A natural variant of the essential host gene MMS21 restricts the parasitic 2-micron plasmid in Saccharomyces cerevisiae

  1. Michelle Hays
  2. Janet M Young
  3. Paula F Levan
  4. Harmit S Malik  Is a corresponding author
  1. Molecular and Cellular Biology program, University of Washington, United States
  2. Division of Basic Sciences & Fred Hutchinson Cancer Research Center, United States
  3. Howard Hughes Medical Institute, Fred Hutchinson Cancer Research Center, United States


Antagonistic coevolution with selfish genetic elements (SGEs) can drive evolution of host resistance. Here, we investigated host suppression of 2-micron (2μ) plasmids, multicopy nuclear parasites that have co-evolved with budding yeasts. We developed SCAMPR (Single-Cell Assay for Measuring Plasmid Retention) to measure copy number heterogeneity and 2μ plasmid loss in live cells. We identified three S. cerevisiae strains that lack endogenous 2μ plasmids and reproducibly inhibit mitotic plasmid stability. Focusing on the Y9 ragi strain, we determined that plasmid restriction is heritable and dominant. Using bulk segregant analysis, we identified a high-confidence Quantitative Trait Locus (QTL) with a single variant of MMS21 associated with increased 2μ instability. MMS21 encodes a SUMO E3 ligase and an essential component of the Smc5/6 complex, involved in sister chromatid cohesion, chromosome segregation, and DNA repair. Our analyses leverage natural variation to uncover a novel means by which budding yeasts can overcome highly successful genetic parasites.


Host genomes are engaged in longstanding conflicts with myriad selfish genetic elements (SGEs, or genetic parasites) (Burt and Trivers, 2008Dawkins, 1976; McLaughlin and Malik, 2017). SGEs propagate within an organism or population at the expense of host fitness (Burt and Trivers, 2008). Many SGEs, including viruses, selfish plasmids, and other pathogens, must coopt the host’s cellular machinery for their own survival: to replicate their genomes, to transcribe and translate their proteins, and to ensure their proliferation by passage into new cells (Burt and Trivers, 2008; McLaughlin and Malik, 2017). If a host variant arises that can suppress SGEs (host restriction), this variant will be favored by natural selection and can rise in frequency in a population. If resistance has no fitness cost, such variants will rapidly fix within host species. Even if these variants are slightly deleterious, such variants could be maintained in quasi-equilibrium in host species (Koskella, 2018; Meaden et al., 2015; Kraaijeveld and Godfray, 1997; Sheldon and Verhulst, 1996).

Studies in diverse biological taxa have leveraged genetic mapping strategies to identify quantitative trait loci (QTL) for host resistance to parasites (Kane and Golovkina, 2019; Cogni et al., 2016; Kelleher et al., 2018; Duffy and Sivars-Becker, 2007). Such studies have revealed that host populations are more likely to harbor variation in resistance to coevolved, rather than to recently introduced, parasites (Duxbury et al., 2019). Studying parasites in their native host context therefore maximizes opportunities to discover host resistance mechanisms. However, it is often difficult to study natural variation in resistance, because hosts and/or parasites are often intractable in the laboratory. Budding yeast provides an ideal system to study host-SGE genetic conflicts, with abundant genetic tools, together with resources for comparative and population genetics. Yeast species harbor a variety of SGEs including retrotransposable elements, RNA viruses and 2-micron (2μ) plasmids (Rowley, 2017; Wickner, 1996; Kelly et al., 2012; Nakayashiki et al., 2005; Krastanova et al., 2005; Bleykasten-Grosshans and Neuvéglise, 2011). Yet, despite its long history as a popular model eukaryote, natural variation in cellular immunity factors against SGEs has been largely uncharacterized in S. cerevisiae and related species (Rowley et al., 2016; Czaja et al., 2019; Scholes et al., 2001; Maxwell and Curcio, 2007; Rowley et al., 2018). Here, we investigated whether S. cerevisiae strains harbor genetic variants that allow them to resist a highly successful SGE: 2μ plasmids.

