple specific Hmo1 binding sites.
Alternatively, the length of the IVR might be more critical for Hmo1 binding than a specific DNA sequence because a correlation was observed between the length of the IVR and Hmo1-binding in our deletion analysis of RPS5-IVR (Supplementary Figure S6B) or in endogenous RPG promoters (22) (our unpublished data). To address this possibility, we constructed several promoter constructs that contained different DNA fragments, e.g. a non-promoter sequence from chromosome V (Chr. V) of S. cerevisiae or pBR322 plasmid, or triplicate RPL10-IVR, between the UAS and Core of RPL10 (Supplementary Figure S6C). Using these modified promoter constructs, the ChIP analysis revealed that while RPS5-IVR (−439 to −260; 180bp) bound similar levels of Hmo1 as full-length RPS5-IVR, two unrelated sequences of similar lengths showed modest (pBR322) or no (Chr. V) Hmo1 binding (Supplementary Figure S6D). Similarly, triplicate RPL10-IVR, which is longer than RPS5-IVR, could not bind Hmo1 (Supplementary Figure S6D). These results suggest that a specific sequence of DNA is more critical than the length of the DNA for Hmo1 binding.
The finding that multiple Hmo1 binding sites exist in RPS5-IVR raised the additional question of whether more than two Hmo1 molecules can bind to an RPS5 promoter simultaneously. To address this question, we conducted sequential ChIP analysis using strains expressing a Hmo1-FLAG tag and/or Hmo1-Pk tag. After sequential immunoprecipitation using anti-FLAG antibody (first) and anti-Pk antibody (second), DNA containing the RPS5-IVR was recovered efficiently (Figure 4D). This result clearly indicated that more than two Hmo1 molecules bind to an RPS5-IVR in vivo, although it is unclear whether they bind to different binding sites.
To confirm the binding of Hmo1 to the RPS5-IVR, we conducted high-resolution ChIP analyses for Hmo1-enriched RPG promoters (RPS5 and RPL27B). The results clearly showed that Hmo1 binds to the IVRs of these promoters (Figure 5A, panels a and c). In addition, similar results were obtained for Hmo1-enriched, non-RPG promoters such as HMO1 (Figure 5A, panel g), suggesting that Hmo1 tends to bind the IVR of its target promoters. Interestingly, Hmo1 even binds to the IVR of the Hmo1-limited RPL10 promoter (Figure 5A, panel e).
Next, we compared the Hmo1 binding site with those of other factors like Rap1, TFIIB (Sua7; a PIC component) and histone H3 (nucleosome) on the same promoters. ChIP analyses showed that Rap1 binds to the UAS of the RPS5, RPL27B and RPL10 promoters but not to any region of the HMO1 promoter (data not shown). This is consistent with our notion that Rap1-dependentcy of Hmo1 binding may differ depending on the locus. Remarkably, ChIP analyses also showed that the RPS5-IVR is nucleosome-depleted (Figure 5A, panel a) and that the PIC assembles at a site between the binding peaks of Hmo1 and a nucleosome (3′-side) on the RPS5 promoter (Figure 5A, panel b). Similar binding properties for Hmo1, TFIIB and histone H3 were observed for the other two Hmo1-enriched promoters (RPL27B and HMO1, Figure 5A, panels c,d and g,h, respectively).
Recent ChIP-seq studies revealed that many class II gene promoters have two well-positioned nucleosomes (−1 and +1) (48). The −1 nucleosome, located 150–300bp upstream of the TSS, regulates the access of transcription factors to this region, while the upstream boundary of the +1 nucleosome lies 10–15bp upstream of the TSS (49,50). As a result, a relatively wide NFR (~140bp) is formed between these two nucleosomes. Intriguingly, RPG promoters have a significantly broader NFR than other promoters, possibly due to the lack of the −1 nucleosome (48). Our results showed that, in Hmo1-enriched promoters, Hmo1 apparently binds to the position occupied by the −1 nucleosome in other promoters.
