Global Gene Expression Analysis of Fission Yeast Mutants Impaired in Ser-2 Phosphorylation of the RNA Pol II Carboxy Terminal Domain

In Schizosaccharomyces pombe the nuclear-localized Lsk1p-Lsc1p cyclin dependent kinase complex promotes Ser-2 phosphorylation of the heptad repeats found within the RNA pol II carboxy terminal domain (CTD). Here, we first provide evidence supporting the existence of a third previously uncharacterized Ser-2 CTD kinase subunit, Lsg1p. As expected for a component of the complex, Lsg1p localizes to the nucleus, promotes Ser-2 phosphorylation of the CTD, and physically interacts with both Lsk1p and Lsc1p in vivo. Interestingly, we also demonstrate that lsg1Δ mutants – just like lsk1Δ and lsc1Δ strains – are compromised in their ability to faithfully and reliably complete cytokinesis. Next, to address whether kinase mediated alterations in CTD phosphorylation might selectively alter the expression of genes with roles in cytokinesis and/or the cytoskeleton, global gene expression profiles were analyzed. Mutants impaired in Ser-2 phosphorylation display little change with respect to the level of transcription of most genes. However, genes affecting cytokinesis – including the actin interacting protein gene, aip1 – as well as genes with roles in meiosis, are included in a small subset that are differentially regulated. Significantly, genetic analysis of lsk1Δ aip1Δ double mutants is consistent with Lsk1p and Aip1p acting in a linear pathway with respect to the regulation of cytokinesis.


Introduction
In Schizosaccharomyces pombe, just as in developmentally complex metazoans, cytokinesis is mediated through the assembly and constriction of a contractile, actomyosin ring [1]. In addition to actin, myosin, and other accessory factors that provide the mechanical force necessary for ring constriction, complex signalling networks also exist to ensure that cytokinesis occurs at the correct location within the cell and at the correct temporal position during the course of the cell cycle [2,3,4,5,6,7,8]. Given the importance of cytokinesis in cellular growth and development, it is not surprising that evidence supporting the existence of a cytokinesis monitoring system has emerged in S. pombe. This system has the capacity to both delay G2/M progression and stabilize the actomyosin ring upon perturbation of the division machinery and thus aids in the faithful and reliable execution of fission yeast cytokinesis [9,10].
Critical components of the monitoring system include the Septation Initiation Network (SIN), the Cdc14p family phosphatase, Clp1p, and the Ser-2 carboxy terminal domain (CTD) kinase, Lsk1p [6,9,10,11,12,13,14]. The SIN is comprised of a complex network of proteins (localizing to the spindle pole body and/or the actomyosin ring) that are required for successful ring constriction following chromosome segregation [6,12,13]. Clp1p, on the other hand, is a Cdc14 family phosphatase that plays an important role in SIN activation. While the loss of clp1 results in only weak cytokinesis defects under normal growth conditions, clp1D cells display a lethal multi-nucleate phenotype when treated with drugs (Latrunculin A) that perturb the normal function of the actomyosin ring. These defects are in part due to the inability of clp1D cells to prolong the duration of SIN signaling [9,10]. Similar to clp1D cells, lsk1D mutants appear normal under typical growth conditions, but display striking cytokinesis defects upon perturbation of the cell division machinery. Furthermore -as demonstrated by genetic analysis with hypo-and hyper-active SIN mutants -Lsk1p promotes SIN activation [11].
Rather surprisingly given its role in regulating cytokinesis, lsk1 encodes a member of a multi-protein complex that specifically phosphorylates the Ser-2 residues present in the twenty-nine heptad repeats (Y 1 S 2 P 3 T 4 S 5 P 6 S 7 ) found at the extreme carboxy terminus of the largest subunit of RNA polymerase II, Rpb1p [15,16,17]. Lsk1p displays significant sequence similarity to human Cdk9p and budding yeast Ctk1. Human Cdk9p, together with cyclin T, forms the p-TEFb complex, while Ctk1 associates with Ctk2 (cyclin sub-unit) and Ctk3 (gamma sub-unit). Both complexes target Ser-2 residues of the RNA pol II CTD and promote transcript elongation [18,19].
Importantly, S. pombe cells bearing rpb1 alleles (rpb1-12XS-2ACTD) that encode 12 copies of a mutant heptad sequence in which alanine has been substituted for serine at the ''two'' position (in order to mimic the non-phosphorylated state) display cytokinesis phenotypes indistinguishable from lsk1D strains [14].
These data are consistent with a model in which the CTD is indeed the critical target of Lsk1p with respect to its effects on cytokinesis, and suggests that upstream kinases (or phosphatases) might act through the CTD to selectively control the transcription of certain gene subsets.
In recent years several studies have provided evidence supporting such a hypothesis. For example, work in both Schizosaccharomyces pombe and Saccharomyces cerevisiae has demonstrated that the impairment of CTD phosphorylation does not necessarily result in generalized defects in transcription or transcriptional control [20,21,22]. On the contrary, discrete sub-sets of genes are found to be differentially regulated in such cases. These results demonstrate, at least in principle, that changes in CTD phosphorylation can result in selective as opposed to general alterations in transcription and phenotype.
In this report, we first demonstrate the existence of a third bona fide subunit of the fission yeast Ser-2 CTD kinase complex, Lsg1p (for latrunculin sensitive CTD kinase gamma subunit). We also analyze the effects of impaired Ser-2 phosphorylation on the expression of genes with roles in regulating cytokinesis and/or the cytoskeleton. Significantly, strains bearing the rpb1-12XS2ACTD mutation display little change with respect to the level of transcription of most genes. However, genes encoding proteins with roles in cytokinesis -including i) the actin interacting protein, Aip1p, ii) the Clp1p interacting protein, Nsk1p, iii) the spindle pole body protein, Cut12p, and iv) Lsk1p itself -are included in a small subset that are differentially regulated. Surprisingly, many genes with roles in the meiotic cell cycle are also significantly affected. This is consistent with recent data demonstrating that Ser-2 phosphorylation is critical for the induction of ste11 transcription during sexual differentiation [23]. The relevance of this finding within the context of understanding the possible relationship(s) between CTD phosphorylation pattern (i.e. a CTD ''code'') and complex phenotypic change is discussed.