2μ plasmids are nuclear SGEs found in multiple, divergent budding yeast species (Blaisonneau et al., 1997; Utatsu et al., 1987; Chen et al., 1992; Peter et al., 2018). They are best characterized in S. cerevisiae, where they are found in high copy numbers:~50 copies per haploid and ~100 copies per diploid cell (Veit and Fangman, 1988; Zakian et al., 1979). 2μ plasmids are stably transmitted through vertical inheritance. However, even if they are lost stochastically, they can be reintroduced via sex and transmitted via non-Mendelian inheritance through meiosis: even if only one haploid parent initially has 2μ plasmids, all four meiotic progeny typically receive plasmids (Futcher and Cox, 1983). Their widespread prevalence in S. cerevisiae and other budding yeast species has raised the question of whether 2μ plasmids might be more commensal than parasitic. In the mid-1980s, two seminal studies showed that S. cerevisiae strains carrying 2μ plasmids (cir+) grew 1–3% more slowly than did their cir0 counterparts under laboratory conditions; thus 2μ plasmids confer a clear fitness defect (Futcher and Cox, 1983; Mead et al., 1986). Recent studies have reinforced the fitness defect associated with carriage of 2μ plasmids (Harrison et al., 2012; Harrison et al., 2014). Furthermore, many mutant yeast strains, which are sick in the presence of 2μ plasmids, can be partially rescued when ‘cured’ of their 2μ plasmids (Dobson et al., 2005; Zhao et al., 2004). For example, nib1 mutants (a hypomorphic allele of ULP1 [Dobson et al., 2005]) form ‘nibbled’ colonies in the presence of 2μ plasmids due to colony sectoring from cells that stop dividing when overburdened with 2μ plasmids, but form smooth (wild-type) colonies in their absence (Dobson et al., 2005). These and other data (Zhao et al., 2004) suggest that 2μ plasmids impose a selective burden on yeast, both under rapid laboratory growth conditions as well as in times of stress. In contrast to bacterial plasmids, which can harbor host-beneficial ‘cargo’ genes, such as antibiotic resistance genes, no such beneficial genes have ever been observed in natural 2μ plasmids (Bennett, 2008). Indeed, there are no known conditions in which 2μ plasmids are beneficial to the host, further supporting that its presence is likely the result of efficient parasitism. Although they are stable in S. cerevisiae, experimental studies show that 2μ plasmids exhibit lower copy number and decreased stability when introduced into exogenous species (Murray et al., 1988). These findings suggest that 2μ plasmids have co-evolved with host genomes to become a successful genetic parasite of yeasts.

To be successful, 2μ plasmids must replicate and segregate with high fidelity into daughter cells during both yeast mitosis and meiosis. Without these capabilities, plasmids risk being lost from the population as their host cells are outcompeted by plasmid-less daughter cells. Yet, 2μ plasmids encode just four protein-coding genes (represented by arrows in Figure 1A) in S. cerevisiae. REP1 and REP2 encode plasmid-encoded DNA-binding proteins that bind to the 2μ STB locus to mediate segregation (Jayaram et al., 1983; Velmurugan et al., 1998; Veit and Fangman, 1988). Mutations in REP1 and REP2 significantly impair segregation fidelity, resulting in failure to transmit plasmid to daughter cells, and subsequent loss from the host population (Murray and Szostak, 1983). If copy number drops below a certain threshold within a host cell, 2μ plasmids activate an amplification mechanism that relies on plasmid-encoded FLP1 (Murray et al., 1987; Som et al., 1988). FLP1 encodes a recombinase that creates plasmid structural rearrangements during host S phase via the FRT sites, facilitating over-replication via rolling circle replication using host replication machinery (Murray et al., 1987; Som et al., 1988; Zakian et al., 1979; Volkert and Broach, 1986; Dobson et al., 1988). RAF1 encodes a protein that regulates the switch to copy number amplification (Murray et al., 1987). Due to this minimal genome, 2μ plasmids rely on host factors for genome replication and segregation during host cell division (Zakian et al., 1979; Rizvi et al., 2018; Prajapati et al., 2017; Ma et al., 2013; Sau et al., 2014; Ghosh et al., 2007; Sau et al., 2015).

