August 15, 2022

Healty

Slick Healthy

Proteotoxicity caused by perturbed protein complexes

Chromosome replacement lines exhibit a transcriptional signature of stress responses

Previously, we constructed 11 chromosome replacement lines between two closely related species, S. cerevisiae (Sc) and S. bayanus var. uvarum (Sb), to dissect the contribution of individual chromosomes to hybrid incompatibility28. Although we identified several cases of strong mitochondrial-nuclear incompatibilities involving two genes28,29, the weak and polygenic nuclear-nuclear incompatibilities widely observed in these lines remained unexplored. Transcriptome analysis can represent a sensitive way of detecting hybrid dysfunctions30,31,32. Therefore, we evaluated the transcriptomic consequences of the presence of foreign chromosomes in all the 11 diploid replacement lines derived from crossing a and α types of the respective replacement lines. We conducted RNA sequencing on total RNA isolated from diploid cell lines grown at 23 °C, a non-stressful temperature for both S. cerevisiae and S. bayanus var. uvarum, and classified a gene as being differentially expressed if the fold-change in expression between the replacement line and the parental line (Sc) was greater than 1.5 with an adjusted p-value < 0.05.

We identified several hundred genes as differentially expressed in each replacement line, even when the genes on the replaced foreign chromosomes were excluded (Fig. 1a and Supplementary Data 1). Why would one or two foreign chromosomes from a closely related species have such a strong influence on the rest of the genome? We found that a large proportion of the differentially expressed genes were commonly up- or downregulated in multiple replacement lines, suggesting that these changes may represent a general response to foreign chromosomes (Fig. 1a and Supplementary Data 1). Moreover, many of these common-response genes encode molecular chaperones or proteins related to stress responses33. Transcriptome analyses of yeast cells grown under diverse exogenous stresses (such as heat shock, osmotic shock, starvation, and oxidative stress) revealed 868 genes that were commonly up- or downregulated under such stresses, termed ESR (environmental stress response) genes33. When we compared the expression profiles of ESR genes between our replacement lines and the cells subjected to various stresses, we observed a positive correlation between most of our replacement lines and stress-treated cells (Fig. 1b and Supplementary Data 2). That ESR signature suggests that our replacement lines harboring foreign chromosomes were physiologically stressed even though the cells were growing in a non-stressful environment. Such a transcriptomic stress response signature is not specific to our replacement lines and has been observed in hybrids of fungi, plants, and animals34,35,36,37,38,39. It suggests a general phenomenon in hybrids to cope up with physiological stress caused by the coexistence of two divergent genomes. Nonetheless, the detailed mechanisms remain elusive.

Fig. 1: Foreign chromosomes induce proteotoxic stress in diploid hybrid cells.

a Hundreds of genes are differentially expressed in diploid chromosome replacement lines. The replacement lines were cultured in rich medium at 23 °C and their transcriptomes were examined using RNA-seq. A gene was classified as differentially expressed if the fold-change of a replacement line gene to its parental S. cerevisiae strain was greater than 1.5 with an adjusted p-value < 0.05. The genes on the replaced chromosomes were excluded from the analysis. Common-response genes refer to the genes commonly up- or downregulated in at least four replacement lines. A complete gene list can be found in Supplementary Data 1. b The expression profile of diploid replacement lines is positively correlated with those of wild-type cells under stress conditions and aneuploid cells. The median expression levels in eleven diploid replacement lines were compared to the environmental stress response (ESR) dataset33 and the aneuploid transcriptome42. Spearman’s correlation coefficients were calculated and are shown in the figure. Details of correlations for individual lines are presented in Supplementary Data 2. c Images of Hsp104-mCherry aggregates in diploid S. cerevisiae (Sc) and replacement cell lines. Hsp104 was diffusely localized in the cytosol when cells were grown at 23 °C, but formed foci upon shifting the temperature to 37 °C. Scale bar: 5 μm. These results were reproducible in eight independent experiments. d Diploid replacement lines take longer to dissolve all aggregates upon heat treatments. Yeast strains containing an HSP104-mCherry-URA cassette were grown to exponential phase in YPD at 23 °C and then shifted to 37 °C. The percentage of cells containing Hsp104-mCherry foci was determined 180 min after shifting the temperature. The 12 L line was not included in this assay since HSP104 is located on replaced Chromosome 12 and S. bayanus var uvarum was excluded as they are cryotolerant and cannot tolerate heat treatments (n = 8; N ≥ 500 cells per time-point). Data are presented as mean values +/− SEM. ***: p-value < 10−3; one-sided Student’s t-test between S. cerevisiae and replacement lines. Source data and detailed statistical information are provided as a Source Data file.