Results
The annotated S. pombe ORF, SPCC4B3.08, defines a putative gamma subunit of the Ser-2 CTD kinase complex Ser-2 CTD kinase complexes in fungi consist of i) a catalytic kinase subunit, ii) a regulatory cyclin subunit, and iii) a regulatory ''gamma'' subunit [24]. A BLAST search using a known gamma subunit as query (budding yeast, Ctk3) revealed that the annotated S. pombe gene, SPCC4B3.08, was a likely candidate to encode the fission yeast orthologue ( Figure S1A). Proteins related to S. pombe SPCC4B3.08 at the primary sequence level exist in the fungal kingdom, but not in metazoans ( Figure S1B and C). The SPCC4B3.08 gene will hereafter be referred to as lsg1 for latrunculin sensitive gamma subunit. lsg1 deletion mutants display cytokinesis defects similar to those displayed by lsk1D and lsc1D strains Deletion of the lsk1 gene in fission yeast confers hyper-sensitivity to low doses of LatA, a drug that inhibits actin polymerization through sequestering actin monomers [25]. At the concentrations used (30-50 times less than that needed to completely depolymerize the actin cytoskeleton) such treatment leads to a Clp1p dependent delay in mitotic entry and the extended activation of the SIN leading to a prolonged cytokinesis-competent state characterized by continuous repair and re-establishment of the actomyosin ring [9,10]. In the presence of the drug lsk1D strains are unable to maintain the integrity of the actomyosin ring leading to cytokinesis failure and the generation of inviable, multi-nucleate cells [11].
In order to test whether lsg1 encodes a bona fide member of the CTD complex we first deleted the lsg1 gene and assayed the mutant for sensitivity to LatA treatment. Ten fold serial dilutions of lsg1D, lsk1D, and a wild-type control strain were spotted onto solid media containing LatA (0.5 mM) or DMSO (drug solvent) and the plates incubated for 3 days at 30uC. In contrast to the wild-type control, both lsg1D and lsk1D cells displayed a similar hypersensitivity to LatA ( Figure 1A).
To further investigate the phenotype of the lsg1 deletion strain at the microscopic level, lsg1D, lsk1D and wild-type cells were grown to mid-log phase at 30uC and treated with LatA for 6 hours. After staining with DAPI and aniline blue (dyes which act as markers for the nucleus and cell wall/septum, respectively) cells were examined by fluorescence microscopy. The cells were classified into five different phenotypic categories: i) uni-nucleate cells, ii) binucleate cells with a morphologically normal septum (i.e. the septum completely bisects the cell), iii) bi-nucleate cells with a fragmented septum (i.e. the septum is non-functional and does not completely bisect the cell), iv) tetra-nucleate cells, and v) cells with greater than four nuclei. As shown in Figures 1B and C, lsg1D, lsk1D and wild-type strains did not display any obvious growth defects when grown in the presence of DMSO. In contrast, approximately 60% of lsg1D and lsk1D cells displayed a tetranucleate phenotype when treated with LatA, whereas only approximately 10% of wild type were tetra-nucleate. Furthermore, cells with greater than four nuclei were only observed in the gene deletion mutants ( Figure 1C). The distribution of phenotypes between the various classes was similar in both lsg1D and lsk1D strains.
Previous studies have shown that Lsk1p and Lsc1p positively regulate the SIN [11,14]. For example, the deletion of either lsk1 or lsc1 rescues the lethal, multi-septate phenotype conferred by the temperature-sensitive cdc16-116 allele (this allele results in hyperactivation of the SIN pathway) [26]. To test the effects of the lsg1D mutation on the SIN, lsg1D cdc16-116 strains were created. As shown in Figure 1D, cdc16-116 cells could not form colonies and died as single undivided cells at 36uC. In contrast, lsg1D cdc16-116 double mutants could easily form viable colonies. Furthermore, when the cells were examined at the microscopic level, cdc16-116 mutants showed a strong multi-septate phenotype at 36uC while the lsg1 cdc16-116 cells exhibited a reduced number of septa and were capable of dividing. This is consistent with a model in which Lsg1p, acts as a positive regulator of the SIN.
Lsg1p physically associates with Lsk1p and localizes to the nucleus As one might expect based on their role in modulating phosphorylation of RNA pol II, both Lsk1p and Lsc1p localize to the nucleus [11,14]. To test if Lsg1p showed a similar localization, we created a C-terminal lsg1-GFP fusion. Analysis of fixed lsg1-GFP cells co-stained with DAPI demonstrated that Lsg1-GFP fusion proteins were targetted to the nuclear compartment (Figure 2A). In a complementary experiment, lsk1D strains expressing GFP-tagged Lsg1p were created to assess the requirement of Lsk1p for the nuclear localization of Lsg1p. This was done since Lsk1p is required for the proper localization of Lsc1p to the nucleus [14]. Analysis of this strain using fluorescence microscopy demonstrated that the Lsg1-GFP signal in lsk1D lsg1-GFP mutants was spread within the cytoplasm and was not exclusively localized to the nucleus (Figure 2A). To ensure that the overall level of expression of Lsg1-GFP was the same in wild-type and lsk1D backgrounds, individual cells were analyzed by thresh- Figure 1. lsg1 deletion mutants display cytokinesis defects similar to those displayed by lsk1D strains. (A) Ten-fold serial dilutions of logarithmically growing wild-type, lsg1D and lsk1D cells were plated onto YES plates containing 0.5 mM LatA or DMSO (solvent control) at 30uC for 3 d. (B) Cells of the indicated genotype were grown to mid-log phase at 30uC and then treated with 0.5 mM LatA for 6 h before being fixed and stained with DAPI (nuclei) and aniline blue (cell wall/septa). Bar, 10 mm. (C) Quantitation of phenotypes observed for cells treated as in B. Two hundred cells were counted for each genotypic class. Error bars indicate s.d. (n = 3) (D, top panels) Cells of the indicated genotypes were grown at 24uC on YES plates, re-streaked to fresh YES media and then incubated at 36uC for 24 h. Bar, 50 mm (D, bottom panels) Cells of the indicated genotypes were olding background subtracted images and quantitatively analyzing the identified particles (n = 20 for each genotype). The mean gray value (the sum of the gray values of all the pixels in the particle divided by the number of pixels) was not significantly different in wild-type (17.4+/21.45) vs. lsk1D mutants (16.5+/21.69) ( Figure 2B). These data demonstrate that Lsk1p is required for the proper localization of Lsg1p to the nucleus, and suggest that it might be important in the intracellular trafficking of the protein.
Next, to determine if Lsg1p physically associates with Lsk1p and Lsc1p in vivo, co-immunoprecipitation experiments using myctagged lsg1 alleles were performed. Consistent with the hypothesis that Lsg1p physically interacts with the other two subunits of the complex, Lsg1-myc fusion proteins could be detected with anti-myc antibodies after immuno-precipitation of extracts from lsg1myc lsk1-HA and lsg1-myc lsc1-HA strains with anti-HA antibodies ( Figure 2C). Taken together, these data strongly suggest that Lsg1p is indeed part of a tri-partite complex with both Lsk1p and Lsc1p.

