4 Discussion

An important question in heterochromatin biology is how its formation is targeted to the right location. In S.cerevisiae,the presence of a set of DNA binding proteins serves as nucleation point for the spreading of heterochromatin. In this work, we identified a new anchor protein for this purpose, the protein Sum1, which was shown to bind a sequence element within the HML-E silencer. Heterochromatin targeting and replication initiation are mechanistically linked by the observation that ORC, the replication initiator, is required for both processes. Here, we show that Sum1 was also required for replication initiation of several origins of replication in the yeast genome. Since the Sum1 protein has previously been identified to be a mitotic repressor for a set of middle meiotic genes, our results notably expand the knowledge about this protein and identify it as a novel regulator of replication and silencing in yeast.

4.1  Sum1 in silencing

Previous studies described a role of the mutant Sum1-1 protein but not the wild-type Sum1 in transcriptional silenicing at both HM loci [Chi and Shore, 1996;Klar,et al., 1985;Laurenson and Rine, 1991]. Sum1-1, through its interaction with the HDAC Hst1 is thereby able to establish an alternative type of heterochromatin that is independent of the Sir-proteins [Rusche and Rine, 2001;Sutton,et al., 2001] (Fig. 4.1B left). In this study we show that the wild-type Sum1 protein is also implicated in silencing of HMLα, though not HMR a. In this role Sum1 binds to a sequence element within the D element, termed D2 at the HML-E silencer.

This finding extends the current knowledge on biological functions of wild-type Sum1 that to date was only known as mitotic transcriptional repressor for a set of middle meiosis specific genes [Xie,et al., 1999]. As a transcriptional repressor, Sum1 often acts in concert with Rfm1 and Hst1 (Fig. 4.1A, left). The histone deacetylase activity of Hst1 is thereby important for the repressive properties of this protein complex. In contrast, we found that the activity of Sum1 as a silencing protein at HML-E was independent of Hst1. This is surprising, especially in light of the observed dependence of Sum1-1 on Hst1 in silencing the HM loci. However, Sum1 as a transcriptional repressor does not always act via Hst1. Microarray analysis of sum1, hst1 and rfm1 strains revealed that of the 146 genes that were derepressed in a sum1 strain only 55 were also derepressed in hst1 and rfm1#SYMBOL#strains [McCord,et al., 2003]. This shows that Sum1 has the potential to repress transcription independently of Hst1. Interestingly, genes that are repressed both by Sum1 and Hst1 are significantly stronger derepressed in a sum1 than in an hst1 strain [McCord,et al., 2003;Xie,et al., 1999]. These data indicate either that Sum1 has intrinsic repressive properties or that Sum1 can interact with an additional factor that has repressive properties (Fig. 4.1A, right). Sum1 interacts with Rfm1, but Rfm1 serves exclusively as bridging protein between Sum1 and Hst1, which was determined genetically and biochemically [McCord,et al., 2003]. In its role as a silencer protein Sum1 may interact with Sir2 or may stabilize the establishment of the Sir-protein complex at HML. In fact one study using a specialized repression system showed an indirect dependence of Sir2 repression to Sum1 presence [Xie,et al., 1999]. However, in numerous protein-protein interaction screens the Sir proteins have not been shown to interact with Sum1. Nevertheless it would be interesting to directly address the question of a Sum1-Sir2 interaction or to search for Sum1 interactors in a yeast two-hybrid screen.

However, Sum1 could also act as silencer protein at HML-E as Abf1 does at HMR-E. Abf1 was so far not found to interact with any of the other proteins implicated in silencing and yet it is important for full repression at HMR. Besides the fact that it is a transcriptional activator elsewhere, Abf1 has been shown to posses nucleosome positioning activity [Lipford and Bell, 2001]. Sum1 could also possess this activity and act comparably to Abf1 at the silencer, thus supporting heterochromatin formation (Fig. 4.1B right).

In its role as transcriptional repressor Sum1 often binds to a sequence element called the MSE upstream of middle sporulation genes [Xie,et al., 1999]. A consensus sequence for the MSE has been determined [Pierce,et al., 2003] but the identified D2 sequence of HML-D does not contain this consensus sequence. This could mean that Sum1 binds to a non-consensus sequence at D2. Perhaps, accessory sequence elements outside D2 also aid in binding. In line with this, in electrophoretic mobility shift assays (EMSA) we observed binding of Sum1 to a sequence that contained the 93 basepair D element but we were unable to detect binding of Sum1 to an oligonucleotide that solely contained the 14 basepair D2.