Chromosome replacement lines display proteotoxic stress

What causes ESR induction in our chromosome replacement lines? Studies in aneuploid cells suggest that ESR can be induced intrinsically by the proteotoxic stress arising from unbalanced chromosome numbers40,41. Notably, the ESR signature in our replacement lines was significantly correlated to that observed for an aneuploid population (ρ = 0.41, p < 2.23 × 10−16, Spearman’s rank correlation, Supplementary Data 2)42. However, the replacement lines are euploid and derived from related yeast species with an almost complete set of orthologous proteins. Thus, the cause of ESR in hybrid replacement lines is likely to be different from that in aneuploid strains.

Hsp104 is a protein disaggregase widely used as a marker for protein aggregation under many conditions of proteotoxic stress43. We assessed if our diploid replacement lines and normal F1 hybrid diploids containing a complete set of the two parental genomes also suffered from proteotoxic stress by analyzing the subcellular localization of Hsp104 during heat adaptation41. We cultured the yeast strains carrying the Hsp104-mCherry fusion protein at 23 °C and then shifted them to 37 °C to induce protein aggregation. We anticipated that, initially, the Hsp104-mCherry signal would be diffused throughout the cytosol, but that after heat treatment it would co-localize with protein aggregate foci to clear the aggregates44. Accordingly, cells suffering from intrinsic proteotoxic stress should have more pronounced protein aggregations and take a longer time to dissolve the aggregates before Hsp104 would disperse throughout the cytosol once again41.

Indeed, our protein aggregation assay showed that all the tested replacement lines as well as the F1 hybrid diploids had a significantly higher proportion of cells with Hsp104 foci at an early time-point (45 min) and took longer to dissolve all aggregates compared to the S. cerevisiae control (Fig. 1c and Supplementary Fig. 1a). At 180 min after the temperature shift, Hsp104 foci were undetectable in most S. cerevisiae control cells, but both the replacement lines and F1 hybrid diploids exhibited significant retention of aggregates (Fig. 1d and Supplementary Fig. 2). These results corroborate our transcriptome data that proteostasis is perturbed in cells carrying foreign chromosomes. Among our diploid replacement lines, S. cerevisiae cells carrying S. bayanus var. uvarum chromosomes 5 and 7 (5+7L), 8 and 15 (8+15L), and 16 (16L) exhibited the slowest adaptive kinetics, suggesting that these lines might display the most severe intrinsic proteotoxic stress. Since we had previously established that the 5+7L replacement line also suffers from mitochondrial-nuclear incompatibility29, we focused our subsequent experimental analyses on the 8+15L and 16L lines.

To rule out the possibility that the defect in heat adaptation presented by hybrid cells was due to cell death or an inability to mount a heat shock response, we measured the cell viability of all lines and the protein abundance of several heat-induced molecular chaperones in the two selected replacement lines exhibiting the most severe phenotypes. Our results show that the replacement lines did not have obvious defects in viability or heat shock response at 37 °C (Supplementary Fig. 1b, c).

A recent study showed that the S. cerevisiae W303 strain was more sensitive to proteotoxicity than other natural isolates due to a defective SSD1 allele in W303 and introducing SSD1 from the oak soil strain YPS1009 into W303 restored proteotoxicity tolerance45. Since our replacement lines were derived from the W303 strain, we checked if the strain background was the primary source of the proteotoxicity by introducing SSD1YPS1009 into the most defective replacement lines. At 180 min after the temperature shift, SSD1YPS1009-containing replacement lines still retained a significant percentage of cells harboring HSP104 foci, suggesting that the observed proteotoxic stress in hybrids was not simply due to the W303 strain background (Supplementary Fig. 3).

