Fission Yeast CSL Transcription Factors: Mapping Their Target Genes and Biological Roles

Background Cbf11 and Cbf12, the fission yeast CSL transcription factors, have been implicated in the regulation of cell-cycle progression, but no specific roles have been described and their target genes have been only partially mapped. Methodology/Principal Findings Using a combination of transcriptome profiling under various conditions and genome-wide analysis of CSL-DNA interactions, we identify genes regulated directly and indirectly by CSL proteins in fission yeast. We show that the expression of stress-response genes and genes that are expressed periodically during the cell cycle is deregulated upon genetic manipulation of cbf11 and/or cbf12. Accordingly, the coordination of mitosis and cytokinesis is perturbed in cells with genetically manipulated CSL protein levels, together with other specific defects in cell-cycle progression. Cbf11 activity is nutrient-dependent and Δcbf11-associated defects are mitigated by inactivation of the protein kinase A (Pka1) and stress-activated MAP kinase (Sty1p38) pathways. Furthermore, Cbf11 directly regulates a set of lipid metabolism genes and Δcbf11 cells feature a stark decrease in the number of storage lipid droplets. Conclusions/Significance Our results provide a framework for a more detailed understanding of the role of CSL proteins in the regulation of cell-cycle progression in fission yeast.


Introduction
Fission yeast, Schizosaccharomyces pombe, has been instrumental in the identification of fundamental principles of cell-cycle control [1]. The cell cycle is regulated at multiple levels by numerous protein factors that integrate proliferative and anti-proliferative signals, both intraand extra-cellular. Key post-translational regulators are the Cdc2 CDK1 cyclin-dependent kinase, and the Wee1 kinase and Cdc25 phosphatase, which regulate Cdc2 activity and mitotic entry in a negative and positive manner, respectively. At the transcriptional level, several transcription factors have been identified that regulate specific clusters of periodically expressed genes during the cell cycle [2][3][4]. These factors include the forkhead proteins Sep1 and Fkh2, and the MADS box protein Mbx1 (active at M phase; role in mitosis and cytokinesis), the zinc-finger protein Ace2 (M/G1 phase; role in cell separation), the MBF complex and its negative regulators Nrm1 and Yox1 (G1/S phase; role in DNA replication), and the GATA-type transcription factor Ams2 (S phase; expression of histone genes) [5][6][7][8][9][10][11][12][13]. Combinatorial regulation by multiple transcription factors has been suggested for many periodically expressed genes [4], and, importantly, for a substantial number of cell cycle-regulated genes the regulator(s) controlling their expression remain unknown. Potential novel transcriptional regulators have been suggested by computational network analysis of gene expression data [14]. Our understanding of the mechanisms driving periodic gene expression during the cell cycle is far from complete.
CSL proteins comprise a family of transcription factors that regulate development and cell fate decisions in Metazoa, mostly via the Notch signalling pathway [15]. Depending on context, Notch/CSL also promote or inhibit cell-cycle progression by regulating the expression of several cyclins and CDK inhibitors [16][17][18][19]. CSL family members are also present in many fungi, but their functions are much less understood [20,21]. We previously identified two antagonistic CSL paralogs in S. pombe, Cbf11 and Cbf12, that function as transcription activators and recognize the canonical metazoan CSL response element on DNA [21][22][23]. Fission yeast CSL proteins have been implicated in the coordination of cell and nuclear division, cell morphology, and regulation of cell adhesion [22,24]. Mutants with altered cbf11 or cbf12 gene dosage have recently been identified in screens for altered cell size and shape, phenotypes associated with aberrant cell-cycle progression [25,26]. Furthermore, the stability of cbf12 mRNA is negatively regulated by the tristetraprolin RNA-binding protein Zfs1 [27], which plays a role in cell adhesion, cell size determination, and the coordination of mitosis and cytokinesis [28,29]. Collectively, these data raise the intriguing possibility of an important role for Cbf11 and Cbf12 in the regulation of cell-cycle progression in fission yeast.
Previously, Chua and colleagues reported expression microarray data for cbf11 deletion and cbf12 overexpression under a single growth condition (rich and minimal medium, respectively), and ChIP-chip data for ectopically overexpressed Cbf12. The authors used the data to identify CSL target genes relevant to cell flocculation and described the regulatory roles of CSL proteins therein [24]. Here we set out to systematically identify CSL target genes under a range of growth conditions. We show that Cbf11 and Cbf12 contribute directly and indirectly to the regulation of distinct sets of genes, including stress-response genes and genes expressed periodically during the cell cycle, and we show that Cbf11 directly regulates a group of lipid metabolism genes. We further show that the function of Cbf11 in regulating cell-cycle progression is affected by nutrients and by protein kinase A (Pka1) and stress-activated MAP kinase (Sty1) pathways.

