Gene expression pattern in ulp2Δ cells is altered after multiple generations
We recently reported that growth and cell cycle defects in ulp2Δ cells were restored after ~ 500 generations (G) of continuous culture with repeated cycles of dilution and outgrowth in fresh medium13,14 (Fig. 1a). Missense mutations in essential SUMO conjugation pathway components (Uba2, Aos1, or Ubc9) were observed in all of the laboratory evolved ulp2Δ cells. In one set of 10 lines evolved in parallel, a single Uba2 cysteine-to-serine point mutation at position 162 (C162S) was observed in all the lines and is presumed to have arisen in the founder colony. However, this did not suppress the growth and cell-cycle impairments and had only a mild suppressive effect on the accumulation of HMW polySUMO conjugates in nascent ulp2Δ cells. In contrast, the other mutants had additional mutations in either Uba2 or Aos1, including uba2C162S, A414P (alanine 414 to proline), and these additional mutations strongly reduced HMW polySUMO conjugate accumulation in the evolved ulp2Δ strains, which mitigated the defects in growth and cell-cycle progression14. Thus, the mechanism of cell adaptation to Ulp2 loss in evolved ulp2Δ cells with only the Uba2-C162S mutation, has remained unclear.
To begin analyzing potential adaptive mechanisms in the ulp2Δ 500G (Uba2C162S) strain, gene expression profiles were analyzed by genomic RNA sequencing (RNA-seq; Fig. 1b, c and Supplementary Data 1). Levels of many transcripts were substantially different (two-fold or more) in the indicated strains compared to expression in the WT. In nascent ulp2Δ cells, the number of genes with increased transcript levels (601) (Fig. 1c) was more than double that of genes with reduced transcript levels (294) (Fig. 1b). Interestingly, the overall changes in transcript levels were much more muted in the two evolved ulp2Δ strains tested; specifically, the ratio of the total number of significantly up- and down-regulated genes in nascent ulp2Δ cells was more than double that of the two evolved strains (ulp2Δ 500G (Uba2C162S; 909 vs. 385) and ulp2Δ 500G (Uba2C162S, A414P; 566 vs. 201)). This indicates a general return to a WT gene expression balance in the evolved ulp2Δ strains despite the irreversible loss of Ulp2.
Energy metabolism genes, including tricarboxylic acid (TCA) cycle, ATP biosynthesis, and respiratory electron transport chain-related genes, were specifically down-regulated in nascent ulp2Δ, whereas ribosomal proteins and SUMO-regulated ribosome biogenesis factors24 were heavily enriched among the genes with a strong increase in expression. In direct contrast to this pattern, genes required for translation, including those involved in rRNA maturation, ribosome assembly and nuclear export, and translation factors, were down-regulated evolved ulp2Δ 500G (Uba2C162S) cells, while genes for energy reserve and cellular carbohydrate metabolic processes were upregulated in this line. In particular, a significant increase was seen in the mRNA levels for enzymes involved in converting glucose to glycogen and, conversely, for enzymes that catalyze catabolic reactions that break down carbohydrates, including hexokinase (Hxk1) in glycolysis, citrate synthase (Cit2) in the TCA cycle, and glycogen debranching enzyme (Gdb1) required for mobilizing glucose reserves from glycogen deposits25–27. These data implied that both the energy storage and consumption processes were markedly activated in ulp2Δ 500G (Uba2C162S) cells (Fig. 1c and Supplementary Data 1). Notably, any significant feature was not observed in the Gene Ontology (GO) analysis for the ulp2Δ 500G (Uba2C162S, A414P) line. In short, our RNA-seq data revealed that the decrease in enzymes for energy metabolism and increase in translation capacity seen in nascent ulp2Δ cells were reversed in the ulp2Δ 500G (Uba2C162S) strain (Fig. 1d) but not in the ulp2Δ 500G (Uba2C162S, A414P) line, suggesting distinct long-term adaptation mechanisms.
