Given the importance of the G2/M transition for cell cycle control, we have screened a near genome-wide fission yeast gene deletion collection 19 to search systematically for gene deletion mutants that divide prematurely, with the objectives of characterizing more comprehensively the components and mechanisms acting in a negative manner at the G2/M control.

We screened 82% of all fission yeast non-essential genes for mutants dividing prematurely at a small cell size, but with minimal effects on growth to avoid mutations influencing cell size indirectly 20. The screening procedure is summarized in Figure 1a and consisted of an initial microscopic visual screen followed by length and width measurements at cell division of candidate mutants (see Material and methods). Fission yeast cells grow by linear extension and therefore cell length correlates with cell volume, facilitating the identification of a relatively subtle size phenotype. We identified 18 mutants that divided at least 1 μ;m shorter than the wild-type strain, which, under the growth conditions used, divided at a length of 14.1 μ;m (Figure 1b; Table 1; Additional file 1; Figure S1 in Additional file 2). We confirmed that cell volume, calculated from length and width at division, was also reduced in the selected mutants (Table 1). The smallest mutant found was wee1Δ, which divided at 7.4 μ;m, around half the cell length of the control strain. The rest of the mutants divided with cell lengths of 75 to 93% of the control strain. In the course of our screen we also observed mutants with significant heterogeneity in cell size at division due to the presence of longer cells. Because these long cells could have arisen from a transient arrest of the cell cycle or delayed mitosis, they were not studied further.

All mutants grew with doubling times essentially similar to wild type (Table 1), except for the wee1Δ and gpa2Δ strains, with doubling times 66% and 40% longer than the wild-type strain. All mutants showed cell cycle phase distributions similar to the wild-type strain (Figure S2 in Additional file 2) except for the wee1Δ mutant, which had an extended G1 phase as previously noted 21. Deletions of five other genes (ypa2, SPAC227.01c, sft1, cpp1 and Lac1) showed cell sizes smaller than wild type but were not analyzed any further because of their sick and slow growing phenotype.

All 18 genes identified are conserved across eukaryotes and most can be grouped into four categories based on their biological functions: (1) regulation of the G2/M CDK activity and cytokinesis (wee1, clp1, pom1, nif1, ppa2, pyp1); (2) glucose sensing/cAMP signaling pathway (pka1, gpa2, git5, git3); (3) mRNA metabolism (zfs1, ski3, pab2); and (4) chromatin structure (snf5, sol1, sgf73). Other genes not found in these categories were SPAC27E2.03c and SPBC19F8.02, with unknown functions. Eleven of the genes identified have been previously reported to be involved in the G2/M control, validating our screen. We cannot give an estimate of the false negative rate of our screen, but it is informative that all gene deletions reported previously to significantly reduce cell size that were present in the set of mutants we screened were found in our study. Our list of mutants does not include several other loss-of-function mutations previously reported to divide at a small cell size. This was because these other mutant strains did not divide at a sufficiently small cell volume to attain the cutoff we used in our growth conditions (YE4S media, 32°C; Table S2 in Additional file 2). Interestingly, we found seven genes for which the small size phenotype has not been previously described; ski3, snf5, sol1, sgf73, pab2, SPBC19F8.02 and SPAC27E2.03c (Tables 1 and 2).

