G1-Cyclin2 (Cln2) promotes chromosome hypercondensation in eco1/ctf7 rad61 null cells during hyperthermic stress in Saccharomyces cerevisiae

Abstract Eco1/Ctf7 is a highly conserved acetyltransferase that activates cohesin complexes and is critical for sister chromatid cohesion, chromosome condensation, DNA damage repair, nucleolar integrity, and gene transcription. Mutations in the human homolog of ECO1 (ESCO2/EFO2), or in genes that encode cohesin subunits, result in severe developmental abnormalities and intellectual disabilities referred to as Roberts syndrome and Cornelia de Lange syndrome, respectively. In yeast, deletion of ECO1 results in cell inviability. Codeletion of RAD61 (WAPL in humans), however, produces viable yeast cells. These eco1 rad61 double mutants, however, exhibit a severe temperature-sensitive growth defect, suggesting that Eco1 or cohesins respond to hyperthermic stress through a mechanism that occurs independent of Rad61. Here, we report that deletion of the G1 cyclin CLN2 rescues the temperature-sensitive lethality otherwise exhibited by eco1 rad61 mutant cells, such that the triple mutant cells exhibit robust growth over a broad range of temperatures. While Cln1, Cln2, and Cln3 are functionally redundant G1 cyclins, neither CLN1 nor CLN3 deletions rescue the temperature-sensitive growth defects otherwise exhibited by eco1 rad61 double mutants. We further provide evidence that CLN2 deletion rescues hyperthermic growth defects independent of START and impacts the state of chromosome condensation. These findings reveal novel roles for Cln2 that are unique among the G1 cyclin family and appear critical for cohesin regulation during hyperthermic stress.

Eco1/ESCO2 family members are essential for cell viability in all model systems tested to date (Skibbens et al. 1999;Tó th et al. 1999;Tanaka et al. 2000;Williams et al. 2003;Vega et al. 2005;Seitan et al. 2006;Kawauchi et al. 2009;Mö nnich et al. 2011;Whelan et al. 2012), with the exception of a clade of parasitic microbes that also lack cohesin-loading complexes (Scc2/NIPBL and Scc4/MAU-2) (Gentekaki et al. 2017) but which are otherwise similarly essential (Rollins et al. 1999;Ciosk et al. 2000;Krantz et al. 2004;Tonkin et al. 2004). One critical role of Eco1 is to establish cohesion between sister chromatids. To accomplish this, Eco1 is recruited to the DNA replication fork by PCNA and other fork associated components (Skibbens et al. 1999, Moldovan et al. 2006, Bender et al. 2020Zhang, Shi, et al. 2017;Zhang, Yeung, et al. 2017). At the fork, Eco1 acetylates the Smc3 subunit of cohesins that are newly deposited onto each sister chromatid. Acetylation blocks cohesin dissociation from DNA (and possibly stabilizes cohesin dimerization or clustering), thereby establishing sister chromatid cohesion. Prior to acetylation, cohesins dissociate from DNA in a process that involves Rad61 (WAPL in humans), which binds the cohesin subunit Pds5 (Gandhi et al. 2006;Kueng et al. 2006;Rowland et al. 2009;Sutani et al. 2009). Cells devoid of Rad61 exhibit robust growth at all temperatures, similar to wildtype, even though they retain elevated levels of chromatin-bound cohesins. Intriguingly, cells that harbor deletions of both ECO1 and RAD61 survive (Rolef Ben-Shahar et al. Sutani et al. 2009), but only within a narrow temperature range (Maradeo and Skibbens 2010;Guacci and Koshland 2012). The temperature sensitive lethality of eco1 rad61 null cells suggest that Eco1 or cohesins perform an activity during hyperthermic stress that is independent of Rad61.

2008;
Here, we report results from an unbiased genome-wide screen that identified the G1 cyclin Cln2 as an essential regulator of this Eco1/cohesin-dependent hyperthermic stress response. Our results reveal that deletion of CLN2 rescues the temperaturedependent growth defects exhibited by eco1 rad61 null cells. While Cln1, Cln2, and Cln3 are functionally redundant in promoting the transition from G1 to S phase (START) (Lew et al. 1991(Lew et al. , 1992Lew and Reed 1993;Levine et al. 1996;Queralt and Igual 2004;Talarek et al. 2017), we find that Cln2 is unique among the G1 cyclins in rescuing eco1 rad61 double mutant cell growth defects. We further provide evidence that the suppression provided by CLN2 deletion occurs independent of START and may act through alterations in chromatin structure.

Materials and methods
Yeast strains, media, and growth conditions All strains used in this study were derived from the S288C background and grown in YPD media unless otherwise noted (Supplementary Table 1). Diploid strains were sporulated in 0.3% potassium acetate and tetrads dissected on YPD agar. The genotypes of the resultant spores were analyzed for each wild type, single and double mutant spore recovered. Phenotypic analyses of isolated spores were performed by generating 10-fold serial dilutions of log phase cultures normalized based on OD 600 as a measure of cell density. Each dilution series was plated on YPD agar plates, or selective medium plates, and grown at a range of temperatures as indicated in each figure. Non-ts revertant strains were tested for conversion to a ts-state after transformation with plasmid harboring either PDS5 (pVG177) or SCC3/IRR1 (pH9.3), kindly provided by Dr Vincent Guacci (Stead et al. 2003), compared to vector alone. pVG177 and pH9 were independently verified by restriction digest. Primers used to delete genes and verify gene deletions are contained in Supplementary Table 2. Synchronization of yeast cultures and assessment of DNA contents by flow cytometry were performed as described previously (Maradeo and Skibbens 2010;Tong and Skibbens 2015).
Site-directed PCR mutagenesis was used to replace the entire CLN2 open reading frame, or a portion of the CLN2 reading frame (retaining sequences that encode for the first 221 residues), within the parental eco1D rad61D YMM828 strain (Maradeo and Skibbens 2010) with a the kanMX selectable marker (Longtine et al. 1998). Appropriate knockouts were confirmed by PCR (primer sequences are contained in Supplementary Table 2).