While in vitro, bacterially expressed and purified full length Sum1 bound HML-E DNA with high affinity, the in vivo ChIP assay showed only a weak binding of tagged Sum1 to HML-E. We used different experimental approaches to improve the in vivo Sum1 binding at HML-E. This included the addition of the c-myc tag to either end of Sum1, prevention of heterochromatin formation across HML by the deletion of Sir4, or ChIPs at different timepoints of the cell cycle, but the enrichment of Sum1 to HML-E remained weak. Possible interpretations are that Sum1 in vivo binds weakly to the non-consensus sequence D2 element, or that Sum1 binds transiently to D2 and is influenced by factors that are not dependent on the cell cycle. Alternatively, using an affinity tag different from the c-myc tag may yield higher enrichment of HML-E in ChIP assays.

	Figure 4.1: The different facets of Sum1
	(A) Sum1 functions as a repressor for middle sporulation (MS) genes either in concert with Rfm1 and Hst1 (Left) or without Rfm1/Hst1 (Right). (B) Sum1 in silencing. (Left) The mutant allele Sum1-1 can repress HMRa in the absence of the Sir proteins. (Right) Wild-type Sum1 aids in the establishment of silencing at HMLα upon binding the D2 element. (C) Sum1 in replication initiation. Sum1 and Hst1 regulate the activity of a subset of origins. The ability of Hst1 to deacetylate histone tails at the surrounding nucleosomes might be involved in this regulation.

A second, unexpected finding of this study is that Sum1 was not only a novel anchor protein at HML-E to silence HML, but also a regulator of replication initiation. In this function, Sum1 may be comparable to Rap1 [Kimmerly,et al., 1988], Mcm1 [Chang,et al., 2004] or Abf1 [Diffley and Stillman, 1988;Eisenberg,et al., 1988] which bind to a subset of yeast origins and are required for efficient initiation. A picture emerges in which yeast replication origins, in addition to ORC, bind an accessory factor that enhances initiation, with different subsets of origins being bound by different modulators.

How does Sum1 promote replication initiation? We showed that Sum1-binding sites in the vicinity of origins and simultaneous presence of Sum1 are important for their replication activity. Data of [Lee,et al., 2002] and our data for HML-E indicated that Sum1 is physically present at origins whose activity is regulated by Sum1. Finally, the ectopic addition of Sum1 binding sites to an inactive origin could render it active, and this activity was also Sum1 dependent. Thus, in regulating origin activity Sum1 is functionally comparable to Abf1 [Diffley and Stillman, 1988;Eisenberg,et al., 1988], which also shares its ability to be a silencing protein [Kimmerly,et al., 1988]. Abf1 binding repositions nuclesosomes both in vivo and in vitro [Lipford and Bell, 2001]. This leads to a region of favourably positioned nucleosomes around the ACS and increases the likelihood of replication initiation. The same could be true for Sum1, though additional experiments will directly have to address this question. Alternatively, Sum1 could be implicated in facilitating the binding of ORC to the ARS site. It is conceivable that Sum1 interacts physically with ORC at the origin such that this interaction is further stabilized by the presence of a Sum1 binding site nearby. In light of this one could re-interpret our co-immunoprecipitation experiment where a Sum1-ORC interaction was abolished upon addition of an agent that can destroy protein-DNA interactions. Abolishing the Sum1-DNA interaction might have weakened the Sum1-ORC interaction to an extent that it was not possible to detect it via co-immunoprecipitation. However, it is also possible, that Sum1 aids in events like the formation of the pre-RC or at the transition of pre-RC formation to replication initiation in that it stabilizes the assembly of involved protein complexes.

We also found that deletion of Hst1 had a negative effect on origin activity at Sum1-regulated origins. Interestingly, in hst1strains the origin function was impaired to a comparable extent as in sum1 strains. This suggests that both proteins affect origin activity by related mechanisms. The fact that HST1 in its role as transcriptional repressor acts almost exclusively via SUM1 [McCord,et al., 2003;Robert, et al., 2004] and that the individual deletions of HST1 or SUM1 are synthetically lethal with mutant orc alleles [Suter,et al., 2004] suggests that Hst1 also interacts with Sum1 at origins. To confirm this hypothesis, the effect of double mutants on origin activity needs to be quantified and compared to that of the individual mutants. Also, it will be interesting to investigate the involvelment of Rfm1, a bridging factor between Sum1 and Hst1, in origin activity.

