3 Results

3.1  Definition of a core region within the D element

The HM loci are flanked by the E- and I-silencers that contain a number of protein binding sites (Fig. 3.1A). At the HML-E-silencer, three functional elements essential for silencing have been defined within a region of 150bp. Deletion experiments uncovered the presence of a Rap1- and an ORC-binding site and the so called D element [Mahoney,et al., 1991] (Fig. 3.1A). In a strain lacking the HML I-silencer, deletion of any one of these regions leads to minor loss but deletion of any two of these regions leads to a complete loss of silencing at HMLα (Fig. 3.1B) [Mahoney,et al., 1991]. Unlike the Rap1- and the ORC binding site, the D element was so far uncharacterized and consisted of a large 93bp sequence stretch. Since all other essential silencer elements contain protein binding sites we hypothesized that the D element might also harbor an unidentified binding site for a silencing protein which we sought to identify in this study.

	Figure 3.1: Silencing properties of the D-element at the HML-E silencer.
	(A) Schematic representation of the HMLα locus on the left arm of chromosome III. Location and elements of the silencers HML-E and HML-I are indicated (RAP = Rap1 binding site, ACS = ORC binding site, D = D element, ABF = ABF binding site). (B) Redundancy of HML-E silencer elements. Loss of HMLα silencing in HML-E silencer deletion mutants was measured as loss of a-mating ability in a patch mating assay. All strains were HML-ΔI.

In a first set of experiments, we asked whether the D element could be narrowed down to a smaller DNA segment, because protein binding motifs usually occupy a sequence stretch of 10 to 20 basepairs. A deletion of this core element should lead to the same silencing defect as deletion of the complete D element. Similar to the original study, we used a strain that carried a deletion of HML-I and a mutation at the ORC binding site (ACS^-) and inserted 6-12bp deletions or mutations in D. In this situation, silencing is compromised such that any further weakening by disruption of D element function should lead to total loss of silencing at HMLα. Expression of HMLα in a MAT a strain leads to simultaneous presence of α- and a information, which results in a non-mating phenotype. Therefore, the HMLα silencing properties of individual D mutants were evaluated as their ability to mate and form diploids with a MATα tester strain. Indeed deletion of a 10bp segment within D termed D2, located 16bp centromere proximal to the ACS (position 133-143, numbering system based on [Feldman,et al., 1984]), mimicked the deletion of the full D element (Fig. 3.2A). Deletion of other segments within D did not alter the silencing properties of the ACS^- strain (Fig. 3.2A). These results suggested that sequences essential for D function were located within the D2 element.

	Figure 3.2: Identification of a D-element core region within the HML-E silencer.
	(A) D2 was the core element of D. Quantitative mating assays were performed to compare the effect on silencing of different D element deletions in a MATa HML-E ACS-ΔI strain background (AEY3395, lane 3). Lane 1: MATa HML-E wild-type (AEY2), lane 2: MATa HML-E ΔI (AEY3388). The mean values of at least three independent experiments are shown. (B) D2 was both necessary and sufficient for HML-E function. Loss of silencing in HML-E ACS-ΔI strains was measured for the 14bp sequence element containing the D2 element (D, AEY3395), mutations in the entire (d2, AEY3426), the first (d2a, AEY3430) or the second (d2b, AEY3434) half of the D2 element, or with the D2 element remaining as the sole D sequence at HML-E (D2, AEY3552).

We next asked whether D2 function could be abrogated by mutating rather than deleting the sequence, because a sequence deletion might not just remove a protein binding site, but could alter silencing by other means, for instance by changing nucleosome position and chromatin architecture. Therefore, we mutated every other basepair by transition in a 14-bp region that contained D2, thus maintaining the purin/pyrimidine composition of the original area. We found that the fully mutated 14-bp area (termed d2) caused HML derepression just like the D2 deletion (Fig. 3.2B). We used the same strategy to individually mutate the first or the second seven bp of this region (termed d2a and d2b). Both d2a and d2b lead to a complete derepression of HMLα (Fig. 3.2B), indicating that sequences necessary for D function were present in both elements.

As the D2 integrity was necessary for D function we next asked whether the presence of D2 alone without accompanying D sequence might suffice for D function. To this end we constructed a HML-E ACS^-ΔI strain that had all D element sequence downstream of D2 removed and assessed its HMLα silencing properties. When compared to a strain carrying the complete D element, no difference in mating capability was dectectable (Fig. 3.2B). This indicated that D2 was also sufficient for the execution of D function.

Taken together, these results showed that the D2 region was the core sequence of the D element. Furthermore, because this element is comparable in length to the Rap1 and ORC binding sites, this suggested that D2 contained a binding site for a protein (complex) essential for silencing.

3.2.1  sum1 caused HMLα derepression

We next sought to genetically identify the hypothesized D binding factor. One prediction for a mutation or deletion in the gene encoding this factor is that it causes derepression when silencing is compromised at HML, but not HMR a, because only HML, but not HMR a, contains a D element. More specifically, this mutation is expected to cause strong derepression only when HML silencing is weakened for instance by mutations in RAP or ACS of HML-E in an HML-ΔI background. In short, removal of the D binding factor is expected to have the same silencing phenotypes as mutation of its binding site in the HML-E silencer and should be epistatic to the binding site deletion.

