W303 studies reported a transcriptomic signature of aneuploidy independent of amplified chromosome identity, but this is not seen in wild aneuploid strains (Torres et al., 2007; Hose et al., 2015). To test dependence on Ssd1, we followed transcriptomes of natural aneuploids and their ssd1Δ derivatives, with or without extra chromosomes. We identified 861 genes with altered expression in both YPS1009_Chr12 ssd1Δ and NCYC110_Chr8 ssd1Δ mutants compared to their isogenic wild-type aneuploids (false discovery rate, FDR < 0.05, Figure 2A). Induced genes showed little change in euploid YPS1009 ssd1Δ but were up-regulated when SSD1 was deleted in the context of Chr12 amplification. NCYC110 showed similar trends, except that in this strain we observed a weak expression signature in euploid ssd1Δ cells that was exacerbated when Chr8 was amplified. Repressed genes included rRNA and tRNA processing and cytokinesis factors, whereas induced genes encompassed the environmental stress response (ESR Gasch et al., 2000), oxidoreductases, carbohydrate and energy metabolism, and genes involved in mitochondrial degradation (p<1e-4, hypergeometric test). This response effectively recapitulates the expression signature seen in W303 aneuploids (Torres et al., 2007, Figure 2A). The response was exacerbated with increasing DNA content in W303 carrying multiple extra chromosomes (Figure 2A). Thus, the previously reported aneuploidy transcriptome signature results from defective Ssd1 function, independent of affected chromosome, exacerbated with additional DNA content, and with some strain-specific nuances.

We wondered if other aneuploidy signatures seen in W303 could be explained by defective Ssd1. In addition to growth delay, aneuploid W303 strains reportedly exhibit larger cell size, altered cyclin Cln2 abundance, delayed G1/S progression, metabolic defects, and signatures of proteotoxicity including protein aggregation and a defect degrading misfolded protein (Torres et al., 2007; Oromendia et al., 2012; Thorburn et al., 2013). While these seminal studies generated important information on aneuploidy toxicity in a sensitized strain, many phenotypes likely result from defective Ssd1. YPS1009_Chr12 ssd1Δ grows slower (Figure 1), produces ~33% higher optical versus cell density indicating larger size, and displays metabolic defects encompassing defective respiratory growth (see Figure 4). We confirmed that Ssd1 binds many cell-cycle transcripts including CLN2 (Supplementary file 2), which is translationally regulated by Ssd1 and explains altered G1/S progression (Ohyama et al., 2010).

Loss of SSD1 also explains proteotoxicity observed for aneuploid strains. First, we followed accumulation of human Von Hippel Lindau (VHL) protein upon over-expression. Because VHL cannot fold in the absence of its interacting proteins in yeast, accumulation of VHL-GFP foci reflects misfolded protein that has yet to be cleared by the proteasome (McClellan et al., 2005; Kaganovich et al., 2008). We found that significantly more YPS1009_Chr12 ssd1Δ cells accumulated VHL-GFP foci compared to wild type (Figure 2B). We also followed the protein disaggregase Hsp104, which binds misfolded and aggregated proteins in discrete protein quality-control centers (Kaganovich et al., 2008; Chernova et al., 2017). Wild-type aneuploids showed no obvious difference in the number of Hsp104-GFP foci compared to euploids (Figure 2C and Figure 2—figure supplement 1), indicating that gross protein aggregation is not a universal feature of aneuploid yeast. However, mutant aneuploids lacking SSD1 showed a higher proportion of cells with Hsp104-GFP foci, and more foci within those cells, compared to wild-type aneuploids (Figure 2C). Ssd1 was previously implicated in protein homeostasis after heat shock (Mir et al., 2009), but our results demonstrate that chromosome amplification in the absence of other stresses is enough to provoke misfolding in ssd1Δ cells. Together, these results show that myriad signatures of W303 aneuploidy can occur due to defective Ssd1.

Ssd1 mutants could be sensitive to specific genes on the amplified chromosomes, or they could have a generalized sensitivity to the burden of extra DNA/protein. To distinguish these models, we transformed YPS1009 strains with a barcoded, high-copy gene over-expression library and measured relative fitness costs after 5 generations of growth (see Materials and methods). Both the euploid and aneuploid ssd1Δ mutants were highly sensitive to the library (Figure 2D): genes that were deleterious in wild type were toxic in the mutant, while many genes with neutral effect in parental strains were deleterious in the absence of SSD1. We cannot exclude a defect maintaining the high-copy 2-micron plasmid (indeed, the mutant cells do not grow well with the empty vector, and Ssd1 is already implicated in plasmid maintenance Uesono et al., 1994). Nonetheless, both the euploid and aneuploid ssd1Δ mutants are highly sensitive to the 2-micron overexpression library.

