Sum1 interacts with Hst1 via Rfm1 14, and this Sum1/Rfm1/Hst1 complex represses a number of midsporulation genes. We therefore asked whether this complex was also involved in Sum1's initiation function, which is reflected in the observation that an orc2-1 mutation is synthetically lethal in combination with sum1Δ 1115. To this end, we investigated whether hst1Δ and rfm1Δ were also lethal with orc2-1. Significantly, an orc2-1 strain with hst1Δ was only able to lose an URA3-marked ORC2 plasmid on counterselective medium (5-FOA) if the strain had previously been provided with an ORC2-carrying plasmid with a different selection marker, but not with a vector control (Fig. 1), which was in agreement with previous work 15. Additionally, we found that rfm1Δ was synthetically lethal with orc2-1 (Fig. 1), indicating that that the whole Sum1/Rfm1/Hst1 complex was involved in the initiation function of Sum1.

In our previous work, we used bioinformatics analysis of genome-wide binding studies to identify regions in the Saccharomyces cerevisiae genome that are bound by both ORC and Sum1 and thus are good candidates for origins of replication that are regulated by Sum1 11. This analysis revealed eight regions that showed binding of both ORC and Sum1, and we have shown for three of these fragments that they are ARS elements and require Sum1 for full initiation capacity. To further validate these putative Sum1-regulated origins, we asked whether the genes downstream of the Sum1 binding sites were repressed by Sum1 and Hst1. To this end, we queried existing microarray data for the expression of these genes in sum1Δ and hst1Δ cells 25. This analysis showed that six of the genes were upregulated upon deletion of SUM1 or HST1 (Table 1), indicating that the localization of Sum1 to these fragments likely recruits Hst1 to repress the neighbouring gene through histone deacetylation. In the other two cases, Sum1 may solely act as a replication factor, because it does not repress the gene next door.

We next asked whether the ARS activity of these putative origins was regulated by Sum1 and Hst1. To test this, in vivo plasmid maintenance of these regions was measured in wild-type, sum1Δ and hst1Δ strains. A detailed description of the ARS fragments, the presence of ACS and Sum1 binding sites and their position relative to neighbouring genes is provided in Additional file 1: Schematic representation of the ARS sequences analyzed in this study. Our earlier study had shown that ARS1013, ARS1223 and ARS1511 depended on Sum1 for full initiation 11. Here, we found that sum1Δ and hst1Δ strains with CEN4-URA3 plasmids containing ARS446, ARS607, ARS1013, ARS1109, ARS1223 or ARS1511 as the sole origin displayed a significantly higher plasmid loss rate than the corresponding wild-type strain (Fig. 2A), showing that both Sum1 and Hst1 were necessary for the ARS activity of these origins. Interestingly, most origins showed a stronger dependence on Sum1 than on Hst1, reflecting the observation that many genes show stronger derepression by sum1Δ than by hst1Δ (Table 1, 25). The effect of sum1Δ and hst1Δ on ARS activity was specific to Sum1-bound origins, because a control origin that is not bound by Sum1, ARSH4, did not show an increased plasmid loss rate 11.

In several cases, the plasmid loss was too high to measure a plasmid loss rate, because primary transformants failed to grow upon restreaking. This was the case for ARS446 in a sum1Δ background and for ARS1013 in sum1Δ and hst1Δ strains (Fig. 2B).

One origin, ARS433, displayed dependence on Sum1, but not Hst1 (Fig. 2A). This reflected the fact that the neighbouring gene, NKP1, was not repressed by Hst1 or Sum1 (Table 1) and suggested that this origin was regulated by Sum1 independently of the Sum1/Rfm1/Hst1 complex.

We further extended our analysis to the ARS606 origin that we had previously identified as a Sum1-regulated origin by searching for co-occurrences of an ACS and a Sum1 consensus-binding site 11. ARS606 was highly unstable in sum1Δ as well as in hst1Δ strains (Fig. 2B), thus precluding the measurement of plasmid loss rates and showing that this Sum1-regulated origin also depended upon Hst1.

In contrast to other intergenic fragments, the region designated ARS447 was not capable of ARS activity, because wild-type and mutant strains transformed with ARS447 plasmids formed pinprick colonies that did not develop into viable cells after restreaking (data not shown). We tested ARS activity of a 0.5-kB as well as a 1.5-kB fragment comprising the putative ORC and Sum1-binding region, but both failed to support autonomous replication. This was in agreement with the fact that this region is designated ARS447 by 26, but not in the Saccharomyces Genome Database (SGD) (see Additional file 1: Schematic representation of the ARS sequences analyzed in this study).

In summary, these data showed that seven origins that were bound by Sum1, required both Sum1 and Hst1 for full initiation activity, suggesting that Sum1 recruited the Sum1/Rfm1/Hst1 complex to these origins, and that histone deacetylation by Hst1 contributed to initiation function.

