ChrI disomy of ulp2Δ cells is lost after extended culture
We first explored how quickly aneuploidy is established. For this we developed an improved inducible AID* degron-tagged ULP2 construct in which the degradation of the mRNA can also be induced through a neomycin-regulated self-cleaving hammerhead ribozyme in the 3′-UTR (Supplementary Fig. 1). We observed that a single ChrI disomy was preferentially generated in cells after depletion of Ulp2, although it took many days in culture. Although we could no longer detect the Ulp2 protein even at the first timepoint (2 days), it is likely that there was still a very low level present that limited the transition to the aneuploid state. Potentially, the transition occurs in a step-wise fashion, with ChrI disomy needed before ChrXII duplication can occur.
We had reported that ChrXII, but not ChrI, disomy was eliminated after ∼100 generations of continuous culture of ulp2Δ cells with repeated cycles of dilution and outgrowth in fresh medium18. To determine if the ChrI disomy would eventually also be lost during laboratory evolution, we monitored the relative copy number of ChrI at various points after the establishment of the ulp2Δ null state (Fig. 1a, b). We previously analyzed copy number for up to 250 cell generations and had not seen loss of ChrI18. By contrast, the extra copy of ChrI in ulp2Δ cells had disappeared by 500 generations. The more rapid elimination of extra ChrXII compared to ChrI, and the eventual loss of both extra copies, was verified by an independent experiment (Supplementary Fig. 2). Yeast ulp2Δ mutants are impaired in cell-cycle progression and chromosome segregation as well as growth17,24. The growth and cell-cycle defects were also fully suppressed in the long-term adapted cells (Fig. 1c, e), which was paralleled by a (partial) reduction in HMW SUMO conjugates (Fig. 1d). Hence, adaptation to loss of Ulp2 leads to a transient aneuploidy, which after culturing for many generations results in reversion to euploidy, nearly WT growth and cell-cycle progression, and reduced accumulation of the HMW SUMO conjugates typical of ulp2Δ cells.
High RP levels in ulp2Δ normalize after multiple generations
Aneuploidy is often associated with responses to stress; the duplicated chromosomes encode genes whose increased levels appear to counteract detrimental effects of the stress3,25. Gene expression imbalances resulting from aneuploidy, however, carry a fitness cost26,27. Aneuploidy may serve as a rapid stopgap response that can be superseded by more sustainable adaptations that somehow counter the stress7. Because aneuploidy disappeared and growth and cell-cycle defects were mitigated by evolving ulp2Δ cells over long periods (Fig. 1), we asked if the evolved strain had acquired any genetic changes.
Genome-wide gene expression profiles of the evolved ulp2Δ cells were analyzed by RNA-seq. In these comparisons, the ulp2Δ cells either carried a WT ULP2 plasmid or empty vector and were passaged for 50 or 500 generations (G). Expression of many transcripts was significantly different among the indicated strains (Supplementary Figs. 3a and 4). Differentially expressed genes and their changes relative to expression in the WT control strain (ulp2Δ + YCplac33-ULP2, 50 G) are shown in Fig. 2a. In nascent ulp2 null cells (ulp2Δ + YCplac33, 50 G), ~80% of the significant changes (twofold or more) involved increases in transcript levels (Fig. 2a and Supplementary Fig. 3b). Genes encoding ribosomal proteins (RPs) and ribosome biogenesis factors were heavily enriched among the genes with strong increases in expression (Fig. 2c (middle) and Supplementary Data 1). Among 136 genes encoding yeast ribosomal subunits, 130 were significantly upregulated in the ulp2 null mutant. By contrast, high expression of RP genes was no longer seen in ulp2 null cells after long-term culture at 30˚C in YPD medium (ulp2Δ + YCplac33, 500 G; Fig. 2c right). Indeed, these long-term adapted ulp2Δ cells had much more overlap in their upregulated transcripts with WT cells than with the nascent ulp2Δ strain (Fig. 2b, c). As predicted from the RNA-seq results, enhanced levels of ribosomal protein Rpl32 in nascent ulp2Δ cells were observed, and this returned to control levels after 500 generations. (Fig. 2d).