We focused on YPS1009 strains to elucidate Ssd1 function in aneuploidy tolerance. Revisiting the YPS1009_Chr12 ssd1Δ transcriptome revealed broader induction of genes, including mRNAs whose proteins localize to subcellular compartments such as mitochondria, ER, vacuole, peroxisome, plasma membrane, and the cell wall (p<1e-4, hypergeometric test). Many of these organelles functionally and physical interact (Scorrano et al., 2019), raising the possibility of broader inter-organelle issues. Consistent with this notion, the ssd1Δ aneuploid also showed transcriptional signatures of the unfolded ER-protein response (Travers et al., 2000) and mitochondrial protein import stress (Weidberg and Amon, 2018) (see Materials and methods).

To test if Ssd1 binds a broader set of mRNAs, we sequenced RNAs recovered from Ssd1 immunoprecipitation (see Materials and methods). The 286 associated mRNAs (FDR < 0.05, Supplementary file 2) were heavily enriched for transcripts encoding cell-wall proteins as expected, but the group was also enriched for cell-cycle regulated mRNAs (including cyclins CLN2 and CLB2/4/5) and those involved in budding, RNA metabolism, and sterol transport (p<1e-4). Myriad other functions were also represented, such as proteins in chromatin regulation, transcription, lipid biogenesis, endocytosis, protein homeostasis, and mitochondrial function, some previously noted (Jansen et al., 2009; Hogan et al., 2008). Interestingly, Ssd1 also bound mRNAs encoding osmotic-response regulators (SLN1, SHO1, MSB2, HOT1), notable since osmotic stress was recently implicated in aneuploidy responses of a different lab strain (Tsai et al., 2019). Ssd1-bound mRNAs were not enriched for those encoded on the amplified chromosome, nor transcripts disproportionately expressed compared to DNA content (Hose et al., 2015), suggesting that the mechanism of aneuploidy tolerance is not through modulation of amplified-gene expression. Most (60%) bound transcripts were not differentially expressed in the aneuploid mutant, downplaying a generalizable role in regulating mRNA abundance. In turn, most mRNAs differentially expressed in the aneuploid mutant are not Ssd1-bound, suggesting widespread secondary responses to the primary defect(s).

Since Ssd1 regulates translation via direct RNA binding (Jansen et al., 2009; Wanless et al., 2014; Kurischko et al., 2011b), we next used quantitative proteomics to measure effects on the cellular proteome. 301 of 3906 measured proteins were more abundant in the aneuploid mutant versus wild type (FDR < 0.05, Figure 3A). Many emerge from induced transcripts; however, a large fraction of proteins was elevated beyond mRNA differences, including cell-wall proteins, nuclear-encoded mitochondrial proteins (p=6e-4, hypergeometric test), and others. Many of these proteins were also elevated in the euploid mutant without significant mRNA changes (Figure 3A). Although there was no enrichment for proteins encoded by Ssd1-bound transcripts, several elevated proteins emerge from Ssd1 targets, including cell-wall transcripts and several nuclear transcripts encoding mitochondrial proteins. For example, UTH1 encoding a mitochondrial protein linked to aging is bound by Ssd1 in our and other studies and is known to be translationally regulated by direct Ssd1 binding at specific mRNA locations (Wanless et al., 2014; Camougrand et al., 2004; Camougrand et al., 2000). Although protein induction was evident in euploid ssd1Δ cells, the defect was exacerbated by Chr12 amplification, with ~4X more Uth1 protein despite little difference in mRNA (Figure 3B). Other mitochondrial proteins emanating from Ssd1-bound mRNAs were also significantly elevated in the ssd1Δ aneuploid, including several mitochondrial ribosomal proteins (Figure 3B). Thus, Ssd1 affects the proteome of aneuploid cells, including from bound transcripts.