Figure 1 with 1 supplement see all
SCAMPR, a novel method to measure 2μ plasmid stability and dynamics.

(A) Schematic of GFP-reporter 2μ plasmid. The endogenous 2μ plasmid encodes an origin of replication (ori), four protein-coding genes (REP1, REP2, RAF1, FLP1) and their interacting DNA loci (STB, and FRT). The GFP-2μ reporter plasmid described here utilizes the full 2μ genome with an additional integrated G418-resistance and GFP expression cassette. (B) A Single Cell Assay for Measuring Plasmid Retention (SCAMPR) utilizes the dual reporter cassette: G418 resistance to ensure plasmid retention while under selection and GFP to facilitate screening of plasmid-positive cells. Cells with the reporter plasmid are kept on G418 selection to ensure the plasmid is present at t = 0 and either released to media without selection or passaged with continued G418 selection. Comparing the GFP intensities of the cell populations with and without G418 selection after 24 hr reveals the plasmid retention dynamics and population heterogeneity of the host genetic background (Figure 1—figure supplement 1). SCAMPR can therefore distinguish between alternate mechanisms of plasmid instability, illustrated in (C) and (D), or the relative contribution of both mechanisms. (C) Gross segregation defects in which plasmids are not distributed to both daughter cells would cause an increase in GFP-negative cells, as well as an increase in ‘super-green’ cells that retain twice as many two plasmids (light shading, dotted histogram). However, we infer that these cells would either be lost or not proliferate due to growth defects associated with high plasmid copy number. As a result of this selection, we expect to see a rapid increase in GFP-negative cells but no dramatic change in the median expression of (surviving) GFP-positive cells. (D) Plasmid instability caused by under-replication or copy number suppression would not cause a precipitous decline in GFP-positive cells as in (C) but would instead lead to a reduction in the median GFP intensity of the GFP-positive cells.

Previous studies have identified host factors required by the 2μ plasmid. For instance, in addition to DNA replication and origin licensing factors, 2μ plasmids require host factors to facilitate proper partitioning into daughter cells, including many spindle-associated proteins (Zakian et al., 1979; Rizvi et al., 2018; Prajapati et al., 2017; Ma et al., 2013; Sau et al., 2014; Ghosh et al., 2007). Furthermore, host-mediated post-translational SUMO-modification of plasmid-encoded proteins has been shown to have a profound effect on 2μ plasmid stability and host fitness. For example, failure to sumoylate the Rep proteins impairs plasmid stability, whereas deficient sumoylation of Flp1 recombinase leads to recombinase overstabilization, resulting in massively increased plasmid copy number and extreme reduction in host cell fitness (Zhao et al., 2004; Burgess et al., 2007; Pinder et al., 2013). Indeed, mutations in SUMO E3 ligases Siz1, Siz2, the SUMO maturase Ulp1, and the SUMO-targeted ubiquitin ligase Slx8 all lead to hyper-amplification and host cell defects (Zhao et al., 2004; Burgess et al., 2007; Pinder et al., 2013). These host-plasmid interactions provide potential means for the host to curb deleterious proliferation of plasmids. However, it is unclear whether gain-of-function alleles exist that restrict or eradicate plasmids.

Until recently, 2μ plasmids have been largely omitted from studies of genetic variation in yeast. Although prior work has predominantly focused on canonical A-type 2μ plasmids (found in laboratory S. cerevisiae strains), recent studies revealed that 2μ plasmids are quite diverse in budding yeast populations (Peter et al., 2018; Strope et al., 2015). These analyses identified C-type plasmids, extremely diverged D-type plasmids and a 2μ plasmid introgression into S. cerevisiae from the closely-related species S. paradoxus (Peter et al., 2018; Strope et al., 2015). Moreover, previously identified B- and newly described B*-type plasmids were shown to be a result of recombination between A and C types (Peter et al., 2018Strope et al., 2015; Xiao et al., 1991a; Xiao et al., 1991b). Furthermore, these studies revealed that there are multiple, distinct strains of S. cerevisiae that do not harbor any 2μ plasmids. Yet, it remains unknown whether 2μ plasmid absence in these strains is the result of stochastic loss or an inherent host trait.