Comparison of our results with the list of budding yeast small size mutants identified in 2223 revealed only limited overlap confined to gpa2/GPA2 and wee1/SWE1 (Table S3 in Additional file 2). The SAGA complex involved in chromatin modification was also present in both lists but represented by different subunits (Sgf73 in fission yeast, Ada1 and Spt3 in budding yeast). The two budding yeast studies differ in the growth conditions used, as Jorgensen et al. 22 scored cell size of exponentially growing strains while Zhang et al. 23 determined cell size from cultures grown to saturation. We specifically examined the cell size phenotype of fission yeast mutants in ortholog genes of the budding yeast genes found in 2223. Thirty-seven genes were identified as fission yeast orthologs to the 45 budding yeast genes that result in small size when deleted, and 23 were contained in the set of mutant strains screened. Only four (GPA2/gpa2, SWE1/wee1, SCH9/sck2 and SLT2/pmk1) genes passed to the liquid screen and finally only GPA2/gpa2 and SWE1/wee1 showed a significant small cell size phenotype in both yeasts. Interestingly, none of the genes identified in our study are directly involved in ribosome biogenesis, which was the major pathway represented in the small size mutants found by Jorgensen et al. 22. This was not because of a low representation of 'ribosome biogenesis' annotated genes in our set of mutant strains, since approximately a third of all S. pombe genes annotated to this Gene Ontology category were present in this set (103 out of 325). The absence of genes involved in ribosome biogenesis from our list of small size mutants could be due to the different strategies used for coordinating cell division with growth in the two organisms, which in budding yeast occurs at G1/S while in fission yeast is usually at G2/M 2. It is possible that the G1/S control could be more sensitive to the ribosome biogenesis than the G2/M control. It is also possible that the small size phenotype of the budding yeast ribosome biogenesis gene mutants results as a response of the cell to the reduction in the growth rate in these mutants rather than to a direct involvement of these genes in cell mass-cell cycle coordination.

Most of the identified mutations had only modest effects on cell size, but we found that combining different mutations reduced cell length further. The quintuple mutant ski3Δ zfs1Δ ppa2Δ snf5Δ clp1Δ divided with a cell length of 7.2 μ;m, 50% smaller than the wild type (Figure 1c). The additive interaction between mutations regarding cell size suggests that these genes define different pathways regulating the G2/M transition. Furthermore, the heterozygous diploid strain ski3⁺/ski3Δ zfs1⁺/zfs1Δ ppa2⁺/ppa2Δ snf5⁺/snf5Δ clp1⁺/clp1Δ was 23% smaller than the control diploid strain (16.6 μ;m versus 21.4 μ;m), establishing that these genes have a quantitative effect on the G2/M transition. Additionally, it has been reported before that an increase in the levels of Wee1, Pka1, Ppa2, Pyp1, Clp1, Pom1 and Nif1 caused cell elongation, which is a sign of mitotic delay or arrest 5142425262728. We tested whether the overexpression of any of the remaining genes identified in our screen (SPAC27E2.03c, zfs1, ski3, SPBC19F8.02, snf5, sgf73, sol1 and pab2) also caused cell elongation, and found that overexpression of ski3 and snf5 significantly increased cell size (Figure 1d), establishing that they act as gene dosage-dependent regulators of the G2/M transition.

We next investigated if the genes identified encoded components of the upstream pathways that regulate the activation of the G2/M CDK. For example, Pom1 and Pyp1 are respectively components of the CGS and the SR pathways 101314. We examined genetic interactions with the regulators Sty1 and Cdr1, which act at the base of each respective pathway (Figure 2a; Table S4 in Additional file 2). The plot in Figure 2a graphically summarizes our results. The sgf73Δ gene deletion in both cdr1Δ and sty1Δ backgrounds (Table S4 in Additional file 2), or in a double mutant cdr1Δ sty1Δ (Figure S3 in Additional file 2), reduced growth rate dramatically and resulted in cells with cytokinesis defects, so this gene was excluded from this analysis. All the remaining double mutants showed cell lengths similar to or smaller than cdr1Δ and sty1Δ single mutants. Approximately half the mutations tested (8 out of 17) did not reduce cell length of the sty1Δ mutant, indicating that the factors encoded by these genes function upstream of Sty1. This group is made up of Pyp1, Pab2, SPAC27E2.03c, SPBC19F8.02 and factors involved in glucose sensing-signaling, Git3, Git5, Gpa2 and Pka1. A connection between the glucose sensing/cAMP signaling pathway and Sty1 has previously been noted 293031 and our work additionally establishes a key role for glucose sensing in the activation of the CDK. Conversely, all deletions reduced the size of the cdr1Δ strain except for pom1Δ as previously shown 1314, indicating that Pom1 is the only component of the CGS pathway in our set of mutants. Interestingly, we also show that Nif1, which physically interacts with and inhibits Cdr1 28, also appears to have a Cdr1-independent role in the G2/M transition.