The spatial arrangement of the PIC, +1 nucleosome and Hmo1 within Hmo1-enriched promoters suggests that Hmo1 and the +1 nucleosome direct assembly of the PIC to a specific site. In this regard, Hmo1 is a novel transcription factor involved in determining the 5′-border of a region available for PIC assembly within the core promoter. In contrast to Hmo1-enriched promoters (RPS5, RPL27B and HMO1), the binding peaks of Hmo1 and PIC overlapped in the RPL10 promoter (Figure 5A, panels e and f). Although there was no evidence to exclude the possibility that Hmo1 and the PIC bind together at the same position in the RPL10 promoter, we assume this is due to limitations of the ChIP resolution in the relatively narrow RPL10-IVR. As an alternative possibility, Hmo1 and PIC could bind to the same position, but in different cell populations.
The binding profiles of Hmo1 and nucleosomes raised the possibility that Hmo1 may inhibit nucleosome formation on the IVR. To test this possibility, we used ChIP analysis to compare histone H3 binding profiles in WT and Δhmo1 cells. The results showed that the RPS5 promoter has a similar NFR in both cell types despite the slight increase in histone H3 binding in Δhmo1 cells (Figure 5B), suggesting that Hmo1 does not play a critical role in the formation and/or maintenance of NFRs.
The results described above suggested that Hmo1 and the +1 nucleosome determine the 5′- and 3′-border, respectively, of the PIC assembly zone. If this is so, Δhmo1 should disrupt the 5′-border of this zone, allowing ectopic PIC assembly on the IVR. An effect that induces ectopic PIC assembly could account for the upstream TSS shift in Δhmo1 cells. To test this possibility directly, ChIP analysis was conducted to determine the binding positions of several PIC components including TFIIB (Sua7), TFIIF (Tfg1), TFIIE (Tfa2) and TFIIH (Tfb3) on the RPS5 promoter in WT and Δhmo1 cells. The binding positions of these factors were shifted upstream in Δhmo1 cells (i.e. from position 8 to 7; Figure 6B, compare panels a, c, e, g with panels b, d, f, h, respectively). Our recent mapping analysis for core promoter elements in the RPS5 promoter revealed that the region corresponding to position 7 cannot bind PIC in the native context in the WT cells, although it can induce PIC assembly when artificially inserted into a different context (51). Therefore, the upstream shift of the PIC assembly site observed in Δhmo1 cells reflects the ectopic PIC assembly in this region (approximately position 7), where PIC intrinsically does not assemble in WT cells. In contrast, a similar upstream shift in the binding position of the PIC component (Sua7) was not observed in tfg1-E346A and Δrpb9 cells (Figure 6C), indicating that the upstream TSS shift in these mutants was caused by a post-PIC assembly defect. Therefore, we concluded that Hmo1 cooperates with the +1 nucleosome to direct PIC assembly to a site between the IVR and the +1 nucleosome by determining the 5′- and 3′-boundaries of a zone available for PIC assembly.
In this study, we aimed to find a function for Hmo1 in the regulation of transcriptional initiation of Hmo1-enriched RPGs by determining how Δhmo1 induces a TSS shift in these genes. The results showed that: (i) the upstream TSS shift in Δhmo1 cells was due to a pre-PIC assembly defect, while the shifts in Δrpb9 and tfg1-E346A cells were caused by a post-PIC assembly defect; (ii) Hmo1 binds over a broad region corresponding to the RPS5-IVR, which is nucleosome-depleted; (iii) multiple Hmo1 molecules bind to the RPS5-IVR; (iv) PIC assembles at a site flanked by Hmo1 and the +1 nucleosome; (v) Hmo1 does not play a critical role in the formation and/or maintenance of NFRs, at least in the RPS5 promoter; and (vi) Δhmo1 causes an upstream shift of the PIC assembly site.
Based on these findings, we proposed a model for a novel function of Hmo1 on its target promoters (Figure 6D). Initially, certain activators on the UAS (e.g. Rap1), remove the nucleosomes around the IVR. In WT cells, Hmo1 then binds to nucleosome-free IVRs to inhibit ectopic PIC assembly in this region, thereby directing PIC assembly to a biologically relevant site in the core promoter (Figure 6D, WT). In contrast, in Δhmo1 cells, activator(s) facilitate PIC assembly at more proximal site(s) within the IVR, which are devoid of Hmo1 and nucleosomes, leading to an upstream TSS shift (Figure 6D, Δhmo1).