In genetic crosses to characterize HML silencing, we observed that a deletion of SUM1 exactly matched the genetic predictions for the D binding factor. This observation was made in experiments to elucidate the role of the N-terminal acetyltransferase NatA at Orc1 [Geissenhoner, et al., 2004]. In the course of this study, double mutants of nat1, a member of the NatA complex, and sum1 were generated and routinely assayed for HM silencing defects. Surprisingly the nat1 sum1 mutant exhibited strong derepression at HML, but not HMR a (nat1, Fig 3.3A). Deletion of NAT1 alone leads to weakened HM silencing [Geissenhoner,et al., 2004] which is a pre-requisite to uncover redundant silencing mechanisms. Since Sum1 is a known DNA binding protein it was possible that the silencing phenotype of sum1 was mediated via the D element.

	Figure 3.3: SUM1 was required for HMLα silencing and was epistatic to the D element.
	(A) Repression of HMLα or HMRa in strains deleted for SUM1, NAT1 or both was measured by patch mating assays. (B) SUM1, but not RFM1 or HST1 was genetically linked to HML-D. HMLα silencing of sum1, rfm or hst1strains in combination with silencer element deletions at HML-E is shown by patch mating assays.

Since NAT1 deletion weakened HM locus repression by compromising ORC function [Geissenhoner,et al., 2004], the additional silencing defect of a SUM1 deletion suggested that Sum1 acted in a parallel pathway to ORC. We therefore investigated the effect of sum1 in the presence of mutations at HML-E. Based on the finding of [Mahoney,et al., 1991], full derepression of HMLα is only expected if any two of the three silencer elements are compromised. This can be achieved either by deleting the binding site in cis, or by mutating the respective gene coding for the binding protein in trans. We thus combined a deletion of SUM1 with individual deletions of any of the E silencer elements and assessed HML silencing. If Sum1 acted via D, the only cis-trans combination without silencing defect should be the one with a D element deletion since then only one silencer element would be impaired. Significantly, sum1 caused a strong loss of HML silencing when RAP or ACS of HML-E were deleted (Fig. 3.3B). However, sum1 did not generally weaken HML silencing, because it did not cause derepression when the D element was deleted. Thus, sum1 affected silencing of HML as predicted for the D binding factor in that it caused derepression when HML silencing was compromised and was epistatic to a deletion of the D element.

Previously, the Sum1 protein has been characterized as mitotic repressor for a set of genes that are upregulated in the middle stages of meiosis [Xie,et al., 1999]. In this function, Sum1 binds a regulatory DNA sequence, the middle sporulation element (MSE), which is present at the promoters of these genes and recruits the histone deacetylase Hst1 via a bridging protein Rfm1 [McCord,et al., 2003;Xie,et al., 1999]. Although wild-type Sum1 has so far not been implicated in silencing a mutant allele of Sum1, Sum1-1 is able to confer silencing to the HM loci in a sir background [Chi and Shore, 1996;Laurenson and Rine, 1991;Livi, et al., 1990]. In this role, Sum1-1 exerts its function by binding ORC and by recruiting Rfm1 and Hst1 [McCord,et al., 2003;Rusche and Rine, 2001;Sutton,et al., 2001].

Our results suggested that normal Sum1 indeed had a role in HML silencing and bound a sequence element within HML-D, although the MSE consensus sequence was not present at the D element. However, it was previously shown that not all Sum1 repressed middle sporulation genes contain an MSE at the promoter [Pierce,et al., 2003], and likewise, not all Sum1 repressed genes require Hst1 and Rfm1 for repression [McCord,et al., 2003]. Nonetheless it was conceivable that Sum1 exerted its silencing function at HML via these two proteins. To address this question we carried out epistasis experiments of RFM1 and HST1 with HML alleles as done with SUM1. However, while deletion of RFM1 showed a general weakening of HML silencing at each of the single silencer deletion strains, the deletion of HST1 did not cause HML derepression (Fig. 3.3B). This indicated that Rfm1 had a role in HML silencing beyond Sum1, and that Sum1 did not cooperate with Hst1 in this context.

3.3.1  In vitro binding of Sum1 to HML-E

To test the notion that the Sum1 protein was the D binding factor, we asked whether Sum1 was able to bind HML-E in vitro. To this end, we performed electrophoretic mobility shift assays (EMSA) with purified full length Sum1 (6xHis-tagged at the N-terminus) from E.coli and HML-E DNA. As a control, the purified 6xHis-Sum1 shifted DNA of a known Sum1 binding sequence, the MSE containing SMK1 promoter [Xie,et al., 1999], towards a slower mobility (Fig. 3.4B, lane 1 and 2).

	Figure 3.4: Binding of Sum1 to HML-E in vitro
	(A) Sum1 bound in vitro to HML-E, but not the INO1 promoter region. (Left) A radioactively labeled 220bp HML-E fragment was incubated without protein (lane 1) or with 0,1µM of bacterially expressed 6xHis-Sum1 (lanes 2-4). For competition experiments, unlabeled DNA of HML-E (specific competitor, lane 3) or a 210bp INO1 fragment (unspecific competitor, lane 4) was added. DNA-protein complexes were resolved on a polyacrylamide gel and labeled DNA visualized by autoradiography. (Right) Sum1 did not bind INO1 DNA, and bacterially expressed 6xHis-β-galactosidase (β-Gal) did not bind HML-E DNA. (B) Competition between SMK1 and HML-E for Sum1 binding. A radioactively labeled double-stranded 19bp fragment containing the MSE site of the SMK1 promoter was incubated without protein (lane 1) or with 0,1µM of bacterially expressed 6xHis-Sum1 (lanes 2-4). For competition experiments, unlabeled DNA of HML-E (specific competitor, lane 3) or HML-E DΔ was added. Upper arrow: protein-DNA complex, lower arrow: free DNA.