Stress resulting from harboring foreign chromosomes causes mitotic and meiotic defects

Our protein aggregation assay revealed that the diploid replacement lines do not adapt to temperature changes as efficiently as the parental strain. However, only some replacement lines exhibited obvious growth defects under normal conditions (Fig. 2a)28. To further understand how the intrinsic proteotoxic stress induced by foreign chromosomes impacts cell fitness, we measured mitotic growth rates in rich medium containing a low dose of the Hsp90 inhibitor, geldanamycin (GdA). Hsp90 is a molecular chaperone essential for maintaining proteostasis and relieving the proteotoxicity caused by stress46. Moreover, several client proteins and protein complexes of Hsp90 are required for mitosis and meiosis47. Thus, Hsp90 inhibition is likely to enhance mild defects in mitosis and meiosis if they already exist in replacement lines. At 23 °C, mild interference of Hsp90 by GdA treatment (50 μM) did not cause obvious growth defects in S. cerevisiae and S. bayanus var. uvarum cells. However, we observed significantly reduced cell growth in all replacement lines as well as the F1 hybrid diploids (Fig. 2a and Supplementary Fig. 2c), suggesting that intrinsic proteotoxic stress further sensitized the cells to even slight perturbation of proteostasis. The growth defects were not specific to GdA or Hsp90 since we observed similar reductions in fitness when the replacement lines were grown at 32 °C, representing mild heat stress that did not affect the growth of wild-type cells (Fig. 2b and Supplementary Fig. 4a).

Fig. 2: Proteotoxic stress-induced incompatibility leads to mitotic and meiotic defects.
figure 2

a Growth rates of diploid replacement lines are significantly reduced when Hsp90 is mildly compromised. The doubling time of S. cerevisiae (Sc), S. bayanus var uvarum (Sb), and replacement lines was measured and compared with or without the Hsp90 inhibitor, Geldanamycin (50 μM GdA), at 23 °C (n = 3). b The growth defect is not specific to compromised Hsp90. When replacement lines were grown in YPD at 32 °C, the doubling time is highly correlated with the data for cells treated with GdA (n = 3, Spearmans ρ = 0.83, p = 0.03). c Several replacement lines exhibit significant meiotic defects when pre-treated with a low dosage of GdA. Cells were grown in pre-sporulation medium with 50 μM GdA and then induced to sporulate under normal conditions without GdA (n = 4). The temperature was always maintained at 23 °C. Sensitivity was calculated as 1 – (sporulation frequency with GdA pre-treatment/sporulation frequency without GdA pre-treatment) and compared to both parents. d, Growth defects of the 16L and 8+15L lines can be partially rescued by providing a complete set of Sc or Sb genomes. The 16L and 8+15L haploid cells were mated with S. cerevisiae (16L × Sc and 8+15L × Sc) or S. bayanus var uvarum (16L × Sb and 8+15L × Sb) to generate heterozygous diploid cells and their doubling time was measured in the presence of 50 μM GdA. The doubling time of homozygous diploid replacement lines was compared to that of the heterozygous diploids (n = 3). e The fitness defect under GdA treatment (50 μM) is significantly correlated with the number of complexes having subunits encoded on the replaced chromosomes (Spearmans ρ = 0.68, p = 0.025). f The percentages of cells containing Hsp104-mCherry foci are significantly correlated with the number of complexes having subunits encoded on the replaced chromosomes (Spearmans ρ = 0.68, p = 0.03). Hsp104 aggregates were counted in cells after having been shifted to 37 °C for 180 min (n = 8, N ≥ 500 cells per time-point). The data are presented as mean values +/− SEM. ***: p-value < 10−3, one-sided Student’s t-test. Source data and detailed statistical information are provided as a Source Data file.

