• Top
• Abstract
• Background
• Results
• Discussion
• Conclusion
• Methods
• Authors' contributions
• Acknowledgements
• References

Research article

Checkpoint independence of most DNA replication origins in fission yeast

Katie L Mickle¹ , Sunita Ramanathan² , Adam Rosebrock¹ , Anna Oliva¹ , Amna Chaudari² , Chulee Yompakdee³ , Donna Scott² , Janet Leatherwood¹ and Joel A Huberman²

¹Department of Microbiology and Molecular Genetics, SUNY at Stony Brook, Stony Brook, New York 11794-5222, USA

²Department of Cancer Biology, Roswell Park Cancer Institute, Buffalo, New York 14263-0001, USA

³Department of Microbiology, Faculty of Science, Chulalongkorn University, Phyathai Rd., Patumwan, Bangkok 10330, Thailand

author email corresponding author email

BMC Molecular Biology 2007, 8:112doi:10.1186/1471-2199-8-112

The electronic version of this article is the complete one and can be found online at: http://www.biomedcentral.com/1471-2199/8/112

Received:	30 July 2007
Accepted:	19 December 2007
Published:	19 December 2007

© 2007 Mickle et al; licensee BioMed Central Ltd.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Background

In budding yeast, the replication checkpoint slows progress through S phase by inhibiting replication origin firing. In mammals, the replication checkpoint inhibits both origin firing and replication fork movement. To find out which strategy is employed in the fission yeast, Schizosaccharomyces pombe, we used microarrays to investigate the use of origins by wild-type and checkpoint-mutant strains in the presence of hydroxyurea (HU), which limits the pool of deoxyribonucleoside triphosphates (dNTPs) and activates the replication checkpoint. The checkpoint-mutant cells carried deletions either of rad3 (which encodes the fission yeast homologue of ATR) or cds1 (which encodes the fission yeast homologue of Chk2).

Results

Our microarray results proved to be largely consistent with those independently obtained and recently published by three other laboratories. However, we were able to reconcile differences between the previous studies regarding the extent to which fission yeast replication origins are affected by the replication checkpoint. We found (consistent with the three previous studies after appropriate interpretation) that, in surprising contrast to budding yeast, most fission yeast origins, including both early- and late-firing origins, are not significantly affected by checkpoint mutations during replication in the presence of HU. A few origins (~3%) behaved like those in budding yeast: they replicated earlier in the checkpoint mutants than in wild type. These were located primarily in the heterochromatic subtelomeric regions of chromosomes 1 and 2. Indeed, the subtelomeric regions defined by the strongest checkpoint restraint correspond precisely to previously mapped subtelomeric heterochromatin. This observation implies that subtelomeric heterochromatin in fission yeast differs from heterochromatin at centromeres, in the mating type region, and in ribosomal DNA, since these regions replicated at least as efficiently in wild-type cells as in checkpoint-mutant cells.

Conclusion

The fact that ~97% of fission yeast replication origins – both early and late – are not significantly affected by replication checkpoint mutations in HU-treated cells suggests that (i) most late-firing origins are restrained from firing in HU-treated cells by at least one checkpoint-independent mechanism, and (ii) checkpoint-dependent slowing of S phase in fission yeast when DNA is damaged may be accomplished primarily by the slowing of replication forks.

Background

The first step in the initiation of DNA replication is the binding to origins of the heterohexameric Origin Recognition Complex (ORC). Subsequently ORC, in collaboration with other proteins, promotes the binding of another heterohexameric complex composed of MiniChromosome Maintenance (MCM) proteins. The combination of ORC, the MCM complex, and certain additional proteins is called a "pre-replication complex" (pre-RC). During S phase, initiation of replication is triggered by Cyclin-Dependent Kinase (CDK) and Dbf4- (or Dfp1-) Dependent Kinase (DDK) at origin-bound pre-RCs (reviewed in [1]).

In the budding yeast, Saccharomyces cerevisiae, replication origins consist of 100–200 bp of DNA, which include an 11-bp AT-rich ORC-binding consensus sequence (reviewed in [1]). In other studied eukaryotic organisms, origins are not so clearly defined. For example, S. pombe origins are larger (500–2000 bp) and have no apparent consensus sequence. However an AT-rich region is necessary for ORC binding [2-4], and asymmetric AT-rich sequence motifs (with A residues primarily in one strand and T residues primarily in the complementary strand) are present and redundantly contribute to origin function [5-10]. In contrast to the high efficiency of many S. cerevisiae origins, S. pombe origins display a wide range of efficiencies, and few, if any, are capable of firing in more than 70% of S phases [11-15]. It appears that replication origins in other eukaryotic organisms share with S. pombe origins the characteristics of absence of a consensus sequence and frequent inefficiency (reviewed in [16,17]). That is, a large number of sequences have the potential to function as replication origins, but only a small subset of such sequences are used by a single cell in any given cell cycle, and the subset selected is highly variable from cell to cell.