The fact that a group of gene deletions (ppa2Δ, sol1Δ, snf5Δ, zfs1Δ, clp1Δ, ski3Δ, nif1Δ) reduced the cell size of both the sty1Δ and cdr1Δ strains indicated that these genes have roles in the G2/M control independently of these two pathways. To confirm the additive phenotype to both the sty1 and cdr1 gene deletions, we deleted these genes in a sty1Δ cdr1Δ strain. The double sty1Δ cdr1Δ mutant was viable and divided with a larger size than any of the parental mutants (Figure 2b; Table S5 in Additional file 2). Neither the ski3 nor nif1 deletion reduced cell length at division of the cdr1Δ sty1Δ mutant, suggesting that Ski3 and Nif1 function upstream of both Cdr1 and Sty1.

The ppa2, sol1, snf5, zfs1 and clp1 gene deletions reduced cell length at division of the sty1Δ cdr1Δ mutant (Figure 2c; Table S5 and Figure S4 in Additional file 2), confirming that their function in the G2/M is independent of both Sty1 and Cdr1. We investigated the genetic interactions within this group of genes (Figure 2d; Table S6 in Additional file 2) and found that, in all cases, mutants carrying pairs of deletions were smaller than the parental single mutant strains, with the one exception of the double mutant snf5Δ sol1Δ, which was similar to the snf5Δ alone. The additive genetic interactions within this group suggest that these genes function in different pathways. The non-additive snf5Δ sol1Δ result is consistent with the fact that Snf5 and Sol1 proteins are two subunits of the same complex 32. Deletion of snf22, which encodes the ATPase subunit of this complex (not present in the deletions screened), also showed an advanced mitosis phenotype similar to the snf5Δ and sol1Δ mutants (cell length at division 12.7 ± 0.6 μ;m), confirming a role of the SWI/SNF complex in the G2/M control.

This analysis has revealed new components in the G2/M control that function upstream of Sty1 (Pab2, SPAC27E2.03c and SPBC19F8.02), has shown that Ski3 and Nif1 function through both Cdr1 and Sty1, and has identified other elements that function in the G2/M transition independently of the CGS and SR pathways (Ppa2, Sol1, Snf5, Zfs1 and Clp1).

We next investigated how ppa2, sol1, snf5, zfs1 and clp1 act at the G2/M transition. It is known that Clp1 regulates Cdc25 stability and consequently CDK Tyr15 phosphorylation 3334. We tested if the other genes of this group also had a role in Tyr15 phosphorylation or in other aspects of CDK activation. We first analyzed if CDK protein levels were altered. It is known that co-overexpression of the mitotic cyclin Cdc13 and CDK Cdc2 advances cells into mitosis 35. However, the levels of Cdc13 and Cdc2 proteins determined both by western blot (Figure 3a) and by single cell fluorescence-activated cell sorting (FACS) analysis (Figure 3b) in the ppa2Δ, snf5Δ and zfs1Δ mutants, and in the double mutant snf5Δ zfs1Δ were similar to or lower than in the control strain. Therefore, the mitotic advancement observed in these mutants cannot be the result of an increase in CDK protein level. We also tested if the effects of these genes on the G2/M transition involve the CDK stoichiometric inhibitor Rum1, which inhibits the CDK during G1 3637. Mutants carrying the rum1 deletion and the zfs1, ppa2 or snf5 deletions were viable, and the lengths at division were similar to the corresponding single mutants (Figure 3c). Therefore, the effects of snf5, zfs1 and ppa2 on the G2/M transition do not act through Rum1.