Yeast culture and transformation
Fission yeast cells were grown according to standard procedures [30] at 30 or 32°C, unless stated otherwise, in either rich yeast extract with supplements (YES) or Edinburgh minimal medium (EMM; Formedium). A list of fission yeast strains used in this study is provided in S1 Table. The lithium acetate method was used for transformation [31]. Overexpression of cbf11 and cbf12 from a plasmid was regulated by the presence (repression) or absence (induction) of 15 μM thiamine in EMM [32]. Routine optical density (OD) measurements of liquid cell cultures were taken using the WPA CO 8000 Cell Density Meter (Biochrom). Growth curves were measured in the VarioSkan Flash instrument (Thermo Scientific) using 12-well dishes and 1.4 ml culture volumes. To assay growth on solid media, exponentially growing cells were 10-fold serially diluted and spotted onto YES plates.

Microscopy
For cell size and septation index measurements, exponentially growing cells were fixed in 10% formaldehyde (in PBS) for 15 min, washed three times with PBS, stained with DAPI (1 μg/ml) and/or calcofluor (50 μg/ml) and subjected to fluorescence microscopy using a Zeiss Axiophot microscope and the OpenLab software (PerkinElmer), or the Olympus CellR system. Cell length at division was determined using the ImageJ 1.45 software [33].
For quantification of the occurrence of catastrophic mitosis, exponentially growing cells were fixed in 70% ethanol, rehydrated in water, stained with DAPI and photographed using the Olympus CellR system.
The procedure for neutral lipid droplet quantification was adapted from [34]. Live cells growing exponentially in YES were stained with Nile red (10 μM; Sigma) for 10 min at room temperature and images were taken using the Olympus CellR system (GFP filter; excitation 475 nm, emission 530 nm). Lipid droplets were counted manually for at least 200 cells per sample and the number of droplets was normalized to cell volume using the ImageJ software (cylindrical approximation of cell shape was used for calculation) [33].

Plasmids and constructs
The lists of oligonucleotides and plasmids used in this study are provided in S2 Table and S3  Table, respectively. Plasmids for inducible overproduction of Cbf11 and Cbf12 under the control of the full-strength nmt1 promoter were constructed as follows. The respective CSL cDNAs were re-cloned from the previously described plasmids pJR08 (cbf11) and pMP32 (cbf12) [22], using the Expand High Fidelity PCR system (Roche) and primer pairs mp59 & mp60 and mp61 & mp62, respectively. The resulting PCR products were cloned into the pGEM T-Easy vector (Promega), verified by sequencing, and inserted into the PstI/NotI sites of the pJR2-3XU vector [32].
The fission yeast knock-in strain expressing C-terminally triple HA-tagged Cbf11 from its endogenous chromosomal locus (MP26) was constructed in a wild-type background (JB32) by standard PCR-mediated one-step gene tagging using the pFA6a-3HA-natMX6 plasmid as template [35]. The mp41 forward primer consisted of 80 nt complementary to the 3' end of the cbf11 ORF (stop codon not included) and 20 nt complementary to the 5' end of the 3HA-natMX6 tagging cassette. The reverse mp55 primer contained 80 nt complementary to the genomic sequence starting 80 nt downstream of the cbf11 ORF (the 80 nt gap was introduced to obtain a primer with a higher melting temperature) and 20 nt complementary to the 3' end of the 3HA-natMX6 tagging cassette.
The strain expressing C-terminally double TAP-tagged Cbf11 from its endogenous chromosomal locus (MP15) was constructed analogously in an auxotrophic background (PN559). The mp41 forward primer and a reverse primer complementary to the region immediately downstream (i.e., without the 80 nt gap as for the mp55 primer) of the cbf11 ORF (mp42) were used for the amplification of the tagging cassette from the pFA6-CTAP4-natMX6 plasmid [35].
To obtain a Δcbf11::natR strain, the natMX6 cassette conferring nourseothricin (clonNAT) resistance was amplified from the pFA6-13Myc-NatMX6 plasmid template [35] using primers mp57 and mp58 (these universal primers are complementary to promoter and terminator regions, respectively, of various derivatives of the MX6 cassette and may be used for other combinations of marker switching). The resulting PCR product was integrated by homologous recombination into an auxotrophic Δcbf11::kanR strain (MP23), and the resulting nourseothricin-resistant G418-sensitive strain (MP34) was verified by PCR.
Deletion of cbf11 in sterile strains was carried out using the pMP91 targeting plasmid based on pCloneNAT1 as described [36]. All other strains used in this study were constructed by standard genetic crosses.

Gene expression analysis by microarrays
For transcriptome profiling of CSL knock-outs, either cells growing exponentially in EMM or YES media (4 × 10 6 cells/ml), or cells reaching early stationary phase (4.5 × 10 7 cells/ml) were used. For CSL overexpression analysis, strains harbouring the required plasmids were grown in EMM with thiamine to late exponential phase, washed twice in EMM without thiamine, diluted and grown further in EMM without thiamine for 12 or 18 hours to induce expression from the nmt1 promoter (during the last step, cells were further diluted as necessary to keep them in the exponential phase of growth). Cells were harvested by centrifugation (1000 g, 3 min, room temperature), suspended in TES buffer (10 mM Tris pH 7.5, 10 mM EDTA pH 8, 0.5% SDS), and total RNA was isolated by phenol-chloroform extraction followed by sodium acetate/ethanol precipitation and column purification using the RNeasy Mini kit (Qiagen).