Importantly, 6xHis-Sum1 also caused a 220bp HML-E fragment to migrate more slowly (Fig. 3.4A, lane 1 and 2), indicating that Sum1 bound to HML-E. This binding was competed away by adding a molar excess of unlabeled HML-E DNA, but not by adding an unspecific 210bp INO1 promoter region fragment, indicating specificity for HML-E (Fig. 3.4A, lane 3 and 4). Sum1 also did not bind the INO1 fragment in an individual binding assay (Fig. 3.4A, lane 5 and 6). Also, the binding ability was unrelated to the 6xHis affinity tag, because 6xHis-tagged β-galactosidase was unable to bind to HML-E DNA (Fig. 3.4A, lane 7 and 8).

Since Sum1 bound two unrelated sequences, SMK1 (containing MSE) and HML-D (without MSE), we were interested to determine whether the two sequences could compete with each other for Sum1 binding. Significantly, the mobility shift of Sum1 with SMK1 DNA was competed away by addition of a molar excess of HML-E, but not by the same amount of HML-E lacking the D element (Fig. 3.4B, lanes 2 to 4), thus showing a competition between the two fragments for Sum1 binding.

	Figure 3.5: D-specific binding of Sum1 in vitro
	(A) Binding of Sum1 to HML-E required the D element. Mutant versions of HML-E were incubated with Sum1(+) or without protein (-) and gel-electrophorezed as in Fig. 3.4. HML-E DNA containing a mutation in the ACS site is termed ACS- (lanes 3,4,7,8), HML-E DNA with deletion of the 93bp D element is termed DΔ (lanes 5, 6, 7, 8). To maintain DNA size in the DΔ derivates, the deleted D element was substituted for the genomic 3´-region of equivalent length. All DNA fragments were ~220bp. (B) Binding of Sum1 to HML-E required the D2 element. Mutant versions of HML-E were incubated with Sum1(+) or without protein (-) as in (A). WT, a 134bp wild-type HML-E fragment containing the ACS and the D element (lanes 1, 2). D2Δ, HML-E without the D2 element (lanes 3, 4). DΔ, a 140bp HML-E fragment lacking the entire D element (lanes 5,6).

To test whether the Sum1-mediated mobility shift of HML-E DNA depended on the D element, we performed a series of EMSAs with mutated HML-E DNA. Whereas a shift was visible both with wild-type HML-E and HML-E with the ACS mutated, it was abolished when either the D element alone or the ACS and the D element together where mutated (Fig. 3.5A, lanes 1 to 8). This showed that Sum1 required the D element in order to bind to HML-E. We also attempted EMSAs of Sum1 with a 14-bp fragment containing the D2 element. However, Sum1 was unable to bind to this short sequence (data not shown), indicating that neighboring sequences within the D element were necessary for full binding of Sum1 in vitro.

To further test the involvement of D2 in Sum1 binding, we determined how the deletion of D2 affected the ability of HML-E to bind Sum1. Whereas a 134bp wild-type HML-E fragment bound Sum1 (Fig. 3.5B, lanes 1 and 2), binding was strongly decreased with a fragment of HML-E lacking 10bp of D2 (Fig. 3.5B, lanes 3 and 4). However, the binding was not as strongly reduced as with a complete deletion of the D element (Fig. 3.5, lanes 5 and 6), indicating that sequences surrounding D2 influenced the binding affinity of Sum1. In summary, these experiments showed that Sum1 bound HML-E in vitro in a D2-dependent fashion.

We next asked whether Sum1 bound to HML-E in vivo. To this end, we performed chromatin immunoprecipitation (ChIP) experiments with 6xmyc tagged Sum1. In the precipitates, we observed a weak, but consistent 2.5-fold enrichment of HML-E DNA in the presence of the α-myc antibody as compared to ChIPs without antibody (Fig. 3.6A) or in strains lacking myc-tagged Sum1 (data not shown). In the same precipitates, the SMK1 promoter, a known binding region for Sum1, was enriched 8-fold (Fig. 3.6A, WT), whereas the unrelated SSC1 gene promoter was not enriched (data not shown). We next tested whether the HML-E enrichment was dependent on the D element. We reasoned that if Sum1 bound the D element in vivo, there should be no antibody specific enrichment of HML-E in a strain deleted for the D element. To this end, we performed ChIP analysis in a strain with an additional deletion of HML-D. In these experiments the difference in HML-E enrichment in the fractions with or without antibody was indistinguishable (Fig. 3.6A, ΔD).

	Figure 3.6: In vivo localization of Sum1 at HML-E
	(A) Sum1 was associated in vivo with HML-E in a D-element dependent manner. ChIPs were performed on sum1 strains containing a 2μ plasmid carrying N-terminally 6xmyc tagged SUM1 under control of its own promoter (pAE1032). WT: wild-type HMLα (AEY2); ΔD: HMLΔDΔI (AEY3391). DNA was immunoprecipitated with (+) or without (-) anti-myc antibody and PCR amplified. A total of 1/50 or 1/100 of the input DNA (lanes 7, 8) or 1/2 (lanes 1, 4), 1/4 (lanes 2, 5) or 1/8 (lanes 3,6) of the immunoprecipitated DNA was analyzed. As a control, the promoter region of the SMK1 gene was PCR amplified. (B) ChIP was performed in sir4strains. Columns indicate the ratio of DNA enrichment with versus without anti-myc antibody (black columns: 6xmyc Sum1; white columns: untagged). The y-axis indicates fold enrichment.