Our past work revealed that wild aneuploid strains down-regulate many nuclear encoded mitochondrial transcripts, a response also seen in Down syndrome models (Helguera et al., 2013; Liu et al., 2017), hinting that mitochondrial regulation is important for tolerating chromosome amplification (Hose et al., 2015). In the current work, multiple lines implicated mitochondrial effects in ssd1Δ aneuploids. To explore this, we tested mitochondrial function in YPS1009 strains. We found a synergistic defect between SSD1 deletion and aneuploidy when cells experienced mitochondrial stress. YPS1009_Chr12 ssd1Δ, NCYC110_Chr8 ssd1Δ (Figure 4A), and aneuploid W303 (Hose et al., 2015) were all sensitive to non-fermentable acetate, beyond what is expected from the additive effects of aneuploidy and reduced respiratory growth rate. This demonstrates Ssd1-dependent respiratory dysfunction autonomous of strain background or affected chromosome. YPS1009_Chr12 ssd1Δ cells were also susceptible to sub-lethal doses of carbonyl-cyanide 3-chlorophenylhydrazone (CCCP), which uncouples mitochondrial membrane potential (Figure 4B). Notably, aneuploid ssd1Δ cells were no more sensitive than expected to cell wall or ER stress (Figure 4—figure supplement 1), indicating a specific interaction with mitochondrial challenge. In the process of this work, we discovered that ssd1Δ aneuploid cells also showed a striking difference in mitochondrial morphology. Wild-type YPS1009_Chr12 grew well but harbored many globular mitochondria compared to the euploid tubular shape (Figure 4C). Although the impact of this morphology is not clear, YPS1009_Chr12 ssd1Δ displayed significant differences, including more tubular forms and increased mitochondria fragmentation (Figure 4D). The West African aneuploid did not display globular mitochondria but did display dysfunction (Figure 4A). Thus, ssd1Δ cells show numerous signs of mitochondrial dysfunction compared to wild-type aneuploid cells.

We wondered if Ssd1 plays a direct role in mitochondrial function, perhaps by localizing nuclear-encoded mitochondrial mRNAs as it does cell-wall mRNAs. We first scored Ssd1 localization by centrifugation-based cellular fractionation. Ssd1 was reproducibly recovered in the organelle-enriched fraction that was depleted of cytosolic actin but enriched for markers of mitochondria (Cox2), ER (Dpm1), and vacuole (Vph1, Figure 5A), which themselves interact. Attempts to separate the components by immunoprecipitation of Ssd1 from the fractions were not successful. We next followed cellular localization of MMR1 transcript, encoding a bud-mitochondria localized protein involved in mitochondrial inheritance (Itoh et al., 2004), by single-molecule RNA FISH (smFISH, Figure 5B). In most wild-type cells, mother-encoded MMR1 was directed to the nascent bud, before the nucleus migrated. Although scoring precise differences in mRNA patterns was challenging, YPS1009_Chr12 ssd1Δ cells displayed twice as many buds lacking MMR1 as wild-type aneuploids (Figure 5C, p<0.02, Fisher’s exact test). In addition, the mutant showed double the cells lacking mitochondria as indicated by Rhodamine staining (Figure 5D), consistent with a defect in mitochondrial inheritance. It is possible that the mutant suffers from delayed dynamics, rather than fully aberrant localization in individual cells. Other nuclear-encoded mitochondrial mRNAs bound by Ssd1 were too abundant to follow by smFISH and will require further delineation. Nonetheless, together our work shows that Ssd1 associates with organelle fractions that include mitochondria (Figure 5A), binds several nuclear-encoded mitochondrial transcripts (Supplementary file 2), and can influence abundance of mitochondrial proteins (Figure 3B) or localization patterns of bound transcript (Figure 5B–C), consistent with the requirement of SSD1 for proper mitochondrial function in aneuploid cells (Figure 4).

A remaining question is why Ssd1 dependence and mitochondrial dysfunction are more severe in aneuploids. We reasoned that underlying ssd1Δ defects are exacerbated by effects of chromosome amplification. One candidate is proteome stress that may emerge from over-production of proteins from the amplified chromosome in the absence of Ssd1-dependent translational silencing, which could tax the proteostatic buffering capacity specifically in the mutant (Donnelly and Storchová, 2015; Oromendia and Amon, 2014; Oromendia et al., 2012). Many recent studies have revealed a connection between mitochondrial function and cytosolic proteome stress: defects in mitochondrial protein import induce cytosolic proteostatic defense mechanisms, and misfolded cytosolic proteins interact with and can even be cleared by mitochondria (Qureshi et al., 2017; Ruan et al., 2017; Wrobel et al., 2015). Furthermore, mitochondrial defects and cytosolic proteostasis stress co-emerge in neurological syndromes, aging, and aneuploidy (Oromendia and Amon, 2014; Helguera et al., 2013; Ruan et al., 2017; Franco-Iborra et al., 2018; Kauppila et al., 2017).