Previously, a small number of S. pombe replication origins was studied intensely by autonomously replicating sequence (ARS) assays, which identify the cis-acting DNA sequences important for origin function when the origin region is relocated to a plasmid. A few S. pombe replication origins were also studied by two-dimensional (2D) agarose gel electrophoretic analyses, which permit measurements of replication fork direction and firing efficiency of origins in their endogenous chromosomal locations. These investigations revealed that most of these previously studied S. pombe origins fire early in S phase. However, a few sequences with positive activity in ARS assays proved to be replicated in late S phase – primarily passively – by forks coming from nearby earlier-firing origins [18]. The studied late-replicating potential origins include ars727, ars2-2, and the telomere-associated sequences at the heterochromatic ends of chromosomes 1 and 2 ([18]; A. Chaudari and J. A. Huberman, unpublished). Surprisingly, most of the heterochromatic regions of the S. pombe genome – centromeres, the silent mating type locus, and ribosomal DNA (rDNA) – proved to replicate in early S phase. The only tested heterochromatic region that replicates in late S phase is the telomeres [18,19].

The conclusions listed in the preceding three paragraphs were based on studies of about 20 origins. To find out whether these conclusions applied to all S. pombe origins or just a subset, two laboratories have recently used genome-wide computer analyses. Analyzing the few known S. pombe origins led Segurado et al. [20] to conclude that origins have an unusually high A+T content. They developed an algorithm based on AT content to generate a list of 387 predicted origins (which they called "AT islands"), each of which contains an unusually high A+T content distributed over a broad region (up to 1 kb). Eighteen of twenty predicted origins that were randomly selected from their list of 387 had detectable in vivo initiation activity when tested by 2D gel electrophoresis, suggesting that this simple computational approach could be a surprisingly accurate predictor of origin locations in S. pombe. Dai et al. [10] noticed that the base composition and sequence properties of known origins (high AT; frequent runs of A or T residues) closely resembled those of the longer and more AT-rich intergenic regions. Indeed, when they tested all of the intergenic regions in a 68-kb stretch on chromosome 2, they found that 14 of the 26 intergenic regions exhibited detectable ARS activity. The remaining inactive intergenic regions were shorter and/or less AT-rich than the active ones. These results suggested that about half of S. pombe intergenic regions (about 2500) may occasionally function as origins. This number is much greater than the previous estimates of 250 [21] to 700 [22,23] origins in the genome. The large number of potential origins in the genome, combined with the fact that 10 of the 14 active ARS elements in the 68-kb studied region displayed only weak activity suggested that origin usage in single cells during single S phases may be determined stochastically [10].

Indeed, the conclusion that replication origins in S. pombe are generally inefficient and fire stochastically, without relationship to the firing frequency of their neighbors and without relationship to their firing in previous generations, was strengthened by the results of another genome-wide approach, DNA fiber fluorography, which directly demonstrated the low efficiencies of individual origins and the lack of coordination between them (i.e., stochastic firing) [13].

Other laboratories have employed microarrays to obtain genome-wide information about the locations and efficiencies of origins in synchronized fission yeast cells entering S phase in the presence of HU, which starves cells for dNTPs and thus slows replication fork movement. Under these conditions, the replication checkpoint is activated. For this reason, all of the microarray analyses have compared results from wild-type cells with results from cells bearing mutations that inactivate the replication checkpoint.

The replication checkpoint responds to stalled replication forks (for example, forks that stall when cells are treated with HU or with a DNA-alkylating agent such as methyl methane sulfonate, MMS). Stalled forks, in combination with other proteins, activate an upstream kinase, Rad3 in S. pombe or Mec1 in S. cerevisiae. Both are homologues of mammalian ATR. The upstream kinase activates a downstream kinase, Cds1 in S. pombe or Rad53 in S. cerevisiae. These are structural homologues of mammalian Chk2 and functional analogs of both Chk2 and Chk1. Indeed, when replication forks stall as a consequence of template DNA methylation by MMS, the replication checkpoint is activated, and progress through S phase is slowed in both budding [24,25] and fission yeast [26-29]. Impeding replication forks also leads to checkpoint-dependent replication slowing in mammalian cells [30-32].

In mammalian cells, checkpoint-dependent slowing of S phase is accomplished by a combination of checkpoint-dependent inhibition of origin firing [16,31-33] and checkpoint-dependent inhibition of replication fork movement [32,33]. However, in budding yeast the slowing of S phase in response to MMS treatment is accomplished entirely by inhibition of replication origin firing; no checkpoint-dependent fork inhibition is detectable [25]. The mechanism by which S phase is slowed in MMS-treated fission yeast cells is not clear. The evidence available before the advent of microarray studies suggested that, as in budding yeast, slowing of replication in fission yeast probably depended on inhibition of origin firing, because, when the replication checkpoint is activated by treating S. pombe cells with HU, the replication of two late potential origins (ars2-2 and telomeres) is retarded in checkpoint-dependent fashion [18]. However, results from the new microarray studies, which are reviewed below, have not been entirely consistent with each other or with these prior studies and have led to greater confusion.

In the first microarray study, Feng et al. [34] took advantage of the fact that replication forks unwind parental strands, generating single-stranded DNA (ssDNA). Thus, if cells enter S phase in the presence of HU, those origins capable of firing in early S phase generate small regions of single-stranded DNA (ssDNA), while regions far from early-firing origins remain double-stranded. Feng et al. [34] looked for regions of ssDNA in both HU-treated S. cerevisiae and in HU-treated S. pombe cells. Furthermore, they examined both wild-type cells and replication-checkpoint-mutant cells.