Five to twenty micrograms of total RNA were labelled by directly incorporating either Cy3or Cy5-dCTP, or Alexa Fluor 555-or 647-aha-dUTP through reverse transcription and used to hybridize either onto in-house glass DNA microarrays containing probes for >99% of all known and predicted fission yeast genes as described [37], or onto Agilent 8x15K or 4x44K custom-made S. pombe expression microarrays according to the manufacturer's protocols. The microarrays were scanned using a GenePix 4000 Laser Scanner (Molecular Devices) and signals were extracted with GenePix Pro 6.0 (Molecular Devices). Initial data processing and normalization were performed using an in-house script [37]. GeneSpring GX (Agilent) and R were used for data evaluation.
The reported values represent gene expression levels for each experimental condition relative to the expression levels in an untreated wild-type or vector-only control from the same experiment. At least two independent biological repeats with dye swap were performed for each experiment (S4 Table). Differentially expressed genes (DEGs) were called based on a conservative fixed cut-off of 2-fold change vs expression levels in the corresponding control sample. We only considered genes with consistent changes across biological repeats (in 2/2 cases, 2/3 cases and 3/4 cases, respectively, depending on the number of repeats available; see S4 Table for details). The significance of overlaps between DEG lists from various conditions was determined by Fisher's exact test with Bonferroni correction for multiple testing in R. For clustering analysis, only those DEGs for which data were available from at least 60% of microarray experiments were used. Clustering was performed based on Euclidean distances using the 'complete linkage' method in R. Five major clusters were obtained by this approach. Upon visual inspection of the algorithmic clustering, Cluster 5 was then manually subdivided into Clusters 5a and 5b. AnGeLi, an in-house software tool employing Fisher's exact test with Bonferroni correction (0.05 significance level), was used for functional enrichment analyses of the microarray results (www.bahlerlab.info/AnGeLi; Bitton et al., in preparation). Microarray data are available in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-2724.
To construct sequencing libraries, DNA fragments were end-repaired using the End-it kit (Epicentre Biotechnologies), purified using the DNA Clean&Concentrator-5 kit (Zymo Research), deoxyadenine-tailed using the 3'-5' exo-Klenow fragment (New England Biolabs), purified again, sequencing adapters from the Illumina Multiplexed PE Sample Preparation Kit were ligated to the DNA fragments using the Quick T4 DNA Ligase (New England Biolabs), samples were purified again, and sequencing libraries were amplified using the appropriate primers from the Illumina Multiplexed PE Sample Preparation Kit. Finally, libraries were purified using the AMPure XP beads (Agencourt), pooled and sequenced on an Illumina Hi-Seq machine (6-fold multiplexing per lane) at the Genomics Core Facility, CR-UK Cambridge Research Institute, UK.
Sequencing reads were de-multiplexed, aligned to the reference fission yeast genome (release 12) using BWA 0.6.1 [39], processed with the samtools 0.1.18 package [40], and inspected using IGV browser 2.0.23 and the igvtools package [41]. Statistically significant peaks of ChIP-seq signals (p 1× 10 −5 ) were called against IP data from an untagged strain using MACS 1.4.1 [42] (when calling ChIP-seq peaks against the respective input DNA data, MACS reported an excess of false positives, likely because the input DNA libraries con-tained~10 times more reads than the IP libraries); only peaks identified in both biological repeats were used in subsequent analyses. Potential binding motifs were searched for within 100 bp under the ChIP-seq peak summits with MEME 4.8.1 [43].
For conventional ChIP-qPCR experiments, 50 ml cultures were used and the immunoprecipitated DNA was purified using 10% Chelex 100 resin (Bio-Rad). For HA-tagged strains, 5 μg of anti-HA tag rabbit polyclonal antibody (ab9110, Abcam; Antibody Registry acc. no. AB_307019) and protein A sepharose beads (GE Healthcare) were used for precipitation. Quantitative PCR was performed using the MESA GREEN qPCR MasterMix Plus for SYBR (Eurogentec) and the LightCycler 480 II instrument (Roche). The primers used are listed in S2 Table. Electrophoretic mobility shift assay (EMSA) The analysis of DNA binding by Cbf11 was described in detail previously [22]. Briefly, cells were harvested at the density of 2.5 × 10 7 cells/ml by centrifugation, washed with STOP buffer (150 mM NaCl, 50 mM NaF, 25 mM HEPES, 1 mM NaN 3 ; pH 8) and kept at -80°C. Native extracts were prepared in Lysis/Gelshift Buffer (25 mM HEPES, 0.1 mM EDTA, 150 mM KCl, 0.1% Triton X100, 25% glycerol, 1M urea, 2 mM DTT, FY protease inhibitors [Serva]; pH 7.6) by breaking the cells with glass beads in a FastPrep24 instrument (MP Biomedicals). Binding to radiolabelled double-stranded DNA probes containing CSL binding sites was detected as slow-migrating bands on a large native 5% polyacrylamide TBE gel. Signal intensities were quantified using the ImageQuant TL 7.0 software (GE Healthcare).