We previously reported several adaptive mutations that provide a selective advantage to evolved ulp2Δ cells14. We searched for other potential gene mutations in ulp2Δ 500G (Uba2C162S) and ulp2Δ 500G (Uba2C162S, A414P) by genome sequencing but did not find any alterations that would lead to changes in protein sequence except for the noted Uba2 mutations (Fig. 1e). To test whether the enhanced expression of the energy metabolism genes observed in ulp2Δ 500G (Uba2C162S) could influence the slow growth of nascent ulp2Δ cells, we transformed the WT strain (ulp2Δ + pULP2) with each of 33 different library plasmids containing identified energy metabolism genes and then analyzed cell growth via serial dilution analysis after evicting the ULP2 cover plasmid (Extended Data Fig. 1). None of the tested high-copy genes by themselves suppressed the growth impairment in the ulp2Δ cells. Altogether, these results imply that the growth recovery of ulp2Δ 500G (Uba2C162S) results from an as yet unidentified (possibly nongenetic) adaptive mechanism that likely requires altered expression of more than a single metabolic enzyme.
Growth rate and RLS are increased in high-passage ulp2Δ cells
Metabolic energy balance is important for cell growth and function; its malfunction is implicated in many complex diseases and aging19. To determine whether the elevated energy metabolism transcripts affect the growth rate of ulp2Δ 500G (Uba2C162S), we analyzed the growth curves of the indicated cells at different temperatures and using different concentrations and sources of carbon (Fig. 2 and Extended Data Fig. 2). With optimal temperature and carbon source (30°C and 2% glucose), typical S-shaped growth curves were observed for both the low passage WT and WT 500G cells. By 12 h, all strains except low passage ulp2Δ cells appeared to have slowed their doubling rate (diaxic shift). Cell density at saturation was slightly decreased in nascent ulp2Δ cells but had increased relative to WT in the evolved ulp2Δ cells. When the culture conditions were changed to low (0.2%) or very high (20%) glucose concentrations, a high temperature (37°C), or a nonfermentable carbon source (2% glycerol), the time needed to reach saturation was delayed and the growth rate changes in the mutants were much greater than under optimal conditions. Remarkably, the two evolved ulp2Δ lines had a clear growth advantage relative to WT cells, especially at 37°C. These findings indicate that laboratory evolution can induce beneficial effects beyond the correction of vegetative growth defects in ulp2Δ cells.
It is commonly assumed that growth rate is negatively coupled with lifespan, but some evidence indicates a positive correlation28,29. The SUMO pathway has been linked with cellular senescence and the organismal aging process. Its substrates include several factors that control cellular senescence, such as p53, RB, and SIRT121. Thus, we next examined the change in RLS caused by the genetic mutation or deletion of the main SUMO pathway proteins (Ubc9, Ulp1, and Ulp2; Fig. 3a). We observed that normal lifespan was significantly shortened in all the tested mutant strains (ubc9-1, ulp1-ts, and nascent ulp2Δ) suggesting that the dynamic regulation of SUMO conjugation and deconjugation is required for maintaining a normal lifespan.
Because Sir2 is a well-known factor that modulates RLS30–32, we subsequently investigated whether Sir2 affects RLS in the ulp2Δ strain (Fig. 3b). Consistent with previous results, RLS was decreased in sir2Δ, and the absence of Ulp2 further reduced sir2Δ lifespan; the latter result implies that Ulp2 plays a role in lifespan modulation separate from that of Sir2. In particular, this finding reveals a novel Sir2-independent pathway that sustains RLS by controlling sumoylation level.