To study the effects of the replication checkpoint on origin firing in S. cerevisiae, Feng et al. [34] compared HU-treated wild-type with HU-treated rad53 mutant cells. They found that 2/3 of S. cerevisiae replication origins are restrained from firing by the replication checkpoint. An even more recent study, which employed copy number measurements rather than measurements of single-stranded DNA, confirmed that 2/3 or more of S. cerevisiae replication origins are checkpoint-restrained [35].

When Feng et al. [34] applied the same procedure to S. pombe and smoothed the results over a 12-kb window, 321 ssDNA peaks (putative origins) were identified in cells lacking the Rad53 homologue, Cds1. Of these, 125 (39%) were specific to cds1Δ cells (apparently checkpoint-restrained). These observations suggested, therefore, that in S. pombe a smaller (but still significant) proportion of origins is restrained by the replication checkpoint than in S. cerevisiae. In both yeasts, ssDNA accumulation was much greater in the checkpoint-deficient strains, presumably due to the roles of Cds1 and Rad53 in stabilizing replication forks (reviewed in [36]).

In the second microarray study, Heichinger et al. [14] measured changes in copy number at each position in the genome as a synchronized population of fission yeast cells entered S phase in the absence or presence of HU. The investigators made similar measurements for checkpoint-mutant (rad3Δ) cells in the presence of HU. The results of all three experiments were in good agreement with each other, and they permitted the identification of 401 relatively strong origins plus 503 putative weaker origins. Surprisingly, only about 2% of these origins appeared to be inhibited by the replication checkpoint – in contrast to the ssDNA measurements of Feng et al. [34], which had suggested that ~39% of fission yeast replication origins are checkpoint-restrained.

The third microarray investigation, by Hayashi et al. [15], provided additional important information but no clear resolution to the apparent conflict between the checkpoint-mutant results of Feng et al. [34] and Heichinger et al. [14]. Hayashi et al. [15] used immunoprecipitation in combination with microarrays to localize pre-RCs (binding sites for both ORC and MCM polypeptides). A total of 460 pre-RCs was found. Then microarrays were used to identify those pre-RCs where significant incorporation of the thymidine analogue, 5-bromo deoxyuridine (BrdU), took place when synchronized cells entered S phase in the presence of HU. Of the 460 pre-RCs, 307 incorporated significant BrdU and were considered to be early-firing and/or strong origins. Little or no BrdU incorporation was detected at the remaining 153 pre-RCs, which were consequently characterized as late-firing and/or weak origins. Hayashi et al. [15] found that there was increased incorporation of BrdU at about 22% of origins (mostly weak/late origins) in checkpoint-mutant (cds1Δ) cells compared to wild-type cells, and such checkpoint-mutant-induced BrdU incorporation was especially frequent in sub-telomeric regions. However, at most of these apparently checkpoint-inhibited origins, the extent of BrdU incorporation in cds1Δ cells was significantly less than for the early/strong class of origins and could only be detected by dividing the signal from cds1Δ cells (small) by the signal from wild-type cells (even smaller) [15].

Thus the three microarray studies appeared to reach rather different conclusions regarding the extent to which fission yeast replication origins are restrained by the replication checkpoint. The two studies that appeared to detect a large fraction (22% or 39%) of checkpoint-restrained origins were both based on comparisons between wild-type and cds1Δ cells. The single study that found only a small fraction (2%) of checkpoint-restrained origins employed rad3Δ cells. Each of the three studies employed a different procedure to measure extents of replication in wild-type and checkpoint-mutant cells. Could these experimental differences explain the different results obtained?

Here we present the results and conclusions of our own microarray-based measurements of DNA replication, as synchronized wild type, cds1Δ and rad3Δ cells enter S phase in the presence of HU. Our results provide an explanation for the differences between the earlier microarray studies. Our findings also lead to interesting conclusions regarding the control of replication timing, the mechanisms by which a subset of origins is restrained from firing in HU-treated cells, the mechanism by which fission yeast cells retard S phase in response to DNA damage, and the relationships between checkpoint regulation of origins and their chromatin structure.

Results

Retardation of DNA replication by hydroxyurea

We synchronized wild-type and checkpoint-mutant (cds1Δ and rad3Δ) cells using the cdc25-22 block-and-release method [37]. After release from the G2 block, these cells normally proceed synchronously through the M, G1 and S phases. In our experiments, however, we added HU (15 mM) at the time of release from the G2 block, so the cells entered S phase in the presence of HU, which dramatically slows replication forks [38-40]. We harvested cells at zero, two, and four hours after the addition of HU.

In HU-treated cells, septum formation proceeds independently of DNA synthesis. We found that septum formation peaked at 60–80% at 100–125 minutes, indicating good synchronization (Figure 1A). We used flow cytometry to measure the amount of replication per cell during the course of our experiment (Figure 1B). At zero hours, most cells were arrested in G2 with a 2C DNA content. At two hours, a 1C peak had started to form as a consequence of cytokinesis. At four hours after HU addition, the major peak was at 1C. In the case of wild-type cells, spreading of this 1C peak toward 2C was evident. This was a consequence of some of the cells progressing into S phase to variable extents, despite the presence of HU. Such spreading was not evident in the checkpoint-mutant cells, consistent with a previous demonstration that replication proceeds more slowly in HU-treated checkpoint-mutant cells than in HU-treated wild-type cells [18].