Hybrid sterility is often observed upon mating between two species. To examine the effect of intrinsic proteotoxic stress on meiosis, we perturbed the proteostasis of replacement lines with the same mild dosage of GdA before sporulation and then measured sporulation rates (see “Methods”). Two replacement lines (7L and 13L) were excluded from this assay since they already exhibited severe sporulation defects due to mitochondrial-nuclear incompatibility28. Among the nine tested replacement lines, six presented significantly higher sensitivity to the perturbation than the parental lines and exhibited significantly reduced sporulation (Fig. 2c). Consistent with our protein aggregation data, the 8+15L and 16L lines showed the most severe defects in both mitosis and meiosis among all tested replacement lines. These results demonstrate that although hybrid cells carrying foreign chromosome(s) only display weak or mild incompatibility, the fitness defects can be easily aggravated by mild environmental perturbations that are tolerable to wild-type cells.

Levels of proteotoxicity are correlated with the number of protein complexes on replaced chromosomes

In our replacement lines, many orthologous proteins expressed from the foreign chromosomes substitute the functions of the respective endogenous proteins. One possible cause for the proteotoxic stress we observed is that the proteostasis-related proteins in S. bayanus var. uvarum are less efficient and more sensitive to environmental perturbations (described as the “weak allele” hypothesis). Alternatively, the proteotoxicity may be induced by misfolded protein complex subunits that have dissociated from unstable chimeric complexes or that failed to assemble48. In a previous study that examined the formation of six stable protein complexes in hybrids between Saccharomyces yeasts (sensu stricto), three of the six formed only species-specific complexes49, indicating that species-specific interactions between complex subunits can evolve quickly, even among closely related species. Since the complex subunits encoded by the replaced chromosome may require subunits encoded on other chromosomes to form protein complexes, any level of species-specific interactions would likely reduce the assembly efficiency or stability of chimeric complexes, resulting in an excess of unassembled or misfolded subunits (described as the “unstable complex” hypothesis).

To test these two hypotheses, we crossed 8+15L and 16L haploid cells with S. bayanus var. uvarum to generate heterozygous diploid cells and examined their fitness. A complete set of the S. bayanus var. uvarum chromosomes were present in these hybrid diploids, so genes on the replaced chromosomes were homozygous. These hybrid strains should still be sensitive to mild Hsp90 perturbations if the replaced chromosomes carry a “weak allele” of proteostasis-related genes. In contrast, if proteotoxicity is caused by unstable protein complexes, the fitness defect should be partially alleviated since the relative abundance of the unstable complexes is reduced. We observed significant rescue of fitness defects when Hsp90 was compromised (50 μM GdA, 23 °C), suggesting that the fitness defect of replacement lines is not due to a “weak allele” from the S. bayanus var. uvarum chromosomes (Fig. 2d). This is further supported in F1 hybrid diploids, where the fitness defect in the presence of one copy of interaction partners was significantly higher than the parental diploids, but reduced considerably when compared to 8+15L and 16L (Supplementary Fig. 2c).

Individual complex subunits often interact with more than one partner in protein complexes. Therefore, compromised interactions between different subunits can result in complex epistatic effects, as observed in complex incompatibility7. Moreover, the loading effect leading to proteotoxicity is cumulative even though the contribution of each unassembled (or misfolded) complex subunit may be mild. We tested if the level of proteotoxic stress in each replacement line is correlated with the number of unique protein complexes encoded by the replaced chromosome (see “Methods”)8. Indeed, we found that fitness defects under GdA treatment (50 μM) at 23 °C were correlated with the number of protein complexes (ρ = 0.68, p = 0.025, Spearman’s rank correlation, Fig. 2e and Supplementary Fig. 4b). Moreover, we wanted to determine if protein complexes encoded by replaced chromosomes were associated with the formation of protein aggregates. Indeed, percentages of Hsp104 foci-containing cells in replacement lines at 180 min were also significantly correlated with chromosomal contributions to the formation of protein complexes (ρ = 0.68, p = 0.03, Spearman’s rank correlation, Fig. 2f). It is interesting to note that the fitness defect is not directly proportional to the length of replaced chromosomes. For example, Chromosome 16 is shorter than Chromosome(s) 6+10, 7, 12, or 13, but 16L is more defective than 6+10L, 7L, 12L, or 13L. We also tested whether the fitness defect is correlated with the ratio of proteins in a complex divided by total proteins on the replaced chromosome. No significant correlation (p > 0.05) was observed between this ratio and fitness defects under GdA or Hsp104 foci. Moreover, several functions and processes related to proteostasis are specifically enriched on protein complex subunits encoded on Chromosomes 16, 8, and 15 (Supplementary Data 3). This suggests that the fitness defect is driven by the nature of protein complex subunits encoded on the replaced chromosome, rather than the number of genes on a chromosome.