The enrichment of Sum1 at HML-E was consistently weaker than that of Sum1 at the SMK1 promoter. One explanation is that the SMK1 promoter likely contains more than one Sum1 binding site [Pierce,et al., 2003], whereas HML-E has only one Sum1 binding site. To test the possibility that Sum1 ChIP at HML-E is sterically hindered due to heterochromatin, we performed co-immunoprecipitations in a sir4 strain. However Sum1 enrichment was not stronger at HML-E in a sir4 strain than in a wild-type. Quantitation showed that HML-E and SMK1 enrichment were 3- and 8-fold, respectively (Fig. 3.6B). Also, adding the 6xmyc tag to the C- rather than the N-terminus did not alter the ability to chromatin-immunoprecipitate Sum1 at HML (data not shown). However, the fact that we observed consistent enrichment, combined with the in vitro binding of Sum1 to HML-E DNA and the effect of sum1 on HMLα silencing strongly suggests that Sum1 bound in vivo to HML-E via the D element.

The presumed Sum1 binding site at the D2 element lies close to the ORC binding site of HML-E. Interestingly, other protein binding sites close to ACS sites of replication origins like one for Abf1 at ARS1 strongly influence the ability of such sequences to initiate replication [Marahrens and Stillman, 1992], raising the question whether Sum1 affected HML-E origin function. In its chromosomal location, HML-E does not initiate replication [Dubey,et al., 1991], because it is inactivated by replication forks emanating from centromere-proximal origins [Sharma,et al., 2001]. However, when removed from this context and placed on a plasmid, HML-E has ARS activity, meaning that it confers autonomous replication to plasmids lacking an origin. Sharma et al. (2001) showed that deletion of a sequence stretch including the D element abrogated the ARS activity of HML-E, indicating that D was required for ARS function.

	Figure 3.7: sum1Δ#SYMBOL#reduced the ARS activity of HML-E.
	Plasmid loss rates were determined in a wild-type (AEY2) and a sum1 (AEY3358) strain. Strains with plasmids carrying ARS H4 (pRS316), HML-E (pAE1119), HMR-E (pAE229) or the HMR-E synthetic silencer SS HMR-E (pAE298) as their sole origins were analyzed. The average loss rates obtained from three independent experiments are shown with corresponding error bars.

We now asked how Sum1 affected HML-E origin activity by measuring the stability of a plasmid carrying HML-E as the sole origin of replication in wild-type and sum1 strains. Notably, the sum1strain exhibited a more than two-fold higher loss rate of the HML-E plasmid than the wild-type strain (Fig 3.7). This suggested that Sum1 was required for full replication initiation efficiency of HML-E on a plasmid. Furthermore, sum1 strains grew more slowly than wild-type strains when selecting for the HML-E plasmid (data not shown), also indicating that plasmid transmission, probably through reduced origin initiation, was impaired. In contrast, sum1 did not affect the stability of plasmids carrying the wild-type or synthetic HMR-E silencers as origins. Also, sum1 did not affect plasmid stability of an ARS H4 plasmid (Fig. 3.7). These results showed that sum1 did not affect other plasmid functions, for instance CEN function. Also, the effect of sum1 was restricted to HML-E, which was predicted because the D element is not found in the other origins tested. Furthermore, it showed that sum1 did not simply impair weak origins of replication (like the synthetic HMR-E silencer). In summary, these results demonstrated that Sum1 showed a specific effect on origin function of HML-E.

The plasmid maintenance defect of sum1 strains with an HML-E-origin plasmid likely reflects a role of Sum1 in replication initiation at this origin. This observation prompted us to ask whether Sum1 might be required more globally for replication initiation and thus might constitute a novel replication initiation factor that aids ORC in initiation at selected chromosomal origins. Significantly, we observed that sum1 caused lethality in strains with mutations in the ORC subunits Orc2 and Orc5, since we were unable to recover double mutants in genetic crosses between sum1 and orc2-1 or orc5-1 strains (data not shown), which was in agreement with (Suter et al., 2004). The orc mutants on their own are temperature sensitive and show reduced firing of chromosomal origins and high plasmid loss [Fox,et al., 1995;Loo, et al., 1995]. sum1 orc2-1 double mutants were able to grow when provided with a URA3-marked plasmidcarrying ORC2. However, they were only able to survive on URA3-counterselective medium (5-fluoro-orotic acid, 5-FOA) when supplemented with plasmids carrying either SUM1 or ORC2 (Fig. 3.8A), showing that the lethality depended on these two genes and that sum1 orc2-1 strains were not inviable due to a germination defect. One interpretation of the synthetic interaction between ORC and SUM1 is that chromosomal replication initiation in the orc mutants is further impaired by the absence of Sum1 such that the cells are unable to survive.

We further assessed genetic interactions between sum1 and mutations in genes encoding other factors required for replication initiation (reviewed in [Bell and Dutta, 2002]).