Genome-wide measurement of extent of replication in hydroxyurea

To determine origin usage in the presence of HU, we employed microarrays to measure the amount of replication as a change in copy number at separate locations along the fission yeast genome. Our microarray probes were constructed by PCR and corresponded to the complete set of fission yeast predicted mRNAs (4,824 probes), other predicted RNAs and intergenic regions (504 probes), and some introns (79 probes). The probes ranged in size from 100 bp to 1200 bp and were primarily directed against the 3' ends of ORFs. Importantly, the spatial positions of the probes on the microarrays were random with respect to the chromosomal locations of the corresponding genes.

We arrested the cdc25-22 mutant strains in G2 by incubation at restrictive temperature and then released them into HU (time 0). We isolated DNAs from 0-hr, 2-hr and 4-hr cells. DNA from 0-hr cells (which were in G2) was used as a control, because all regions of the genome have the same copy number in G2 phase. The DNAs from 2-hr or 4-hr cells were labeled with Cy3 fluorescent dye, while the control DNA was labeled with Cy5. Cy3-labeled DNAs were mixed with Cy5-labeled DNAs and simultaneously hybridized to the microarray. We determined the ratio of Cy3 to Cy5 hybridization for each probe in the microarray. These ratios were then normalized to a genome average value of 1.0. The results from multiple hybridizations were averaged together, and the average values were plotted against the position of each probe in its chromosome. Complete collections of plots at a resolution of 150 kb per plot are available online as Additional Files 1, 2 and 3, for chromosomes 1, 2 and 3, respectively. The normalized, averaged values employed in the plots are also available online as Additional Files 4, 5 and 6 (tables of data for chromosomes 1, 2 and 3, respectively). The raw microarray data have been deposited in the ArrayExpress public repository [41] under accession number E-MEXP-1127.

A sample plot, from the right arm of chromosome 1, is shown in Fig. 2 at higher resolution (45 kb from the left to the right end of the plot). The symbols are explained in the keys above and below the figure and in the figure legend. Because the probes mostly correspond to sequences within genes, they are not uniformly spaced along the DNA. In Fig. 2 and elsewhere we present unsmoothed results to prevent possible information loss, and also because occasional large intergenic gaps between probes (several such gaps are visible in Fig. 2) would make smoothing misleading.

In Fig. 2, a relative copy number of 1.0 indicates an average amount of replication at that time point. Because all experiments were carried out with cells entering S phase in the presence of HU, the extent of replication of the total genome at each time point is, in most cases, very small (Fig. 1B). Therefore, probes located at or very close to origins that fired in 100% of cells in the presence of HU should have relative copy numbers close to 2.0. In contrast, probes that didn't replicate at all under these conditions should have relative copy numbers close to 0.5. Since both the 2-hour and 4-hour data sets were normalized to 1.0, those probe sequences that replicated more than average between 2 hours and 4 hours increased in value between 2 and 4 hours; conversely, the probe sequences that replicated less than average decreased in value between 2 and 4 hours. This behavior is a mathematical consequence of our normalization procedure and does not indicate that DNA was lost from any region of the genome between 2 and 4 hours.

As is evident in Fig. 2, in most cases neighboring probes generated similar copy number values. Furthermore, the values for the wild-type, cds1Δ and rad3Δ strains were also frequently similar. However, in a few cases single probes in single strains generated values out of line with their neighbors. We suspect that these unusual values reflect experimental noise, but it is also possible that some of the apparently anomalous values may be significant. Additional investigations are needed to distinguish between these possibilities.

The results shown in Fig. 2 demonstrate the generally good agreement between the published genome-wide studies. Hayashi et al. [15] mapped pre-RCs using a tiled, high-resolution (250-bp) microarray. In contrast, the other microarray analyses, including ours, employed lower-resolution microarrays with probes located at ORFs. In the case of Heichinger et al. [14], probes were also located in intergenic regions. Consequently, the pre-RC locations determined by Hayashi et al. [15] are probably more accurate than the origin locations determined in the other studies, especially since the other studies (except ours) used peak-detection algorithms that employed sliding windows analyzing smoothed data. For these reasons, we interpret the new results and findings from previous studies shown in Fig. 2 to indicate that all studies are in agreement that (i) there are one or two replication origins, active in HU, located at or near AT1129, and (ii) there is another HU-active origin located at or near AT1130. In addition, our data suggest the possible presence of a weaker origin that fires later in wild-type cells (signals stronger at four hours than two hours) at approximately 4.005 Mbp, close to probe SPAC27E2.02. Hayashi et al. [15] identified a pre-RC close to 4.010 Mb (dark blue dot), but did not detect significant BrdU incorporation at this position in the presence of HU (which is why the circle is dark blue rather than red; Fig. 2 legend), consistent with the other microarray studies, including ours, which also did not detect much replication at this position.