Genomic sequencing
Genomic DNA was isolated from pools containing 6 individual segregants. Briefly, 6 segregants for each revertant were grown in YPD broth and cell densities normalized (based on OD 600 ) prior to pooling. For each pooled sample, cell lysis was performed by physical agitation (Mini-Beadbeater, BioSpec) in the presence of 0.5 mm glass beads (BioSpec) and lysis buffer (1% SDS, 100 mM NaCl, 10 mM Tris pH8.0; 2% Triton X-100) and phenol: chloroform: isoamyl (25:24:1). Lysate was centrifuged (15,000 rpm, 5 min at 4 C), the aqueous supernatant incubated in RNAse (Sigma) and PCI re-extracted, genomic DNA obtained by centrifugation using Phase Lock Gel Light columns (Quanta), prior to precipitation by addition of 100% EtOH. DNA quality and concentration were assessed, following manufacturer's instructions, using the Quant-it BR Assay kit for dsDNA (Invitrogen), prior to sending to the Microbial Genome Sequencing Center (Pittsburgh, PA, USA) for sequencing on the Illumina NextSeq 500 instrument. FASTQ files were aligned to the S288C reference genome (Engel et al. 2013) using the Burrows-Wheeler Aligner with Maximal Exact Matches (BWA-MEM) algorithm version 0.7.15 with default settings (Li and Durbin 2010). Variants were called using FreeBayes version 1.1.0 with default settings and the "-pooled-continuous" argument (Garrison and Marth 2012). Calls were annotated using SnpEff version 4.3T with default settings (Cingolani et al. 2012).

Microscopy and condensation assay
Chromatin condensation assays were performed as described previously Skibbens 2017, 2020). DNA masses and rDNA structures were detected by DAPI staining. Chromatin images were captured using a Nikon Eclipse E800 microscope equipped with a cooled CD camera (Coolsnapfx, Photometrics) and IPLab software (Scanolytics). rDNA loops were often indistinguishable from chromatin masses in eco1 rad61 double mutant cells. Attempts to include rDNA structural analyses resulted in the omission of a significant, if not majority, number of eco1 rad61 double mutant cells. To accommodate the severity of this defect, we instead assessed genome-wide condensation. A region of interest (ROI) was determined that matched the majority of wildtype genomic mass areas. Specifically, we experientially determined the size of a circular ROI or mask that was filled by wildtype DNA masses or in which the majority (approximately 3/ 4) of the edges coincided with wildtype DNA masses. The resulting mask (62 pixel circle) was then superimposed over each DNA mass to individually assess whether the DNA mass (excluding rDNA loop extensions) filled or failed to fill the ROI mask or contact the majority of the mask edges. This strategy enabled the inclusion of nearly 100% of all DNA masses (overlapping masses, in which genomic boundaries were obscured, were omitted from quantification).

Flow cytometry and cell cycle progression
Cell synchronizations and cycle progression were monitored using propidium iodide and flow cytometry (BD FACScan) as described previously (Maradeo and Skibbens 2010;Tong and Skibbens 2015). Briefly, log phase cultures maintained at 30 C were normalized to an optical density typically between 0.1 and 0.4. For chromosome condensation assays, log phase cells were placed in fresh YPD supplemented with nocodazole (5 mg/ml final concentration) and maintained at 37 C for 3 h prior to fixation. For cell cycle progression experiments, log phase cells were first synchronized in early S phase by maintaining growth at 30 C in fresh YPD supplemented with hydroxyurea (HU) (0.2 M final concentration) before washing cells and resuspending them in fresh YPD supplemented with nocodazole (5 mg/ml final concentration). Cells were then shifted to 37 C and samples collected every 30 min. In either case, cells were spun and resuspended in fixing solution (0.2M Tris 70% EtOH solution). Cells were then treated with RNase (Roche) and proteinase K (Roche) solutions to remove RNA and protein, respectively. To analyze DNA content, cells were stained with a 0.0001% propidium iodide (Sigma) solution. Prior to use, the stock solution was diluted 10 ml þ 990 ml in 0.2M Tris solution for each milliliter of sample. Cells were sonicated and DNA content quantified by flow cytometry using a BD FACSCanto II.

Statistical analyses
A chi-square test of independence was used to assess the relationship of gene mutations (eco1D rad61D double mutant cells and eco1D rad61D cln2D triple mutant cells) on chromatin condensation levels, standardized using wildtype cells. For each strain, the number of observed genomic masses that matched or exceeded the template were compared to expected (all genomic masses in the field of view) number of genomic masses. A statistical significance P-value at or below 0.05 was used to demonstrate gene mutation dependence. We also compared chromatin condensation effects across the 3 strains. Here, the percentage of genomic masses, within a given field, that matched or exceeded the template were used for subsequent analysis, performed using a 2-tailed Student's t-tests to assess statistical significance using a P-value at or below 0.05 between each group.