Finally, we investigated if these genes alter the phosphorylation levels of Cdc2 at residue Tyr15. The levels of phosphorylated Cdc2 in ppa2Δ, snf5Δ, zfs1Δ and the double mutant snf5Δ zfs1Δ were similar to those in the wild-type strain (Figure 3a), suggesting a role in the G2/M transition independent of Tyr15 phosphorylation regulation. To further support this observation, we tested if the effect of these gene deletions was also observed in a background containing a non-phosphorylatable Cdc2 mutant protein. We used a strain expressing a mutant Thr14Ala Tyr15Phe Cdc2 kinase (Thr14 is occasionally phosphorylated 38) fused to the cyclin Cdc13, which is well tolerated by the cell 18 contrary to the non-fused mutant CDK 8. Cells with this Cdc13-L-Cdc2 fusion protein have a wild-type doubling time, cell length and cell cycle distribution. In agreement with the roles of the SR and CGS pathways regulating the G2/M transition through CDK Tyr15 phosphorylation, the non-phosphorylatable CDK fusion protein and not the wild-type fusion protein specifically abolished most of the effects on mitotic onset of sty1 and cdr1 gene deletions (Figure 3d; Table S7 and Figure S5 in Additional file 2), establishing that this system can be used for testing if Snf5, Sol1, Ppa2 and Zfs1 act on the G2/M control through CDK Tyr15 phosphorylation. We deleted these genes individually in strains with Cdc13-L-Cdc2 or Cdc13-L-(T14A Y15F)Cdc2 fusion proteins. Deletions reduced cell length at division of the strain carrying the Cdc13-L-Cdc2 fusion protein in a similar way to that observed in the wild-type background (Figure 3d, upper panel; Table S7 and Figure S6a in Additional file 2). The deletion of ppa2 in the Cdc13-L-(T14A Y15F)Cdc2 background rendered cells inviable, similar to the lethal phenotype of the double mutant wee1-50 ppa2Δ at restrictive temperature 25. We measured cell length at division of the remaining viable strains and found that cells harboring these deletions were shorter than the control strain, although the CDK could not be phosphorylated on Tyr15 (Figure 3d, bottom panel; Table S7 and Figure S6b in Additional file 2). The snf5Δ and sol1Δ deletions were not additive in the Cdc13-L-(T14A Y15F)Cdc2 background, while snf5Δ and zfs1Δ were additive, reducing cell length by 23%. These results show that the premature mitosis of snf5Δ, sol1Δ and zfs1Δ mutants is independent of Tyr15 phosphorylation and establishes that there must be additional regulatory mechanisms acting at the G2/M transition.

This systematic screen of more than 80% of fission yeast non-essential genes has identified a significant proportion of the genes acting negatively at the G2/M transition. The 18 genes identified are listed in Table 2 together with their connection to the G2/M control. We found that most of these genes function through CDK Tyr15 phosphorylation. Eight of these genes function upstream of sty1, and of these, three, pab2, SPAC27E2.03c and SPBC19F8.02, are described here for the first time as negative regulators of mitotic onset and define new components of the SR pathway. Only one gene, pom1, acts solely in the CGS pathway. However, our data indicate that ski3 and nif1 function in both the SR and CGS pathways, suggesting a cross-talk between these two pathways previously thought to act independently.