Ser-2 phosphorylation is reduced in lsg1D mutants
The Rpb1p CTD can be phosphorylated on both Ser-2 and Ser-5 residues of the heptad repeats [27]. Lsk1p and Lsc1p are required for Ser-2 phosphorylation of the CTD, but do not affect the phosphorylation status of Ser-5 residues [14]. We thus chose to examine phosphoserine-2 levels in lsg1D mutants relative to wild-type cells. To test if this was the case, western blotting using phosphospecific antibodies was performed to determine the effects Figure 2. Lsg1p localizes to the nucleus and physically associates with Lsk1p. (A) Cells of the indicated genotypes were grown to mid-log phase at 30uC in YES media, fixed, and then stained with DAPI (nuclei) and observed using the DAPI and GFP filter sets. Bar, 5 mm. (B) Mean gray level intensity measurements of wild-type (top) or lsk1D cells (bottom) expressing the Lsg1-GFP fusion protein. Background subtracted images were thresholded and the identified particles quantitatively analyzed using ImageJ. (C) Cells expressing the indicated epitope tagged fusion protein were grown to mid-log phase in YES, lysed under native conditions, and subjected to anti-HA immunoprecipitations. Both total lysates (2% input) and immunoprecipitates were resolved by SDS-PAGE and immunoblotted with antibodies specific for the myc epitope. Note the presence of the nonspecifically reacting peptide (present in all immunoprecipitates) that acts as a convenient loading control. doi:10.1371/journal.pone.0024694.g002 grown to mid-log phase at 24uC and then shifted to 36uC for 4 h. Cells were subsequently fixed and stained with aniline blue (cell wall/septa) and DAPI (nuclei). Bar, 10 mm. doi:10.1371/journal.pone.0024694.g001 of the lsg1 deletion on the phosphorylation of the CTD. A set of three commercially available phosphospecific antibodies was used: 8WG16 (preferential recognition of unphosphorylated CTD), H14 (preferential recognition of Ser-5 phosphorylated CTD), and H5 (preferential recognition of phospho-Ser-2 and phospho-Ser-2,5) [27,28]. The level of Ser-2 phosphorylation in lsg1D and lsk1D strains was reduced in comparison to wild-type cells. However, there was no significant change in the level of Ser-5 phosphorylation or in the levels of unphosphorylated CTD. Antibodies specific for a-tubulin were used to show that equal amounts of protein were loaded for each sample ( Figure 3A).
Interestingly, Karagiannis and Balasubramanian (2007) and others [23] reported that lsk1D mutations completely abolish Ser-2 phosphorylation. In contrast, the above experiment demonstrated only a reduction in the levels of Ser-2 phosphorylation. It was thus of interest to identify the origin of this discrepancy. Since the original protocols did not include phosphatase inhibitors, the experiment was repeated with and without this treatment. Similar to the previously published work, bands were completely absent from lsk1D and lsg1D samples when no phosphatase inhibitors were included in the extraction buffer ( Figure 3B). We thus conclude that the inclusion of phosphatase inhibitors in the extraction buffer was the source of the discrepancy in results. Importantly, these data suggest the existence of at least one other kinase capable of phosphorylating Ser-2 residues in the CTD.