A spontaneous suppressor screen identifies potential regulators of Eco1 function
Budding yeast Eco1/Ctf7 (Eso1 in S. pombe, ESCO1,2 in vertebrates) is the founding member of a highly conserved family of acetyltransferases that are essential regulators of cohesin function (Skibbens et al. 1999;Tanaka et al. 2000;Ivanov et al. 2002;Rolef Ben-Shahar et al. 2008;Unal et al. 2008;Zhang, Shi, et al. 2008). While mutations in a small number of cohesin pathway genes (PDS5, SCC3/IRR1, and RAD61) were initially reported to bypass Eco1 function, the resulting double mutant cells (for instance, eco1D rad61D double mutated cells) are viable within only a narrow temperature range (Rowland et al. 2009;Sutani et al. 2009;Maradeo and Skibbens 2010;Guacci and Koshland 2012). These findings suggest that Eco1 and cohesin perform essential functions in response to hyperthermic stress that are independent of Rad61. We focused on the suppression provided by RAD61 deletion since Rad61 is not a core cohesin component and RAD61 loss is sufficient to impact chromatin structure in the absence of other cohesin gene mutation (Gandhi et al. 2006;Kueng et al. 2006).
Prior to performing a genome-wide suppressor screen, we first independently documented the hyperthermic sensitivity of eco1D rad61D double mutant cells. Similar to wildtype cells, rad61D single mutant cells exhibited robust growth at all temperatures tested (Fig. 1a). In contrast, eco1 ctf7-203 mutant cells exhibited severe growth defects at 30 C and were inviable at 37 C ( Fig. 1a) (Skibbens et al. 1999). eco1D rad61D double mutant cells exhibited robust growth at 30 C but were inviable at 37 C ( Fig. 1a), in support of earlier findings that eco1D rad61D cells exhibit severe hyperthermic sensitivity (Sutani et al. 2009;Maradeo and Skibbens 2010;Guacci and Koshland 2012).
To identify novel suppressors of eco1D rad61D cell hyperthermic growth defects, 2 independently derived eco1D rad61D double mutant cells, YMM828 and YMM829 (Maradeo and Skibbens 2010), were grown to log phase before plating 5 Â 10 7 cells onto YPD medium. After 4 days growth at 37 C, the plates were replica plated onto YPD agar and incubated at 37 C for an additional 2 days. From each plate, we isolated 10 moderate-to-large-sized colonies, each containing spontaneous mutations that support growth under hyperthermic stress. From here on, these 20 strains will be referred to as revertants since each spontaneous mutation reverts temperature-sensitive (ts) growth to a non-ts growth status. Since our goal was to identify novel suppressors of eco1 mutant cell defects, each of the 20 triple mutant revertants (eco1D rad61D and the reverting mutated gene) was transformed with plasmid alone or plasmid harboring either PDS5 or SCC3/IRR1. Resulting PDS5 or SCC3/IRR1 transformants that exhibited cell inviability at 37 C were excluded from further analyses.
Spontaneous gene mutations that rescue eco1D rad61D mutant cell ts growth defects likely arise concordantly with other gene mutations that are unrelated to the revertant phenotype. To reduce the number of unrelated mutations, the remaining revertants were back-crossed 2 times to wildtype cells. In each case, the reverting mutation segregated according to Mendelian genetics (data not shown), suggesting that a single gene mutation was responsible for the revertant phenotype. To aid in the identification of the suppressor mutations, multiple segregants, derived from the same revertant, were pooled together based on the logic that only the suppressor mutation would be present in all segregants (present at a frequency of 1.0 in the pool). In contrast, nonrevertant mutations were expected to segregate randomly and thus be present at intermediate or low frequencies. After pooling the segregants for each revertant (revertant A and B), genomic DNA was harvested and whole genome sequencing performed. Each of the 2 segregant pools (Pool A and B) contained 2 mutations at a frequency of 1.0 (Table 1). Of the 2 mutations (CLN2 and SMC2 for Pool A; CLN2 and RPL31A for Pool B), only the CLN2 mutation (frameshift at threonine 222), was present at a frequency of 1.0 in both pools ( Table 1), suggesting that this allele might be the suppressor mutation and arose independently in these revertant strains.
CLN2 deletion suppresses eco1D rad61D cell growth defects Despite the high frequency of the CLN2 mutation in all segregants derived from the 2 revertants, it became important to independently confirm that the CLN2 mutation is responsible for the revertant phenotype. In the first of 2 strategies, the complete coding sequence of CLN2 was deleted from our parental eco1D rad61D strain. In the second strategy, we constructed a partial deletion of CLN2 (in which only the first 221 amino acids of Cln2 are retained in the genome), to replicate the allele obtained in our spontaneous suppressor screen (Table 1). In both cases, site-directed PCR mutagenesis of CLN2 was confirmed by PCR in the resulting transformants (Longtine et al. 1998). Log phase cultures of the parental eco1D rad61D double mutant strain, and 2 independent transformants each of eco1D rad61D cln2D and eco1D rad61D cln2 1 À 221 triple mutant strains, were serially diluted, plated onto YPD agar, and incubated at either 30 C or 37 C. As expected, all strains exhibited robust growth at 30 C while eco1D rad61D double mutant cells were inviable at 37 C. Both isolates of the eco1D rad61D cln2D triple mutant cells, as well as the eco1D rad61D cln2 1 À 221 triple mutant cells, exhibited substantial growth at 37 C (Fig. 1b). These results confirm that CLN2 loss-of-function mutations rescue the ts growth defect otherwise present in eco1D rad61D double mutant cells.
Cln2 is unique among G1 cyclins in suppressing eco1D rad61D cell growth defects Cln1, Cln2, and Cln3 are functionally redundant such that any single G1 cyclin, bound to cyclin-dependent kinase (CDK), can promote START. Cln3-CDK, however, is the earliest inactivator of Whi5, a key inhibitor of START, that acts in both larger mother cells and smaller newly budded cells (reviewed Fisher 2016;Ewald 2018;Li et al. 2021). We thus tested whether deletion of CLN3, using site-directed PCR, would suppress eco1D rad61D double mutant cell ts growth defects, similar to deletion of CLN2. As expected, eco1D rad61D double mutant cells and eco1D rad61D Fig. 1. Identification of the G1 cyclin, Cln2, as a regulator of Eco1-Rad61 pathways. a) eco1D rad61D double mutant cells are viable within a narrow temperature range. Ten-fold serial dilutions of wildtype, rad61D and eco1 ctf7-203 single mutant cells, and eco1D rad61D double mutant cells at the indicated temperatures. b) CLN2 deletion, and truncated Cln2 1-221 (CLN2 1-221D ), both suppress eco1D rad61D double mutant cells growth defects. Ten-fold serial dilutions of wildtype cells, 2 independent isolates of eco1D rad61D cln2D triple mutant cells, and 2 independent isolates of eco1D rad61D cln2 1-221 cells grown at the indicated temperatures.
cln3D triple mutant cells (3 isolates shown) exhibited robust growth at 30 C. Surprisingly, and unlike eco1D rad61D cln2D cells, all isolates of eco1D rad61D cln3D triple mutant cells were inviable at 37 C and provided no growth advantage at 23 C (Fig. 2a). Thus, CLN3 deletion does not phenocopy the revertant growth defect provided by CLN2 deletion during hyperthermic stress, revealing a novel distinction between these G1 cyclins.
Cln1 and Cln2 are roughly 60% identical at the amino acid level and bind identical target-docking sites. Cln1-CDK and Cln2-CDK are also functionally redundant in regulating Whi5  inactivation in smaller bud cells that arise due to the asymmetric division in budding yeast. Given the highly conserved roles of Cln1 and Cln2, we predicted that the deletion of CLN1 would suppress eco1D rad61D double mutant cell ts growth defects, similar to the deletion of CLN2. We used site-directed PCR to generate eco1D rad61D cln1D triple mutant cells and assessed these for growth at 23 C, 30 C, and 37 C. Under all conditions, growth of eco1D rad61D cln1D triple mutant matched that of eco1D rad61D double mutant: each strain grew at 30 C but were inviable at 37 (Fig. 2b). In combination, these results reveal that Cln2-CDK performs an activity that is distinct from both Cln1-CDK and Cln3-CDK.