To test the model that synergistic dysfunction underlies aneuploidy sensitivity in ssd1Δ aneuploid yeast, we applied nourseothricin (NTC), among the aminoglycoside drugs that induce mistranslation and protein misfolding (Ling et al., 2012). We confirmed that NTC treatment increased the number of Hsp104-GFP foci in the aneuploid wild type, and discovered that CCCP produced an even stronger effect even though wild-type aneuploids grew well in the drug (Figures 6 and 4B). Wild-type aneuploids were slightly sensitive to NTC, but the mutant was significantly more sensitive, beyond the expected additivity of aneuploidy and NTC response, revealing a synergistic defect induced by the drug in combination with aneuploidy (Figure 6A). The NTC sensitivity suggests that wild-type aneuploids with full length SSD1 can largely buffer proteostasis upon chromosome amplification but may exist near capacity. Mitochondrial defect, protein over-abundance, and mislocalized transcripts/proteins resulting from SSD1 deletion may simply push cells over the edge (see Discussion).

This raised an important prediction: if synergistic defects in mitochondrial function and proteostasis sensitize ssd1Δ cells to chromosome amplification, then combinatorial drug treatment to mimic these defects should selectively target wild-type aneuploids. In fact, this was the case: wild-type euploid and aneuploid cells tolerated short-term CCCP and NTC individually, but when combined aneuploid growth was significantly delayed beyond the euploid strain and the expectation of additive effects (Figure 6C). Longer-term combinatorial drug treatment limited growth of euploid YPS1009 but selectively blocked proliferation in YPS1009_Chr12 (Figure 6D). The effect was persistent across strains and chromosome amplifications: combinatorial treatment halted over-night growth of W303 with duplications of Chr8, Chr10, and Chr16 and NCYC110 carrying extra Chr 8 (although in this strain CCCP was actually protective against NTC toxicity in the euploid cells at the doses used)(Figure 6D).

Strains used in this study are listed in the Resource Table. W303_Chr8, W303_Chr8-Chr15, and W303_Chr8-Chr10-Chr16 were generated using the method of Chen et al. (2012), passaging 16 generations in 20 μg/mL radicicol (A.G. Scientific) and plating on 8 or 16 μg/mL fluconazole to select for Chr8 aneuploidy. Karyotype was determined by array-comparative genomic hybridization and sequencing. W303 strains shown in Figure 1E were grown in SC-his + G418 to maintain marked copies of Chr12 (or corresponding markers in the otherwise isogenic wild type Torres et al., 2007). In general, deletions were generated by homologous recombination of relevant makers (e.g. KAN-MX or HIS3) into the designated locus, followed by diagnostic PCR to confirm correct integration and absence of the target gene. Because ssd1Δ cultures lose extra chromosomes (perhaps simply due to overtaking of the culture by stochastic euploid revertants), deletions were generated in wild-type strains that were then crossed to YPS1009_Chr12 ssd1Δ, followed by tetrad dissection to isolate aneuploid spore clones with desired genotypes. In all cases, aneuploidy was confirmed and periodically checked through diagnostic qPCR of one or two genes on the affected chromosomes (AAT1 and SDH2) normalized to a single-copy gene elsewhere in the genome (ERV25 or ACT1) – normalized ratios close to two reflect gene duplication, and ratios between 1.2–1.8X indicated partial loss of aneuploidy in the cell population. GFP-tagged genes were generated by integrating a GFP-ADH2terminator-HIS3 cassette (Huh et al., 2003) via homologous recombination into strain series AGY1504-1507 in which HIS3 was previously deleted by replacement with KAN-MX marker. In the case of SSD1-GFP strains, a cassette consisting of GFP followed by the native SSD1 terminator, 337 bp downstream of SSD1^YPS1009, was generated by PCR sewing with the NAT-MX marker. In all cases, cloned or tagged genes were confirmed by sequencing. SSD1^YPS1009 plus 1000 bp upstream and 300 bp downstream was cloned into a pRS-derived CEN plasmid for complementation. Because YPS1009_Chr12 ssd1Δ cannot tolerate the 2-micron plasmid, the VHL-GFP gene plus promoter and terminator sequences were cloned from pESC-LEU-GFP-VHL (ADDGENE #21053) into a pRS-derived CEN plasmid. Human VHL cannot fold without cofactors but is typically cleared from cells through proteasome activity (McClellan et al., 2005). Accumulation of VHL-GFP foci is generally taken as an inability to clear misfolded proteins.