Extensive correspondence between predicted origins identified by five genome-wide approaches

We used our data to evaluate the extent of replication in HU, under our experimental conditions, of origins predicted by the four previous genome-wide analyses – the AT islands defined by Segurado et al. [20], the putative origins identified by Feng et al. [34] and Heichinger et al. [14], and the pre-RCs found by Hayashi et al. [15]. We also identified 22 additional putative origins (regions of clear replication activity according to our data) that were not detected in the previous studies. We compared the locations of AT islands, putative origins and pre-RCs with our HU copy number results (as in Fig. 2), and we classified them into six categories – strong, medium, weak, very weak, below detection limit, and ambiguous – according to the criteria detailed in Methods. Briefly, these criteria were based on the fact that our probes were specific for ORFs, but origins are located in the intergenic regions between ORFs. Consequently, each putative origin was flanked by two probes (or more, in cases where the location of the putative origin was poorly defined). For each probe, there were two time points (two hours and four hours) and three cell strains. For an origin to be classed as "strong", we required that either the wild-type signal or the signals for both checkpoint-mutant strains had to be greater than 1.5 at one or more of the four possible position/time combinations (left probe, 2 hours; left probe, 4 hours; right probe, 2 hours; right probe, 4 hours). For an origin to be classed as "very weak", the signal for the wild-type and/or both checkpoint-mutant strains needed to exceed 1.1 at one or more position/time combination(s). Our definitions of "weak" and "medium" origins are intermediate between those of "very weak" and "strong" (see Methods for details).

The results of our comparisons for the full lengths of all three chromosomes are shown in tabular form in Additional File 7. As an example of the information provided by Additional File 7, in Table 1 we show a small subset of that information – just our classifications of the putative origins within the region on chromosome 1 that is pictured in Figure 2. We encourage readers to examine the complete data for all three chromosomes in Additional File 7.

Additional File 8 is a table summarizing the classifications of origin strengths (from Additional File 7) by chromosome and by type of origin. One of the interesting results in this table is that at least 83% of every type of origin was classified as functional ("very weak", "weak", "medium", or "strong") under our experimental conditions, implying extensive agreement between our results and those of the other studies [20,34,14,15].

Additional File 9 summarizes the extents of overlap between the origin locations predicted by the other studies [20,34,14,15]. Figure 3 displays some of the combined results for all evaluated origin positions from Additional Files 7, 8, 9 in graphic form. The rectangular Venn diagrams in Fig. 3A show – in pair-wise combinations – the extents to which the origins predicted by the other studies co-localize with each other. The proportion of co-localization (73%) proved to be highest for the 320 origins identified by Feng et al. [34] in cds1Δ cells, when compared with the 401 predicted by Heichinger et al. [14]. In other cases, the proportion of co-localization was less (Fig. 3A).

Since all of the previously predicted origins co-localize by 83% or greater with chromosomal positions identified as functional in our studies (Additional File 8), we conclude that the 655 putative origins that we identified as functional under our conditions (Additional Files 7 and 8) represent a more complete picture of mapped fission yeast replication origin locations than obtained from any single one of the previous studies. Note, however, that Heichinger et al. [14] detected 503 weaker origins in addition to their 401 stronger origins. We were not able to compare our results with their 503 weaker origins, because the coordinates of their weaker origins were not provided in their publication [14]. It is likely, however, that our 655 functional origins are mostly a subset of their 904 (401 + 503) origins.

We have also compared our classification of origin efficiencies (from very weak to strong) with the quantitative measurements of origin efficiency reported by Heichinger et al. [14]. For each of their 401 published origins [14], we plotted (Fig. 3B) the strength of the origin according to our rough classification (Additional File 7) against the efficiency of the origin during mitotic S phases as determined by Heichinger et al. ([14]; these data are also listed in Additional File 7). In many cases, two or more origins had the same classification in our results and the same efficiencies according to Heichinger et al. [14]. In these cases, the black circle representing the data point in Fig. 3B was enlarged in proportion to the number of origins having the indicated characteristics. The results in Fig. 3B indicate a fairly good correlation between our rough efficiency classifications and the efficiency measurements of Heichinger et al. [14].

On the basis of extent of incorporation of BrdU in the presence of HU, Hayashi et al. [15] divided the pre-RCs that they identified into two classes: strong/early (identified by red circles and squares in our Figs. 2 and 6) and weak/late (identified by dark blue circles and squares in our Figs. 2, 6 and 8). As shown in Fig. 3C (based on results in Additional File 7), these two classes corresponded well with our five classes. The origins that we classified as below limit or very weak were mostly weak/late according to Hayashi et al., while the origins that we classified as medium or strong were almost entirely strong/early by their classification [15].

The results in Fig. 4 show that there is also a good correlation between the AT islands identified by Segurado et al. [20] and our classifications. AT islands tend to be stronger rather than weaker origins. The behavior of AT islands is clearly different from that of all the non-AT-island positions in the genome, which most frequently score as below limit or very weak. The different behaviors of AT islands and non-AT-island positions are not the consequence of a random distribution of copy number signals of varying strengths along the genome, because when our copy number signals are randomized with respect to position, distributions like the one illustrated by the cream-colored bars in Fig. 4 are obtained. The cream-colored distribution is, in fact, the average of approximately 1000 different randomizations of the copy number signal data with respect to position.