Flow cytometry
Cells were fixed with 70% ethanol, rehydrated in 50 mM sodium citrate, treated with RNase and stained with propidium iodide (4 μg/ml) as described [44]. Highly flocculating strains were treated with 20 mM EDTA and mildly sonicated to reduce cell clumping. DNA content was measured using a BD LSR II instrument (BD Biosciences) or CyAn ADP Analyzer (Beckman Coulter); at least 10,000 cells were measured for each sample. Data were analysed in WinMDI 2.9 (The Scripps Research Institute, San Diego, CA, USA) and FlowJo (FlowJo, LLC). Gating of singlet cells was performed as described [45] where required.

Results and Discussion
Antagonism between Cbf11 and Cbf12 at transcriptome level Cbf11 and Cbf12 bind to DNA in a sequence-specific manner in vitro [21,22], and also in vivo [23,24]. Furthermore, both CSL proteins can drive gene expression from a CSL response element-containing minimal promoter on a plasmid-based reporter system in vivo [23]. Thus, as in Metazoa [15], Cbf11 and Cbf12 likely function as transcription factors [23]. We therefore applied expression microarrays to determine which genes are potentially regulated by CSL proteins, and whether altered gene expression can explain the mutant phenotypes associated with CSL gene manipulation [22].
We analysed the transcriptomes of Δcbf11 and Δcbf12 single and double deletion mutants under various growth conditions (S4 Table), comprising rapid proliferation in rich YES or minimal EMM media, and entry into stationary phase (YES), where cells cease to proliferate. Transcription factor overexpression, called 'phenotypic activation', can be powerful for identifying target genes [46]. We therefore also assayed cells during strong cbf11 and cbf12 overexpression (12 and 18 hours post induction) in EMM. Differentially expressed genes (DEGs) in CSL-manipulated cells relative to wild-type cells were determined for each experimental condition based on a 2-fold cut-off, resulting in 524 DEGs overall (S5 Table). Filtering and hierarchical clustering revealed 6 major clusters, comprising 340 genes that showed differential expression in CSL-manipulated strains under at least one condition tested (Fig 1A and S6  Table).
We then performed functional enrichment analyses in these gene clusters (Fig 1B). Clusters 1-5a were significantly enriched for core environmental stress response (CESR) genes, which are induced or repressed under various stress conditions, primarily by the mitogen-activated protein (MAP) kinase Sty1 p38 and its downstream effector, the transcription factor Atf1 [49]. In addition, Clusters 1, 2 and 4 were also enriched for genes induced during nitrogen starvation, and late during sexual differentiation ('late meiotic' genes) [48]. The latter two gene lists largely overlap with the CESR genes. The significance of CESR gene deregulation in Δcbf11 and cbf12-overexpressing cells is discussed below. Notably, Clusters 4 and 5b were also enriched for genes showing periodic expression during the cell cycle [47]. The deregulation of these genes is consistent with the CSL cell-cycle phenotypes described previously [22]. Finally, genes coding for transmembrane and GPI-anchored proteins were enriched in Clusters 5a and 5b, suggesting that their deregulation underlies the altered cell adhesion in CSL-manipulated cells [22,24]. Furthermore, GPI-anchored proteins have been implicated in cell shape specification in fission yeast [26] and their aberrant expression might therefore contribute to the cell morphology defects of Δcbf11 cells [22].
We conclude the following from these results: 1. Only few DEGs (<10) were evident in the transcriptomes of Δcbf12 cells under any condition studied (S1 Fig, panel A). This finding is consistent with the scarcity of mutant phenotypes associated with Δcbf12 cells under standard growth conditions [22].  . Different data columns under the same condition represent independent biological repeats. The mRNA levels at each condition relative to the levels in wild-type control cells are colourcoded as indicated at bottom, with missing data in grey. CSL binding to the upstream intergenic regions of the respective DEGs, as detected by ChIP-seq from cells growing exponentially in YES, and periodicity of gene expression during cell cycle [47] are indicated at right (dark bars signify CSL binding/periodic expression). Six major DEG clusters ('1'-'5a/b') are indicated by the right-most colour bars. (B) Functional enrichment analysis of the DEG clusters from (A). Numbers in the matrix represent fold enrichment in the cluster compared to all other genes, and the enrichment significance is denoted by the colour of the cell background. 'N starvation'-genes induced upon nitrogen removal [48]; 'late meiotic'-genes induced after meiotic divisions [48]; 'CESR'-core environmental stress response genes [49]; 'top 500 periodic'-top-ranking 500 genes expressed periodically during cell cycle [47]; 'transmembrane'-genes coding for transmembrane proteins [50]; 'GPI anchor'-genes coding for GPI-anchored proteins [51]. where medium-strength overexpression of cbf11 resulted in no apparent cell phenotypes while overexpression of cbf12 was toxic [22].