	Figure 3.8: Genetic interactions between SUM1 and replication initiation components.
	(A) Synthetic lethality of orc2-1 and sum1. An orc2-1 sum1 strain carrying an URA3 marked ORC2 plasmid (pRS316-ORC2) was transformed with a SUM1 (pAE1032) or an ORC2 (pAE53) plasmid or the corresponding empty vectors. Its ability to lose the pURA3-ORC2 plasmid was tested on 5-FOA medium. (B) Synthetic growth defects of cdc6-1, cdc7-1 or cdc45-1 with sum1. Serial dilutions of several segregants from each cross were plated and incubated at the semi-permissive temperature of the respective cdc single mutant. For cdc6-1, strains AEY600, 3358 and AEY3537 to 3541, for cdc7-1 strains AEY3542 to 3546 and for cdc45-1 stains AEY373 and AEY3548 to 3551 were used. Incubation was 3 days for cdc6-1 and cdc7-1 and 6 days for cdc45-1. Cdc6-1 marked with an asterisk indicates the parental strain, which was not isogenic to the sum1strain.

Cdc6 is required in early G1 for chromatin binding of MCM proteins and formation of the pre-replicative complex (pre-RC) at origins of replication. Cdc7 is part of the DKK (Dbf4 dependent kinase) that is required for the G1/ S-phase transition, perhaps by phosphorylating MCM proteins. Cdc45 plays an important role in the transition from initiation to replication. It is required for association of the DNA polymerases with chromatin and colocalizes with the polymerases at the replication fork (see also chapter 1.6.2). We found that double mutant strains of sum1 with temperature-sensitive alleles of CDC6, CDC7 and CDC45 were viable, but showed a growth defect as compared to the single mutants at their respective semi-permissive growth temperature (Fig. 3.8B). Since these mutations impair replication initiation, our findings further supported the notion that Sum1 played a global role in initiation.

	Figure 3.9: Test for physical interaction of Sum1 and Orc2.
	(A) Co-immunoprecipitation of Sum1 and Orc2. Strains AEY1558 (-) and AEY3474 (6xHis-Orc2, +) carried a 6xmyc-Sum1 2μ-plasmid (pAE1032) and a HMLα (pAE1123) 2μ-plasmid. Precipitates were analyzed by SDS-PAGE and immunoblotting using anti-myc-antibody. Input (lanes 1,2), Immunoprecipitation (IP, lanes 3, 4), Supernatant (sup, lanes 5,6). (B) Co-immunoprecipitation of Sum1 and Orc2 in the presence of ethidiumbromide. Strains and experimental procedures were as described in (A). Where indicated, ethidiumbromid to a concentration of 100μg/ml was added to the lysate (Input, lanes 2, 4) prior to the addition of antibody (IP, lane 6, 8).

Our observation of a role for Sum1 in replication initiation and the genetic interaction between sum1and orc mutations might be due to a direct interaction of Sum1 and ORC. A previous study reported a weak interaction between Orc3 and Sum1 in vivo [Sutton,et al., 2001]. We also found a weak interaction between Orc2 and Sum1 by co-IP (Fig. 3.9A, lane 4). However this interaction is possibly mediated via DNA since it was abrogated upon addition of high concentrations of ethidiumbromide, which is thought to disrupt protein-DNA interactions (Fig. 3.9B, lane 5,6). This indicated that Sum1 may not interact directly with ORC. However it could also mean that subpopulations of Sum1 and ORC are located close to each other on DNA sequences.

3.6.1  Identification of genomic sites of combined Sum1 and ORC binding

A global role for Sum1 in replication initiation predicts a significant number of replication origins that are also bound by Sum1. To search for such sequences, we used the data from two previous studies that identified genomewide Sum1 and ORC binding sites using ChIP mediated microarray analysis [Lee,et al., 2002;Wyrick,et al., 2001].

	Figure 3.10: Search for intergenic regions that bind both Sum1 and ORC.
	Plot of p-values for Sum1 binding (Lee et al. 2002), p<0.01 versus ORC binding (Wyrick et al. 2001), p<0,05. Inserted diagram: all data points. The origin function of intergenic regions marked with an asterisk was tested below.

In both experiments, the authors used the same error model to convert the observed Cy5/Cy3 intensity ratios into p-values (the probability that such a ratio or larger could be observed from a non-binding event). In their large-scale analysis, they imposed a strict prescription (p<0.001) to reduce the number of wrong binding predictions (false positives) at the expense of a higher false negative rate (discarding true binding events). For our purposes, we considered those eight intergenic regions where p(Sum1)<0.01 and p(ORC)<0.05 (Fig. 3.10; Table 3.1).

Table 3.1: Genomic loci that are bound by Sum1 at p<0.01 and ORC at p<0.05

Of these, five were located upstream of a gene that is derepressed in sum1Δ more than 2,5fold as determined by [Bedalov,et al., 2003], suggesting that they constitute true Sum1 binding regions (Table 3.1, column 6).

In a complementary approach we used a binding-motif based sequence search to find origins that require Sum1 for full activity. Initially, we used the consensus sequence for ORC binding (WTTTAYRTTTW) [Broach, et al., 1983] and the sequence of the identified D2 element (TTTTCGGCACGGAC) and searched the genome for co-occurrance of the two sequences within an distance of 200bp or less. One problem in this approach was that there was no consensus sequence available for D2, so sequence variations in other possible D2 elements could only be considered upon allowing a certain number of random mismatches. Except one candidate (Table 3.2, line 1), we found a high number of hits that did not pass subsequent refining steps (see below).