Cln2 antagonizes Eco1 functions independent of Rad61
RAD61 deletion does not fully bypass the essential function of Eco1, such that eco1 rad61 cells are inviable at 37 C ( Fig. 1) (Maradeo and Skibbens 2010;Guacci and Koshland 2012). Given that CLN2 deletion fully rescues eco1D rad61D cell growth under hyperthermic conditions, it became important to test whether deletion of CLN2 would fully bypass Eco1 function or only partially suppress Eco1 function, similar to deletion of RAD61. To differentiate between these 2 possibilities, CLN2 was deleted from eco1 ctf7-203 ts cells (Skibbens et al. 1999). Log phase cultures of wildtype cells, eco1 ctf7-203 single mutant cells, and eco1 ctf7-203 cln2D double mutant strains were serially diluted, plated onto YPD plates, and maintained at either 23 C, 30 C, or 37 C. As expected, all strains exhibited robust growth at 23 C (Fig. 3). Consistent with prior studies, eco1 ctf7-203 single mutant cells exhibited severe ts growth defects at 30 C and were inviable at 37 C. eco1 ctf7-203 cln2D double mutant strains (3 isolates shown) exhibited significant improvement in growth at 30 C, compared to eco1 ctf7-203 single mutant cells but were inviable at 37 C (Fig. 3). Our finding that the individual deletion of either RAD61 or CLN2 deletion fails to rescue eco1 mutant cell ts growth (Figs. 1 and 2), but that the combined deletion of RAD61 and CLN2 provides for robust growth at elevated temperatures, suggests that each exhibits a distinct activity on either cohesins or Eco1-dependent regulation of cohesins (see Discussion).
CLN2 deletion-dependent suppression of eco1D rad61D cell growth defects appears distinct from coupling cohesion establishment to DNA replication The partial suppression of eco1D cell ts growth defects by mutation of either CLN2, PDS5, or SCC3/IRR1 (Rolef Ben-Shahar et al. 2008;Sutani et al. 2009;Rowland et al. 2009;Maradeo and Skibbens 2010; current study) likely arises due to a coordinated reduction of opposing activities that rebalance some inherent cohesin activity. In contrast, overexpression of wildtype genes that suppress eco1 mutant cell growth defects are more likely gene products that augment Eco1 function. For instance, elevated levels of PCNA (POL30) suppresses eco1 ctf7-203 ts growth defects (Skibbens et al. 1999), a finding that revealed that the tethering together of sister chromatids is intimately linked to DNA replication. It thus became important to ascertain whether eco1D rad61D cell growth defects might be suppressed by PCNA overexpression, possibly linking the effects of the CLN2 deletion to a DNA replication-based mechanism. To test this possibility, we overexpressed PCNA from a high-copy plasmid that harbors POL30 (Skibbens et al. 1999). eco1D rad61D double mutant cells, harboring vector alone or vector driving elevated PCNA levels, exhibited strong growth at 30 C but were inviable when maintained at 37 C (Fig. 4). Thus, the suppression that results from CLN2 deletion is likely independent of the beneficial effect of coupling cohesion establishment to DNA replication.