Fig. 5, which is based on data in Additional Files 7 and 8, reveals that, although the proportions of strong, medium, weak and very weak origins are similar in chromosomes 1 and 2, origins in chromosome 3 are more likely to be strong or medium and less likely to be weak, very weak or below-limit. This difference between chromosomes may be related to the fact that chromosome 3 is smaller than the other two chromosomes.

To test whether the two larger chromosomes contain regions that are as rich in strong origins as chromosome 3, we displayed our origin classifications (Additional File 7) according to the position of each origin within its chromosome. Fig. 6 shows that the results from previous studies (light blue circles, Heichinger et al. [14]; magenta circles, Segurado et al. [20]; orange circles, the cds1Δ strain of Feng et al. [34]; purple circles, the wild-type strain of Feng et al. [34], and dark blue and bright red circles or squares, the pre-RCs identified by Hayashi et al. [15]) frequently resemble each other (as indicated by lining up vertically) and frequently correspond to regions that replicated in HU in our studies (vertical lines). It is also evident that the frequencies of regions significantly replicated in HU vary along the genome. Consistent with the pie graph in Fig. 5, chromosome 3 (Fig. 6C) contains high proportions of strong and medium origins (taller, redder lines). Regions with similarly high proportions of strong and medium origins are also evident in chromosomes 1 and 2 (indicated by pale green background). Unlike chromosome 3, however, the larger chromosomes also contain regions with reduced frequencies of strong and medium origins (indicated by pale yellow background). Note that the difference between chromosome 3 and the other chromosomes extends across the whole of chromosome 3 and does not simply reflect differences between subtelomeric regions at the ends of chromosomes 1 and 2 and the rDNA-adjacent regions at the ends of chromosome 3. The regional variations evident in Fig. 6 are similar in position and extent to the regional variations noted by Heichinger et al. [14] and by Hayashi et al. [15].

Extensive correspondence between origins detected by microarrays and previous 2D gel and molecular combing studies of origin function

In Additional File 10 we list the 54 regions of which we are aware that have previously been tested by 2D gel electrophoresis for origin activity in a chromosomal context. 51 of these regions are in positions where our probe density is sufficient to permit evaluation of potential origin function. Of these, 43 regions appeared to be functional (though usually weak) origins by 2D gel electrophoresis, because they displayed bubble arcs as well as Y arcs. The same 43 regions were classified as potentially functional origins (relative copy number > 1.1) according to our microarray data. In three other cases (AT1022, ars2-2, and telomere-associated sequences) the opposite occurred – the potential origin was non-functional according to our results, and it was also non-functional by 2D gel electrophoresis (no detectable bubble arc). Thus in 46 cases out of 51 (90%), there was good correspondence between our results and 2D gel tests of origin function.

Two of the apparent exceptions are AT2067 (Fig. 7C) and Ori6 (Fig. 7D), both on chromosome 2 (see Additional File 10 for references). These did not generate copy numbers > 1.1 in our experiments but did display bubble arcs in 2D gel analyses. The origin at AT1045 (Fig. 7B) behaved similarly. It replicated only very weakly (Additional File 10 and Fig. 7B) but nevertheless displayed a readily detectable bubble arc in 2D gel analyses (Fig. 7A). For the experiment in Fig. 7A, we employed an HU-block-and-release synchronization protocol. We treated log-phase wild-type (checkpoint-competent) cells with HU for 3 hours. The HU treatment caused replication forks to stall, and most cells were arrested in very early S phase, with replication forks stalled a short distance away from early-firing origins. Then, at the 0-minute time point, we removed HU, and replication forks were able to resume moving. At the 0-minute time point, only a few cells in the population contained replication forks within the ScaI fragment that encompasses AT1045 (see the diagram under the 2D gel panels in Fig. 7A). These forks came from neighboring origins, and they generated Y-shaped replication intermediates as they moved through the restriction fragment. These produced a faint "Y arc" after 2D gel electrophoresis (white arrow in the 0-minute time point of Fig. 7A). Fifteen minutes after HU removal, the Y arc signal became stronger. At 30 and 45 minutes after HU removal, the Y arc signal became even stronger, and a "bubble arc" appeared (gray arrows in Fig. 7A). The bubble arc indicated that, in some of the cells, AT1045 functioned as an origin after cells were released from the HU block. Thus, AT1045 is a late-firing origin. These results explain why our microarray analyses detected only very weak replication at AT1045. The HU-treated cells used for our microarray experiments corresponded to the HU-treated cells analyzed at the 0-minute time point in Fig. 7A. In the microarray experiments, the region encompassing AT1045 would have replicated in only a small fraction of the cells. The same rationale (origin firing only after removal of the HU block) can explain the exceptional case of AT2067 (Additional File 10; Fig. 7C), and it certainly explains the exceptional case of Ori6 (Additional File 10; Fig. 7D). A 2D gel analysis of an Ori6-containing restriction fragment was carried out by D. D. Dubey and colleagues (personal communication; submitted for publication) on cells synchronized by HU-block-and-release, as in Fig. 7A. Results similar to those in Fig. 7A were obtained. The restriction fragment containing Ori6 showed maximum signals from bubble and Y arcs at 45–60 minutes after release from the HU block, but showed very little signal from such arcs at 0 minutes. These examples therefore suggest that S. pombe cells may contain many late-firing origins that escape detection when analysis is confined to HU-blocked cells. Since all of the microarray experiments, ours included, employed HU-blocked cells, the current set of microarray experiments is likely to have overlooked those origins that fire only in late S phase.