Earlier genome-wide analysis of SUMO modifications in yeast revealed that SUMO preferentially localizes to genes involved in translation, such as tRNA and RP genes28, and both RPs and ribosome biogenesis factors undergo SUMOylation29. Interestingly, sumoylation of the Rap1 transcription factor cooperates with the target of rapamycin complex 1 (TORC1) kinase pathway to promote transcription of genes encoding RPs28. We therefore tested whether the elevated RP transcripts seen in nascent ulp2Δ cells could potentially overcome inhibition of TORC1 by rapamycin, (Fig. 2f). Despite growing more poorly on rich medium alone (YPD), the nascent ulp2Δ strain grew better than the matched WT strains on rapamycin (at 50G; top two rows). After long-term passage, all the strains grew better on the drug, although not as robustly as the nascent ulp2 null strain. These data are consistent with the model that TORC1 and sumoylated Rap1 function in parallel to promote transcription of RP genes28.
We found that sumoylated Rap1 is a substrate of Ulp2 (Supplementary Fig. 5a). Association of Rap1 with RP genes was not significantly affected by loss of Ulp2 (Supplementary Fig. 5b), consistent with data suggesting sumoylation of Rap1 functions downstream of Rap1 binding to RP promoters28. The elevated levels of SUMO-conjugated Rap1 in ulp2Δ may therefore help drive RP gene transcription and rapamycin resistance.
Numerous snoRNAs are elevated in high-passage ulp2Δ cells
Increased levels of several categories of transcripts, such as those for fatty acid beta-oxidation and ammonium transport, were seen in both ulp2Δ cells containing a ULP2 plasmid (WT) and those with an empty vector after 500 generations (Fig. 2c). These might reflect a general response to long-term growth under these laboratory conditions. A much more specific response of long-term adapted ulp2 null cells was the increase of many snoRNAs, which are critical for rRNA modification (Fig. 2c, right). Of 77 annotated snoRNAs, expression of over half (42) was significantly elevated, compared with WT (Fig. 2e and Supplementary Data 2). The snoRNAs contribute both to the processing of the pre-rRNA into 18s, 5.8s, and 28s rRNAs, and to methylation and pseudouridylation of rRNA30. Novel snoRNA functions are also being uncovered, including some important for oncogenesis and different stress responses31,32,33,34. Notably, a multiple myeloma-associated snoRNA, ACA11, downregulates transcription of RP genes35.
Although we could not simultaneously upregulate dozens of snoRNAs, as observed in high-passage ulp2Δ cells, we could test whether increased levels of specific snoRNA gene(s) could ameliorate any of the ulp2 mutant defects that characterize low-passage cultures. Yeast 2-µm genomic library plasmids containing one or more snoRNA genes were transformed into WT (MHY1379) cells, and the URA3-marked ULP2 plasmid was then evicted. For these tests, we selected 13 plasmids bearing snoRNAs that displayed comparatively large increases in levels in high-passage ulp2Δ cells (Fig. 2e and Supplementary Data 2). None of the high-copy plasmids rescued the growth impairment, aneuploidy, or accumulation of HMW SUMO-conjugated proteins typical of low-passage ulp2Δ cells (Supplementary Fig. 6). However, plasmids containing certain snoRNA genes suppressed the elevated transcripts of the tested RPL12B and RPS18B RP genes, with eight and five, respectively, of the 13 plasmids suppressing to significant degrees (Fig. 2g). Furthermore, 11 of the 13 plasmids as well as an additional plasmid with three clustered snoRNA genes from ChrXII (see below) reduced excess accumulation of Rpl32 protein upon loss of Ulp2 (Fig. 2h). Whether the identified snoRNAs directly regulate RP expression cannot be inferred from these data, but the inverse correlation in expression between a cohort of snoRNAs and RP genes is supported. Inasmuch as snoRNAs are essential for ribosome formation, it is possible that enhanced ribosome biogenesis under these snoRNA overexpression conditions partially relieves the requirement for higher RP gene expression.
Ulp2 modulates SUMO modification at RP and snoRNA genes
Past studies have implicated mammalian SUMO proteases in numerous transcriptional processes16, but there was no evidence that yeast Ulp2 is directly involved in transcription. To determine whether Ulp2 is recruited to RP or snoRNA genes, we carried out chromatin immunoprecipitation (ChIP) assays using cells with the chromosomal ULP2 gene tagged with a sequence encoding a Flag epitope. We used the Rap1 transcription factor as a positive control as it is known to associate with the RP genes and some snoRNA genes28. The ChIP analysis indicated that Ulp2 associated with the promoter and coding sequence (CDS) of both RP and snoRNA genes (Fig. 3a); the protein deacetylase Hst3 and the transcriptional repressor Nrg2, whose genes were significantly downregulated in ulp2Δ cells after extensive passaging, were not associated with these genes. These data suggest that Ulp2 could be directly involved in transcription of these gene classes.