Genes upregulated in
Δcbf11 cells growing rapidly in YES significantly overlapped with genes upregulated after 18 hours of cbf12 overexpression (S1 Fig, panel A; 34 genes, p = 3.8 × 10 −17 ). Previous analysis of phenotypes associated with cbf11 deletion and cbf12 overexpression indicated that Cbf11 and Cbf12 function in an antagonistic manner [22]. The microarray data now reveal that this is also true at the level of transcription regulation. Furthermore, the number of genes affected by Cbf11 and Cbf12 in opposite directions is likely an underestimation due to the stringent cut-offs used for DEG calling. When the average expression of each gene per experimental condition was plotted for all individual DEG clusters, a clear antagonistic trend in regulation by the two CSL paralogs was evident (S1 Fig, panel B). This antagonism between Cbf11 and Cbf12 is likely physiologically relevant since as little as~4-fold cbf12 mRNA upregulation, which occurs under some growth conditions [22], triggered Δcbf11-like phenotypes [27].

Cbf11 and Cbf12 bind to overlapping set of genomic loci in vivo
Expression microarrays provide only limited insight to distinguish direct from indirect regulation of gene expression. To identify direct CSL target genes, we performed chromatin immunoprecipitation followed by deep sequencing (ChIP-seq) for TAP-tagged Cbf11 and Cbf12 from cells grown exponentially in YES-one of the conditions with the highest number of CSL DEGs. We found 121 (for Cbf11) and 380 (for Cbf12) genes with significant and reproducible peaks of ChIP-seq coverage in their upstream intergenic regions, where promoters are likely located (Fig 1A and 1C). The amplitude of the ChIP-seq peaks was generally lower for Cbf11 in both biological repeats, increasing the probability of missing some binding events. A selection of CSL ChIP-seq peaks was validated by ChIP followed by quantitative PCR (qPCR), using strains with TAP-tagged or HA-tagged Cbf11 and Cbf12. The two ChIP methods showed good agreement (S2 Fig). The overlap between the CSL ChIP-seq targets (91 and 320 genes for Cbf11 and Cbf12, respectively, for which expression data were available in our microarray dataset) and the overall 524 DEGs identified by microarrays (which include the DEGs filtered out for clustering analysis) was limited: only 21 genes (~23% of ChIP-seq targets) for Cbf11 and 79 genes (~25% of ChIP-seq targets) for Cbf12, yet these overlaps were statistically significant (Fig 1C). This finding suggests that most changes in transcript levels observed upon CSL manipulation could have been brought about indirectly, perhaps through a regulatory cascade as in the case of metazoan CSL family members [52]. Intriguingly, the putative direct CSL targets included several genes coding for characterized or predicted regulators [50], such as transcription factors (Rsv2, SPAC2H10.01, Gsf1, SPCC1393.08), kinases and phosphatases (Byr2, Srk1, Oca2, Pyp2), and RNA-binding proteins (Mei2, Mug24), some of which with known roles in cellcycle regulation [53][54][55]. It is possible that additional direct CSL targets could be identified using ChIP-seq under different growth conditions. Also, some CSL binding events detected in cells growing exponentially in YES might only result in gene expression regulation upon a specific stimulus that was not tested in this study. Another potential explanation for the limited overlap between CSL ChIP-seq targets and DEGs is the stringency we applied for DEG calling. Indeed, Cbf11-bound genes tend to be more differentially regulated in Δcbf11 cells growing exponentially in YES compared to non-bound genes, and significant differential regulation occurs also for the set of Cbf12-bound genes under cbf12 overexpression (S3 Fig). The ChIP-seq targets of Cbf11 and Cbf12, on the other hand, showed a highly significant overlap (71 shared genes, p = 4 × 10 −69 ). However, most of these common CSL-bound genes did not show differential expression upon CSL manipulation (Fig 1C). It is possible that the regulatory crosstalk between the two CSL paralogs is mediated through a mechanism that does not involve direct binding of Cbf11 and Cbf12 to the same, antagonistically regulated promoters; however, 16 out of the 21 Cbf11-bound DEGs (~76%; mostly from Clusters 4 and 5a) were also bound by Cbf12 (Fig 1C). The detailed mechanism of CSL cross-regulation thus remains to be elucidated by future experiments.
Chua and colleagues previously reported expression microarray data for cbf11 deletion (YES medium) and cbf12 overexpression (EMM medium), and ChIP-chip data for ectopically overexpressed Cbf12 (EMM medium) [24]. There is significant overlap between the Chua dataset and the corresponding subset of our results (S4 Fig). However, Chua and colleagues typically report many more target genes. There are several possible reasons for this discrepancy. First, we performed ChIP-seq using tagged endogenous CSL alleles expressed close to physiological levels, as opposed to ectopic overexpression employed by Chua. Second, we conducted more biological repeats of both transcriptomic and ChIP analyses and applied stringent cutoffs, limiting the number of reported target genes. Third, different genomic platforms were used in the two studies. Thus, we believe that our data validate, and further refine the CSL target genes reported by Chua et al. [24].
We conclude that Cbf11 and Cbf12 bind to an overlapping set of target genes, including genes with important roles in cell cycle regulation, while the detected overlap with genes showing altered expression in CSL-manipulated cells is much more limited.