Host cells could influence 2μ plasmid fitness by affecting their copy number, stability, or population heterogeneity. However, these parameters are not captured in traditional plasmid loss assays, which measure either plasmid copy number averaged across the entire population, or plasmid presence versus absence independent of copy number. To quantitatively capture all of these parameters, we developed a new high-throughput, single-cell, plasmid retention assay, SCAMPR (Single-Cell Assay for Measuring Plasmid Retention). We identified three yeast strains that naturally lack 2μ plasmids and reproducibly show a high rate of mitotic instability of 2μ plasmids upon plasmid reintroduction. Focusing on one resistant strain, we used SCAMPR to show that resistance is a dominant, multigenic trait. Using QTL mapping by bulk segregant analysis, we identified one significant genomic locus that impairs 2μ mitotic stability. A candidate gene approach within this locus showed that a single amino acid change in MMS21 contributes to plasmid instability. MMS21 is a highly conserved E3 SUMO ligase and an essential component of the Smc5/6 complex, which has not previously been implicated in 2μ biology. Thus, our study reveals a novel pathway by which 2μ resistance has arisen and persists in natural populations of S. cerevisiae.


SCAMPR: Single-Cell Assay for Measuring Plasmid Retention

To determine if there is heritable natural variation in 2μ plasmid stability in S. cerevisiae strains, we needed an assay to measure plasmid maintenance at the single-cell level. Traditionally, plasmid loss dynamics have been measured by two types of assays. The first of these is the Minichromosome Maintenance (MCM) assay, in which strains containing plasmids with selectable markers are assessed for plasmid presence versus absence by counting colonies on both selective and non-selective media over time (Maine et al., 1984). Due to the labor intensiveness of the assay, MCM is low-throughput since different dilutions need to be tested to recover and reliably count 30–300 colonies per plate. Furthermore, as only a single copy of a selectable marker is required for viable cell growth, substantial variation in plasmid copy number can go undetected by the MCM assay.

A second type of assay traditionally used to measure plasmid stability uses molecular methods, such as quantitative PCR (qPCR) or Southern blotting, to assess mean plasmid copy number, relative to genomic DNA, across a population of cells (Lee et al., 2006). Compared to the MCM assay, qPCR has the advantages of being high-throughput and not requiring a selectable marker in the plasmid of interest. However, qPCR (or Southern blotting) can only measure the average copy number of a plasmid in a population. Any heterogeneity in plasmid presence or copy number would be undetectable by qPCR. Even a combination of the MCM and qPCR assays lacks the resolution to determine the distribution or variability of plasmid copy number within a host population.

We therefore designed a single-cell assay using a reporter 2μ plasmid. To ensure that this plasmid closely resembles endogenous plasmids, we eschewed the use of the yEP multi-copy plasmids commonly used to express yeast ORFs, because they contain only a small portion of the natural 2μ plasmid. Instead, we built a new 2μ reporter plasmid, which contains both a selectable marker (G418 resistance) as well as a screenable (eGFP) marker, each under a constitutive promoter (Breslow et al., 2008). Previously, others described recombinant 2μ plasmid construction and identified a site that does not impact 2μ plasmid stability when less than 3.9 kb DNA is integrated (Ludwig and Bruschi, 1991). We therefore integrated the marker cassette (2703 bp) into the endogenous plasmid at this location using yeast assembly (Figure 1ALudwig and Bruschi, 1991; Gibson et al., 2008). Importantly, we chose this insertion location because it should not impact typical plasmid function: replication, segregation and copy number amplification should proceed as with the unaltered endogenous plasmid (Ludwig and Bruschi, 1991). This allows us to monitor natural plasmid functions relative to variable host compatibility.