Multiple protein complexes are destabilized in the most defective replacement lines, 16L and 8+15L

Our genetic experiments and protein complex correlation analysis suggested that unstable chimeric protein complexes are the major cause of proteotoxic stress underlying complex incompatibilities. To directly test the “unstable complex” hypothesis, we characterized native protein complex formation in the most defective replacement lines, 8+15L and 16L, and compared it with the parental Sc strain. We predicted that the replacement lines should display more unstable protein complexes than Sc and these would be enriched with subunits encoded by replaced Chromosomes 8, 15, and 16.

We monitored the formation of soluble protein complexes in yeast cells using native size-exclusion chromatography (SEC) followed by mass spectrometry (Fig. 3a, see Methods for details). The elution patterns of subunits in a protein complex through SEC represent the stability of the protein complex (Fig. 3a and Supplementary Fig. 3). For a stable protein complex, most subunits are expected to elute as a few continuous high-molecular-weight fractions. However, if the protein complex becomes unstable, the subunits dissociate and elute as low-molecular-weight fractions, resulting in a different elution pattern (Fig. 3). To accurately quantify the elution pattern difference (EPD), we incorporated stable isotope labeling by amino acids in cell culture (SILAC) into our experimental procedures to enable an independent direct comparison of the top two defective replacement lines (8+15L and 16L) to the parental strain (Sc) (Fig. 3a and Supplementary Fig. 5).

Fig. 3: Multiple protein complexes are destabilized in 8+15L and 16L cells.
figure 3

a Workflow of the SEC-based analysis of the formation of protein complexes. SEC, SILAC, and mass spectrometry were combined to compare the formation of protein complexes in Sc and 16L cells (see Methods for details). The dotted red boxes highlight a stable complex in the Sc line but partially disassembled in 16L cells. The subunits of this complex are expected to elute in different fractions for the Sc and 16L cells, leading to distinct elution patterns. b The protein complexes with subunits on Chromosomes 8+15 and Chromosome 16 are less stable than non-Chromosome 8+15 and 16 subunit-containing complexes. Elution pattern difference (EPD) values were calculated for every protein in Sc-heavy/Sc-light (Sc/Sc), Sc-heavy/16L-light (Sc/16L), and Sc-heavy/8+15L-light (Sc/8+15L) sets (also see Supplementary Fig. 5). The distributions of EPD values for different groups of proteins of the Sc/Sc, Sc/16L and Sc/8+15L sets are shown in box plots and the protein numbers of each group are 2342 (total), 1359 (complex), 983 (non-complex), 735 (Chr16 complex), 624 (non-Chr16 complex), 866 (Chr8+15 complex), and 493 (non-Chr8+15 complex). Distributions with the same letter (above each boxplot) are not significantly different from each other (Dunn’s pairwise tests with Bonferroni correction, p-values > 0.05, see Supplementary Data 11 for the p-values of all Dunn’s pairwise tests). n.s.: not significant. c Twenty-three protein complexes are significantly destabilized in 8+15L and 16L cells. A protein complex was defined as unstable if the EPD values of the complex subunits in the Sc/16L or Sc/8+15L sets were significantly higher than those in the Sc/Sc set. Here, 7 protein complexes are destabilized in both 8+15L and 16L cells (black), 3 unstable complexes are specific to 16L (green) and 13 complexes are unstable only in 8+15L (blue). Also, see Supplementary Fig. 6. Boxplots indicate median (middle line), 25th and 75th percentile (box), and min and max (whiskers). The summary of boxplots is provided in Supplementary Data 12 and source data are provided as a Source Data file.