cln2D-dependent suppression of eco1D rad61D cell growth defects occurs independent from a delay in START
Given the role of Cln2-CDK in promoting the transition from G1 to S phase, we wondered whether the cln2D-dependent suppression of eco1D rad61D cell growth defects might be due to slowed cell cycle progression. We tested this possibility in the following 2 ways. First, log phase cultures of wildtype cells and 2 independent isolates each of eco1D rad61D double mutant cells and eco1D rad61D cln2D triple mutant cells were grown at 30 C (permissive for all 5 strains) prior to shifting to 37 C. The DNA content in each condition was quantified by flow cytometry (Fig. 5). As expected, wildtype cells exhibited log phase growth characteristics (indicated by roughly equivalent G1 and G2/M DNA peaks), such that pre-and postreplication peaks of DNA were unaltered even after prolonged growth at 37 C (Fig. 5). Independent isolates of eco1D rad61D double mutant cells exhibited a G2/M bias in their DNA profiles both at 30 C and 37 C. This cell cycle delay is consistent with the cohesion and condensation defects that persist in these cells and that likely trigger a mitotic checkpoint (Skibbens et al. 1999;Guacci and Koshland 2012). In contrast, the DNA profiles of eco1D rad61D cln2D triple mutant cells appeared unaffected by growth at 37 C, exhibiting similar G1 and G2/M peaks as wildtype cells (Fig. 5). Thus, loss of CLN2 does not result Fig. 3. Deletion of CLN2 suppresses eco1 mutant cell ts growth, similar to but distinct from deletion of RAD61. Ten-fold serial dilutions of wildtype cells, eco1 ctf7-203 mutant cells, and 3 independent isolates of eco1 ctf7-203 cln2D double mutant cells grown at the indicated temperatures.
in the accumulation of cells in a G1/S transition state but instead appears to promote cell cycle progression. We next mapped cell cycle progression of wildtype cells, eco1D rad61D double mutant cells, and eco1D rad61D cln2D triple mutant cells at greater temporal resolution. Log phase cultures of each were arrested in preanaphase at 30 C, a temperature permissive for all of the strains, by the addition of nocodazole into the medium. After 2 h, cells were washed free of nocodazole, incubated at 37 C in fresh medium and samples harvested every 30 min for analysis by flow cytometry. Wildtype cells exhibited a strong G2/ M (postreplication) DNA content in nocodazole-arrested cells, but within the first 30 mi-timepoint, approximately half the cells exited mitosis and contained a G1 content of DNA. Within the first hour, almost all wildtype cells had transitioned to a G1 (pre-DNA replication) state (Fig. 5). In contrast, neither isolate of eco1D rad61D double mutant cells exhibited significant cell progression at either the 30 min or 1 h timepoint (Fig. 5). The eco1D rad61D cln2D triple mutant cells exhibited a slight delay in cell cycle progression, compared to the wildtype cells, but far outpaced the progress of eco1D rad61D double mutant cells. Thus, loss of Cln2 promotes cell cycle progression, suggesting that the Cln2-CDK activity required for growth under hyperthermic conditions occurs independent of START.