Some of the additional apparent exceptions to the general correlation between our microarray results and 2D gel electrophoretic results are dealt with in the figure of Additional File 11, and some further examples of correlation between late-replicating origins and microarray results are provided in Additional File 12. The extensive correlation between molecular combing analyses on chromosome 3 [13] and microarray results is described in Additional File 13. The general conclusions from Figure 7 and Additional Files 10, 11, 12, 13 are: (i) the correspondence between 2D gel measurements and microarray measurements is excellent and may reach as high as 98%, (ii) most apparent exceptions to this correlation appear to be a consequence of the failure of microarray measurements on HU-blocked cells to detect certain origins that are active only in late S phase, regardless of whether those origins are examined in wild-type cells or in checkpoint-mutant cells, and (iii) the correlation between microarray measurements and molecular combing measurements is also excellent.

Extensive correspondence between replication in HU-treated wild-type and checkpoint-mutant cells

It is evident in Figs. 2 and 7 and in Additional Files 11, 12, 13 that for most probes the signals produced by wild-type cells were similar to those generated by checkpoint-mutant (cds1Δ and rad3Δ) cells. Clearly, for the regions displayed in these figures, at least, our results for S. pombe are strikingly different from the observations in S. cerevisiae that 2/3 or more of replication origins fire in HU-treated rad53-mutant cells but not in HU-treated wild-type cells [34,35]. To measure the frequency of similar checkpoint-restrained origins in S. pombe cells, we examined each putative origin (Fig. 6, Additional File 7) for significant differences between the wild-type signal and the checkpoint-mutant signals. Since our hypothesis was that origins would be regulated by a checkpoint pathway dependent on both Rad3 and Cds1, we required that the signals from cds1Δ and rad3Δ cells both differ from the wild-type signal by more than 20 percent, consistently for at least two of the four (or more) signals that we evaluated for each origin (see Methods for details).

We found only 21 origins (3.2% of the 657 origins classified as "very weak" or stronger) for which the signals in checkpoint-mutant cells were, by these criteria, higher than the signals for wild-type cells (Additional Files 7 and 8). Twenty of the 21 checkpoint-restrained origins found in our study were present on chromosomes 1 and 2; only one was located on chromosome 3 (Additional Files 7 and 8). In Fig. 6, positions indicated by the letter "C" show the locations of these checkpoint-restrained origins. The distribution is non-random, with checkpoint-restrained origins frequently being weak or very weak and being more abundant near chromosome ends. Since chromosome 3 contains a smaller proportion of weak and very weak origins and has ends that are different from those of chromosomes 1 and 2 (the ends of chromosome 3 contain tandem rDNA repeats and are located in nucleoli), it is not surprising that we found only one checkpoint-dependent origin on chromosome 3. An example of a checkpoint-restrained origin near the right end of chromosome 1 is shown in part A of Additional File 14.

We were surprised to discover that a larger number of predicted origins (36, or 5.5% of the total; Additional Files 7 and 8) function significantly better in wild-type cells than in checkpoint-mutant cells. These wild-type-dependent (more accurately, checkpoint-dependent) origins are indicated by the letter "W" in Fig. 6. Checkpoint-dependent origins are found in all three chromosomes, tend to be medium or strong, and are more frequent in chromosome interiors than near chromosome ends. An example of two adjacent checkpoint-dependent origins is provided in part B of Additional File 14.

Correspondence between telomeric heterochromatin and replication-checkpoint-dependent restraint of replication in HU

Figs. 6A and 6B show that there is a higher frequency of checkpoint-restrained origins near the ends of chromosomes 1 and 2 than in their interior. We wondered whether these checkpoint-restrained origins might correlate with the presence of heterochromatin in telomeric regions. Cam et al. [42] have used chromatin immunoprecipitation and high-density tiled microarrays ("ChIP on chip") to measure the abundance of heterochromatin markers, including methylated lysine 9 in histone H3 (H3K9me), across the fission yeast genome. They found H3K9Me and other heterochromatin markers at telomeres, centromeres, the silent mating type region, rDNA and, at lower levels, at certain genes. We have downloaded the point-by-point measurements of Cam et al. [42] from their Internet site [43] and have graphed (Fig. 8) the abundance of H3K9me for approximately 50 kb near each telomere of chromosomes 1 and 2 together with our own copy number measurements. The relative enrichments of H3K9me [42] are shown by the golden lines at the bottom of each panel in Fig. 8. The other components of Fig. 8 are similar to those of Figs. 2, 7, and Additional Files 11, 12, 13. To interpret Fig. 8, it is important to know that – due to technical problems caused by multiple repeated motifs – the nucleotide sequences for chromosomes 1 and 2 do not extend to the true chromosome ends. Ten kb or more of nucleotide sequence are missing from each end. In addition, although the nucleotide sequences for stretches of 5–15 kb at the true chromosome ends have been determined [44,45], it is not yet known which variation of these true telomere sequences (the variants are extremely similar to each other) is associated with which chromosome end. In this study, in addition to ORF probes we used PCR probes corresponding to true telomere sequences within the 800-bp stretch previously employed as a probe in our 2D gel studies of telomere replication timing [18]. These probes detect telomere-associated sequences immediately adjacent to the simple-sequence telomere repeats at all four ends of chromosomes 1 and 2. The averaged results obtained with these probes are plotted in each panel of Fig. 8 to the left or right of the known left-end or right-end sequences (respectively) of these two chromosomes.