Table 3.2 : Genomic loci of ORC and Sum1 consensus motif co-occurrence.

In a subsequent approach we searched genomewide for areas <200bp that contained the consensus motif of the MSE (DSYGWCAYWDW), a well characterized Sum1 binding site [Pierce,et al., 2003], and the ORC binding site. To exclude random sequences, we checked the resulting ~100 candidates for whether they were in vivo targets of ORC and Sum1. As above we used the data from two previous studies that identified genomewide Sum1 and ORC binding sites using ChIP mediated microarray analysis (see Table 3.2, (1) and (3)) [Lee,et al., 2002;Wyrick,et al., 2001]. Additionally, we utilized data on replication timing by [Raghuraman,et al., 2001] to find possible in vivo origins (2). Finally we checked whether expression of ORFs adjacent to the loci was upregulated in a sum1 strain (3) [Bedalov,et al., 2003]. Loci that fullfilled at least two of the above mentioned requirements were scored as possible candidates.

	Figure 3.11: Sum1 influenced origin function of ARS1013
	(A) Schematic representation of ARS1012 and ARS1013 located at the ORF YJL038C on chromosome X. The location of Ndt80 and Sum1 consensus sites (Pierce et al. 2003) and ACS matches is indicated. Bold lines represent fragments whose ARS function was tested. (B) SUM1 was required for ARS activity of ARS 1013 on plasmids. Strains AEY2 (WT) and AEY3358 (sum1Δ) were transformed with URA-CEN4 plasmids carrying either ARS1012 (pAE1076) or ARS1013-3 (pAE1081) as their sole origin. Transformants obtained upon transformation of ARS1013-1 or ARS1013-2 -URA-CEN4 plasmids (pAE1078, pAE1080) were not restreakable.

Among the known ARS sites that had been identified in the Sum1-ORC binding screen was ARS1013 (Fig. 3.10, Table 3.1) which mapped to the intergenic region upstream of ORF YJL038C [Wyrick,et al., 2001]. We asked whether ARS activity of ARS1013 was affected by Sum1 by testing ARS function of three overlapping ARS1013 fragments [Wyrick,et al., 2001] in wt and sum1Δ strains (Fig. 3.11A). Two fragments (ARS1013-1, -2) formed pin-prick transformants that failed to grow upon restreaking (data not shown). In contrast, ARS1013-3, which contains several Sum1 bindings sites, formed small transformants in wild-type strains and pin-prick transformants in sum1Δ strains. Furthermore, the wild-type transformants formed colonies upon restreaking, whereas the sum1 transformants did not (Fig. 3.11B). This demonstrated that ARS function of ARS1013 was improved by the presence of Sum1 binding sites and depended on SUM1. Another ARS adjacent to ARS1013, ARS1012, is an active origin of replication [Raghuraman,et al., 2001], but does not contain Sum1 binding sites closeby (Fig. 3.11B). When tested for plasmid maintenance, ARS1012 transformants grew equally well in wild-type and sum1 strains (Fig. 3.11B). Taken together, these experiments showed that Sum1 binding sites within a replicator improved origin function.

To further test this notion, we next determined whether Sum1 sites other than those naturally present at ARS1013 could improve ARS function of a weak origin. This was achieved by adding ectopic Sum1 binding sites from HML-D (4xD2 or HML-D) or the SMK1 promoter to ARS1013-2 and testing ARS function in wt and sum1Δ strains.

	Figure 3.12: Addition of Sum1 binding sites improved the ARS function of ARS1013-2.
	Strains AEY2 (WT) and AEY3358 (sum1Δ) were transformed with URA-CEN4 plasmids either carrying ARS1013-3 (pAE1081) or variants of ARS1013-2 containing additional fragments of HML-E (4xD2, pAE1159 and HML-E ACS-, pAE1160) or the SMK1 promoter (pAE1161) upstream of the ARS1013-2 fragment.

Addition of HML-D or the SMK1 promoter significantly improved ARS function, and the improvement was completely dependent on Sum1 (Fig. 3.12), which showed that Sum1 sites from alternative sources had the ability to increase replication initiation of a plasmid origin. Addition of four D2 elements barely increased initiation, suggesting that the D2 element was too minimal for Sum1 binding in this context.

Our observation that Sum1 influenced plasmid stability suggested that it might also affect chromosomal replication initiation of Sum1-binding origins. To investigate this, we measured origin firing of ARS1013 in its native chromosomal location in wt and sum1Δ strains by performing two-dimensional origin mapping gels [Fangman and Brewer, 1991].

	Figure 3.13: Sum1 was required for chromosomal origin activity of ARS1013
	(A) Schematic representation of hybridization patterns of DNA harboring a chromosomally active or inactive origin after two dimensional gel electrophoresis (Y-arc: passively replicated fragment, bubble-arc: fragment carrying an active origin of replication). (B) Sum1was required for chromosomal origin activity of ARS1013. The appearance of bubble shaped replication intermediates indicative of chromosomal initiation (arrows) was measured by 2-D gel electrophoresis and Southern hybridization in a wild-type (AEY2) and sum1Δ (AEY3358) strain.