Gene expression analysis
To this point all data were supportive of Lsg1p functioning in a complex with Lsk1p and Lsc1p to phosphorylate Ser-2 residues within the heptad repeats of the Rpb1p subunit of RNA polymerase II. Furthermore, genetic analysis of lsg1D, lsk1D, and lsc1D mutants clearly demonstrated that all three sub-units were important for the reliable execution of cytokinesis. Since misregulation of Ser-2 phosphorylation might alter the control of transcription, we decided to analyze the global gene expression profiles of mutants impaired in Ser-2 phosphorylation of the RNA polymerase II CTD. We hypothesized that the transcription of genes involved in cytokinesis or the cytoskeleton might be selectively affected by impairment of Ser-2 phosphorylation. Two strains were used in these studies: a mutant strain bearing the rpb1-12XS2ACTD mutation and a control strain of the genotype rpb1-12XWTCTD.
The rpb1 gene in the rpb1-12XS2ACTD strain has been mutated so as to encode a total of 12 mutant heptad repeats (Y 1 A 2 P 3 T 4 S 5 P 6 S 7 ) in which alanine residues have replaced serine residues at the ''two'' position. Interestingly, this mutant phenocopies lsk1D, lsc1D, and lsg1D strains in terms of the strain's inability to complete cytokinesis upon LatA treatment, and in its capacity to rescue the temperature sensitive cdc16-116 mutation [11,14]. The rpb1 gene in the rpb1-12XWTCTD strain, on the other hand, encodes a molecule bearing 12 wild-type heptad repeats (Y 1 S 2 P 3 T 4 S 5 P 6 S 7 ). This strain displays no obvious growth or morphological phenotypes and does not exhibit any cell cycle or cytokinesis defects. These strains can thus be employed as helpful tools in understanding the effects of the impairment of Ser-2 phosphorylation on cellular physiology and transcription.
To obtain global gene expression profiles, rpb1-12XCTD and rpb1-12XS2ACTD strains were grown to mid-log phase at 30uC, and then treated with either DMSO (solvent control) or LatA for three hours. Total RNA was then extracted and used in microarray hybridizations using Yeast Genome 2.0 Gene Chips purchased from Affymetrix. Three replicates of each strain (rpb1- 12XCTD or rpb1-12XS2A) under each growth condition (plus or minus LatA) were obtained for a total of 12 samples. After data processing (see Materials and Methods), the twelve samples were grouped according to two parameters: genotype (rpb1-12XCTD or rpb1-12XS2ACTD) and drug (LatA treated or DMSO treated). In this way, four experimental classes were created: 1) rpb1-12XCTD, DMSO treated, 2) rpb1-12XCTD, LatA treated, 3) rpb1-12XS2ACTD, DMSO treated, and 4) rpb1-12XS2ACTD, LatA treated. The complete data set, showing log 2 normalized intensity values for all genes is included in File S1.
We first examined the gene expression data at the ''macro'' level to determine if there were any gross differences in the global expression profiles between strains and/or LatA treatment. Frequency distributions of the data revealed only modest changes in expression; all four data sets had similar mean expression levels, standard deviations, as well as similar percentile values ( Figure 4A, Table S1). Furthermore, scatterplots comparing rpb1-12XCTD and rpb1-12XS2ACTD strains treated with either DMSO or LatA ( Figure 4B) showed a strong correspondence of expression for the vast bulk of genes. Thus, on a global scale, loss of Ser-2 phosphorylation resulted in only modest changes in the control of gene expression.

Stress response genes
Next, in order to provide a non-biased assay of the biological veracity of the hybridization data, we examined the core environmental stress response (CESR) genes. The expression levels of these genes is strongly affected by a wide variety of different stresses that include heat shock, cold shock, osmotic stress, oxidative stress, as well as a variety of different DNA damaging agents [29]. The expression of these genes can thus be used as ''markers'' for exposure to environmental stress. Since one would expect LatA treatment to constitute such a stress, we reasoned that the expression levels of the CESR genes could be used as an informative control to ensure that the expression arrays were providing biologically meaningful data.
To analyze the data, the CESR gene set was divided into two individual groups, the CESR-Up group (comprised of genes that are up-regulated upon exposure to environmental stresses) and the CESR-Down group (comprised of genes that are down-regulated upon exposure to environmental stress). The expression levels of these two groups of genes in DMSO and LatA treated rpb1-12XCTD cells, as well as DMSO and LatA treated rpb1-12XS2ACTD cells, were compared using volcano plot (a scatterplot of the -log 10 -transformed p-values from gene-specific t-tests vs. the log 2 fold change). As seen in Figure 5A the CESR-Up genes (blue) are generally up-regulated in response to LatA treatment, while the CESR-Down genes (red) are generally down-regulated in both genetic backgrounds. Furthermore, regression lines fitted to the CESR-Up and CESR-Down genes, result in lines of positive and negative slope respectively, as expected. These results indicated that the microarray data was providing accurate and biologically relevant expression data.
Since it was conceivable that the LatA hyper-sensitivity observed in rpb1-12XS2ACTD strains was due to defects in mounting an appropriate transcriptional response to stress, we  also compared CESR gene expression between genotype. As expected, no significant differences in CESR gene expression were observed between DMSO treated rpb1-12XCTD and rpb1-12XS2ACTD strains (data not shown). Furthermore, in response to LatA treatment, only three (out of 136) CESR-UP genes, and six (out of 104) CESR-DOWN genes were differentially regulated between genotype ( Figure 5B an C). Lastly, when we expanded the analysis to include any gene annotated by the Gene Ontology Consortium as having a role in the cellular response to stress, we reached similar conclusions; only 8 out of 561 genes in the set were differentially regulated between genotype upon LatA treatment while no genes in the set were differentially regulated upon DMSO treatment ( Figure S2). These results strongly suggest that the LatA sensitivity conferred by rpb1-12XS2ACTD mutations does not not stem from any gross failings in the control of expression of stress response genes.