The amount of nucleotide sequence that is missing from each chromosome end is unknown, but it is thought that the known sequence at the left end of chromosome 1 is closest to the true telomere, and it is thought that approximately 10 kb of sequence are missing in this case [42,46]. The microarray probes used by Cam et al. [42] detected elevated enrichment (mostly > 10-fold) of H3K9me in the leftmost 20 kb of chromosome 1 (Fig. 8A). It is remarkable that, within the same 20-kb stretch, every one of our six probes detected higher copy number levels in the cds1Δ and rad3Δ strains than in the wild-type strain at the 4-hour time point. In the stretch between 20 and 36 kb from the left end of chromosome 1, Cam et al. found reduced enrichment of H3K9me. In this stretch, we found that five of our nine probes detected elevated copy numbers in checkpoint-mutant strains at 4 hours. Rightward of 36 kb, the enrichment of H3K9me was at background level (approximately 1), and this background level was maintained in most of the rest of chromosome 1, with the exception of the centromere and right telomere [42]. Nowhere else in the chromosome except at the right end (Fig. 8B) – even at the origins identified as checkpoint-restrained (marked "C" in Fig. 6) – is there a stretch with so many contiguous probes showing significantly higher copy number at 4 hours in the checkpoint-mutant strains (Additional File 1). At the right end of chromosome 1 (Fig. 8B) our measurements are consistent with those at the left end (Fig. 8A). In the stretch from 5.572 to 5.580 Mb, enrichment for H3K9me is high, and all three probes in this stretch detected elevated copy numbers after 4 hours in HU in the checkpoint-mutant strains. From 5.540 to 5.572 Mb, enrichment of H3K9me is less, and 7 of the 12 probes in this stretch detected elevated copy numbers in checkpoint-mutant strains at the 4-hour time point. To the left of 5.540 Mb, H3K9me is not enriched. Although the probe in this portion of Fig. 8B detected higher copy numbers in the checkpoint-mutant strains at 4 hours, the vast majority of probes further to the left (Additional File 1) did not.

Similar results were obtained at the left and right ends of chromosome 2 (Figs. 8C and 8D). Cam et al. [42] detected only moderate enrichment of H3K9me in the stretch from 0 to 15 kb at the left end of chromosome 2. Presumably the absence of higher enrichment is a consequence of the absence of unambiguous nucleotide sequence further to the left. In the stretch moderately enriched for H3K9me, six of nine probes detected enhancement of checkpoint-mutant copy numbers relative to wild-type at the 4-hr time point. In the stretch within Fig. 8C to the right of 15 kb, where H3K9me is not enriched, only one of five probes detected significant checkpoint-mutation-dependent signal enhancement. At the right end of chromosome 2, in the highly H3K9me-enriched stretch from 4.518 to 4.542 Mb, all four probes detected checkpoint-mutation-dependent signal enhancement (Fig. 8D). However, in the moderately enriched stretched from 4.499 to 4.518 Mb, four of the 10 probes were not significantly affected by checkpoint mutations. Although the leftmost probe in Fig. 8D, which is the only probe in this panel from a region not enriched in H3K9me, detected modest copy number enhancement at 4 hours in checkpoint mutant strains, the vast majority of probes further to the left did not (Additional File 2). Thus there is a striking correlation between the distribution of H3K9me at the ends of chromosomes 1 and 2 and the consistency and extent of checkpoint-mutation-dependent replication in HU.

The induction of replication at the 4-hour time point in the heterochromatic subtelomeric regions of the checkpoint-mutant strains (Fig. 8) seemed so strong that we thought it likely that we would be able to find evidence for subtelomeric origin activation in checkpoint-mutant strains at the 4-hour time point by the appearance of bubble arcs in 2D gel electrophoretic analyses. To test this possibility, we analyzed a NdeI-EcoRV restriction fragment approximately centered on the peak of A+T near the highly-induced region at nucleotide position ~4.523 Mb at the right end of chromosome 2 (Fig. 8D, Fig. 9A, Additional File 8). As shown in Fig. 9A, the 2D gels confirmed the impressive induction of replication in this region at the 4-hour time point in the checkpoint-mutant strains. However, although Y arcs were easily visible in the checkpoint-mutant strains at 4 hours, no bubble arcs were evident. This failure to detect the anticipated bubble arcs could be a consequence of (i) the origin responsible for replicating this region being located near the ends or outside of this restriction fragment, in which case only Y arcs would be visible, or (ii) breakdown of bubble structures within this restriction fragment during the four-hour incubation in the presence of HU in the absence of the replication checkpoint, one of whose functions is to stabilize replicating DNA.