The principle of origin mapping by using two-dimensional gel-electrophoresis is based on the fact that actively replicating DNA migrates differently from passively replicating DNA in an agarose gel. To visualize this difference, DNA fragments carrying a suspected origin of replication are generated by restriction enzyme digestion of genomic DNA and separated in a two-dimensional gel run. After transfer to a membrane via Southern blot, the fragment is detected with a radioactively labeled DNA and the migration pattern is visualized by autoradiography. If an active origin is present on the fragment, two replication forks are emanating outwards from the origin thus creating a bubble shaped structure. Passively replicated DNA shows only one replication fork migrating through the fragment creating a characteristic Y-shape (Fig. 3.13A). Since the tested DNA is obtained from a pool of cells, the fragments represent a pool of possible replication stages ranging from not yet replicated DNA to almost replicated DNA with doubled DNA content. A 2D-migration pattern resembling a strongly bent arc therefore represents passively replicated Y-DNA (Fig. 3.13A left) while actively replicated bubble-DNA is indicated by a lightly bent arc pattern (Fig. 3.13A right). Simultaneous presence of the two migration patterns indicates that the examined ARS does not initiate replication in all cells.

Upon testing ARS1013 we observed a weak signal indicative of bubble-shaped replication intermediates along with strong signal for Y-arc shaped replication intermediates in the wild-type strain (Fig. 3.13B, arrow), indicating that ARS1013 was only active in a fraction of cells and thus, that it was an inefficient chromosomal origin. This was expected, because ARS1013 lies close to ARS1012, which has stronger ARS activity than ARS1013 and therefore probably initiates in the majority of cell cycles and inactivates ARS1013. However, this signal was absent in the sum1Δ strain (Fig. 3.13B). This showed that Sum1 was required for replication initiation of ARS1013 in its chromosomal environment.

We also determined the plasmid maintenance properties of other intergenic regions from our dataset (see Table 3.1). The intergenic regions iYLR307W (ARS1223) and iYOL024W (ARS1511) were both designated “proposed ARS” (pro-ARS) by Wyrick et al (2001) due to their ability to bind ORC and Mcm proteins. However, their ARS activity so far has not been tested. We selected these regions, because they colocalize with probable in vivo origins of replication as determined by genome-wide density transfer experiments [Raghuraman,et al., 2001]. In a plasmid maintenance assay, we found that ARS1223 and ARS1511 indeed conferred autonomous replication to an origin-less plasmid, and that they displayed a significantly increased plasmid loss rate in sum1 cells as compared to wild-type cells (Fig. 3.14A). This showed that the replication capacity of these origins depended on Sum1.

Additionally we tested ARS606 that had been identified by our consensus sequence search (see Table 3.2). When assaying plasmid stability of plasmids carrying ARS606 as the only origin, we observed that ARS activity of ARS606 strongly depended on Sum1. sum1 transformants containing this ARS did not grow upon restreaking, whereas wild-type transformants did (Fig. 3.14B).

	Figure 3.14: Sum1 affected ARS activity of selected origins of replication.
	(A) ARS1223 and ARS1511 required SUM1 for full ARS activity. Plasmid loss rates were determined in a wild-type, WT (AEY2) and a sum1 (AEY3358) strain. Strains with URA-CEN4 plasmids carrying ARS1223 (pAE1130) or ARS1511 (pAE1135) as their sole origins where analyzed. The average loss rates obtained from three independent experiments are shown with corresponding error bars. The loss rate in sum1 strains was approximately 2-fold (ARS1223) and 5.5-fold (ARS1511) higher than in wild-type strains. (B) ARS activity of ARS606 was dependent on SUM1. Strains AEY2 (WT) and AEY3358 (sum1 ) were transformed with URA-CEN4 plasmids carrying ARS606 (pAE1126) as their sole origin and streaked on a -Ura plate.

Since ARS1511 and ARS606 exhibited a strong dependence on SUM1 if tested on plasmids it was of interest to test the SUM1 dependence of these origins at their native chromosomal location. To this end we measured origin firing of ARS606 and ARS1511 in wt and sum1 strains by performing two-dimensional origin mapping gels (Fig. 3.15). ARS607 was additionally chosen, because it located to an intergenic region that also co-localized with a known Sum1 binding site (Fig. 3.10, iYDR523C). In a first approach we measured origin activity of ARS606 in strains grown at normal temperature (30°C). However the amount of origin firing of ARS606 was indistinguishable in wild-type and sum1 strains. Signal quantification revealed that both strains exhibited a comparable ratio of bubble shaped to Y-arc shaped replication intermediates (Fig 3.15A, 30°C and Table 3.3). Since sum1 strains show a mild growth defect at low temperature (data not shown) we reasoned that an existing difference in origin activity might be enhanced at lower temperature.

	Figure 3.15: Chromosomal origin activity of selected ARS sites in wild-type and sum1 strains.
	(A) SUM1 improved chromosomal origin activity of ARS606. The ratio of bubble- to Y-arc shaped replication intermediates was compared by 2-D gel electrophoresis and Southern hybridization in a wild-type (AEY2) and sum1Δ (AEY3358) strain grown at 30°C or at 18°C. Bubble shaped replication intermediates in sum1Δ are marked with an arrow. (B) Chromosomal origin activity of ARS607 and ARS1511 in wild-type (AEY2) and sum1Δ (AEY3358) strains grown at 18°C. Bubble shaped replication intermediates of the sum1Δ autoradiogram are marked with an arrow.