Since Hst1 is an HDAC [Sutton,et al., 2001], the influence of Hst1 on initiation may be exerted through chromatin deacetylation (Fig. 4.1C). In fact, HDAC activity at an origin had previously been linked to its activity. Nucleosomes adjacent to OriP, the replication origin of the Epstein Barr virus, were shown to be deacetylated, presumably by human HDAC1/2, at the G1/S transition but not at other times of the cell cycle [Zhou, et al., 2005]. How histone deacetylation around origins promotes their replication initiation capacity is not yet clear but [Zhou,et al., 2005] showed that the chromatin remodeling complex SNFh2 was bound to the origin concurrently to HDAC1/2 and was important for origin activity. Thus it is conceivable that deacetylation of origins by Hst1 is also a prerequisite for the association of a chromatin remodeling factor which in turn leads to increased DNA accessibility and facilitates pre-RC assembly. In line with this, it has previously been shown that acetylation of lysine 16 at histone H4 influences the function of ISWI, another chromatin remodeler [Corona, et al., 2002]. However, although Hst1 is a histone deacetylase, another possibility is that it might deacetylate a non-histone protein such as ORC, other pre-RC components or a regulator of initiation thereby activating replication initiation. Hst1 is a homologue of Sir2, and mammalian Sir2 homologues have been shown to deacetylate non-histone proteins such as p53 [Luo, et al., 2001].

Interestingly, two other HDACs have been implicated in replication initiation: The absence of Rpd3 deacetylation causes late origins to fire early [Aparicio,et al., 2004;Vogelauer,et al., 2002], and Sir2 has a negative role in initiation at selected origins [Pappas,et al., 2004]. This is an apparent paradox, since (histone) deacetylation by Hst1 causes increased firing at the origins we tested. However the authors investigated origins unrelated to the set of origins identified in this work. Perhaps there are different classes of origins which are also subject to differential regulation by histone deacetylases. Regulation of origin initiation is known to be context dependent, and it would be interesting to test our set of origins in rpd3 strains. To test the effect of Sir2 on origins of our selection one could target Sir2 fused to a Gal4 binding-domain to an origin that carries Gal4 binding sites instead of Sum1 binding sites. When targeted to our set of origins Sir2 may behave much like Hst1 since there are several indications that Sir2 under some circumstances can substitute for Hst1. For example, one study showed that Hst1 in vitro is able to deacetylate K16 at histone H4 just as Sir2 [Sutton,et al., 2001]. Another study, that used the MSE of SMK1 fused upstream of a LacZ reporter gene showed that expression of this gene was repressed under wild-type conditions but activated in a sum1#SYMBOL#or hst1#SYMBOL##SYMBOL#strain [Xie,et al., 1999]. Overexpression of Sir2 in the hst1#SYMBOL#strain partially re-established repression of LacZ, indicating that Sir2 could take over the task of Hst1 to a certain extent [Xie,et al., 1999].

The observation of synthetic phenotypes between orc and cdc mutations and sum1 suggests that SUM1 may have a global role in replication initiation. SUM1, although it is not lethal if deleted [Chi and Shore, 1996], may have a supportive function at a number of origins. The observed synthetic phenotypes of sum1and mutant alleles of replication proteins suggest that deletion of SUM1 may compromise replication initiation such that it is incompatible with reduced initiation ability. A reason why the synthetic phenotype of sum1and orc2-1 was more severe than the one of sum1 and the cdc mutations could either be a direct function of Sum1 with ORC, but could also be due to differences in severity of the mutant alleles.

Hst1 might have a comparably important role in replication since independent observations found a synthetic lethal phenotype between orc mutations and hst1 [Suter,et al., 2004]. However, it is also possible that sum1 and hst1 additionally affect other processes that become essential in orc mutants, for instance sister chromatid cohesion [Suter,et al., 2004].

An increased plasmid loss as observed in sum1andhst1can also result if factors involved in sister chromatin cohesion are impaired. This could be caused by inefficient function at CEN sequences on the plasmid. We can rule this out for Sum1 because we used identical plasmid backbone sequences (including CEN) in some of our ARS assays and also scored origins, whose activity was equally high in wild-type and sum1strains (i.e. ARS1012). This indicates, that Sum1 did not affect CEN function.