These dual markers ensured the reporter plasmid could be both introduced and retained in plasmid-lacking strains. In addition the GFP reporter allows strains to be easily assayed for plasmid presence, absence, and copy number (because GFP intensity scales with copy number in yeast [Suzuki et al., 2012; Labunskyy et al., 2014; Lauer et al., 2018; Zhu et al., 2015]). Coupling this reporter with flow cytometry allows us to assay single cells to better understand the dynamics and mechanism of plasmid loss. Our analyses revealed that GFP intensity for this 2μ reporter plasmid is roughly normally distributed across single cells (Figure 1—figure supplement 1) indicating that GFP signal did not saturate the detector at high copy number. The endogenous 2μ plasmid loss rate is ~10−5 per cell per generation as estimated by colony-hybridization Southern blots (Futcher and Cox, 1983). Based on this prior estimate which relied on total loss events, we infer that the stability of the GFP-2μ reporter plasmid is lower than that of the endogenous 2μ plasmid. This could either reflect the difference in precision of plasmid stability measurements or be due to the cost of constitutive expression of the dual markers. Nevertheless, we conclude that the reporter is well suited for comparative stability studies using the same plasmid in different host backgrounds.

We used this 2μ reporter plasmid with flow cytometry analyses (Figure 1B) to simultaneously infer both total plasmid loss events by measuring the proportion of GFP-negative cells, as well as changes in the median plasmid copy number based on GFP intensity (Figure 1C–D). Importantly, we could also assess the population distribution of GFP intensity, revealing the inherent cellular heterogeneity in plasmid copy number and loss. This assay is also higher throughput than traditional methods. We refer to this assay as SCAMPR (Single-Cell Assay for Measuring Plasmid Retention).

2μ plasmid instability in natural yeast isolates is rare and heritable

2μ plasmids are nearly universally present in laboratory strains of S. cerevisiae. However, recent studies of natural isolates have revealed a diversity of plasmid types in natural populations, and even strains lacking 2μ plasmids altogether (Peter et al., 2018; Strope et al., 2015). We were particularly interested in plasmid-free strains as these might harbor genetic variants that actively inhibit plasmid stability. To this end, we surveyed a panel of 52 natural S. cerevisiae isolates for plasmid presence versus absence via PCR analyses (see Materials and methods). From this panel, we identified three strains (representative gel in Figure 2—figure supplement 1A) that do not contain the 2μ plasmid: Y9 (from ragi, millet), YPS1009 (from oak exudate), and Y12 (from palm wine) (Supplementary file 1). To rule out the possibility that PCR surveys were confounded by 2μ polymorphisms, we also tested these strains via Southern blotting (Figure 2—figure supplement 1B), which supported our conclusion of plasmid absence. Wild diploid strains are homothallic, and capable of mating type switching and self-diploidizing following sporulation. To create stable haploid lines for subsequent analyses, we deleted HO endonuclease in the natural isolates before sporulating to produce stable heterothallic haploid strains from each of the three plasmid-free natural yeast isolates (Supplementary file 2).

Although these three strains lack detectable 2μ plasmids, this absence could be either the result of stochastic loss or host genetic variation that inhibits plasmid stability. Stochastic loss could occur because of rare bottlenecking events in wild populations or during laboratory passaging (Kelly et al., 2012; Nakayashiki et al., 2005). We predict that such losses would not protect these strains from re-introduction of natural 2μ plasmids via sex and subsequent propagation (Futcher and Cox, 1983). Therefore, if absence were due to stochastic loss, we would expect our reporter 2μ plasmids to be stable in these strains. Alternatively, if the absence of 2μ plasmids reflects true host genetic variation conferring resistance, our reporter 2μ plasmid would be mitotically unstable in these strains. To test these two alternatives, we transformed the GFP-2μ reporter plasmid into haploid cells from these three natural isolates and tested for mitotic plasmid loss using a qualitative colony sectoring assay. As a control, we examined reporter stability in the permissive lab strain BY4742 that was ‘cured’ of its endogenous 2μ plasmid (Tsalik and Gartenberg, 1998) (see Materials and methods). These analyses showed a clear difference in GFP sectoring (plasmid loss) between the BY4742 laboratory strain and the three natural isolates (Figure 2A).

Figure 2 with 2 supplements see all
Plasmid instability is a heritable trait in three natural S. cerevisiae isolates.