Cytokinesis genes
Since rpb1-12XS2ACTD strains display defects related to cytokinesis we next analyzed genes annotated as having roles in cytokinesis or in controlling the cytoskeleton (a total set of 333 genes) (see File S1 for a complete list). To identify differentially expressed genes we filtered the data using volcano plot analysis to screen for statistically significant changes (p-value,0.05) greater than 1.5 fold in magnitude. When DMSO treated rpb1-12XCTD and rpb1-12XS2ACTD strains were compared -conditions where the two strains show no obvious phenotypic differences -there were no cytokinesis genes found to be differentially regulated. When LatA treated rpb1-12XCTD and rpb1-12XS2ACTD strains were compared only 4 out of 333 genes were differentially regulated ( Figure 6). While this low number was somewhat surprising, the identities of the genes were highly notable. The first of the three up-regulated genes was lsk1 itself, suggesting the existence of positive feedback between low phosphoserine-2 levels and the control of lsk1 expression. The second of the up-regulated genes, nsk1, encodes a known physical interactor of Clp1p, a previously characterized regulator of the cytokinesis monitoring system [9,10]. The third of the up-regulated genes, cut12, encodes a spindle pole body protein with known roles in regulating mitosis, cytokinesis, and the SIN [30,31,32]. Intriguingly, the only downregulated gene in the set, aip1, encodes an actin interacting protein whose human orthologue has a known role in preventing cytokinesis failure in HeLa cells [33].

Genetic Analysis
To explore whether any of these differentially regulated genes might be relevant to the LatA sensitivity of rpb1-12XS2ACTD strains, we chose to focus on aip1 (analysis of nsk1 and cut12 will be presented elsewhere). To examine the role of Aip1p we analyzed aip1D, aip1D clp1D and aip1D lsk1D mutants and assayed their LatA sensitivity. aip1D clp1D were analyzed since loss of clp1 exacerbates the cell division defects of a variety of diverse cytokinesis mutants. Moreover, mutants with weak cytokinesis defects often display strong phenotypes in clp1D backgrounds. While aip1D single mutants showed only modest defects, aip1D clp1D double mutants were exquisitely sensitive to LatA treatment. Even at very low concentrations -where both single mutants and the wild-type were insensitive -double mutants were unable to complete cytokinesis and accumulated multiple nuclei (Figure 7). Also of significance is the observation that no such synthetic interaction was observed in aip1D lsk1D (Figure 7) or rpb1-12XS2ACTD aip1D (data not shown) strains i.e. the double mutants were no more sensitive than lsk1D or rpb1-12XS2ACTD single mutants. These results demonstrate that Aip1p does indeed influence sensitivity to LatA in fission yeast. Furthermore, the fact that loss of Aip1p increased sensitivity in clp1D strains, but had no effect in lsk1D strains, is consistent with aip1 acting in a parallel pathway to clp1 and in a linear pathway with lsk1. These data are thus consistent with a model in which the control of aip1 transcription is part of a transcriptional ''program'' affected by Lsk1p mediated phosphorylation of the CTD.

Meiosis genes
We next examined the global set for genes significantly affected by genotype. Upon LatA treatment we found that only 152 out of 4875 genes (approximately 3%) were differentially regulated (foldchange .1.5, p,0.05) when comparing rpb1-12XCTD and rpb1-12XS2ACTD strains (see File S1 for a complete list). In this set were a diverse set of genes involved in cellular processes including metabolism, membrane transport, and cell wall architecture. Despite the apparent diversity, a clear pattern in these genes was noted upon careful analysis of the data. Of the gene set, 47 (approximately 31%) were found to be part of the transcriptional program initiated upon entry into the meiotic cell cycle. By cross referencing to data [34] made freely available from the Bahler lab (http:// www. bahlerlab. info / projects / sexualdifferentiation/ meiosis/) we found that 41 of these genes were induced upon entry into the meiotic cycle, whereas 6 were repressed ( Figure S3). Consistent with this result, Ser-2 phosphorylation was recently shown to be important for the induction of ste11 transcription during sexual differentiation. Moreover, homothallic lsk1 gene  deletion mutants were also found to display meiotic defects [23]. Taking all data together, these results are consistent with a model in which the modulation of CTD phosphorylation results in selective changes in a limited set of genes and in discrete phenotypic alterations.