A global effect for sum1 on replication initiation suggests that the number of Sum1-modulated origins must be sufficiently high to cause cell death in orc2-1 sum1 mutants, but our predicted set of possible Sum1-affected origins shows only few such origins. However, our mode of prediction was quite stringent: For our first approach, in addition to a requirement to be bound by both ORC and Sum1, we only scored origins upstream of genes that were derepressed in a sum1 strain [Pierce,et al., 2003]. Thus, several parameters restricted here our origin identification: 1) The ChIP-on-chip analysis for ORC binding sites has probably not identified all sites, since [Breier,et al., 2004] found sequences by computational analysis that were not in the ORC binding data set [Wyrick,et al., 2001] but were active origins in the ARS assay. This is also reflected by the fact, that we found another Sum1-dependent ARS, ARS606, by an independent search for Sum1 and ORC binding site colocalization. 2) Similarly, the p-value prescription of the binding experiment may also exclude intergenic regions with real binding of Sum1. For instance, one known Sum1 binding site, the MSE within the SMK1 promoter [Xie,et al., 1999], was bound according to [Lee,et al., 2002] at a p-value of 0.22 which was more than one decimal power beyond our cutoff p-value of 0.01 and hence did not score in our search. Also, microarray analysis may only be sensitive enough to find locations with multiple Sum1 binding sites, as is the case for many Sum1-regulated genes [Pierce,et al., 2003], whereas origins may contain only one Sum1 binding site, as is the case for HML-E. 3) There may be Sum1 binding sites that do not regulate the neighboring gene, but may be part of an origin. 4) The Sum1 binding site may be at a longer distance from the ACS, and 5) origins with co-occurrence of ORC and Sum1 binding may also lie within coding regions.

In our second approach to find Sum1-ORC binding site co-occurrences in silico we initially obtained ~100 possible candidates. In several stringent refinement steps we applied much of the above mentioned large scale data to evaluate whether the selected loci might be in vivo loci of co-occurrance. Thus many of these shortcomings also affect this selection and might explain why we only obtained 10 possible candidates of which we tested ARS606. Taken together, it seems likely that several more Sum1-regulated origins exist that await identification.

So far, Sum1 was solely considered a repressor of meiotic genes. Our work now demonstrates that Sum1 has a global function in replication initiation. One notable aspect about the involvement of Sum1 in replication is its regulation during meiosis. While constant throughout the mitotic cell cycle, Sum1 protein levels dramatically decrease during the early stages of meiosis, probably concomitantly with premeiotic S phase, and are lowest in the middle stages [Lindgren,et al., 2000]. This raises the question how Sum1-affected origins initiate in premeiotic replication. Perhaps the absence of Sum1 leads to a delayed or a reduced firing rate at selected origins, and thus origin usage may be reduced in meiotic cells, which is in agreement with the observation that sum1 diploids progress slightly slower than wild-type into meiosis [Lindgren,et al., 2000]. The decrease of Sum1 levels is most probably accomplished by targeted protein degradation since mRNA levels remain constant throughout meiosis [Lindgren,et al., 2000]. It is necessary to determine the exact kinetics of this degradation since meiotic replication initiates in the early meiotic stages. Sum1 levels must be low enough at the time of pre-RC formation to inhibit this process. An experimental approach to address this question would be to have a synchronyzed cell population proceed into meiosis and to determine in vivo origin activity of a Sum1-regulated origin in premeiotic S-phase.

In contrast to Sum1, expression of HST1 and RFM1 is significantly increased during that time. Since Sum1 is the targeting factor for Hst1, availability of Hst1 and/or Rfm1 might lead to interaction with another cofactor to reestablish repression of genes that were specifically induced during the earlier stages of meiosis [Chu,et al., 1998].