(A) A colony sectoring assay qualitatively measures GFP-2μ reporter plasmid loss on solid media. Whereas the majority of colonies in the BY4742 background express GFP, only a small fraction of cells in colonies from wild isolates Y9, Y12, and YPS1009 express GFP. (B) The MCM assay quantifies the frequency of 2μ loss events in different yeast strains. Haploid cells from three wild isolates (Y9, Y12, YPS1009) have significantly lower plasmid retention than haploid cells from the laboratory BY4742 strain. ***p<0.0001, Kruskal-Wallis test. (C) SCAMPR assays confirm that a significantly smaller fraction of Y9 strain haploid cells retain the GFP-2μ reporter plasmid after 24 hr, relative to haploid BY4742 cells. ***p<0.0001, Kruskal-Wallis test.

To quantify this difference in plasmid stability between the permissive lab strain and the non-permissive natural isolates, we next measured plasmid stability of the GFP-2μ plasmid over a 24 hr period (~12 generations) using a traditional MCM assay (Figure 2BMaine et al., 1984). Consistent with the colony sectoring assay, we determined that the reporter GFP-2μ plasmid is significantly less stable in the naturally cir0 wild isolates than in the plasmid-permissive laboratory strain. For example, the Y9 strain maintained 2μ plasmids in only ~5% of cells on average, whereas the BY4742 lab strain maintained plasmids in ~60% of cells. Even after normalization for phenotypic lag (see Materials and methods), we concluded that Y9 and BY4742 strains retain plasmids at 20% versus 70% frequency, respectively. The other two wild strains showed similar plasmid loss frequencies, with the YPS1009 strain exhibiting more variability between replicates than the other two strains. Taken together, these data suggest that 2μ plasmids are mitotically unstable in these three natural isolates. Thus, the absence of endogenous 2μ plasmids in these strains is the result of host genetic variation rather than stochastic plasmid loss.

Dominant 2μ plasmid instability in the Y9 strain

Of the three natural isolates in which we observed 2μ plasmid instability, the Y9 strain isolate had the least variable plasmid loss phenotype. Furthermore, in a broad analysis of yeast strains, the Y9 strain was found to be phylogenetically close to the Y12 strain (Hyma and Fay, 2013; Liti et al., 2009). Based on this phylogenetic proximity, we hypothesized that Y9 and Y12 strains may share the same genetic basis for host-encoded plasmid instability, which might make this genetic determinant easier to identify. We therefore decided to focus on further understanding the phenotypic and genetic basis of plasmid instability in the Y9 strain.

We tested whether growth disadvantages could explain the plasmid instability observed in Y9. This possibility was suggested by our colony sectoring assays (Figure 2A) in which some GFP-positive colonies were smaller than GFP-negative colonies. We therefore compared growth rates of BY4742 and Y9, each with and without the 2μ reporter plasmid (Figure 2—figure supplement 2). We determined that BY and Y9 haploid strains have similar growth rates to one another without the reporter plasmid. Both Y9 and BY4742 exhibited a similar decrease in growth rate when grown under G418 selection to retain the reporter plasmid. This growth difference could be due to either the presence of the selective drug G418, or due to the fitness cost imposed by the reporter plasmid itself. However, in either condition, Y9 and BY4742 showed similar growth rates to one another. We therefore conclude that plasmid carriage cost is not the predominant cause of the different plasmid instability seen in BY4742 and Y9 strains.

We characterized the putative mechanism of 2μ plasmid instability in the Y9 strain using the SCAMPR assay (Figure 1, Figure 2CFigure 1—figure supplement 1B–C). We measured the change in distribution of GFP intensity (inferring plasmid copy number changes) among single cells, and total loss events (determined by increase in GFP-negative cells). If the 2μ plasmid were undergoing systematic under-replication due to defects in replication, we might expect an overall and homogenous decrease in median plasmid copy number across the population (Figure 1D). Alternatively, if the 2μ plasmid were being missegregated, we might instead see increasing population heterogeneity, with some cells inheriting no plasmid, and their sister cells inheriting twice the number of plasmids as the original mother cell (Figure 1C). Others have shown that when cells experience over-amplification of 2μ plasmids those cells stop dividing due to the massive fitness cost, as in the case of nibbled and similar phenotypes (Dobson et al., 2005; Zhao et al., 2004; Zhao and Blobel, 2005). This fitness defect explains why we may see an increase in plasmid-negative cells at the population level, without the corresponding increase of super-green high-plasmid cells (dotted line).