We identified a total of 2432 proteins common to the experimental Sc-heavy/8+15L-light (Sc/8+15L, 3432 proteins), Sc-heavy/16L-light (Sc/16L, 2742 proteins) and control Sc-heavy/Sc-light (Sc/Sc, 2856 proteins) sets. Among them, 1359 proteins were subunits of 463 previously identified protein complexes (Supplementary Data 4 and 5)8. The control Sc/Sc set did not display a significant difference in EPD values for complex and non-complex proteins, nor between any other groups (Fig. 3b; Kruskal–Wallis H test, H(6) = 0.42, p = 0.65), indicating that our experimental procedures were robust. In contrast, there was a significant difference in the EPD values between different groups in the experimental Sc/8+15L and Sc/16L sets (Fig. 3b; Kruskal–Wallis H test, 8+15L: H(4) = 112.16, p = 2.2 × 10−16 and 16L: H(4) = 67.12, p = 9.2 × 10−14). Pairwise tests between individual groups showed that the EPD values (median) of complex subunits were significantly higher than non-complex proteins (Dunn’s pairwise tests with Bonferroni correction, adjusted p-values = 9.9 × 10−12 (Sc/8+15L) and 5.9 × 10−9 (Sc/16L)). To rule out the possibility that proteins encoded by Sb chromosomes might have different elution patterns contributing to observed EPDs, EPD values were recalculated after removing the proteins encoded by replaced chromosomes. We observed similar results (Supplementary Fig. 6). These data provide direct evidence that many protein complexes had become destabilized in 8+15L and 16L cells.

If the complex instability is due to incompatibility between S. cerevisiae and S. bayanus var. uvarum protein subunits (i.e., species-specific interactions), we anticipated that the complexes having subunits encoded by foreign Chromosomes 8, 15, and 16 should be less stable than those lacking such subunits. Indeed, Chr8+15 and Chr16 complex proteins (i.e., proteins from the complexes containing subunits from Chromosomes 8+15 and 16, respectively) presented significantly higher EPD values compared to non-Chr8+15 and non-Chr16 complex proteins (i.e., proteins from the complexes lacking any Chromosome8+15-encoded and 16-encoded subunit, respectively) (adjusted p-values = 1.0 × 10−10 (Sc/8+15L) and 3.1 ×10−4 (Sc/16L), Fig. 3b, Supplementary Fig. 6a, b). Furthermore, there was no significant difference between the EPD values of non-Chr8+15 and non-Chr16 complexes and non-complex proteins, respectively (adjusted p-values = 1 (Sc/8+15L) and 0.06 (Sc/16L)). Thus, the presence of foreign complex subunits was likely the primary contributory factor for the instability of protein complexes in 8+15L and 16L cells.

In order to understand the general features of complex incompatibility, we endeavored to identify protein complexes that were destabilized as a whole rather than in few individual subunits. A protein complex was deemed unstable in the 8+15L or 16L cells if the EPD values of the complex subunits in the Sc/8+15L or Sc/16L sets were significantly higher than those in the Sc/Sc set (Wilcoxon signed-rank tests, Benjamini–Hochberg FDR corrected p-values < 0.05, see “Methods” for details). While 20 complexes were unstable in Sc/8+15L, 10 protein complexes were unstable in Sc/16L (Fig. 3c and Supplementary Data 6a). Among them, complexes with subunits encoded by the replaced chromosomes (i.e., the chimeric protein complexes) were prone to be destabilized (9 out of 10 in 16L and 19 out of 20 in 8+15L, p-value < 0.05, Fisher’s exact test, Supplementary Data 6a). We also performed the complex stability analysis after removing the complex subunits encoded by replaced chromosomes and observed similar results (Supplementary Fig. 6c, d and Supplementary Data 6b). Interestingly, most of the unstable chimeric complexes are involved in basic cellular functions, including transcription, translation, and respiration. These data further support that species-specific interactions between complex subunits can evolve rapidly, even within complexes having essential cellular functions.