Remarkably, lifespan was greatly extended in both evolved ulp2Δ strains, with RLSs as much as 80% above WT (Fig. 3c). As noted above, these same strains also exhibited higher growth rates relative to WT (Fig. 2). High rates of cell growth generally correlate with a shorter lifespan, but there are many exceptions. For instance, the loss of the global transcriptional regulator Sus1 or ubiquitin-specific protease Ubp10 simultaneously reduces growth and RLS33,34. Previously, we reported that several Uba2 double mutants, which included the C162S point mutation, suppressed the growth and cell-cycle defects seen in nascent ulp2Δ cells by reducing HMW SUMO-conjugated protein levels14. The C162S mutation by itself did not. To determine whether this reduction caused by Uba2 mutation affects RLS and growth, we analyzed the RLS and growth changes in uba2Δ and ulp2Δ strains expressing Uba2, Uba2C162S, A414P, Uba2C162S, or Uba2A414P (Fig. 3d and Extended Data Fig. 3). The expression of Uba2C162S, A414P eliminated the shortened lifespan and proportionally rescued the growth defects in ulp2Δ; these effects were not seen with in ulp2Δ cells carrying the singly mutated Uba2C162S or Uba2A414P. These findings suggest that Uba2 activity must be reduced to below a certain threshold, in this case requiring two Uba2 mutations, before RLS or growth impairment in ulp2Δ is rescued. The Uba2 double mutation may also be crucial for the RLS change in ulp2Δ 500G (Uba2C162S, A414P).
Taken together, the growth curve and RLS analyses showed that a normal limitation on cellular replication ability was overcome in both of the tested laboratory-evolved ulp2Δ strains, suggesting that adaptive mechanisms for overcoming Ulp2 loss might extend cellular lifespan.
Extended RLS linked to altered energy metabolism in ulp2Δ 500G (Uba2C162S)
Based on the RNA-seq experiments (Fig. 1b, c) we next addressed the impact the apparent elevated energy metabolism had on the lifespan of ulp2Δ 500G (Uba2C162S) (Fig. 3e and Extended Data Fig. 4a). Mitochondrial DNA (mtDNA) is necessary for mitochondrial respiration and energy metabolism35, leading many researchers to investigate RLS in cells lacking mtDNA (rho0 cells) to determine whether mitochondrial respiration affects RLS. Previous results show that the impact of mtDNA loss on cell longevity depends on the yeast strain36–38.
We found that the RLS of WT MHY1379 cells which had been rendered rho0 was decreased by about 20% compared to that of the WT, which is consistent with previous results37. Moreover, the extended RLS of ulp2Δ 500G (Uba2C162S) was abolished by loss of mtDNA, whereas ulp2Δ 500G (Uba2C162S, A414P) rho0 cells retained a significantly higher RLS, suggesting that mitochondrial function is a key element to prolonged RLS in ulp2Δ 500G (Uba2C162S) and that another pathway, perhaps SUMO regulation, is required for the extended lifespan of ulp2Δ 500G (Uba2C162S, A414P). The addition of antimycin A, which specifically blocks mitochondrial respiration and reduces RLS37, also more significantly impaired RLS in ulp2Δ 500G (Uba2C162S) than in ulp2Δ 500G (Uba2C162S, A414P cells) (Fig. 3f and Extended Data Fig. 4b). Thus, our results suggest that extended RLS in ulp2Δ 500G (Uba2C162S) cells results from a shift in energy metabolism.
Ubc9 auto-sumoylation is increased in ulp2Δ 500G (Uba2C162S)
Because the global profile of sumoylated proteins was altered in the evolved ulp2Δ strains (Extended Data Fig. 4c), we speculated that changes in SUMO substrates may contribute to cellular transcriptome changes relevant to RLS increase. We performed affinity purification in the evolved ulp2Δ strains expressing N-terminal 6His-FLAG (HF)-tagged Smt3 followed by mass spectrometry analyses to identify SUMO-modified proteins (Fig. 4, Extended Data Fig. 5, and Supplementary Data. 2). In agreement with previous findings39,40, HF-Smt3 was linked with multiple proteins involved in transcription, translation, metabolism, cytokinesis, and sumoylation in the WT (Fig. 4a). As expected, the initial Ulp2 loss led to broad changes in the SUMO-conjugate profile (Fig. 4a), which supports the conclusion that the Ulp2 SUMO protease has many target proteins required for diverse pathways, including transcription, the cell cycle, and ribosome biogenesis9. Both evolved ulp2Δ strains tended to deplete rather than increase SUMO-conjugated proteins; among them were proteins involved in energy metabolism and translation, including key enzymes need for glycolysis and gluconeogenesis (e.g., Adh1, Gpm1, Pdc1, Pgk1, and Eno241–43), several ribosome subunits, translation control factors (Gis2 and Tef1)44,45, and Ksp1, a kinase involved in TOR signaling (Fig. 4b, c)46.