While Δhmo1 does not have a severe effect on NFR in the RPS5 promoter, it causes a slight increase in histone H3 binding to the core promoter region of RPS5 (Figure 5B, compare region 8 between WT and Δhmo1). Because this region is occupied by PIC but not by Hmo1 in WT cells (Figure 5A), we assume that this slight increase in histone H3 binding in Δhmo1 cells may be caused by a decrease in PIC binding. However, we currently cannot exclude the opposite possibility that the increase in histone H3 binding may cause an upstream shift of the PIC assembly site. In this case, Δhmo1 would somehow allow invasion of the +1 nucleosome into the core promoter, pushing the PIC towards a more upstream site(s) (Figure 6D, Δhmo1).
While the finding that Hmo1 specifically binds to the NFR on Hmo1 target genes suggests that a nucleosome-free state is a pre-requisite for Hmo1 binding, we have not yet been able to identify specific cis-element(s) for Hmo1 binding in the mapping analysis, possibly because there are multiple binding sites for Hmo1 in RPS5-IVR. Although the IFHL motif and/or GGY(n) repeat were proposed as a binding site for Hmo1 by bioinformatic approaches (22,52), not all Hmo1-enriched RPG promoters contain these element(s) (our unpublished data). Furthermore, deletion of the IFHL motif reduced Hmo1 binding only modestly (22) (our unpublished data). Previously, Hmo1 was isolated in a yeast one-hybrid screen as a CAG repeat binding protein (53). Notably, the CAG repeat and IFHL motifs [or GGY(n) repeat] are GC-rich. In addition, the GC-content in the IVR of Hmo1-enriched RPGs (48.6%) is significantly higher, on average, than in Hmo1-limited RPG (36.1%) (our unpublished data). Therefore, we speculate that Hmo1 may recognize sequences with a relatively high GC-content, as represented by the IFHL motif within the NFR, and conversely that Hmo1 may be excluded from core promoter regions, which are very AT-rich. This speculation seems to be consistent with the results in Supplementary Figure S6C and D.
In previous studies, spt was identified as a suppressor of defects associated with insertions of the Ty1 transposon or δ sequence into the HIS4 or LYS2 promoters (54–56). Despite the lack of direct evidence, it is likely that relocation of the TSS from the Ty1 or δ sequence to the original HIS4 or LYS2 promoters is caused by a shift in the PIC assembly site in some spt mutants. It is possible that mutations in TBP (SPT15), SAGA (SPT3, SPT7, SPT8 and SPT20) and Mediator (SPT13) might cause this phenotype through changes in the sequence specificity of PIC components, not due to the unmasking of the NFR. Notably, SPT2 encodes a yeast HMG-like protein and participates in the repression of cryptic transcription in the coding region (57,58), showing similarities to Hmo1. However, it is likely that the spt phenotype is due to destabilization of nucleosomes in spt2 cells (57), and such a defect has also been described in spt6 and spt16 cells (59). Similarly, mutations in histones (SPT11 and SPT12) and in the transcriptional regulators of histone genes (SPT1, SPT10 and SPT21) may produce the spt phenotype by a similar mechanism. In contrast, Hmo1 does not severely affect the position/stability of nucleosomes, but rather functions as if it replaces the function of nucleosome to inhibit the access of various transcription factors. Consistently, Δhmo1 did not show the spt phenotype (our unpublished data). Considering these differences, we propose that Hmo1 is a novel type of transcription factor that directs PIC assembly to a physiologically relevant site by a novel mechanism.