Unless otherwise noted, strains were cultured for ~3 generations into log phase in rich YPD medium at 30°C, with the exception of microscopy experiments where cells were grown in low-fluorescence synthetic-complete medium and imaged live. Induction of VHL-GFP was performed by growing cells in YP with 2% raffinose + 2% galactose for 4 hr. Wild-type strains shown in Figure 6C–D were grown over-night in log-phase before addition of 1 ug/mL nourseothricin (Werner BioAgents, Jena, Germany) or 25 uM CCCP (Millipore-Sigma, St. Louis, MO). Growth rates were calculated by exponentially fitting changes in optical density. Relative final OD in Figure 6D was measured in biological triplicate after 24 hr growth of YPS1009, NCYC110, and W303 strains exposed to 25 uM CCCP with 0.5 ug/mL (NCYC, W303) or 1 ug/mL (YPS1009) NTC, respectively. Aneuploidy was periodically verified through diagnostic qPCR as described above. Expected growth rates under an additive model were estimated based on the fold-defect in one condition (e.g. aneuploidy versus euploidy) multiplied by the fold-defect in a second condition (e.g. NTC sensitivity in the euploid); significant differences in observed versus expected data were assessed with replicate-paired T-tests. Unless otherwise noted, all studies used at least biological triplicates with data represented as the average and standard deviation (except count data in Figures 2B–C, 4D, 5C–D and 6B in which average and standard error of the mean across biological replicates is shown).

Haploid strain AGY736 (YPS1009_Chr12 Mat alpha ho::HYGMX) was crossed with AGY768 (W303_Chr12 Mat a) and the resulting diploid sporulated and dissected evenly on agar plates. Colony diameter after 72 hr was scored for 208 spores; 76 spores ranking in the smallest ~40% of the distribution were scored for their propensity to lose Chr12 within 20 culture generations of growth: each spore was passaged for 2 days in liquid YPD, after which genomic DNA was isolated and Chr12 abundance scored by diagnostic PCR as described above. Loss of Chr12 signal was taken as aneuploidy sensitivity (20 spores, Pool A1) whereas cells that maintained Chr12 signal were taken as enriched for aneuploid tolerant cells (40 spores, Pool B1). Spore sp100 (Mat alpha ADE2 HIS3 LEU2 trp- URA3) that was prototrophic for influential markers was selected, its aneuploidy status verified by qPCR, and it was backcrossed to AGY735 (YPS1009_Chr12 Mat a ho::HYGMX). 37 segregants were scored only for their propensity to lose aneuploidy after 2 days of passage, generating an aneuploidy-sensitive pool (10 spores, Pool A2) and a pool enriched for aneuploidy-tolerant strains (25 spores, Pool B2). A control cross of euploid hYPS1009 X euploid W303 was generated and phenotyped for colony size as above. 46 and 50 spores were taken as ‘small’ (colony diameter <437 square pixels) or ‘large’ (colony >591 square pixels) for Pool D and Pool F, respectively.

Each clone was grown to saturation, an equal volume of each culture pooled appropriately, and genomic DNA isolated (Qiagen, Germantown, MD) from ach pool. Pooled genomic DNA was sequenced using NEBNext Ultra DNA Library Prep Kit for Illumina on an Illumina HiSeq 2000 to an average of 20M 100 bp reads per pool. To avoid potential mapping biases, an artificial reference genome was created where single nucleotide polymorphisms (SNPs) between the two parental genomes were substituted for a third allele not present in either genome. Reads from sequenced pools were aligned to the artificial reference using bwa-mem (Li and Durbin, 2010). A pileup at known parental SNPs was created using samtools (Li et al., 2009), and allele counts at each SNP were calculated. SNP positions were filtered to retain SNPs with at least 15X coverage, both parental alleles scored, and allele frequency between 0.1–0.9 to eliminate false signals during bulk segregant analysis. Bulk-segregant analysis was performed using MULTIPOOL (v 0.10.1) (Edwards and Gifford, 2012) run across ~60,000 SNPs in contrast mode using the default recombination fraction (3300 cM) with -N set to the number of segregants in the aneuploidy-sensitive/small-colony pool in each cross (A1 = 20; A2 = 10; D = 46). Potential QTLs were identified at loci where allele frequency varied the greatest between the two pools. SSD1 was validated as the causal locus through gene deletions and complementation as shown in Figure 1. Sequencing data for each pool are available in the Short Read Archive (SRA) under access number PRJNA548343, and MULTIPOOL output files are available in Supplementary file 1, as described in Edwards and Gifford (2012).