In our SCAMPR analyses, we find that even under G418 selection, Y9 cells do not maintain the reporter 2μ plasmid as efficiently as BY strains (52% GFP-positive versus 90% respectively) (Figure 1—figure supplement 1). Moreover, upon removing pressure to maintain the plasmid (no G418 selection), the proportion of Y9 cells with no GFP (no 2μ plasmid) increases significantly, from 48% to 83%. However, the median GFP intensity (and inferred 2μ plasmid copy number) of plasmid-bearing Y9 cells remains largely unchanged (Figure 1—figure supplement 1B); even with G418 selection, GFP intensity (2μ plasmid copy number) is lower in Y9 than BY4742. We therefore conclude that 2μ plasmid loss in Y9 haploid cells occurs primarily via abrupt, complete loss of plasmids from cells in the population rather than a steady decrease in copy number (Figure 1—figure supplement 1B). This observed pattern of plasmid loss is consistent with plasmid segregation failure during mitosis, rather than a copy number suppression mechanism or plasmid under-replication. As a result of this segregation failure, ‘non-permissive’ Y9 haploid cells lose the 2μ reporter plasmid substantially more quickly than the permissive BY4742 laboratory strain (Figure 2C), mirroring our observations from colony sectoring assays (Figure 2A).

Next, we investigated whether mitotic instability of the 2μ plasmids in the Y9 strain is genetically recessive or dominant by examining heterozygous diploids of permissive and non-permissive strains. To ensure that mitotic instability was not influenced by ploidy itself, we first tested whether the plasmid instability phenotype we observed in haploid strains persists in homozygous diploid BY4742 and Y9 strains. We found an even bigger difference in plasmid instability between homozygous diploid Y9 and BY4743 strains than between haploid strains (Figures 3A and 2B). We generated a heterozygous diploid strain by crossing the GFP-2μ plasmid-containing permissive BY4742 lab strain to the non-permissive Y9 haploid strain. If plasmid loss in Y9 cells were due to inactivating mutations within a host ‘permissivity’ factor required for 2μ mitotic segregation, we might expect plasmid instability to be recessive, with the BY4742 allele providing rescue in the heterozygote. Alternatively, if plasmid instability in Y9 cells were due to a host-encoded, gain-of-function ‘restriction’ factor that impairs mitotic stability of 2μ plasmids, we would expect mitotic instability of 2μ plasmids to be dominant; heterozygous diploids would also exhibit plasmid instability. We found that heterozygous BY4742/Y9 diploid cells rapidly lose the plasmid after G418 selection is removed (Figure 3A). These findings could result from haploinsufficiency of a permissivity factor, or dominance of plasmid restriction factors in the Y9 genome. We therefore considered both possibilities in subsequent analyses.

Figure 3 with 1 supplement see all
Genetic architecture and dominance of the Y9 plasmid instability phenotype.

(A) Compared to homozygous BY4742 diploids, heterozygous BY4742/Y9 diploid cells display low plasmid retention after 24 hr, similar to homozygous Y9 diploids. This suggests that the plasmid instability of Y9 cells is a dominant trait. All strains were analyzed with the SCAMPR plasmid retention assay. **p<0.001, ***p<0.0001, Kruskal-Wallis test; n.s. = not significant. (B–C) Phenotype distribution across ~600 random progeny strains (C) shows that most have an intermediate phenotype between that of the parental haploids (B). All strains were analyzed in triplicate with the SCAMPR assay. We selected the bottom ~20% (‘non-permissive’) and top ~20% (‘permissive’) of strains from this distribution for bulk sequencing and segregant analysis.