We found that snf5, sol1, zfs1, ppa2 and clp1 function independently of both sty1 and cdr1, and that snf5, sol1 and zfs1 act on mitotic onset independently of CDK Tyr15 phosphorylation. The advanced mitotic phenotype of their deletions, described for first time for snf5 and sol1, was not due to changes in CDK protein level or Rum1 deregulation, indicating that they represent components of uncharacterized rate-limiting controls acting at the G2/M transition. We suggest that the lethality of ppa2Δ when combined with the Tyr15 mutant CDK could be due to a role in the G2/M transition also independent of Tyr15 phosphorylation. These proteins could be involved in regulating the dephosphorylation of CDK substrates given that, in Xenopus laevis eggs, PP2A phosphatase controls cell cycle progression by counteracting the CDK-dependent phosphorylation of mitotic substrates 39, and in S. cerevisiae, Cdc14 (homolog of Clp1) dephosphorylates CDK substrates 40. Alternatively, these proteins could function in some other way on CDK targets or could inhibit the CDK by unknown regulatory pathways. However, they would not be expected to have a role in the assembly of the CDK complexes given they still exert effects in the presence of the fusion protein Cdc13-L-Cdc2 (Figure 3d; Table S7 and Figure S6b in Additional file 2). Alternatively, these proteins might be involved in the cellular localization of the complex and their absence could facilitate access of the CDK to its substrates. Another possibility is that these proteins are involved in an as yet uncharacterized posttranslational modification of the CDK. The elucidation of the molecular details of the mechanism of action of these proteins on the G2/M transition will require further study.

S. pombe media and methods are described in Moreno et al. 41. Strains used are listed in Table S8 in Additional file 2. Experiments were carried out in yeast extract complex media supplemented with 0.15 mg/ml L-Histidine, L-Leucine, Adenine and Uridine (YE4S) at 32°C.

The viable set of a near genome-wide S. pombe haploid deletion collection 19 was screened for mutants that divide prematurely with a smaller size than the wild-type strain. In total, we screened the cell size phenotype of 2,969 different gene deletions (82% of fission yeast non-essential genes). Our strategy consisted of an initial microscopic visual screen of mutants growing on agar plates of complex media (YE4S) followed by measurements of cell size at division of candidate mutants growing exponentially in liquid culture. Mutant strains were arrayed in 96-well plates and 150 μ;l of YE4S per well was inoculated and incubated at 32°C for >16 h with shaking. We then inoculated a solid YE4S media plate using a pin tool and incubated this plate for 12 to 20 h until small colonies (40 to 60 cells) formed. Each mutant strain was inoculated in quadruplicate and was compared with the control strain PN558 (ade6-M210 leu1-32 ura4-D18 h+) growing in the same plate . The visual screen for cell size phenotypes was carried out using a Zeiss Axioskop 40 microscope equipped with a 20×/0.4 NA objective and an additional 1.8× magnification. From this first screen, we selected 239 different mutant strains for a second screen in liquid culture, in which growth conditions were better controlled. Candidate mutants were grown in individual flasks containing 15 ml of YE4S media at 32°C, and cell size was screened when the culture was growing in exponential phase (0.35 to 0.45 OD₅₉₅). Cell wall and septum were stained with Blankophor (1:100,000; MP Biochemicals, Solon, OH, USA) and cells were observed with a Zeiss Axioskop microscope, equipped with a QICam Fast camera and using a 63×/1.4 NA objective. We selected 68 mutant strains selected at this stage, which were grown again under the same conditions, and the width and length of dividing cells measured from pictures. We selected mutants that divided at least 1 μ;m shorter in cell length than the control strain. Wide mutants that showed reduced cell length but wild-type volume were discarded as these mutants are altered in morphology rather than in the control of the cell cycle. Finally, mutants were backcrossed with the wild-type strain to confirm the co-segregation of the phenotype with the deletion marker and to remove auxotrophies, and the specificity of the gene deletion was verified by PCR. Genes in our set of small size mutants were deleted for more than 91% of the ORF 19. Cell size phenotypes of the new small size mutants identified in our screen have been annotated in PomBase 42.

Cell length and width were measured from pictures of live Blankophor-stained cells using the PointPicker plug-in of ImageJ (National Institutes of Health). Cell volume was estimated from the length and width values by considering the shape of a fission yeast cell as a cylinder topped with two hemispheres. Average cell length and width values were determined from 60 to 150 cells, and comparison between strains was done using a two-tailed unpaired Student's t-test. In box-and-whisker plots, boxes are delimited by the first quartile, median and third quartile, and whiskers mark maximum and minimum values within a range of up to 1.5 standard deviations. Values outside this range are displayed as individual dots.