RNA-seq was done as previously described (Jovaisaite et al., 2014) using total RNA isolated from log-phase cultures. Illumina reads were mapped to the S288c genome substituted with SNPs from YPS1009, NCYC110, or W303 as called in Sardi et al. (2018), using bwa-meme. In general, data represent the average of biological triplicate, with the exception of h-YPS1009 strains shown in Figure 2 done in quadruplicate and W303_Chr8-Chr15 and W303_Chr8-Chr10-Chr16 done in duplicate. Replicates for each strain suite were paired on the same day, enabling replicate-paired statistical analysis, done in edgeR (Robinson et al., 2010). Genes in Figure 2 were selected by considering both YPS1009_Chr12 ssd1Δ versus YPS1009_Chr12 wild type and NCYC110_Chr8 ssd1Δ versus NCYC110_Chr8 wild type. Hierarchical clustering was performed using Cluster 3.0 (Eisen et al., 1998) and visualized in Java Treeview (Saldanha, 2004). Functional enrichment of GO terms was performed using the program SetRank (Simillion et al., 2017). Activation of the UPR was inferred from enrichment of Hac1 targets among induced genes (p<1e-4, hypergeometric test, compiled in Chasman et al., 2014), and signatures of mito-CPR was indicated as up-regulation of Pdr3 targets including CIS3 as reported in Weidberg and Amon (2018). Sequencing data are available from the GEO database under accession number GSE132425. Processed data are also available in Supplementary file 2.

RNA-immunoprecipitation (RIP) was performed similar to previously described (Jansen et al., 2009) with the following modifications: Cell lysate was treated with RQ1 RNase-free DNase (Promega, Madison, WI) for 15 min at room temperature, an aliquot was removed as the input material, and RNA was immunoprecipitated using GFP-Trap Magnetic Agarose (Chromotek, Planegg-Martinsried, Germany) against Ssd1-GFP from euploid and aneuploid lysate for 1 hr at 4°C (due to Ssd1-GFP degradation with longer incubation). An identical procedure was performed with untagged YPS1009 cells as a mock-RIP. Recovered RNA was subjected to Illumina sequencing as described above. RIP-seq was performed in duplicate for euploid and for aneuploid cells; bound transcripts were identified through combined edgeR (Robinson et al., 2010) analysis of the four RIP-seq samples, contrasting RIP to input for each sample and then to mock-IP normalized to its own input. Bound transcripts were taken as those with FDR < 0.05 (Supplementary file 2). Sequencing data are available from the GEO database under accession number GSE132425.

The suite of YPS1009_Chr12 strains (AGY731, AGY735, AGY1503, AGY1517) was transformed with Moby 2.0 high-copy expression library (Magtanong et al., 2011) and an aliquot removed as the starting pool. Cells were grown in biological triplicate for five generations in YPD medium, plasmid DNA collected from the starting and ending pools, and barcodes sequenced as previously described (Magtanong et al., 2011). Pools were normalized by total barcode reads per sample and fitness costs taken as the log2(fold change) in barcode abundance after versus before outgrowth. Data in Figure 2D represent the distribution of replicate-averaged data.