Unstable chimeric protein complexes have lower soluble protein abundances in 8+15L and 16L cells

Disassembly of protein complexes can result in degradation or insoluble aggregate formation of dissociated protein subunits50,51. If cells contain many unstable protein complexes, their systems governing proteostasis may be overwhelmed and further compromised. We examined if there was a decrease in the abundance of the destabilized protein complex subunits in 8+15L and 16L cells. Further analysis of our SILAC data revealed significantly lower levels of complex subunit proteins compared to non-complex proteins in 8+15L and 16L cells (Fig. 4a and Supplementary Fig. 7). Consistent with our EPD data, the subunit abundance of Chr8+15 and Chr16 complexes was significantly lower than for non-Chr8+15 and non-Chr16 complexes (Fig. 4a, Supplementary Fig. 7, and Supplementary Data 7).

Fig. 4: Proteotoxic stress in the replacement lines can be partially relieved by up-regulating the protein degradation machinery.
figure 4

a, Protein abundances of complexes having subunits encoded on Chromosome 16 are significantly reduced in 16L cells and Chromosomes 8+15 are reduced in 8+15L cells when compared to non-Chr16/non-Chr8+15 complexes, respectively (also see Supplementary Fig. 7). The distributions of log2-transformed SILAC ratios of proteins in different groups are shown in boxplots. Distributions with the same letter (above each boxplot) are not significantly different from each other (Dunn’s pairwise tests with Bonferroni correction, p > 0.05, see Supplementary Data 11 for the p-values of all Dunn’s pairwise tests). n.s.: not significant. b Growth defects of the 16L and 8+15L lines are partially rescued by deleting the UBP6 gene. The ubp6Δ mutants exhibit enhanced proteasomal degradation activity, thereby facilitating the removal of destabilized complex subunits. Diploid cell lines were grown in YPD with 50 μM Geldanamycin (GdA) at 23 °C and their doubling times were compared (n = 3). c Protein aggregate load of the 16L and 8+15L lines is alleviated in ubp6Δ mutants. The Hsp104 aggregate data were obtained after the cells had been shifted from 23 to 37 °C for 180 min (n = 8; SEM, N ≥ 500 cells per time-point). d Growth defects of the 16L and 8+15L lines are aggravated in the heterozygous RPN6/rpn6Δ mutants. Rpn6 is an essential component of proteasomes, and proteasomal degradation activity is mildly compromised in RPN6/rpn6Δ mutants. Diploid cell lines were grown in YPD at 28 °C and their doubling times were compared (n = 3). The data are presented as mean values +/− SEM. ***: p-value < 10−3, one-sided Student’s t-test. Boxplots indicate median (middle line), 25th and 75th percentile (box), and min and max (whiskers). The summary of boxplots is provided in Supplementary Data 12. Source data and detailed statistical information are provided as a Source Data file.

Since the replaced chromosomes also impacted global gene expression (Fig. 1a and Supplementary Data 1), we tested whether the observed reduction of complex subunit abundance was simply due to decreased gene expression by computing the Spearman’s correlation coefficient between the protein (SILAC-ratios) and transcript levels. We found that the protein and transcript levels were least correlated in both Chr16 and Chr8+15 complex groups compared to the other groups (Supplementary Data 8). Fisher’s r-to-z transformation, which can compare the significance in the difference between correlation coefficients52, also showed that the Spearman’s correlation coefficients in the Chr8+15 complex group (ρ = 0.24) and the Chr16 complex group (ρ = 0.19) were significantly lower than that of the non-Chr8+15 complex group (ρ = 0.35) and non-Chr16 complex group (ρ = 0.45), or the non-complex group (8+15L: ρ = 0.39, 16L: ρ = 0.46) (p-values < 0.05). On the other hand, we did not find the correlation coefficients for the non-complex and non-Chr8+15 complex groups or non-complex and non-Chr16 complex groups respectively to be significantly different (p-value = 0.31). These data suggest that gene expression change is not the only factor contributing to the observed reduction of subunit protein abundance and translational or post-translational regulation is involved.