Among the proteins characterized by changes in SUMO modification in the evolved strains, we were especially intrigued by the large increase in SUMO-conjugated Ubc9 accompanied by a corresponding decrease in its unconjugated state in ulp2Δ 500G (Uba2C162S) cells (Fig. 5a). Conversely, SUMO-modified Ubc9 almost disappeared in the ulp2Δ 500G (Uba2C162S, A414P) strain, presumably due to the extremely low levels of SUMO conjugates. Previous work demonstrated that the sumoylation of human Ubc9-Lys14 (K14) inhibits its ability to sumoylate the well-known target RanGAP1 but promotes modification of another substrate, Sp10022. In S. cerevisiae, Ubc9 auto-sumoylation is also observed, in this case on two residues, K153 and K157; this negatively regulates septin sumoylation, which is required for maintaining normal cell morphology23. Thus, the unusually high levels of Ubc9 auto-sumoylation seen in the in vitro evolved ulp2Δ 500G (Uba2C162S) strain and potentially the reduced levels of such modification in ulp2Δ 500G (Uba2C162S, A414P) cells is predicted to rewire cellular SUMO target discrimination.
Ubc9 auto-sumoylation increases RLS in a Sir2-independent manner
It was previously reported that Ubc9 auto-sumoylation was nearly abolished in a ubc9-K153/157R (ubc9-RR) mutant; this did not strongly affect vegetative growth or general SUMO conjugation but caused a severe reduction of global cellular sumoylation during meiosis47. We confirmed that cell growth did not differ between the WT and ubc9-RR strains in two different genetic backgrounds (W303 and MHY500; Fig. 5b); however, global SUMO conjugates were comparatively reduced in the ubc9-RR cells (Fig. 5c). Because the ulp2Δ 500G (Uba2C162S) strain displayed enhanced Ubc9 auto-sumoylation and extended RLS, we tested the effect of Ubc9 auto-sumoylation on RLS (Fig. 5d). The RLS of the ubc9-RR mutant was indeed shorter than that of WT cells, suggesting that Ubc9 auto-sumoylation is required for maintaining a normal lifespan. Moreover, the lifespan of a sir2Δ ubc9-RR double mutant was significantly shorter than that of either single mutant, indicating an additive effect on lifespan when these mutations are combined (Fig. 5e). As noted in Fig. 3b, Ulp2 loss shortens RLS in a Sir2-independent manner, and the results in Fig. 5e suggest that Ubc9 auto-sumoylation also contributes to a normal RLS through a Sir2-independent mechanism. Thus, our data indicate that the SUMO pathway contributes to yeast lifespan in a way unrelated to Sir2 function.
Ubc9 auto-sumoylation is required for RLS extension in ulp2Δ 500G (Uba2C162S)
To assess the effect of Ubc9 auto-sumoylation on global sumoylation in ulp2Δ 500G (Uba2C162S) cells, we measured SUMO-conjugate levels by anti-SUMO immunoblotting of extracts from ulp2Δ 500G (Uba2C162S) cells also bearing the ubc9-RR mutant allele (Fig. 5f). The ubc9-RR allele led to a sharp decline in HMW SUMO conjugates in the ulp2Δ 500G (Uba2C162S) background, while a more moderate reduction was seen in nascent ulp2Δ cells, suggesting that Ubc9 auto-sumoylation is an important factor in the maintenance of cellular SUMO conjugates in evolved ulp2Δ (Uba2C162S) cells.
While Ubc9K153/157R expression only weakly reduced polySUMO conjugate accumulation in ulp2Δ cells expressing WT Uba2 (Fig. 5g, lanes 3–4), Ubc9K153/157R expression in the ulp2Δ strain expressing Uba2C162S efficiently suppressed accumulation of the excess HMW polySUMO conjugates (Fig. 5g, lanes 7–8). This finding implies that the C162S mutation of Uba2 enhances suppression of the ulp2Δ defect by the auto-sumoylation-resistant Ubc9K153/157R protein. To determine whether Ubc9 auto-sumoylation influences long-lived ulp2Δ 500G (Uba2C162S) cells, we examined RLS in evolved ulp2Δ cells into which the ubc9-RR allele was introduced (as the only source of Ubc9) (Fig. 5h). The ubc9-RR allele dramatically reduced RLS in the ulp2Δ 500G (Uba2C162S) strain but not in ulp2Δ 500G (Uba2C162S, A414P) cells. This implies that Ubc9 auto-sumoylation is necessary for extended RLS in ulp2Δ 500G (Uba2C162S).