Subset of cell cycle-regulated genes is deregulated upon CSL manipulation
The CSL DEG clusters 4 and 5b were enriched for genes expressed periodically with the fission yeast cell cycle [47] (Fig 1B). Notably, similar enrichment was also found in the 34 DEGs commonly upregulated in both Δcbf11 and cbf12-overexpressing cells (11 periodic genes, p = 0.02). Overall, 81 periodic genes were either induced or repressed in response to altered CSL levels (Fig 2 and S7 Table). The distribution of peak expression times of these genes during the wildtype cell cycle shows only slight enrichment around the S/G2 transition, compared to all periodically expressed genes (S5 Fig); no further cell-cycle phase-specific enrichment was detected by separate analyses of periodic genes up/downregulated in Δcbf11 cells grown exponentially in YES.
Interestingly, for 31 of these 81 periodic genes (~38%), Cbf12 binding to the corresponding upstream intergenic regions was detected by ChIP-seq, while the overall overlap of CSL DEGs with Cbf12 binding sites was just~25%. Previous genome-wide studies identified novel DNA motifs possibly involved in the regulation of periodic transcription [2][3][4]; however, none of these motifs is related to the CSL response element sequence [22,56,57]. Cbf11 binding was only detected at the promoters of 6 of the 81 periodic CSL DEGs.
While our expression microarray and ChIP-seq data were obtained using unsynchronized cell populations, prohibiting direct assessment of CSL role in the temporal control of gene expression, they nevertheless provide useful indications which genes and processes are affected by CSL proteins during the cell cycle: One of these periodic DEGs is eng1, encoding an endo-1,3-beta-glucanase required for primary cell septum disassembly, which was downregulated 1.5-fold in Δcbf11 cells and 2.2-fold in cbf12-overexpressing cells. Absence of Eng1 results in the inability of daughter cells to separate after cytokinesis [8], reminiscent of the cell separation phenotype observed upon CSL manipulation [22].
Other periodic DEGs include cgs1, encoding the regulatory subunit of protein kinase A, which was upregulated on average 2-fold in Δcbf11 cells and 1.7-fold in cbf12-overexpressing cells. In low glucose, Cgs1 binds to and inhibits Pka1, the catalytic subunit of PKA, which regulates the cell cycle in response to nutrients [58]. Notably, cells lacking PKA activity are small [26,59]. Another example is srk1, which was upregulated 2.4-fold in Δcbf11 cells and 1.9-fold in cbf12-overexpressing cells, and its promoter is bound by Cbf12 (Fig 2). Srk1 is a MAP kinaseactivated protein kinase that acts downstream of the Sty1 MAP kinase and is involved in negative regulation of G2/M transition of the mitotic cell cycle. Cells overexpressing srk1 are long [53]. Such overexpression of proteins with important regulatory effects on cell-cycle progression might well contribute to the cell-cycle phenotypes observed in cells lacking Cbf11 or overproducing Cbf12 ( [22] and Fig 3).  Fig 1A) that belong to top-ranking 500 genes expressed periodically during the cell cycle [47]. CSL binding to promoters of respective DEGs, as detected by ChIP-seq from cells growing exponentially in YES, is indicated at right (dark bars signify CSL binding). Cluster membership is indicated by right-most colour bars (colour coding as in Fig 1A).
doi:10.1371/journal.pone.0137820.g002 Additional experiments are needed to determine any direct contributions of Cbf11 and Cbf12 towards regulating the expression of periodically expressed genes, some of which can explain the cell-cycle progression and cellular morphology defects associated with altered CSL protein levels.

Cbf11 and Cbf12 affect specific aspects of cell-cycle progression
Initial characterization of Cbf11 and Cbf12 revealed that cells with altered CSL levels (cbf11 deletion, cbf12 overexpression) display defects in septation and impaired coordination of cell and nuclear division, resulting in catastrophic mitosis and the 'cut' phenotype (Fig 3A and 3B and [22]; typically 10-20% of Δcbf11 cells are 'cut'). The previously reported cell cycle-related defects triggered by the overexpression of cbf12 were observed in cells grown in the MB medium [22], which is nutrient-poor compared to the more widely used EMM. Also, overexpression driven by the inducible medium-strength nmt1 promoter variant used previously might not have been sufficient to trigger a full phenotypic response to increased gene dosage. Therefore, we analysed cultures grown in EMM where Cbf11 or Cbf12 were overexpressed from the full-strength nmt1 promoter. At 17 hours post induction, the 'cut' phenotype was detected both in cells overexpressing cbf12 (~5% 'cut') and in cells overexpressing cbf11 (~6% 'cut'; Fig 3C).
Prompted by the discovery of deregulated expression of some periodic genes in CSL mutants (Fig 2), we tested whether other aspects of cell-cycle control are also affected by CSL manipulation. Increased or decreased cell lengths at division are hallmarks of mutations in important cell-cycle regulators such as cdc2 CDK1 , wee1, or cdc25 [26]. We measured the cell length at division in wild-type and CSL single and double mutant strains growing exponentially in YES medium, and in CSL-overexpressing cells growing in EMM (Fig 3D). Deletion of cbf11, either alone or in combination with deletion of cbf12, typically resulted in decreased maximum length of septated cells ('wee' phenotype [26]). In addition, a minor subpopulation of Δcbf11 cells showed increased length of septated cells compared to wild type. Furthermore, a subpopulation of cbf11-overexpressing cells displayed severely reduced maximum cell length. These results point to a possible role for Cbf11 in the timing of the G2/M transition. While no change in septated cell length was detected in Δcbf12 cells, there was gross deregulation of maximum cell length upon cbf12 overexpression (Fig 3D). Cell size alteration upon cbf12 overexpression has also been reported by others [25].