CLN2 deletion suppresses the hypercondensation defect in eco1D rad61D double mutant cells
The positive effect of CLN2 deletion appears to be independent of both START regulation (Fig. 5) and the coupling of cohesion establishment to DNA replication (Fig. 4). ECO1 mutation not only abrogates the establishment of cohesion during S phase, but also produces condensation defects of mitotic chromosomes (Skibbens et al. 1999;Gard et al. 2009;Zuilkoski and Skibbens 2020b). Rad61/WAPL is another regulator of chromatin condensation, but loss of Rad61/WAPL results in hypercondensed chromosomes (Lopez-Serra et al. 2013;Tedeschi et al. 2013;Wutz et al. 2017). Since both Eco1 and Rad61 impact condensation, we tested the extent to which CLN2 deletion impacted chromosome condensation in eco1D rad61D double mutant cells. Notably, even wildtype cells respond to hyperthermic stress by increased compaction along the rDNA locus (Shen and Skibbens 2017; Matos-Perdomo and Mach ın 2018). The rDNA locus, which forms a pufflike structure during G1 but condenses into a tight loop or linear Fig. 4. Elevated expression of PCNA (POL30) does not suppress eco1D rad61D mutant cell growth defects during thermic stress. The growth, 2 temperatures, of 2 independent isolates of eco1D rad61D double mutant cells transformed with vector alone compared to 3 independent isolates of eco1D rad61D double mutant cells transformed with vector that directs elevated POL30 expression. structure during mitosis, is a prominent feature analyzed across numerous condensation studies (Guacci et al. 1997;Sullivan et al. 2004;D'Ambrosio et al. 2008;Lopez-Serra et al. 2013;de Los Santos-Velá zquez et al. 2017;Shen and Skibbens 2017;Matos-Perdomo and Mach ın 2018;Srinivasan et al. 2018;Lamothe et al. 2020). Thus, we performed a condensation assay, modified from FISH protocols, that provides for high-resolution analyses of both genomic mass areas and rDNA structures in response to elevated temperatures (Guacci et al. 1994(Guacci et al. , 1997Shen and Skibbens 2017;Srinivasan et al. 2018;Boginya et al. 2019;Shen and Skibbens 2020).
To first assess the effect of CLN2 deletion on chromatin condensation, log phase cultures of wildtype cells, eco1D rad61D double mutant cells, and 2 independent isolates of eco1D rad61D cln2D triple mutant cells, were released into fresh medium supplemented with nocodazole and incubated at 37 C for 3 h. Cell cycle arrest was monitored by flow cytometry (Fig. 6a) and chromatin condensation assessed by fluorescence microscopy as described previously (Shen and Skibbens 2017). As expected, preanaphase wildtype cells contained easily discernible rDNA that formed discrete loops that emanated from the bulk of the genomic mass (Fig. 6b). In eco1D rad61D double mutant cells, however, rDNA loops appeared reduced in length (Fig. 6b), but most often were indiscernible such that attempts to analyze rDNA structures resulted in the exclusion of a significant population of cells. Despite this, easily discernible rDNA loops were clearly evident in eco1D rad61D cln2D triple mutant cells, qualitatively suggesting that CLN2 deletion rescue the hypercondensation defects otherwise observed in eco1D rad61D double mutant cells.
We thus focused on quantifying genomic mass areas for each of the 4 strains to assess the impact of CLN2 deletion on chromosome condensation. The area of each genomic mass (excluding rDNA loops) was individually queried for all cells in the field of view, and compared to a standard area typical of DNA masses in wildtype cells (see Materials and Methods). Indeed, the results show that the majority (89%) of DNA masses from wildtype cells matched the template implemented for this analysis (Fig. 6, b and c). In contrast, the genomic mass area of eco1D rad61D double mutant cells appeared reduced and thus more highly condensed (Fig. 6b), with only 30% of DNA masses matching that of the template (Fig. 6c). The majority of genomic masses from eco1D rad61D cln2D triple mutant cells returned to a near wildtype level of condensation ( Fig. 6b) with 66% matching that of the template (Fig. 6c). While the increase of genomic mass hypercondensation in eco1D rad61D double mutant cell was statistically significant, compared to wildtype cells, the level of genomic mass hypercondensation of eco1D rad61D cln2D triple mutant cells was also significantly different from that of wildtype cells. These results suggest that rescue of hypercondensation defects is not the only mechanism through which CLN2 deletion promotes cell viability. Toward this end, we note that double rDNA rings, indicative of cohesion defects (Guacci et al. 1997;Shen and Skibbens 2017), were seldom observed in eco1D rad61D cln2D triple mutant cells.
Our cell cycle experiments suggest that the positive effect of CLN2 deletion occurs independent of the role for Cln2-CDK in START (Fig. 5). If true, then triple mutant cells shifted to the restrictive temperature of 37 C after START should result in normal Fig. 5. Deletion of CLN2 does not delay, but instead promotes, cell cycle progression in eco1D rad61D mutant cells. a) DNA contents assessed using flow cytometry of log phase wildtype cells, 2 independent isolates of parental eco1D rad61D double mutant cells, and 2 independent isolates of eco1D rad61D cln2D triple mutant before and after growth at 37 C for 3 h. b) DNA contents of the cells, described in (a), released from a mitotic arrest (NZ, nocodazole) at the permissive temperature of 30 C then shifted to 37 C. Samples harvested every 30 min were analyzed for DNA content by flow cytometry.
condensation of the genomic masses. To test this prediction, log phase cultures of wildtype cells, eco1D rad61D double mutant cells, and 2 independent isolates of eco1D rad61D cln2D triple mutant cells were incubated at 30 C in fresh medium that contained HU, which arrests cells after START in early S phase. The resulting S phase cultures were washed, released into fresh medium supplemented with nocodazole, and incubated at 37 C for 3 h. Cell samples were harvested at each synchronization step and DNA contents analyzed by flow cytometry (Fig. 6d). As in the prior experiment, wildtype cells contained large and uniformly stained genomic DNA masses (Fig. 6e) in which the majority (80%) matched the template area (Fig. 6f).
Discrete loops of rDNA were also easily discernible (Fig. 6e). The genomic mass areas obtained from eco1D rad61D double mutant cells, however, appeared more condensed (Fig. 6e), such that only 33% of the DNA masses matched the template (Fig. 6f). When distinguishable, the rDNA loops appeared shorter, compared to those obtained from wildtype cells (Fig. 6e). Despite shifting to 37 C after START, genomic masses of eco1D rad61D cln2D triple mutant cells exhibited increased areas (Fig. 6e), with Fig. 6. Deletion of CLN2 rescues the chromatin hypercondensation defect otherwise present in eco1D rad61D cells under thermic stress. a) and d) DNA content of wildtype cells, parental eco1D rad61D double mutant cells, and 2 independent isolates of eco1D rad61D cln2D triple mutant at either log phase (a) or synchronized in S phase using HU (d), prior to growth for 3 h at 37 C in medium supplemented with nocodazole (NZ). b) and e) Micrographs of DNA masses and rDNA obtained from preanaphase cells as described in (a) and (d), respectively. Chromosomal masses and rDNA loop structures were detected using DAPI. White arrows indicate rDNA loops, which typically reside proximal to the genomic mass. c) and f) Quantification of genome mass areas, relative to template, for cells described in (a) and (d), respectively. For graph c: wildtype (n ¼ 38), eco1D rad61D (n ¼ 105), eco1D rad61D cln2D (2 replicates: n ¼ 237 and n ¼ 176 to produce an n total ¼ 413). For graph f: wildtype (n ¼ 117), eco1D rad61D (n ¼ 189), eco1D rad61D cln2D (2 replicates: n ¼ 381 and n ¼ 83 to produce an n total ¼ 464). Chi-squared tests were subsequently used to assess the dependence of gene mutations (* indicates P-value at or below 0.05) on chromatin condensation levels. For graph c: wildtype, P-value ¼ 0.999; eco1D rad61D, P-value ¼ 0.0000272; eco1D rad61D cln2D, Pvalue ¼ 0.00104. For graph f: wildtype, P-value ¼ 0.983; eco1D rad61D, P-value ¼ 3.74 Â 10 À14 ; eco1D rad61D cln2D, P-value ¼ 4.45 Â 10 À14 . Not shown: we also compared chromatin condensation effects across the 3 strains. The percentage of genomic masses, within a given field, that matched or exceeded the template was calculated and those values used to determine statistical significance: wildtype to eco1D rad61D, P ¼ 0.00068; eco1D rad61D to eco1D rad61D cln2D (both replicates combined), P ¼ 0.0066. 56% matching the template and thus reverting to near wildtype (Fig. 6f). In addition to the rescue in genomic mass area, the rDNA appeared similar in length and paired as in those of wildtype cells (Fig. 6e). These results suggest both that Cln2-CDK antagonizes a chromatin condensation activity in cells under hyperthermic stress and that this activity occurs independent of START regulation.