To this end we performed the experiment with DNA from strains grown at 18°C. In fact the sum1 strain showed a slightly reduced ratio of bubble to Y-arc signal compared to the wild-type (Fig. 3.15A, 18°C and table 3.1), indicating that at 18°C ARS606 was less often active in the sum1 strain. However signal intensity ratios for ARS607 and ARS1511 were indistinguishable between wild-type and sum1 strains (Fig. 3.15B, table 3.1).

This either indicated that the difference was only prominent enough at ARS606 to be visualized by the method of 2D gel electrophoresis mapping, or that ARS607 and ARS1511 are not affected by Sum1 in their chromosomal location. However, the observed influence of Sum1 on chromosomal origin activity of ARS1013 and ARS606 along with the synthetic phenotypes of sum1 with replication initiation proteins (see chapter 3.5) showed that Sum1 acted as a global initiation factor.

Table 3.3 : Sum1 affected chromosomal origin activity of ARS606 at low temperature.

In our search for replication origins that are bound by Sum1 we mostly selected candidate origins that were located upstream of genes that are derepressed in a sum1Δ strain (see chapter 3.6). Since Sum1 acts in concert with Hst1 in a subset of these genes (Table 3.1(7)) [Bedalov,et al., 2003;McCord,et al., 2003], we asked whether Hst1 might also affect the ability of our selected origins to initiate replication. In a plasmid maintenance assay, we found that ARS1223 and ARS1511 displayed a significantly increased plasmid loss rate in hst1 cells as compared to wild-type cells (Fig. 3.16A). Notably the magnitude of plasmid loss in hst1 cells (ARS1223: 1.9fold; ARS1511: 5.4fold) was almost identical to that in sum1 cells (ARS1223: 2fold; ARS1511: 5.5fold, Fig.3.14A). This suggested that these two proteins acted in concert at the tested origins.

We also tested the involvement of Hst1 in origin activity of ARS1013 which was shown to be dependent on Sum1 (see chapter 3.6 and Fig. 3.11B and 3.13B). When assaying the ARS function of overlapping ARS1013 fragments in hst1cells we found - as with Sum1 - that fragment 1013-2 formed pin-prick transformants (data not shown). ARS activity of fragment 1013-3, which contains several Sum1 binding sites, however, was completely dependent on Hst1, since hst1 transformants failed to grow upon restreaking whereas wild-type transformants did (Fig. 3.16B). This demonstrated that ARS function of ARS1013 was both dependent on the presence of Sum1 binding sites and Hst1.

ARS606 was identified independently of our first search (see Fig.3.10), and ORFs in the vicinity of this origin are only weakly affected in sum1 cells [Bedalov,et al., 2003] (Table 3.2). Nevertheless, its activity in plasmid loss assays was strongly dependent on Sum1 (Fig. 3.14B). This prompted us to ask whether Hst1 also affected the activity of this ARS. Indeed we observed that origin activity of ARS606 strongly depended on Hst1, since hst1 transformants containing this ARS did not grow upon restreaking, whereas wild-type transformants did (Fig. 3.16C left).

	Figure 3.16: Hst1 affected ARS activity of Sum1 modulated origins of replication.
	(A) ARS1223 and ARS1511 required HST1 for full ARS activity. Plasmid loss rates were determined in a wild-type, WT (AEY2) and a hst1Δ (AEY1499) strain. The experimental procedure was as described in Fig. 3.14A. The loss rate in hst1Δ strains was approximately 1,9 fold (ARS1223) and 5,4 fold (ARS1511) higher than in wild-type strains. (B) HST1 was required for ARS activity of ARS 1013 on plasmids. Strains AEY2 (WT) and AEY1499 (hst1Δ) were transformed with a URA-CEN4 plasmid carrying ARS1013-3 (pAE1081) as its sole origin. Transformants obtained upon transformation of a ARS1013-2 -URA-CEN4 plasmid (pAE1080) were not restreakable. (C) ARS activity of ARS606 was dependent on HST1 but not on SET3. (Left) Strains AEY2 (WT) and AEY1499 (hst1 ) or (Right) strains BY4741 (WT) and BY4741 set3Δ (set3Δ) were transformed with URA-CEN4 plasmids carrying ARS606 (pAE1126) as their sole origin and streaked on a -Ura plate.

Hst1 was previously shown to occur in two different protein complexes. In one case, it is in complex with Sum1 and Rfm1 [McCord,et al., 2003;Pijnappel,et al., 2001] and the other case, it is in the Set3 complex [Pijnappel,et al., 2001]. The multisubunit Set3 complex acts as a histone deacetylase and functions as a repressor of genes during the early/middle stages of the sporulation program [Pijnappel,et al., 2001]. Therefore it was conceivable that the Set3 complex might be involved in modulating the activity of Sum1-regulated origins. Since the protein Set3 is the central component of the multi-subunit Set3 complex, we assayed ARS606 activity in a set3 strain. However, growth of set3 transformants was indistinguishable from that of wild-type transformants (Fig. 3.16C right), indicating that the Set3 complex was not involved in origin activity of Sum1/Hst1 affected ARS sites.

Considering that hst1, like sum1, is synthetically lethal with orc2-1 and orc5-1 [Suter,et al., 2004] these results indicate that Hst1 also plays a significant role in replication initiation of a notable set of replication origins.