The timing of cell-cycle events also manifests in the percentage of cells dividing in an unsynchronized population at any given time (septation index; Fig 3B and 3C). While neither CSL deletion mutant showed any marked changes in the septation index, it was increased in both cbf11and cbf12-overexpressing cells (24.5% and 16.6%, respectively) compared to wild type (10.6%).
Taken together, these findings indicate important roles for CSL proteins in the cell's decision to divide, and support our previous notion that balanced Cbf11 and Cbf12 activities are required for proper cell-cycle progression [22]. Notably, pleiotropic cell cycle and morphology phenotypes, similar in several aspects to Δcbf11 cells (Fig 3 and [22]), have also been observed when the fkh2 transcription factor gene, which regulates M-phase genes, was deleted or overexpressed [9,10]. cbf11 interacts genetically with sty1 and pka1 pathways As mentioned above, our results (Figs 1 and 2) suggest possible functional links between CSL and two antagonistic systems important for cell-cycle progression-the PKA (Pka1) and stressactivated MAP kinase (Sty1) pathways [60]. Therefore we constructed and characterized double mutants of Δcbf11 with Δpka1/Δsty1 to assess any genetic interactions related to cellcycle control. As shown in Fig 4A, deletion of sty1 strongly suppressed the growth defect of Δcbf11 cells; moderate suppression was also evident in Δcbf11 Δpka1 cells. Microscopic analysis of these double mutants revealed diminished Δcbf11-associated defects in nuclear integrity and septum formation (Fig 4B and 4C); this was also confirmed by FACS analysis of DNA content (Fig 4D). Furthermore, a fraction of septated Δcbf11, Δpka1, and Δcbf11 Δpka1 cells actually formed multicellular filaments. This phenotype was not additive in the double mutant, indicating common defects in daughter cell separation in Δcbf11 and Δpka1 cells (Fig 4C and 4E).
There were also striking effects of pka1 and sty1 mutations on cell size at division in Δcbf11 cells (Fig 4E). Sty1 has a role in promoting entry into mitosis and Δsty1 cells are elongated [61]. Such elongation was uniformly observed in the Δcbf11 Δsty1 strain. The absence of Sty1 thus completely suppressed the 'wee' phenotype found in most Δcbf11 cells. Pka1, on the other hand, is a negative regulator of mitotic entry and Δpka1 cells are short [26,59], about the same length as most Δcbf11 cells. Notably, the 'wee' phenotype was further potentiated in Δcbf11 Δpka1 cells, and the subpopulation of large cells, present in the Δcbf11 single mutant, was diminished.
Thus, there is indeed crosstalk in the regulation of cell-cycle progression between the CSL transcription factors, and the PKA and stress MAP kinase pathways in fission yeast. Since many Sty1 target genes are upregulated in Δcbf11 cells (Fig 1A and 1B), Cbf11 might act as a negative regulator of Sty1.

Defects of Δcbf11 cells are diminished in minimal medium
Since both Pka1 and Sty1 pathways are sensitive to nutrient availability [61][62][63][64], we tested the influence of growth media composition on Δcbf11-associated defects. Wild-type cells grow slower in minimal medium (EMM) than in rich medium (YES). In contrast, the slow growth phenotype of Δcbf11 cells observed in YES was partially alleviated in EMM (Fig 5A). The Δcbf11 growth rate showed a further slight increase in a 1:1 mixture of YES and EMM, suggesting that both media contain specific substances (or different concentrations thereof) that are limiting for the growth of Δcbf11 cells. However, the contribution of EMM seemed to be more important as the Δcbf11 doubling time increased in a dose-responsive manner with decreasing EMM content in the mixed medium ( Fig 5B). Furthermore, the cell morphology and cell separation defects, and the 'cut' phenotype observed in Δcbf11 cells growing exponentially in YES were largely absent in EMM (Fig 5C and 5D). Also, FACS analysis of DNA content confirmed diminished nuclear integrity defects in Δcbf11 cells grown in EMM (Fig 5E).
Taken together, we propose that Cbf11 is less critical during growth on EMM. This differential requirement is also consistent with the small number of DEGs detected in Δcbf11 cells grown in EMM (Fig 1A and S1 Fig, panel A). Notably, Cbf11 showed decreased DNA-binding activity in vitro (Fig 5F; note that the change in bandshift intensity is less pronounced in wild-type cells compared to Cbf11-tagged cells). This result could reflect lower Cbf11 expression or altered posttranslational modifications of Cbf11 in EMM. Indeed, western blot analysis revealed lower Cbf11 amounts in EMM compared to YES. Furthermore, Cbf11 was present in two isoforms, and their relative quantities were different in YES and EMM (Fig 5G). It remains to be established whether these changes in protein levels and isoforms also occur with untagged Cbf11. Thus, our data suggest that Cbf11 regulates gene expression mainly during rapid cell proliferation in rich medium. The precise nature of the nutrient(s) or other factor(s) critical for Cbf11 activity is currently under investigation. Indicates whether a gene has been deemed to be periodically expressed, and the cell-cycle phase of peak expression.