The fact that Sum1 is repressed in meiosis, which in yeast is induced by depriving cells of glucose, and that Sum1 is required for HML silencing, is in agreement with an earlier, elegant observation that silencing can be made dependent on environmental conditions [Shei and Broach, 1995]. In this study, HM silencers transposed to the MAT locus could repress MAT if grown on glucose-containing medium, but this silencing was relieved on non-fermentable carbon sources such as are used to induce meiosis. In light of our results, one interpretation of this observation is that Sum1 is no longer present under these conditions, so that silencing is abrogated. However, in the original study, not only HML-E, which contains a Sum1 binding site, but also HMR-E, which lacks Sum1 binding sites, showed this effect. Perhaps there are as yet unrecognized binding sites for environmentally regulated proteins at HMR-E that aid in HMR silencing. However reestablishment of silencing after a shift back to glucose-containing medium exhibited a long lag in HML-E [Shei and Broach, 1995]. Thus one might hypothesize that upon shift to conditions favourable for SUM1, a certain lag time would be expected until Sum1 protein is present again and HML-E can fully exert its repressive properties. Interestingly, in the early/middle stages of meiosis the protein kinase Ime2, one of the general meiosis regulators, negatively regulates Sum1 repression at a promoter [Pak and Segall, 2002]. It was speculated that Ime2 marks Sum1 for targeted degradation by phosphorylation of Sum1 [Pak and Segall, 2002]. Therefore it would be worthwile to test this hypothesis by observing mitotic Sum1 levels in a strain that carries an inducible copy of Ime2 on plasmids. If Sum1 levels decreased upon induction of Ime2 it would be interesting to test the influence of targeted Sum1 degradation on HML-E silencer activity in the experimental setup done by Shei and Broach [Shei and Broach, 1995].

Interestingly, other silencer binding proteins like Rap1 and Abf1 function as transcriptional activators rather than repressors elsewhere in the genome [Halfter, et al., 1989;Shore and Nasmyth, 1987]. This situation is paralleled in higher eukaryotes in that the recruitment of Polycomb group complexes to Polycomb response elements (PREs) to maintain homeotic gene repression involves proteins like GAGA and Pho that can function as transcriptional activators as well as repressors [Brown, et al., 1998;Kerrigan, et al., 1991].

On a broader perspective, the finding that a factor, whose expression is regulated by the cell program (i.e. meiosis vs mitosis), influences replication initiation and silencing in yeast, can be compared to the way multicellular organisms exercise control over replication and heterochromatin formation during development. Metazoans use differential origin patterns to replicate a given chromosomal area depending on the cell type. For example, Drosophila embryonic cells have a much broader use of origins than cells of later stages, probably in order to complete the early cell cycles faster than in more differentiated cells, which must accommodate their cell cycle to the respective tissue environment [Sasaki, et al., 1999]. Also, the spacing between meiotic origins in the newt Triturus cristatus is much longer than in mitotic cells, and accordingly, premeiotic S phase is substantially longer than the mitotic S phase [Callan, 1974]. The function of Sum1 at yeast origins may be analogous to that of Drosophila Myb at replication origins in the chorion loci of follicle cells, where Myb is required for site-specific DNA replication leading to gene amplification [Beall, et al., 2002].

On an amino-acid sequence level there are no homologues of Sum1 in budding yeast and in more complex eukaryotes. Syntenic homologues in two other fungi, AAL045C of Ashbya gossypii (ATCC 10895) [Dietrich, et al., 2004] and an unnamed ORF in Kluyveromyces lactis have been found by sequence comparison but their role is unknown to date. The Sum1 protein exhibits only few distinct domains. Two AT-hook domains indicating a DNA binding protein had been identified at pos. aa 204-216 and aa 326-338 [Aravind and Landsman, 1998]. Also a coiled coil domain was predicted at pos. aa 155-170. In light of the absence of sequence homologues in larger eukaryotes perhaps other eukaryotic replication modulators exist that are functionally related to Sum1. They might be expressed in the early stages of development and, in cooperation with ORC, activate origins that are silent in their absence. The down-regulation of these hypothesized factors would reduce origin usage, thus contributing to the lengthening of the cell cycle by increasing the distance between origins. Conversely, origins could be activated differentially in specialized cell types or in meiosis by regulating the expression of origin accessory factors. In summary, the modulation of heterochromatinization and replication initiation by regulating an accessory factor could constitute an economical way for an organism to control origin usage and heterochromatin formation during development and differentiation.

In conclusion, we propose a model for the regulation of origin choice and usage as well as heterochromatin formation during meiosis and differentiation. We present data that a factor that is repressed in meiosis is required for replication initiation at several origins and for gene silencing in yeast. We propose that larger eukaryotes use this mechanism of regulating an accessory factor to differentially control replication and the chromatin state of their genome during different stages of development. A future challenge will be to identify such eukaryotic regulators and to investigate how they integrate the processes of replication initiation and heterochromatin formation.