Considering the fundamental importance of nucleosomal structures for cell growth, it is difficult to determine the specific roles of −1/+1 nucleosomes in PIC assembly. Nevertheless, the precise positioning of these nucleosomes suggests that they would determine the 5′- and 3′-boundaries, respectively, of the zone for PIC assembly (i.e. NFR). In a subset of RPG promoters, Hmo1 binds to a region that is usually occupied by the −1 nucleosome in other promoters, while Δhmo1 allows invasion of PIC assembly in this region. These results indicate that Hmo1, instead of the −1 nucleosome, would determine the 5′-boundary of the zone for PIC assembly, at least in Hmo1-enriched RPG promoters, whereas the +1 nucleosome would still determine the 3′-boundary even in these promoters. To our knowledge, this is the first experimental evidence to show that there is indeed a 5′-boundary at the zone for PIC assembly and that the boundary is formed by a protein other than histones, at least in some promoters. Currently, PIC components are thought to bind to core promoters via the recognition of defined (or ill-defined) cis-elements and/or histone modifications of −1/+1 nucleosomes (49). In this context, the mechanism described above, by which certain factor(s) restrict the zone for PIC assembly, should provide an additional layer of specificity to the system, ensuring that PIC can be assembled only at a physiologically relevant site.
Besides a role in directing PIC assembly to an appropriate site, Hmo1 may also promote PIC assembly itself, since the binding of PIC components decreased significantly in Δhmo1 cells (Figure 6B). In such a role, Hmo1 could facilitate PIC assembly by recruiting and/or stabilizing TFIID on its target promoters, as proposed for the −1 and +1 nucleosomes (49), either via direct interaction with TFIID subunits (23), or through Fhl1/Ifh1 coactivators (2,22). In fact, some yeast (Nhp6a/b) and human (HMGB1/2) HMGB proteins are known to activate transcription by stabilizing the TBP/TFIID–TFIIA promoter complex (60,61).
As another mechanism, Hmo1 might promote transcription by bending or looping a promoter DNA (62). A recent study, using cryo-electron microscopy, revealed that Rap1 on the UAS associates with TFIIA/TFIID at the core promoter, resulting in the looping-out of the region between the UAS and core promoter (i.e. IVR) (63). Intriguingly, the IVRs of Hmo1-enriched RPGs are significantly longer (approximately twice) than those of Hmo1-limited RPGs (22) (our unpublished data). Therefore, Hmo1 might promote and/or stabilize loop-formation, thereby helping Rap1 to affect TFIIA/TFIID efficiently from a distance in RPG promoters containing long IVR. Consistent with this, a TFIIA mutant, toa1-2 (K255A R257A K259A), which has defect in loop-formation (63) or in TFIIA–TFIID interaction (64), showed synthetic growth defects with Δhmo1 (23). At present, although there is no evidence to show that Hmo1-mediated loop-formation occurs in vivo, DNA-looping mediated by Top2 and Hmo1 was proposed to prevent chromosome fragility of variously transcribed intergenic regions in S-phase (21).
In summary, we found that Hmo1 plays a novel role in transcription by forming the 5′-boundary (instead of −1 nucleosome on many other promoters) for the PIC assembly zone on a subset of RPG promoters. Intriguingly, Hmo1 binds to 35S rDNA but only to nucleosome-free (i.e. actively transcribed) repeats in this gene (26). These alternate localizations of Hmo1 and nucleosomes to both RPG and rDNA loci indicate that Hmo1 may have specialized functions that cannot be replaced by nucleosomes. An attractive hypothesis is that the novel function of Hmo1 may play an important role in the coordinated synthesis of RP and rRNA under various environmental conditions. Further studies will be required to define more precisely the roles of Hmo1 in transcription at these two loci.
Supplementary Data are available at NAR Online.
The 2009 Strategic Research Project (No. W2104) of Yokohama City University, Yokohama Academic Foundation, Japan Society for the Promotion of Science; the Ministry of Education, Culture, Sports, Science and Technology of Japan; CREST, Japan Science and Technology Corporation. Funding for open access charge: The Ministry of Education, Culture, Sports, Science and Technology of Japan.
Conflict of interest statement. None declared.
We would like to thank Drs H. Iwasaki, T. Wada, M. Imashimizu and other members of our laboratory for advice and helpful discussions. We also thank Drs R. Young and F. Winston for supplying the yeast strains; S.W. Ki and K. Ohtsuki for making yeast strains and plasmids; H. Takahashi for making anti-Sua7 antibody; and H. Hirano for kindly allowing us to use the real-time PCR machine.