To further validate our proteomics data, we examined one of the unstable complexes, RNA polymerase III, which was identified in both 8+15L and 16L and could be biochemically purified from total cell extracts53. We calculated a close to two-fold reduction (49%) in the abundance of RNA polymerase III subunits in 16L cells, even though the same total amounts of protein lysates from 16L and Sc cells were used for the pull-down experiment (Supplementary Fig. 8a). Together, these data indicate that once subunits have dissociated from destabilized complex, they are quickly degraded or form insoluble aggregates.

Up-regulating the protein degradation machinery alleviates the fitness defects of replacement lines

Since individual incompatible loci only contribute to a small proportion of the fitness defects, it is almost impossible to confirm their effect specifically. Instead, we tested whether the fitness defect could be relieved by downregulating global proteotoxicity. The ubiquitin-proteasome machinery is a primary pathway known to regulate unbalanced multi-protein complexes. Cells exhibit intrinsic proteotoxic stress when proteasomes are overwhelmed by extensive perturbations51,54. In ubiquitin-mediated proteasomal degradation, multiple ubiquitin molecules are covalently linked to candidate substrates and act as recognition motifs for 26S proteasomes55. Ubp6, a ubiquitin-specific protease, has a dual role in ubiquitin recycling and regulation of proteasomal degradation. Proteasomal degradation activity is accelerated in the absence of Ubp656,57,58. We postulated that if excess amounts of destabilized complex proteins in the replacement lines overburdened proteasomes to induce fitness defects, an absence of UBP6 should allow the proteasomes to degrade these proteins more efficiently and improve cell proliferation by the replacement lines. We used ubp6Δ mutants of the 8+15L and 16L lines exhibiting the greatest proliferative defects to perform growth assays under the treatment of 50 μM GdA at 23 °C. As anticipated, the fitness defect was indeed rescued in our ubp6Δ mutant replacement lines (Fig. 4b), suggesting that proteotoxic stress due to overburdening of proteasomes was responsible for the growth defects. In addition, we performed protein aggregation assays in both mutant lines and observed that the number of cells harboring Hsp104-mCherry foci decreased significantly in the ubp6Δ mutants (Fig. 4c and Supplementary Fig. 8b), further confirming that ubp6 deletion partially relieved intrinsic proteotoxic stress in the 8+15L and 16L cells.

If 26S proteasomes are the most crucial degradation machinery controlling protein complex homeostasis in our replacement lines, we expected hybrid cells to display severe growth defects when their proteasomes are compromised, even under normal growth conditions. We constructed heterozygous deletion mutants of Rpn6, an essential lid component of the 26S proteasome, from our 8+15L and 16L lines, and measured their fitness at 23 and 28 °C. Heterozygous RPN6/rpn6Δ mutation only partially compromised the activity of proteasomes and had mild impacts on the fitness of Sc cells at both temperatures. In contrast, the RPN6/rpn6Δ mutants of the 16L and 8+15L lines revealed a significant fitness defect at 28 °C (Fig. 4d and Supplementary Fig. 8c).

Lastly, we confirmed that the proteotoxic stress observed in the replacement line was not due to an ineffective proteasome system. Proteasomal activity assays showed that the activity of endogenous 26S proteasomes in 16L cells did not differ from that of Sc cells (Supplementary Fig. 8d). Together, our results demonstrate that destabilized protein complexes in hybrid cells often increase the burden of proteasomes, a key regulator of proteostasis. Depending on the number and abundance of chimeric complexes, that overburdening results in differential levels of hybrid incompatibility.