Cell pellets were resuspended in 6 M guanidine HCl and boiled for 5–10 min; proteins were precipitated with methanol up to 90%, spun 5 min at 15 K g, and resuspended in lysis buffer (8 M urea, 100 mM Tris, pH = 8.0, 10 mM TCEP, 40 mM chloroacetamide). Samples were diluted to 1.5 M urea and digested overnight at room temperature with LysC (Wako Chemicals, USA) and for 3 hr with trypsin (Promega, USA) at 1:50 enzyme to protein ratio. Samples were desalted using Strata X columns (Phenomenex Strata-X Polymeric RP, USA). For LC-MS/MS, samples were resuspended in 0.2% formic acid and separated via reversed phase (RP) chromatography. 2 µg of tryptic peptides were injected onto a capillary RP column prepared in-house and packed with 1.7 μm diameter Bridged Ethylene Hybrid C18 particles as described in Shishkova et al. (2018). Columns were installed onto Dionex nanoHPLC (Thermo, Sunnyvale CA) and heated to 50°C using a home-built column heater. Mobile phase buffer A was composed of water and 0.2% formic acid, mobile phase B - 70% ACN and 0.2% formic acid. Samples were separated over a 120 min gradient at flow rate of 325 nl/min. Peptide cations were converted into gas-phase ions via electrospray ionization and analyzed using a Thermo Orbitrap Fusion Lumos (Thermo, San Jose CA) mass spectrometer, according to the previously published methods (Hebert et al., 2018). Raw data were searched using MaxQuant (v. 1.6.1.0) against Saccharomyces cerevisiae database (SGD, downloaded 10.15.2018). Searches were performed using precursor mass tolerance of 27 ppm and a product mass tolerance of 0.3 Da. Proteins were identified and quantified via MaxLFQ using default settings with enabled ‘Match between runs,’ requiring LFQ ratio of 1, and MS/MS spectra not required for LFQ comparisons. Raw data are available in the PRIDE database (Project accession # PXD013847). Prior to publication, reviewers can access the files using the following credentials: Username: reviewer95858@ebi.ac.uk, Password: 6w9IaMi3. Processed data and a list of proteins shown in Figure 3 are available in Supplementary file 2, and normalized protein abundance data are available in Supplementary file 3.

Organelle-enriched and -depleted fractions were generated for euploid (AGY1446) and aneuploid YPS1009_Ch12 SSD1-GFP (AGY1447) and untagged SSD1 cells as a control, using Mitochondrial Yeast Isolation Kit (Abcam, Cambridge, United Kingdom) according to manufacturer protocol with slight modifications to minimize protein degradation. Western blots were developed using anti-GFP ab290 (Abcam), anti-Actin MA1-744 (Thermo Fisher Scientific), anti-Cox2 ab110271 (Abcam), anti-Vph1 ab113683 (Abcam), and anti-Dpm1 ab113686 (Abcam) on a Li-COR Odyssey instrument (Model 9120). Cells for microscopy were plated on plain or poly-L-Lysine coated slides and either single images (HSP104-GFP in Figure 6) or z-stack images (all other microscopy) every 0.5 µm were acquired with an EVOS FL Auto two equipped with an RFP EVOS light cube. Z-stacks were collapsed into a single image with EVOS software for publication. Mitochondria in Figure 4C were visualized with Rhodamine B Hexyl Ester (ThermoFisher, R648MP) according to manufacturer’s protocol; images represent an overlay of the bright-field image onto the fluorescence image to highlight cell boundaries. Cells were scored by marking total cells in bright-field images and the scoring presence or absence of Rhodamine B Hexyl Ester signal. A minimum of 380 cells were scored per strain across three biological replicates. Very similar results were obtained tracking Pet123-GFP signal.

smFISH was performed as previously described (Gasch et al., 2017) except performed on an EVOS FL Auto two and with transcripts detected manually in FIJI (Schindelin et al., 2012). FISH probe sets were designed against MMR1 (conjugated to Quasar 670) and mitochondrial 15 s and 21 s rRNAs (conjugated to Quasar 570, Stellaris, Middlesex, United Kingdom). Mitochondrial morphology in Figure 3D was quantified using mitochondrial rRNA probes, which produced images very similar to Rhodamine staining but enabled visualization independent of mitochondrial membrane potential. Morphology was scored in each cell manually using the multi-point tool in FIJI, recording the number of cells with any tubular, any globular, or only globular morphologies. Cells with fragmented mitochondria were defined as those with at least three discontinuous fragments from the tubular structure or having at least three fragments in addition to the largest globular focus. >200–400 cells were scored for all microscopy experiments and across multiple biological replicates per strain. MMR1 localization was scored by identifying buds (scored as cells lacking DAPI or containing bar nuclei by DAPI staining) and scoring those either containing or lacking MMR1 transcripts.