Discussion
In this report, we identify the gene product of the S. pombe ORF, SPCC4B3.08, as a bona fide member of the fission yeast Ser-2 CTD kinase complex. This conclusion is based both on biochemical evidence demonstrating physical interaction between Lsg1p and the Lsk1p and Lsc1p sub-units ( Figure 2C), together with genetic evidence showing that lsg1 gene deletion mutants exhibit cytokinesis phenotypes indistinguishable from lsk1D strains (Figure 1). We also demonstrate that Lsg1p localizes to the same intracellular compartment as Lsk1p, and moreover that this intracellular localization is dependent on Lsk1p (Figure 2A and B). This last observation highlights the interdependence of the Ser-2 CTD kinase complex and suggests that Lsg1p is targeted to the nucleus through its association with Lsk1p. Alternatively, Lsk1p mediated phosphorylation of Lsg1p might be important for intracellular trafficking of the Lsg1p protein independently of its association with Lsk1p.
While clearly required for the Ser-2 specific phosphorylation activity of the kinase complex (Figure 3), the mechanism by which ''gamma'' subunits in general, and Lsg1p in particular, affect the catalytic kinase component of the complex remains poorly understood. While the stability of budding yeast Ctk3 is known to be regulated through the presence of a PEST domain [35], Lsg1p does not possess an orthologous sequence ( Figure S1). This observation thus precludes the modulation of Lsg1p stability through a PEST domain as a possible means of regulatory control over Ser-2 CTD kinase activity. In other tri-partite CDK complexes such as the Mat1-cyclin H-CDK 7 complex, the noncyclin regulatory sub-unit is known to stabilize the cyclin and kinase components and thus to promote the formation of a catalytically active CDK [36]. Regardless of the molecular mechanism(s) regulating the complex, it is interesting to speculate as to whether the Lsg1p and Lsc1p sub-units might be sensitive to distinct cellular signals. In this way the control of CTD phosphorylation could be finely tuned to integrate information from a wide variety of diverse signalling pathways.
When considering such models it is intriguing to note the variety of potential phosphorylation sites available to act as substrates for regulatory kinases and/or phosphatases (29 copies of the heptad sequence exist in fission yeast, whereas 52 repeats exist in humans). Thus, differential phosphorylation of the CTD may define specific ''codes'' which can be ''read'' by CTD binding proteins to influence transcriptional control [15,16,37,38]. This potential complexity has led some to speculate that the CTD is crucial for the intricate and selective regulation of genetic networks in eukaryotes. For example, Stiller and Hall (2002) have hypothesized that acquired CTD/protein interactions have been crucial in the evolution of complex patterns of gene expression characterizing developmentally complex eukaryotes. The importance of the CTD is also borne out by the observation that it is required for viability, but not for basal transcriptional activity in vitro [16,17,38,39]. This key observation strongly suggests that, while the CTD is not catalytically essential, it must perform other crucial functions within eukaryotes [15,37].
In addition to these arguments, recent empirical evidence has also appeared suggesting complex roles for the CTD in the regulation of discrete genetic pathways within eukaryotic cells.
First, work in S. pombe has clearly shown that alterations in CTD phosphorylation do not result in generalized defects in the overall level of transcription. For example, Lee et al., (2005) demonstrate that the reduction of Ser-5 phosphorylation (through mutation of the cdk7 orthologue, mcs6) leads to the reduction of a sub-set of transcripts involved in the regulation of septum dissolution at the end of cytokinesis. Furthermore, temperature-sensitive mcs6 mutants die with a phenotype characteristic of mutants that are unable to fully degrade the septum after cytokinesis.
Even more intriguing is the observation that the phosphorylation status of the CTD can be affected by environmental conditions. For example, the level of Ser-2 phosphorylation increases both upon heat shock, and during the diauxic shift in Saccharomyces cerevisiae. In addition, phosphoserine-2 levels are also seen to increase upon exposure to DNA damaging agents in a manner dependent on the Ser-2 CTD kinase, Ctk1p [21,22]. Interestingly, ctk1D mutants are sensitive to these very same DNA damaging agents. Moreover, genes involved in DNA repair are amongst the ctk1 dependent genes that are transcriptionally altered by exposure to DNA damage. These observations suggest that CTD phosphorylation plays a role in the transcriptional control of genes involved in the DNA damage response.
In this report we provide further support for a model in which the CTD influences the selective modulation of gene expression. We clearly show that impairment of Ser-2 phosphorylation does not lead to gross changes in transcriptional profile. Both rpb1-12XCTD and rpb1-12XS2ACTD strains display a similar distribution of gene expression levels ( Figure 4A) as well as similar mean expression and percentile rankings (Table S1). In addition, scatterplots show a strong correspondence of expression for the vast bulk of genes ( Figure 4B). Indeed, only 152 out of 4875 S. pombe genes (and four out of 333 ''cytokinesis'' genes) were found to be differentially regulated between genotype upon LatA treatment.
While such a small number was somewhat unexpected, the identities of the cytokinesis related genes were striking. The first of the three up-regulated genes was lsk1 itself. While not informative with respect to the cytokinesis phenotype, the identification of Lsk1p is significant since it suggests the existence of positive feedback between low phosphoserine-2 levels and the control of lsk1 expression. That is to say, it is conceivable that low phosphoserine-2 levels in the rpb1-12XS2ACTD strain might be sensed by the cell and trigger increased lsk1 expression to compensate. The second of the up-regulated genes, nsk1, encodes a Clp1p interacting protein. Given that Clp1p is a critical regulator of the cytokinesis monitoring system in S. pombe, it is plausible that changes in nsk1 expression could be related to the LatA sensitivity of rpb1-12XS2ACTD strains. The third of the upregulated genes, cut12, encodes a spindle pole body protein with known roles in regulating the Plo1p kinase which in turn acts as a positive regulator of the SIN [30,31,32]. The only down-regulated gene in the set, aip1, encodes an actin interacting protein. Intriguingly, when human orthologues of this gene are knocked down in HeLa cells, both cytokinesis failure and an accumulation of multinucleate cells are observed [33].
While we favor a model in which the subset of differentially regulated genes represents the output of a transcriptional ''program'' -where multiple genes contribute to phenotypewe were able to provide evidence that one of the identified cytokinesis genes, aip1, does indeed play a role in promoting the successful completion of cytokinesis (Figure 7). While no severe defects were detected in otherwise wild-type backgrounds, aip1 gene deletion mutants displayed strong synthetic interactions with clp1D mutants. This was not surprising since weak cytokinesis mutants have been observed to display strong phenotypes when placed in backgrounds where the cytokinesis monitoring system is non-functional [9,10]. Moreover, whereas the LatA hypersensitivity of clp1D mutants was exacerbated by the aip1 deletion, the LatA sensitivity of lsk1D cells remained unaffected in aip1D backgrounds. This result suggests that Lsk1p and Aip1p act in a linear pathway and would thus be consistent with a model in which Lsk1p mediated changes in CTD phosphorylation selectively affects the transcription of genes affecting LatA hypersensitivity.
Lastly, a close inspection of the data also revealed the selective mis-regulation of genes involved in meiosis. Of the differentially regulated genes, 47 (approximately 31%) were found to be trascriptionally up-or down-regulated upon entry into the meiotic cell cycle ( Figure S3). Remarkably, this is consistent with recent data showing a requirement for Lsk1p in the transcriptional program initiated upon sexual differentiation [23]. These results bring up the question of whether the cytokinesis phenotype might be related to defects in the meiotic transcriptional program. In this respect it is interesting to note that the disruption of components of the MAPK signalling pathway (Ste20p, Ste11p, Ste7p) leading to the activation of the budding yeast Ste12p transcription factor (the functional equivalent to fission yeast Ste11p), as well as loss of function mutations in budding yeast ste12 itself, cause defects in the expression of genes involved in cell wall integrity during vegetative growth [40]. This result sets a precedent in establishing a role for meiotic specific regulators during the vegetative cell cycle, and moreover, brings to mind the possibility that ste11 dependent genes might play a morphogenetic role during cytokinesis. In any event, taking all results together, our data are consistent with a model in which alterations in CTD phosphorylation result, not in drastic global changes in transcription, but rather in selective changes in a limited set of genes resulting in discrete phenotypic changes. Regulatory modules that function through the fine-tuned modulation of CTD phosphorylation might thus represent a previously underappreciated mechanism of genetic control in eukaryotes.