Association of the Ulp2 SUMO protease with chromatin sites would be predicted to decrease local protein sumoylation, as occurs at the RP genes28. We compared levels of yeast SUMO (Smt3) at the RP and snoRNA genes in WT and ulp2Δ cells (Fig. 3b). His6-tagged Smt3 was used for the ChIP analysis. Loss of Ulp2 was associated with increased levels of SUMO/Smt3 in both the promoter and CDS regions at most of the tested loci.
Taken together, our data reveal that Ulp2 associates with RP and snoRNA loci where it limits local accumulation of SUMO (presumably SUMO-conjugated chromatin proteins such as Rap1 or histones). This suggests that Ulp2 has a direct role in the expression of RP and snoRNA genes.
Increased Ccr4 in ulp2Δ limits expression of snoRNA and rRNA
Many rDNA silencing factors are targets of sumoylation, and at least the Net1, Tof2, Cdc14 and Fob1 nucleolar proteins are substrates of Ulp236,37. Ulp2 localizes to rDNA loci, and Ulp2 mutations strongly impair association of Net1, Tof2 and Fob1 with rDNA36, suggesting Ulp2 generally promotes rDNA silencing. We therefore assumed that rRNA levels would be altered in ulp2Δ cells undergoing laboratory evolution; however, rRNA levels were unchanged in cells passaged for 50 or 500 generations (Fig. 3c).
We previously reported that ChrI aneuploidy in ulp2Δ cells is suppressed by increased dosage of two ChrI genes, one encoding Ccr4, a subunit of the multifunctional Ccr4–Not complex, and the other Cln3, a G1-phase cyclin18. Therefore, raising either Ccr4 or Cln3 levels is sufficient to prevent the ChrI duplication. Notably, the Ccr4-Not complex limits RNA Pol I binding to rDNA loci and inhibits rRNA synthesis38. Moreover, a wide variety of snoRNAs are over-expressed in mutants of the Ccr4-Not complex39,40.
We confirmed that lack of Ccr4 resulted in higher expression of rRNA (18s and 25s), SNR30, and SNR60 (Fig. 3d, e); RP transcript levels were unaffected. The ulp2Δ mutation by itself did not alter expression of these genes, but in a ulp2Δ ccr4Δ strain, these RNAs, and not RP transcripts, appeared to be modestly increased relative to ccr4Δ. Ccr4 protein levels rose sharply as a result of the ChrI disomy in nascent ulp2Δ cells (Fig. 3f); this increase was accompanied by a specific enrichment of Ccr4, but not the Not5 subunit of the Ccr4-Not complex, at snoRNA gene promoters as well as the 35s gene (Fig. 3g).
Taken together, our data and previous results suggest that Ulp2 may generally limit transcription of the RP, snoRNA and rRNA genes, whereas Ccr4 blocks the expression of snoRNA and rRNA genes, but not of RP genes (Fig. 3h, left). Loss of Ulp2 leads to increased Ccr4 as a result of the ChrI disomy, which in turn limits the transcription of snoRNA and rRNA genes (Fig. 3h, right). The net effect is a specific increase in RP expression in incipient ulp2Δ cells.
Exogenous ChrXII snoRNAs suppress the ChrXII ulp2Δ disomy
Next, we screened for genes on ChrXII that could prevent the ChrXII disomy upon ULP2 deletion (Fig. 4a). Each of 135 plasmids containing individual ChrXII fragments41 was transformed into ulp2Δ cells carrying a URA3-marked ULP2 plasmid (MHY1379), and the ULP2 plasmid was then evicted. Ratios of ChrI and ChrXII to ChrIII, a chromosome that is not duplicated in ulp2Δ cells, were analyzed by a qPCR-based ploidy assay (Supplementary Fig. 7).
Transformants with 22 different library plasmids reduced the ratio of ChrI/ChrIII or ChrXII/ChrIII from ∼2 to 1, suggesting possible suppression of ulp2Δ aneuploidy. We repeated the ploidy assay on five chromosomes with the candidate plasmids in nascent ulp2Δ cells (Fig. 4b). Four plasmids covering two separate regions of ChrXII continued to suppress disomy of ChrXII but not of ChrI, (Fig. 4b, c). They did not affect the slow growth and accumulation of polySUMO conjugates (Supplementary Fig. 8). To determine the genes within these clones required for suppression of the ulp2Δ aneuploidy, 15 DNA fragments were individually cloned into library vector pGP564 (Fig. 4c), and their effect on ploidy was tested (Fig. 4d). Extra copies of CCW12 or a cluster of snoRNA genes (SNR61, SNR55, and SNR57) specifically prevented the development of the ChrXII disomy. Ccw12 is a mannoprotein important for cell wall integrity, and its cellular level is tightly regulated42,43,44. The possible contribution of Ccw12 to ulp2Δ aneuploidy was not pursued further here.