Discussion
Eco1/ESCO2 and cohesins are required for an impressive array of cellular processes: cohesion establishment coupled to DNA replication during S phase, chromosome condensation and DNA repair during G2/M phase, nucleolar integrity, chromosome orientation within the nucleus, and dynamic transcriptional regulation during G1 so that cells can respond appropriately to changing internal and external cues. The identity of Cln2 as a key regulator of cohesin-based functions represents an important step forward in deciphering chromatin biology. Beyond Cln2, only Rad61 appears to directly antagonize cohesin function (Rolef Ben-Shahar et al. 2008;Sutani et al. 2009). The antagonistic roles played by Rad61/WAPL include dissociating cohesin from DNA during S phase and altering DNA extrusion loop lengths during G1 (Guacci and Koshland 2012;Lopez-Serra et al. 2013;Tedeschi et al. 2013;Gassler et al. 2017;Haarhuis et al. 2017;Wutz et al. 2017;Bloom et al. 2018). Other gene mutations that suppress phenotypes associated with ECO1 mutation occur in a more predictable fashion. For instance, smc3 K113N is an acetyl-mimic allele, providing a straight-forward rationale for the suppression of defects exhibited by eco1 mutant cells (Rolef Ben-Shahar et al. 2008;Sutani et al. 2009). Some eco1 mutant cell phenotypes are suppressed by specific mutations within PDS5, but Pds5 serves as a bridge between Rad61. These PDS5 mutations, however, appear to have little effect on cell growth or cohesion (Sutani et al. 2009). An intragenic allele of ECO1 also suppresses eco1-dependent growth defects, likely through promoting Eco1 dimerization (Sutani et al. 2009). Note that several studies now point to the importance of both Eco1 and cohesin dimerization/clustering during the establishment of sister chromatid cohesion and DNA loop formation, although other studies suggest that at least cohesins may act in monomeric form (Skibbens et al. 1999;Zhang, Kuznetsov, et al. 2008;Kulemzina et al. 2012, Tong andSkibbens 2014;Eng et al. 2015;Cattoglio et al. 2019;Kim et al. 2019;Liu et al. 2020;Xiang and Koshland 2021). As a final class of suppressors, ELG1 deletion or PCNA overexpression, both of which result in elevated levels of chromatin-bound PCNA, suppress eco1 mutant cell phenotypes through increased Eco1 recruitment to the DNA replication fork (Skibbens et al. 1999;Moldovan et al. 2006;Skibbens 2009, 2010;Parnas et al. 2010;Song et al. 2012;Johnson et al. 2016;Bender et al. 2020;Zuilkoski and Skibbens 2020a). Cln2 appears distinct from each of these suppressor pathways and thus creates a new category of cohesin regulators.
A second major revelation of this study is the unique role, among G1 cyclins, that Cln2 exhibits on cohesin pathways. G1 cyclins are critical for the irreversible transition from G1 into S phase, or START. Cln1-Cln3 are functionally redundant such that (1) any single G1 cyclin can promote START and (2) CLN1-CLN3 triple mutants are rescued by the expression of any one of several human G1 cyclins (Richardson et al. 1989;Koff et al. 1991;Lew et al. 1991Lew et al. , 1992Levine et al. 1996;Queralt and Igual 2004). G1 cyclins, however, can exert temporal differences in START regulation. For instance, Cln3 in association with cyclin-dependent kinase (Cln3-CDK) is an early inactivator of Whi5 (reviewed Fisher 2016;Ewald 2018;Li et al. 2021). Whi5, the yeast homolog of human retinoblastoma protein (Crane et al. 2019), is an inhibitor of the SBF-dependent transcription of genes that are required for the transition into S phase (Talarek et al. 2017;Teufel et al. 2019). SBF (and MBF) activation results in Cln1 and Cln2 upregulation, leading to Cln1-CDK and Cln2-CDK complexes that further inactivate Whi5 and provide a positive feedback mechanism to promote START. Due to the asymmetric division that occurs in Saccharomyces cerevisiae, the additional growth phase required by small buds (as opposed to the larger mother cells) that result from cell division appears more dependent on the shared activities of Cln1-CDK and Cln2-CDK. Moreover, Cln1 and Cln2 proteins are roughly 60% identical at the amino acid level and exhibit nearly identical target docking-site motifs (Pope et al. 2014;Bhaduri et al. 2015;Bandyopadhyay et al. 2020). Little if any evidence differentiates the activities of Cln1-CDK and Cln2-CDK. Thus, our findings that deletion of CLN2, but not that of either CLN1 or CLN3, suppresses the growth defect of eco1D rad61D double mutant cells highlight an important distinction between G1-CDKs and reveal a previously unknown role for Cln2-CDK. The identification of Cln2-CDK substrates that render eco1D rad61D double mutant cells inviable under thermic stress remains unknown but likely involves cohesin modifications given that both Eco1 and Rad61 regulate cohesins. In contrast, heat shock/chaperone proteins Hsp80 and Hsc82 also regulate hyperthermicinduced rDNA hyper-condensation, although this condensed chromatin state does not overtly extend to the remaining genomic mass (Shen and Skibbens 2020). Condensin complexes (condensin I and II in vertebrate cells) represent another set of factors through which chromosome condensation is regulated during G2/M and thus could be targets of Cln2-CDK (reviewed in Davidson and Peters 2021;Skibbens 2019;Paulson et al. 2021). Testing which of these factors (cohesins, HSPs, condensins, or other complex components) are targeted by Cln2-CDK during hyperthermic stress (as revealed in eco1D rad61D mutant cells) represents a significant undertaking for future studies, but the findings presented here regarding the dramatic changes in chromatin structure provide clues critical for directing those efforts.
The revelation that CLN2 deletion fully rescues eco1 rad61 mutant cell ts growth defects (and suppresses eco1 mutant ts growth defects) is somewhat surprising. For instance, eco1 mutation renders cells inviable when combined with mutated alleles of either CAK1 (encoding the CDK-activating kinase Cak1) or CDC28 (encoding the catalytic subunit of all cyclin-dependent kinases in budding yeast) (Brands and Skibbens 2008). Results from that study further revealed that CDKs indeed promote cohesion. A priori, these findings predict that cyclin mutation (or deletion) would be lethal in combination with eco1 mutations, in contrast to the rescue reported here. Resolving this apparent conundrum likely involves differentiating between the roles of mitotic CDK (deficiencies which are lethal in eco1 mutant cells) from Cln2-CDK (deficiencies which suppress eco1 mutant cell growth defects). The intersection of Eco1 and CDKs, however, is even more complex (Fig. 7). Elegant studies revealed that Eco1 is a phosphoprotein and CDK substrate (Ubersax et al. 2003), which is part of a signal that promotes Eco1 degradation during G2 and limits cohesion establishment activities to S phase (Skibbens et al. 1999;Lyons and Morgan 2011;Lyons et al. 2013;Seoane and Morgan 2017). This CDK/phosphorylation-dependent mechanism appears highly conserved, with subsequent ubiquitination provided by the Skp1-Cullin-F-box (SCF) ubiquitin ligase and augmented by Cullin Ring-like (CRL4) ubiquitin ligase (Lyons et al. 2013;Minamino et al. 2018;Sun et al. 2019). The mechanism through which CLN2 deletion suppresses eco1D rad61D mutant cell ts phenotypes, however, is demonstrably distinct from the CDK-dependent mechanism through which Eco1 activity is limited-given that ECO1 is fully deleted from these cells.
What is the cellular process through which CLN2 deletion rescues eco1D rad61D mutant cell ts lethality? While the current study focused on chromosome hypercondensation, structural changes in chromatin are likely to also impact gene transcription. Notably, transcriptional dysregulations that arise due to mutation of either ESCO2 or cohesin genes are typically lethal in humans (see below). We also note the relative absence of cohesion defects, as evidenced by lack of obvious rDNA double rings in eco1D rad61D cln2D triple mutant cells. Thus, our findings do not exclude the possibilities that Cln2, through CDK activation, impacts numerous cohesin functions through which cohesion, condensation, transcription, and DNA repair are mediated. Our results do reveal, however, that regulating the various cohesin functions are parsed to different factors, consistent with prior observations (Guacci and Koshland 2012;Zuilkoski and Skibbens 2020a).
Finally, the identification of CLN2 deletion as a suppressor of eco1 rad61 mutant cell ts growth defects provides a new framework for interpreting results from studies regarding the role of cyclins in development and cancer. In humans, severe developmental defects, collectively referred to as cohesinopathies, result from mutation of ESCO2 (homolog of yeast Eco1/Ctf7), most cohesin genes (SMC1A, SMC3, PDS5, RAD21) or cohesin regulators (NIPBL or HDAC8) (reviewed in Banerji et al. 2017a;Sarogni et al. 2020;Mfarej and Skibbens 2020b;Vega et al. 2020). All of these cohesin-pathway genes are essential, such that the syndromes that arise from ESCO2 mutation (RBS) or cohesin/regulator mutations (CdLS) occur infrequently-likely due to early and spontaneous pregnancy terminations. Individuals with RBS and CdLS exhibit defects in craniofacial structures, limb growth, GI and respiratory tracts, heart formation, and even intellectual abilities (Smithells and Newman 1992). Of interest here is that SMC1A mutated CdLS patient cells often exhibit downregulation of a G1 cyclin (cyclin D1, Ccnd1). While this observation prompted those authors to link reduced Ccdn1 levels to apoptosis, cell cycle delay, and tumorigenesis (Fazio et al. 2016), our findings formally raise the possibility that CdLS patient cell survival may result in part from Ccnd1 reduction, a model supported by findings that nipblb/smc1a reduction does not necessarily produce Ccnd1 down regulation (Fazio et al. 2016). On the other hand, the upregulation of Ccnd1 is tightly correlated with melanoma, breast, and other cancers (Ramos-Garc ıa et al. 2019;Gonzá lez-Ruiz et al. 2021). Future studies focused on the intersection between human Cln2 homologs, Esco2/cohesin pathways, and changes in chromatin structure will likely provide new insights into both tumorigenesis and birth defects.