Taken together, our results suggest that Ubc9 auto-sumoylation-dependent SUMO target discrimination is a fundamental aspect of lifespan extension in ulp2Δ 500G (Uba2C162S). This may be connected to changes in energy metabolism, as suggested by our RNA-seq and mitochondrial DNA elimination data, while ulp2Δ 500G (Uba2C162S, A414P) increases RLS via a distinct route.
Ubc9 auto-sumoylation affects genome-wide SUMO binding to chromatin.
SUMO has both positive and negative impacts on transcription48,49, and its level in chromatin is dynamically regulated by the Ulp2 protease50. Transcription factors provide abundant SUMO substrates39,51, and SUMO is enriched in numerous loci, primarily the promoter regions of constitutively activated genes and genes involved in translation, such as ribosomal protein and tRNA genes49,52,53. Because ubc9-1 mutation strongly diminishes the association of SUMO with constitutive genes, ribosomal protein genes, and tRNA promoters49,52, we examined the role of Ubc9 auto-sumoylation in SUMO occupancy on chromatin via chromatin immunoprecipitation (ChIP)-seq analysis with strains expressing HF-tagged Smt3 (Fig. 6 and Supplementary Data 3). The SUMO ChIP-seq data sets contained 776 peaks in the WT and 758 peaks in the ubc9-RR mutant. Almost all the SUMO peaks in the two strains overlapped (Fig. 6a). SUMO was almost always located near the promoter regions (99.7% in WT and 99.8% in ubc9-RR; Fig. 6b), and of the SUMO-occupied genes, ~ 70% are protein-coding genes and ~ 30% are noncoding RNA gene (Fig. 6c), in agreement with previous results52,53. In particular, the ubc9-RR mutation contributed both positively and negatively to SUMO localization at diverse gene promoters, including LYS1, CIT2, TGL2, POL32, and NOC4 (Fig. 6d, e, f). KEGG pathway analysis revealed significantly increased and decreased SUMO enrichments in the ubc9-RR strain in genes involved in various pathways, including glycolysis, gluconeogenesis, ribosomal translation, and some metabolic pathways (Fig. 6g, h). This finding suggests that Ubc9 auto-sumoylation dynamically regulates chromatin sumoylation at multiple genes related to metabolic and ribosomal pathways.
Similar to the RNA-seq data (Fig. 1b, c), the general chromatin targets of auto-sumoylated Ubc9 were genes involved in energy metabolism and translation. Although SUMO both activates and represses transcription, previous studies suggest that sumoylation is usually associated with transcriptional repression48,49. Ubc9 lacking SUMO modifications enhances sumoylation of the glycolysis and gluconeogenesis gene loci, which may result in low levels of their transcripts. By contrast, highly auto-sumoylated Ubc9 in the ulp2Δ 500G (Uba2C162S) strain correlates strongly with increases in ribosomal gene expression levels, likely via decreased SUMO-chromatin binding at these sites. Thus, the higher ribosomal protein gene transcripts might be attributable to low levels of SUMO at these loci in ulp2Δ 500G (Uba2C162S).