Yeast Methods
All Schizosaccharomyces pombe strains used in this study (Table S2) were either obtained from the Karagiannis lab collection, created during the course of this study, or in the case of the aip1 gene deletion strain, purchased from Bioneer Corporation (Alameda, CA). Schizosaccharomyces pombe cells were cultured in either YES or Edinburgh Minimal Media (EMM) (Forsburg and Rhind 2006) with the appropriate supplements (Adenine, Histidine, Leucine, or Uracil). Liquid cultures were grown with shaking (200 rpm) at 30uC. Genetic crosses were performed using standard methods [41]. In experiments involving Latrunculin treatment, S. pombe cells were grown to mid log phase (O.D. 0.2) and treated with 0.2-0.5 mM of Latrunculin A (Enzo Life Sciences International, Plymouth Meeting, Pennsylvania) dissolved in DMSO. Cells were grown at 30uC with shaking at 200 rpm for 3-6 hrs, before being fixed. Plasmid vectors were transformed into S. pombe using the lithium acetate protocol according to Forsburg and Rhind [41].

Cloning Methods
To create the lsg1 gene deletion mutant, two separate fragments corresponding to sequences upstream and downstream of the lsg1 open reading frame (SPCC4B3.08) were PCR amplified from wild type genomic DNA using primers RS1, RS2, RS3 and RS4 (Table  S3). The 226 bp fragment generated using RS1 and RS2 primers was cloned upstream of the ura4 + gene (used as a selectable marker) in the pSKURA4 plasmid using KpnI and XhoI restriction sites. The 310 bp fragment generated using RS3 and RS4 primers was cloned downstream of the ura4 + using XbaI and SacI restriction sites. Digestion with KpnI and SacI liberated a linear dsDNA fragment composed of the 226 bp upstream fragment, the ura4+ selectable marker, and finally the 310 bp downstream fragment. This linear dsDNA fragment was then used to transform S. pombe strain MBY1343. Ura + transformants were isolated and subject to colony PCR using primers RS10, RS11, ME12 and ME28 to identify clones in which the linear dsDNA fragment had replaced the endogenous lsg1 gene via homologous recombination.
To create the Lsg1-GFP integrant strain, a C-terminal fragment of the lsg1 gene was PCR amplified from genomic DNA using primers RS5 and RS6 (Table S3). The fragment was then cloned upstream and in frame to the GFP gene located in the pJK210-GFP vector. This construct was then transformed into a ura4-D18 S. pombe strain. Ura + transformants were then isolated and subjected to colony PCR to identify clones in which the construct had integrated into the genome via homologous recombination.
To create the Lsg1-myc integrant strain, a C-terminal fragment of the lsg1 gene was PCR amplified from genomic DNA using primers RS8 and RS14 and cloned into the SmaI and KpnI restriction sites of the pJK210-myc plasmid. The plasmid construct was then transformed to S. pombe strain MBY1343. Ura + transformants were then isolated and subjected to colony PCR using primers RS9 and RS17 to identify clones in which the construct had integrated into the genome via homologous recombination. To create the Lsg1-HA integrant strain, a Cterminal fragment of the lsg1 gene was PCR amplified from genomic DNA using primers RS9 and RS16 and cloned into the SmaI and EcoRI restriction sites of the pJK210-HA plasmid. The plasmid construct was then transformed to S. pombe strain MBY1343. Ura + transformants were then isolated and subjected to colony PCR using primers RS9 and RS16 to identify clones in which the construct had integrated into the genome via homologous recombination.