The three snoRNA genes in the cloned cluster were all C/D box snoRNAs. Besides suppressing the ChrXII disomy, high-copy expression of the cluster suppressed the excess expression of RP genes RPL12B and RPS18B normally observed in nascent ulp2 mutant cells (Fig. 4e); the much higher dosage of these three snRNA genes compared to ChrXII disomy presumably accounts for the suppression of RP transcripts only in the former case. As predicted from Fig. 3a, Ulp2 was directly recruited to both promoter and coding sequences of SNR57 (Fig. 4f), and SUMO levels at these sites increased in ulp2Δ cells (Fig. 4g), suggesting that Ulp2 is involved in the transcription of SNR57 as well as other snoRNA genes. These data support the hypothesis that enhanced levels of snoRNAs help circumvent the cellular stress caused by loss of Ulp2, first through a ChrXII disomy that raises the dosage of these C/D box snoRNAs, and then after many generations, through a broad increase in expression of many snoRNAs, which is linked to suppression of excess RP transcripts.
High-passage ulp2Δ cells acquire adaptive mutations
Cellular mechanisms for adaptation to stress include gene mutation, aneuploidy, and transcriptional or translational changes, depending on the nature and length of the stress3,4. In a previous study, aneuploidies developed under heat or high pH stress were superseded over longer periods by gene mutations and refinements of gene expression presumed to carry lower fitness costs than aneuploidy7. We asked if aneuploid ulp2Δ yeast undergo an analogous evolutionary process using whole genome DNA sequencing (DNA-seq) to identify potential gene mutations arising during repeated cell passaging (Table 1). Mutations were detected in both coding and noncoding DNA sequences; we focused on those predicted to cause changes to protein sequences.
Interestingly, BUB1, MAD3, and BUB3 genes, which are involved in the spindle checkpoint of the cell cycle45, and SPO23 gene, encoding a protein that interacts with meiosis-specific protein Spo146, carried missense mutations in evolved ulp2Δ cells that either continuously carried or had reintroduced the ULP2 plasmid. Although these ulp2Δ strains had a CEN plasmid with the WT ULP2 gene, plasmid instability might reduce Ulp2 expression in these cultures17. This could cause cell-cycle arrest in response to checkpoint activation47, a defect that may be overcome by mutations in the checkpoint genes. The adaptive significance of the Spo23 mutation is unknown.
In the ulp2Δ strain that had been passaged for ~500 generations without ULP2 plasmid (ulp2 null cells), two missense mutations were found (Table 1). One was in IFM1, encoding mitochondrial translation initiation factor 2; this was not investigated further. The other mutation was in UBC9, which codes for the SUMO-conjugating E2 enzyme. The predicted K27N substitution is in an N-terminal domain that contains binding sites for the E1 and E3 enzymes48. Accumulation of polySUMO conjugates in these late-passage cells had attenuated as well (Fig. 1d).
Analysis of ulp2Δ descendants after long-term passaging
To assess the generality of the altered mutant phenotype of evolved ulp2Δ cells observed in Figs. 1 and 2, we grew ten separate ulp2Δ lines from a single nascent ulp2Δ colony and evaluated them after 250 and 500 generations. ChrXII disomy, but not that of ChrI, had disappeared by 250 generations in all the evolved lines, as had their growth defects (Fig. 5a, b), consistent with Fig. 1b, c and our previous report18. After an additional 250 generations, the disomy of ChrI was also eliminated. Analysis by FACS showed the recovery from the initial ulp2Δ cell-cycle defect at both 250 and 500 generations (Fig. 5c). Also, after both 250 and 500 generations, all the ulp2Δ strains showed resistance to high temperature and oxidative stress (Supplementary Fig. 9). Consistent with RNA-seq data obtained from ulp2Δ at 500 generations in Fig. 2, all ten of the high-passaged ulp2Δ strains showed reversion of the RPL12B and RPS18B transcripts to normal levels along with elevated transcripts of snoRNAs, SNR30 and SNR60 (Fig. 5d). Thus, all ulp2Δ strains eventually move away from aneuploidy and adopt alternative adaptive strategies that enhance fitness.
As noted earlier, after 500 generations of continual passaging, the original ulp2Δ strain tested no longer accumulated excess SUMO conjugates (Fig. 1d). We also measured SUMO conjugate levels by anti-SUMO immunoblotting in the ten separately evolved ulp2Δ strains after 250 and 500 generations (Fig. 5e). PolySUMO-conjugated proteins significantly declined in five of the strains after 250 generations (strains 3, 4, 8, 9, and 10). After 500 generations, conjugate levels had also declined in strains 1 and 7.