The dependence of origin function on the HDAC Hst1 suggested that histone deacetylation was necessary for efficient initiation activity. We therefore determined whether acetylation levels at these origins increased in the absence of Sum1 or Hst1. hst1Δ has previously been shown to moderately increase H3 and H4 acetylation 27, but Hst1 specificity so far has not been determined. For this purpose we performed chromatin immunoprecipitations (ChIP) with antibodies against different histone H4 acetyl-lysine residues. Importantly, we found that acetylation of H4 K5 was significantly increased at most ARS in the absence of Sum1 and of Hst1 (Fig. 3A and Additional file 2: Significance levels for increased H4 acetylation in sum1Δ and hst1Δ strains). These results indicated that H4 K5 was a target for deacetylation by Hst1. In contrast, H4 K5 acetylation was not increased at a control region not bound by Sum1, ARSH4 (Fig. 3A). CDC20 served as an additional control region that showed an increase in acetylation with the H4 K5 antibody, which was only significant in one of the two experiments (Additional file 2: Significance levels for increased H4 acetylation in sum1Δ and hst1Δ strains), but not with the other antibodies (Fig. 3).

Furthermore, this observation suggested that Hst1 directly affected replication initiation at these origins by deacetylating histone H4 K5. There were three exceptions to this scenario. ARS433 plasmid maintenance was independent of Hst1, but ARS433 showed an Hst1-dependent increase in H4 K5 acetylation. Furthermore, Sum1 was required for full ARS activity of ARS606 and ARS1109, but apparently did not affect their acetylation state (Fig. 3A). Thus, there seem to be scenarios where the relationship between Sum1, histone deacetylation and initiation is more complex. In contrast to H4 K5, H4 K12 acetylation was not significantly increased at Sum1- and Hst1-dependent ARS elements. Whereas ARS1223 and ARS1511 showed a higher amount of H4 K12 acetylation in hst1Δ (P < 0.05 Additional file 2: Significance levels for increased H4 acetylation in sum1Δ and hst1Δ) cells, the effect for ARS607 was only significant in the sum1Δ strain (P < 0.05 Additional file 2: Significance levels for increased H4 acetylation in sum1Δ and hst1Δ strains). The other origins showed no increase in acetylation (Fig. 3B). This suggested that H4 K12 was not a major target for deacetylation by Hst1, and that this site did not contribute to the regulation of initiation.

Similarly, our analysis indicated that H4 K16 was not a general target of deacetylation by Hst1. Only ARS1223 and ARS1511 showed a significant higher H4 K16 acetylation level in both sum1Δ and hst1Δ yeast strains (Fig. 3C, Additional file 2: Significance levels for increased H4 acetylation in sum1Δ and hst1Δ strains).

ChIP analysis with a poly-acetyl-H4 antibody showed overall increases in the acetylation levels at these origins in sum1Δ and hst1Δ strains. Whereas ARS1223 and ARS1511 showed a several fold higher acetylation, the effect was weaker for ARS447 and ARS607 and only significant for the sum1Δ strain. ARS433, ARS446 showed no significant increase in H4 acetylation, and ARS606 and ARS1109 had the same state of acetylation in all three strains (Fig. 3D). This was consistent with the notion that H4 K5 was the main target of the histone deacetylase Hst1, whereas other histone H4 lysine residues were minor or no targets of Hst1.

In summary, this analysis suggested that Sum1 regulated initiation at selected origins by recruiting Hst1 to these regions.

The previous experiments raised the question whether increases in histone acetylation at origins by sum1Δ or hst1Δ were responsible for the loss of origin activity. To test this, we asked whether the activity at these origins was decreased in a strain in which the acetylatable lysine residues of the H4 N-terminus (K5, 8, 12 and 16) were mutated to glutamine. These mutations mimic a constant acetylation state of the respective histone H4 residue as is the case in the absence of the HDAC Hst1. We then analyzed the effect of this mutation on plasmid stability of the two plasmids that were most strongly affected by Sum1 and Hst1, ARS606 and ARS1013. Significantly, strains with ARS606 or ARS1013 plasmids showed a reduced growth rate in an H4 mutant strain as compared to wild-type (Fig. 4), indicating a loss of plasmid stability. In contrast, ARSH4, whose acetylation level did not alter in the absence of Hst1 or Sum1, did not show a reduced growth rate in the H4 mutant strain (Fig. 4). It has previously been reported that a simultaneous mutation of histone H4 K5, 8, 12, 16 to glutamine lengthens the cell cycle 28. However, the observation that we did not detect a difference in the growth ability of mutant cells compared to wild type with the control plasmid (ARSH4) indicated that the growth differences with ARS606 and ARS1013 were specific to these origins and not due to a generalized growth defect of the H4 mutation. In summary, this suggested that increases in the acetylation level in sum1Δ and hst1Δ cells were responsible for the reduced origin activity of these ARS elements.

The yeast strains and plasmids used in this study are listed in Tables 2 and 3, respectively. All yeast and E. coli manipulations were carried out according to standard protocols 35. The hst1Δ and rfm1Δ gene disruptions were performed using the KanMX cassette according to the guidelines of EUROFAN 36 and verified by PCR.