doi:10.1371/journal.pone.0137820.t001 The mRNA levels at each condition relative to the levels in wild-type control cells are colour-coded as indicated at top right, with missing data in grey; for description of cultivation conditions see legend to Fig 1A. (B) The sequence motif identified by MEME-ChIP [65] in the Cbf11-bound promoter regions of genes from (A) resembles closely the canonical metazoan CSL response element, 'Su(H)'. (C) Previous transcriptome profiling of the genes from (A) over two cell cycles of elutriation-synchronized wild-type cultures suggest moderate periodic oscillations in the expression of some of these genes. Data taken from [2]. (D) Probes used for EMSA experiments (sense strand shown); CSL response elements are in bold and introduced mutations are in lower-case and underlined. 'RBP', 'KSHV', 'MUT' and 'DEL' probes were derived from metazoan/viral CSL-responsive genes [22]. (E) Representative competitive EMSAs with wild-type cell extracts and radioactively labelled RBP probe; unlabelled cut6, ptl1 Cbf11 regulates a group of lipid metabolism genes As shown above, the lists of promoters bound by Cbf11 and Cbf12 in unsynchronized cells overlapped significantly, but not fully (Fig 1C). To get further insight into any specific role of Cbf11, we analysed the promoters bound exclusively by Cbf11, which were significantly associated with lipid metabolism genes (7 genes in total, p = 0.019; Table 1 and Fig 6A). Notably, a sequence motif closely resembling the canonical metazoan CSL response element [56,57] was found in the promoters of all these 7 genes (Fig 6B). Furthermore, 6 of these genes were downregulated in Δcbf11 cells grown exponentially in YES, i.e., under the same conditions as for the ChIP-seq experiment (Fig 6A). Some of these lipid metabolism genes show weakly periodic expression during the cell cycle ( Fig 6C, data taken from [2]). We tested whether Cbf11 can bind in vitro to double-stranded DNA oligonucleotides containing the in silico-identified putative Cbf11 binding sites from ptl1 and cut6 promoters. In competitive EMSA assays, Cbf11 indeed bound to these sequences specifically, with somewhat lower affinity compared to the previously tested probes derived from known metazoan CSL binding sites (Fig 6D-6F). We conclude that Cbf11 is likely a direct activator for transcription of lipid metabolism genes. Interestingly, the colonies of the Δcbf11 strain overproduce a layer of wax-like material on their surface [22], which may result from a possible dysregulation of lipid metabolism when Cbf11 is missing. To assess the impact of cbf11 deletion on neutral lipid content in cells growing exponentially in YES, we quantified the number of storage lipid droplets using Nile red staining. Compared to wild type, there was a striking depletion of lipid droplets in Δcbf11 cells, supporting a role of Cbf11 in lipid metabolism. Curiously, in many apparently live Δcbf11 cells, no lipid droplets were detected at all (Fig 6G and 6H).
Notably, it has been recently reported that mammalian cells remodel their membrane lipid composition during the cell cycle, and these periodic changes are important for proper cellcycle progression [66]. Importantly, the cut6 gene is required for proper coordination of cell and nuclear division, and cut6 mutants show the 'cut' phenotype [67], which also occurs in the Δcbf11 strain [22]. This finding raises the possibility that the 'cut' phenotype of Δcbf11 is mediated by cut6.

Conclusions
CSL transcription factors affect multiple aspects of cell-cycle progression in fission yeast ( [22,25] and this study). Our genomic analyses uncovered sets of genes regulated directly or indirectly, and often antagonistically, by Cbf11 and Cbf12, and highlighted specific biological processes in which CSL proteins play a distinct role. We showed that many genes expressed periodically during the cell cycle are deregulated upon CSL deletion/overexpression, which could explain some of the cell-cycle defects associated with CSL manipulation. We further showed that cbf11 interacts genetically with the nutrient-responsive cell-cycle regulatory pathways controlled by protein kinase A (Pka1) or stress-activated MAP kinase Sty1 p38 . Interestingly, Cbf11 expression is affected by media composition, and Cbf11 activity is required mainly under rapid proliferation in rich medium. Finally, we identified Cbf11 as a novel regulator of lipid metabolism genes. Our results provide a basis for a more detailed understanding of CSL Selected CSL binding sites identified from ChIP-seq data by peak calling algorithm MACS [42] were validated by ChIP-qPCR. The ChIP-qPCR enrichment of target loci DNA was normalized to a control locus where no CSL binding was detected by ChIP-seq and plotted against integrated ChIP-seq signal from the region ±100 bp from qPCR amplicon centre (area under peak, AUP).   Table. Genes showing differential expression in CSL deletion or overexpression (from Fig 1A). Table. Clusters of genes showing differential expression in CSL deletion or overexpression (from Fig 1A). Table. Periodically expressed genes showing differential expression in CSL deletion or overexpression (from Fig 2). (XLS)