Ubc9 auto-sumoylation influences sumoylation of proteins involved in translation and energy metabolism
Ubc9 auto-sumoylation can alter substrate selection by the SUMO-conjugation pathway22. To determine whether Ubc9 auto-sumoylation facilitates a broad transition in substrate targeting that might be connected to the genome-wide changes of SUMO-chromatin association (Fig. 6), we compared SUMO-conjugated substrates between the WT and ubc9-RR strains expressing HF-Smt3 using affinity purification and mass spectrometry (Fig. 7 and Supplementary Data 4). From cells grown in rich media to mid-exponential phase, we identified 245 SUMO-modified proteins only in WT cells, 34 substrates in both the WT and ubc9-RR strains, and none that were exclusively seen in ubc9-RR cells. This indicated that the number of sumoylated proteins had greatly declined in the auto-sumoylation mutant (Fig. 7a). KEGG enrichment analysis showed that proteins functioning in translation and metabolism constituted the largest classes of proteins that were detected only in the WT sumoylome (Fig. 7b); these categories were also enriched among the sumoylated proteins found in both WT and ubc9-RR cells (Fig. 7c).
Translation and metabolism are complex processes. To establish the 80S initiation complex for protein synthesis, the eIF4F complex facilitates mRNA association with the 43S preinitiation complex, which leads to formation of the 48S initiation complex, to which the 60S subunit binds following start codon selection (Fig. 7d)54. For the metabolic breakdown of glucose, glycolysis produces pyruvate in the cytoplasm, which is either converted ultimately to ethanol during fermentation or imported into mitochondria for respiration via the TCA cycle (Fig. 7e). SUMO is known to act as a key regulatory factor in both of these pathways24. Our data reveal that multiple translation initiation and termination factors, as well as ribosomal proteins, are only detectably sumoylated when Ubc9 can be auto-sumoylated, and the same is true for multiple glycolytic and TCA enzymes. These results suggest Ubc9 auto-sumoylation provides a mechanism to control the sumoylation and presumably levels or activity of proteins directly involved in protein synthesis and energy metabolism.
Ubc9 auto-sumoylation is increased in aged cells
Aging induces markedly increased sumoylation levels, which contribute to changes in mitochondrial dynamics and mitophagy in C. elegans55. To look for possible changes in yeast Ubc9 auto-sumoylation in response to aging, we first measured the SUMO conjugate levels in replicatively aged cells (Fig. 8a). Old cells were obtained by isolating biotin-labeled mother cells after three rounds of sorting. We found that the average bud scar number was more than twenty per cell (Fig. 8a, right panel). Similar to C. elegans, global SUMO conjugates were greatly increased in replicatively aged S. cerevisiae cells. By contrast, the ubc9-RR mutation blocked much of the age-linked accumulation of SUMO conjugates. This indicates that Ubc9 auto-sumoylation is needed to sustain enhanced sumoylation as yeast age. Notably, Ubc9 auto-sumoylation also increased substantially in aged cells compared to younger cells, and its unconjugated form was almost fully depleted (Fig. 8b). Therefore, it is possible that Ubc9 auto-sumoylation elicits the enhanced SUMO modification of targets in aged cells, and this in turn may promote the survival of aged cells, which we tested next.
Ubc9 auto-sumoylation is required for RLS extension by CR
In several organisms, CR leads to increased mitochondrial function to boost endogenous energy production and repress ribosome biogenesis and translation via the downregulation of TOR signaling, resulting in a prolonged lifespan56–59. Our results (Fig. 1b, c) showed CR-mediated changes in cellular physiological and metabolic characteristics that were also observed in ulp2Δ 500G (Uba2C162S) that had not undergone CR. Additionally, Ubc9 auto-sumoylation dramatically modulates the sumoylation of proteins in pathways affected by CR (Figs. 6 and 7). To investigate the relationship between Ubc9 auto-sumoylation and CR, we measured RLS in a ubc9-RR strain (Fig. 8c). The RLS of WT cells grown on SC plates was slightly lower than those grown on YPD rich medium. As expected from previous work60, RLS was clearly extended by reducing glucose content in the media from 2–0.5%. This starvation-induced RLS enhancement was lost in ubc9-RR cells. We repeated the RLS analyses three times and concluded that the increased ratio of RLS by lowering glucose concentration from 2.0–0.5% was significantly decreased by the inhibition of Ubc9 auto-sumoylation (Fig. 8d and Extended Data. 6). We suggest that Ubc9 auto-sumoylation regulates both transcriptional and post-translational targets of the SUMO pathway, particularly those that are involved in energy metabolism and protein translation to create a longevity-enhancing state that can also be reached through CR (Fig. 8e).