ARS fragments (length approx. 500 bp; ARS446 and ARS447 1500 bp) containing ARS and Sum1 consensus sequences (ACS: WTTTAYRTTTW; SUM1: DSYGWCAYWDW) were amplified via PCR from genomic DNA of wild-type cells and subcloned into the pGEM-T Easy vector (Promega) (for details, see Additional file 1: Schematic representation of the ARS sequences analyzed in this study). The vector pAE1076, which contains ARS1012, was used as a backbone for construction of the CEN4-URA3-ARS plasmids. It was digested with EcoRI and HindIII to release the ARS1012 fragment, and the new ARS fragments with compatible overhangs were ligated into the vector. The final constructs were verified by sequence analysis. Primer sequences are available from the authors upon request.

Plasmid loss rates were determined for wt (AEY2), sum1Δ::HisMX (AEY3358) and hst1Δ::KanMX (AEY1499) carrying CEN-URA3 plasmids containing the different ARS elements as follows. Yeast transformants were grown to stationary phase in liquid minimal medium lacking uracil, and cultures were used to inoculate YPD supplemented with adenine, histidine, leucine, lysine, tryptophan and uracil. Cells were grown for at least 12 doublings at 30°C with shaking. Before and after the incubation, equal amounts of the cultures were plated on minimal medium with or without uracil. The plasmid loss rate (L) was determined by measuring the fraction of cells containing the plasmid before (F_i) and after (F_f) incubation in full medium as 1–10^x with x = [log(F_i) - log(F_f)]/number of doubling times 37. The loss rate is therefore equivalent to the fraction of daughter cells that have received no plasmid during the previous cell division.

The following antibodies were used in this study (rabbit antiserum, upstate): Anti-acetyl-Histone H4 (Lys5 Catalog # 07-327 Lot # 30417); Anti-acetyl-Histone H4 (Lys12 Catalog # 07-595 Lot # 28885); Anti-acetyl-Histone H4 (Lys16 Catalog # 07-329 Lot # 32214); Anti-acetyl-Histone H4 (polyclonal antiserum Catalog # 06-866 Lot # 20667).

100 ml of yeast cells were grown to an OD₆₀₀ of 1 and cross-linked with 1% formaldehyde for 30 minutes with shaking at room temperature. Cells were harvested by centrifugation (3 minutes, 4,000 × g) and washed twice in 1× TBS. Pelleted cells were resuspended in 100 μl ice-cold lysis buffer (50 mM HEPES (pH7.5), 1 mM EDTA, 140 mM NaCl, 1% (v/v) Triton X-100 and 0.1% sodium deoxycholate) containing protease inhibitors and disrupted with glass beads. The cell supernatant was sonicated four times for ten seconds with 200 ms impulses, centrifuged, and the protein concentration was adjusted with lysis buffer. One aliquot was taken as an input control for the quantitative real-time PCR. Aliquots were precleared with protein G-agarose (Sigma) for 2 h at 4°C and incubated over night with 3 μl antibody. After incubation, the lysates were treated with protein G-agarose-beads. The immunoprecipitates were washed with 1 ml of the following buffers (ice-cold): 1. low salt solution (0.1% (v/v) SDS, 1% (v/v) Triton X-100, 2 mM EDTA, 20 mM Tris (pH 8.1) 150 mM NaCl), 2. high salt solution (0.1% (v/v) SDS, 1% Triton (v/v) X-100, 2 mM EDTA, 20 mM Tris (pH 8.1) 500 mM NaCl), 3. LiCl buffer (0.25 M LiCl, 1% (v/v) Nonidet P-40, 1% (w/v) sodium deoxycholate, 1 mM EDTA, 10 mM Tris pH 8.1), twice 1× TE. The samples and the input DNA were subsequently treated with elution buffer (1% (v/v) SDS, 0.1 M NaHCO₃), incubated over night at 65°C to reverse cross-linking and incubated 1 h with proteinase K (Roche). The DNA was extracted with chloroform-phenol-isoamylalcohol and precipitated with ethanol.

The ChIP and the input samples were used in different dilutions as templates for the PCR with SYBR Green RealMasterMix (Eppendorf). The reference dilutions were used to generate a standard curve that was taken to determine the DNA amount of the ChIP samples. Fragments of 211 to 356 bp in size were amplified (see Additional file 1: Schematic representation of the ARS sequences analyzed in this study). The real-time PCR setup was as follows: An initial denaturation step at 94°C for 2 minutes, followed by 45 cycles of denaturation at 94°C for 15 seconds, annealing at 56°C for 30 seconds and elongation at 68°C for 40 seconds. After cooling to 40°C for 2 minutes and then 1 minute at 50°C, the temperature was raised every 5 seconds in 1°C intervals up to 95°C. The template amount of the immunoprecipitated samples was measured as the mean value of three dilutions relative to the computer-calculated standard curve of the input reference. Evaluation of the data comprised two independent ChIPs with standard error of the mean (six data values). Separate P-values were calculated for changes in acetylation levels of the mutant strains in comparison to the wild-type from the three samples of each independent ChIP (see Additional file 2: Significance levels for increased H4 acetylation in sum1Δ and hst1Δ strains).