Hence, all of the independently evolved ulp2Δ strains eventually reestablished euploidy and near-normal cell growth and cell-cycle progression, but even after 500 generations, several of the strains still accumulated excess HMW SUMO conjugates, suggesting alternative long-term mechanisms for adapting to Ulp2 loss.
Point mutations in the SUMO E1 enzyme in evolved ulp2Δ
In the original evolved ulp2Δ strain, a mutation in the UBC9 gene had been uncovered (Table 1), presumably because reduced E2 activity suppresses accumulation of polySUMO conjugates. No UBC9 mutations were found in the ten new ulp2Δ strains grown without ULP2 for 250 and 500 generations (Table 2). Since most of the evolved ulp2Δ strains showed reduced accumulation of polySUMO conjugates (Fig. 5e), we checked for mutations in all other genes known to be required for normal SUMO conjugation in vegetative cells, specifically those encoding the E1 (Aos1/Uba2 dimer), E3s (Siz1, Siz2 and Mms21), and SUMO protease Ulp1.
Interestingly, we found several missense mutations in UBA2, which encodes the catalytic E1 subunit. All ten strains shared a mutation in the UBA2 ORF, causing a C162S change in the encoded protein (Fig. 5f). This mutation was probably acquired during the original 5-FOA counter-selection to evict the covering ULP2 plasmid because all ten ulp2Δ strains originated from the same colony after growth for ~30 generations on 5-FOA. Additional UBA2 missense mutations were observed in six of the evolved ulp2Δ strains (Fig. 5f). All these ulp2Δ strains expressed a doubly mutated Uba2 and suppressed accumulation of polySUMO conjugates (Fig. 5e). Strain 8, which also suppressed polySUMO-conjugate levels yet had no additional Uba2 mutations, picked up a mutation in Aos1 (E346V). Interestingly, while the uba2-C162S mutation was insufficient to suppress polySUMO-conjugate levels in strains 1 and 7 at 250 generations, the additional Uba2 mutations seen at 500 generations were associated with a strong decrease in conjugate levels. Conversely, three evolved ulp2Δ strains (2, 5, and 6) had no detectable changes in the tested SUMO pathway components, consistent with their continued high levels of polySUMO conjugates (Fig. 5e).
The Uba2 missense mutations occurred throughout the protein (Fig. 5f). Only C162S is located within the domain bearing the catalytic Cys173 residue. The other mutations occur in the adenylation domain (AD) or the ubiquitin-like domain (UFD) required for recruiting E249. When provided as the only source of Uba2 in otherwise WT cells, all the Uba2 point mutants, including the Uba2C162S single mutant and the six Uba2 double point mutants, reduced SUMO conjugation (Fig. 5g) without impairing cell growth (Supplementary Fig. 10). Cells expressing Aos1E346V continued to accumulate conjugates, but the aos1-E346V uba2-C162S double mutant did not.
Taken together, our results suggest that long-term loss of Ulp2 SUMO protease activity favors countervailing partial loss-of-function mutations in genes encoding SUMO ligation pathway enzymes, particularly the essential Ubc9 (E2) and Uba2/Aos1 (E1) enzymes.
UBC9 and UBA2 mutations prevent aneuploidy in nascent ulp2Δ
To test the hypothesis that ulp2Δ aneuploidy could be directly suppressed by these balancing mutations in SUMO-ligating enzymes, we measured chromosome copy number, as well as cell growth, cell-cycle progression, and SUMO conjugates, in nascent ulp2Δ cells that carried ubc9, uba2, or aos1 missense mutations (Fig. 6a). Ubc9K27N expression efficiently suppressed the development of aneuploidy and the other mutant traits in low-passage ulp2Δ cells (Fig. 6b–e). Notably, even though provision of the single point mutant Uba2C162S did not suppress the growth and cell-cycle defects of nascent ulp2Δ cells, it slightly reduced levels of HMW SUMO-conjugated proteins and prevented the development of ChrXII (but not ChrI) disomy (Fig. 6f–i). Aneuploidy and the other defects of cells newly depleted of Ulp2 were also prevented by all tested uba2 double point mutants and the uba2C162Saos1E346V double mutant (Fig. 6f–m). These suppression results with nascent ulp2 null cells strongly support the contention that the E1 and E2 missense mutations isolated through long-term adaptation to Ulp2 loss are directly responsible for the recovery from aneuploidy and the suppression of the growth and stress resistance defects.