Fluorescence Microscopy
S. pombe cells expressing Lsg1-GFP fusions, were fixed using ethanol fixation [41] and stored in PBS pH 7.4. To observe nuclei and cell wall/septa material, cells were mixed with 0.02 mg/mL 4'6,-diamidino-2-phenylindole (DAPI) and 1 mg/mL aniline blue. Fluorescent images were obtained using Zeiss Axioskop 2 microscope driven by ImageJ 1.41 software (National Institutes of Health) and Scion CFW Monochrome CCD Firewire Camera (Scion Corporation, Frederick Maryland) using DAPI and GFP filter sets. Gray level linescan measurements were performed using the ''plot profile'' function of ImageJ. Mean gray level measurements were performed using the ''threshold'' and ''analyze particles'' functions of ImageJ.

Biochemical and Immunological Methods
Cells of the indicated genotype were grown up to the mid-log phase at 30uC, collected by centrifugation, and resuspended in STOP buffer (10 mM Tris-HCl pH 8.0, 150 mM NaCl, 50 mM NaF, 10 mM EDTA, 1 mM NaN 3 ). Cell pellets were stored at -80uC up to a maximum of 6 months. Cell pellets were subsequently thawed, and lysed using vortexing with glass beads in extraction buffer (1% IGEPAL CA630 (tetr-Octylphenoxy polyethanol), 150 mM NaCl, 50 mM Tris-HCl pH 8.0, 2 mM EDTA, 1 mM PMSF (phenylmethanesulphonylfluoride), 2 mM Benzamidine, 50 mM NaF, 0.1 mM Na 3 VO 4 , 50 mM Bglycerophosphate, 15 mM p-nitrophenyl phosphate, J Tablet Sigma Protease Inhibitors). Twenty micrograms of total cell extracts were subjected to SDS-PAGE, transferred to PVDF membranes and immunoblotted using one of the following three antibodies: 8WG16 (Sigma), H14 (Sigma) or H5 (Sigma). 8WG16 and H14 were used at 1:5000 dilutions and H5 was used at a dilution of 1:1000. Peroxidase conjugated anti-mouse IgG (Sigma) at a dilution of 1:10000 was used against 8WG16 whereas peroxidase conjugated anti-mouse IgM (Sigma) at a dilution of 1:10000 was used to recognize H5 and H14 antibodies. In co-immunoprecipitation experiments 2 mg of cell extracts obtained as described above were immunoprecipitated following the manufacturer's protocol using Protein G Dynabeads (Invitrogen).

Global Gene Expression Analysis
Strains bearing the rpb1-12XCTD and rpb1-12XS2A mutations were grown to mid-log phase and treated with 0.5 mM LatA or DMSO for 3 hours at 30uC. Total RNA extraction was performed using the RiboPure Yeast Kit (Ambion) according to the manufacturer's instructions. Fifteen micrograms of total RNA was processed into labelled cRNA and hybridized to GeneChip Yeast Genome 2.0 Arrays (Affymetrix) by the London Regional Genomic Centre according to the Affymetrix Gene Profiling Reagents User Guide (P/N 702749 Rev. 1) available from the Affymetrix website (http://www.affymetrix.com/support/ technical/manuals.affx). Three biological replicates were performed. Global gene expression data was subsequently obtained from the London Regional Genomic Center as CEL computer files containing the raw hybridization signal intensities. Analysis was performed using the Genespring GX 10.0.1 (Agilent Technologies Inc). The CEL files were first normalized with RMA (robust multi-array analysis) algorithm. RMA uses PM (perfect match) probes from the data and corrects the background by fitting a model that is the addition of an exponentially distributed signal and a normally dispersed background [42]. The normalization procedure is helpful in eliminating non-biological sources of variation across individual samples acquired form, for example, differences in hybridization efficiency that varies in different genechips. This generated hybridization intensity data for all twelve samples that were used for further analysis. For quality control purposes Affymetrix includes probes for the bioB, bioC, bioD and cre genes on all genechips. During the hybridization protocol different concentrations (1.5, 5, 25, and 100 pmoles) of bioB, bioC, bioD and cre transcripts respectively, are ''spiked'' into the hybridization cocktail in order to assay for hybridization efficiency. As expected, when the data for these genes were inspected, it could be observed that the cre gene showed the highest strength of hybridization, followed by bioD, bioC, and bioB. This indicated that the hybridizations had been performed successfully, and that the data could be confidently used in further statistical analysis. The data were grouped under two different categories; genotype, representing rpb1-12XWTCTD or rpb1-12XS2ACTD, and drug, representing LatA or DMSO treatments. Based on this, all of the data could be classified as 1) rpb1-12XWTCTD, DMSO treated, 2) rpb1-12XWTCTD, LatA treated, 3) rpb1-12XS2ACTD, DMSO treated, and 4) rpb1-12XS2ACTD, LatA treated. To identify differentially regulated genes, data for these genes were inspected and data were filtered using Volcano plot analysis employing the Benjamin-Hochberg multiple testing correction. Genes showing statistically significant differences (p-value,0.05) and fold changes greater than 1.5 were identified as differentially regulated. Microarray experiments were performed in compliance with MIAME guidelines. Data has been deposited into the GEO database (Accession Number GSE31369).  Figure S3 A sub-set of genes differentially regulated in rpb1-12XS2ACTD strains are part of the transcriptional program initiated upon entry into meiosis. Using data [34] freely available from the Bahler website (http://www.bahlerlab. info/projects/sexualdifferentiation/meiosis/), the expression levels of the indicated genes were plotted versus time after meiotic induction. Data is presented as the ratio between expression level at the indicated times and expression level in vegetative cells. Genes were grouped according to the level and timing of induction/repression. (TIF)