Gene deletion was performed as in 43. Gene targeting oligos were designed using the tool 'Gene deletion' from the Bähler Lab web site 44. Overexpression of genes was done using the multicopy plasmid REP3X under the thiamine-repressible nmt1 gene promoter 45. ORFs were amplified by PCR with specific oligos carrying restriction enzyme sites for cloning into REP3X. In overexpression experiments, cells were grown in synthetic media (EMM) containing 5 μ;g/ml thiamine and expression of the gene was induced upon thiamine removal by filtration and incubation in thiamine-free medium for 20 h.

Total protein extracts were prepared from 10⁸ cells, collected by centrifugation and resuspended in the same volume of HB buffer (60 mM β-glycerol phosphate, 15 mM MgCl₂, 15 mM EGTA, 1% (v/v) Triton X-100, 150 mM NaCl, 25 mM MOPS-NaOH pH7.2), containing protease inhibitor (1 mM phenylmethylsulfonyl fluoride (PMSF and Roche Complete Mini protease inhibitor cocktail) and phosphatase inhibitors (Roche PhosStop, Roche, Indianapolis, IN, USA). Cell suspensions were boiled for 5 minutes, and then transferred to a tube containing 1.2 ml of glass beads (0.4 mm; Sigma, St Louis, MO, USA). Cells were disrupted in a FastPrep cell disruptor (ThermoSavant, Waltham, MA, USA) for 3 × 20 s. HB buffer plus inhibitors (50 μ;l) was added and the crude extract was recovered and mixed with 5× sample buffer (20% (v/v) β-mercaptoethanol, 20% (w/v) SDS, 0.05% (w/v) bromophenol blue, 25% (v/v) glycerol, 300 mM Tris-HCl pH6.8). Finally, extracts were boiled for 5 minutes and centrifuged at 13,000 rpm for 1 minute. In western blots, Cdc13 was probed with rabbit polyclonal SP4 antibody 9 (1:3,000); Cdc2 with commercial rabbit polyclonal anti-PSTAIRE (1:250; Santa Cruz Biotech, Santa Cruz, CA, USA); phoshorylated Tyr15 Cdc2 with commercial rabbit polyclonal (1:500; Cell Signaling Technology, Danvers, MA, USA); and Atb2 with monoclonal TAT1 antibody (1:5,000; gift from K Gull 46). Horse radish peroxidase-conjugated goat anti-mouse or goat anti-rabbit IgG (Pierce-Thermo, Waltham, MA, USA) were used at a dilution of 1:10,000 as secondary antibodies.

DNA content per cell was determined from 10⁴ cells that were fixed with 70% (v/v) ethanol and then washed with 1 ml 50 mM sodium citrate. Cells were resupended in 0.5 ml 50 mM sodium citrate containing 0.1 mg/ml RNase A and incubated at 37°C overnight. DNA was stained with 2 μ;g/ml propidium iodide and samples were sonicated before analysis in a BD FACSCalibur instrument. Single cell analysis of CDK protein levels was performed from strains expressing yellow fluorescent protein (YFP)-tagged Cdc13 or Cdc2 proteins under their native promoters. Cells were grown in YE4S at 32°C and 1 ml of culture at 0.2 OD₅₉₅ (approximately 4 × 10⁶ cells) was fixed with 1% (w/v) formaldehyde for 15 minutes; then cells were washed and resuspended in 1 ml phosphate-buffered saline. Cells were briefly sonicated prior to measuring fluorescence signal in a FACSCalibur instrument (BD Biosciences, San Jose, CA, USA) equipped with a 488 nm excitation laser and a 530 nm bandpass filter. Autofluorescence from a non-YFP tagged strain was subtracted from the YFP fluorescent signal.