Data availability
Strains and plasmids are available upon request. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and tables. The short read sequencing data have been deposited in the NCBI BioProject database (PRJNA836598).
Supplemental material is available at G3 online.

Fig. 7.
Multifaceted and opposing roles for CDK regulation of Eco1 activity. In this current study, we provide evidence that Cln2-CDK regulates chromatin condensation reactions in combination with Eco1 and Rad61. Cln2-CDK promotes chromatin hypercondensation after START (possibly during S phase) and in response to thermic stress. Also during S phase, Eco1 establishes cohesion between cohesins newly deposited (via Scc2,4) onto nascent sister chromatids after passage of the DNA replication fork. Eco1 acetylation of cohesin (stabilizing cohesin dimers) blocks Rad61-dependent cohesin dissociation activities. After S phase, M-CDKs phosphorylate Eco1, resulting in Eco1 degradation through both G2 and M phases. In response to damage, however, Eco1 expression is significantly upregulated during G2/M to promote DNA damage-induced cohesion (not shown), even as Eco1 degradation continues. Intriguingly, eco1 is synthetically lethal in combination with either cdc28 (CDK) or cak (CDK activating kinase), revealing that Eco1 and M-CDK perform complementary roles to maintain genomic integrity (which are likely to include chromosome condensation, cohesion maintenance, and transcription regulation). During G1, Eco1 is critical for regulation gene transcription and genomic architecture within the nucleus. The role of G1-CDKs in this portion of the cell cycle remain largely unknown.
Samantha Sarli, Rachel Sternberg, Divya Sirdeshpandi, Krupa Patel, and Anne Smolko) that were present over the course of this work and contributed their thoughts and support. We are deeply indebted to Drs Gregory Lang and Vincent Guacci for the sharing of reagents and expertise.

Funding
This work was supported by awards to R.V.S. from the National Institutes of Health [R15GM110631 and R15GM139097]. Any opinions, findings, and conclusions or recommendations expressed in this study are those of the authors and does not necessarily reflect the views of the National Institutes of Health.