Genetic architecture underlying 2μ plasmid instability in the Y9 strain

Previous studies have shown that 2μ plasmids efficiently propagate via non-Mendelian inheritance through meiosis in laboratory strains of S. cerevisiae (Harrison et al., 2014; Sau et al., 2014; Brewer and Fangman, 1980; Hsiao and Carbon, 1981). Because BY4742/Y9 heterozygous diploids exhibit dominant plasmid loss, we maintained G418 selection up to and during sporulation to enrich for tetrads in which all four spores retained 2μ reporter plasmids. We then measured plasmid instability phenotypes among meiotic progeny of BY4742/Y9 heterozygous diploids to understand the genetic architecture underlying the Y9 strain’s plasmid instability.

If a single genetic locus were responsible for 2μ plasmid instability, we would expect tetrads to exhibit a 2:2 segregation pattern, with half of the spores phenotypically resembling the permissive BY4742 parent and the other half resembling the non-permissive Y9 parent. Of the 60 tetrads examined, approximately 20% of 4-spore tetrads exhibited a roughly 2:2 segregation pattern, and the remaining 80% tetrads exhibited more complex patterns of inheritance (Figure 3—figure supplement 1). Our results indicate that plasmid instability is heritable but not monogenic. Based on these findings, we used the Castle-Wright estimator (Lande, 1981; Zeyl, 2005) to estimate that 2μ plasmid instability is encoded by at least 2–3 independently segregating large effect loci in the Y9 genome, although these estimates could be affected by stochastic variability of the 2μ plasmid stability phenotype.

Next, we performed quantitative trait locus (QTL) mapping using bulk segregant analysis (BSA) to identify genetic loci that contribute to the Y9 strain’s 2μ plasmid instability phenotype (Figure 4—figure supplement 1Ehrenreich and Magwene, 2017; Lander and Botstein, 1989). We selected 600 random spores resulting from a heterozygous BY4742/Y9 diploid containing our reporter 2μ plasmid and used SCAMPR to phenotype plasmid stability (Amberg et al., 2005). Most of these progeny exhibit intermediate 2μ plasmid stability between haploids of the parental BY4742 and Y9 strains (Figure 3B–C). We then pooled and bulk-sequenced 132 ‘non-permissive’ progeny strains that represented ~20% of progeny with the lowest 2μ plasmid stability and 126 ‘permissive’ strains that represented the ~20% of progeny with the highest 2μ plasmid stability (Figure 3C).

In addition to the progeny pools, we also sequenced the genomes of the three 2μ plasmid-negative strains we identified (Y9, Y12, and YPS1009), as well as UC5, a 2μ plasmid-containing strain (by PCR) that is closely related to Y9 and Y12 (Hyma and Fay, 2013; Cromie et al., 2013). We mapped reads from these strains back to the S. cerevisiae reference genome and created de novo assemblies for each strain (see Materials and methods). Unexpectedly, whole genome sequencing revealed that the haploid Y9 parent strain was disomic for chromosome XIV (Figure 4—figure supplement 2A), with the aneuploid chromosome segregating in the Y9 x BY4742 cross. The homothallic Y9 diploid was euploid for chromosome XIV by qPCR and shows a similar plasmid loss phenotype as a homozygous diploid Y9 that has an additional chromosome XIV (Figure 4—figure supplement 2B). These data demonstrate that the aneuploidy for chromosome XIV is not a large contributor to Y9’s plasmid instability phenotype (Pavelka et al., 2010). We therefore disregarded the segregating chromosome XIV disomy in our subsequent analyses.

We identified genomic differences between the Y9 and BY4742 strains (see Materials and methods), then compared allele frequencies between the ‘permissive’ and ‘non-permissive’ pools of meiotic progeny from BY4742/Y9 heterozygotes (Figure 3C, Figure 4—figure supplement 1A). We identified genomic regions in which inheritance of the Y9 allele is significantly more common in the non-permissive progeny pool than the permissive pool (Figure 4A) using the MULTIPOOL algorithm to generate likelihood-based ‘LOD’ (logarithm of the odds) scores (Lander and Botstein, 1989; Edwards and Gifford, 2012). We identified loci that are likely linked to the plasmid stability phenotype. Although there are a few genomic regions with moderate LOD scores of ~4 (Figure 4B), the most striking LOD score of 9.996 was seen for a high-confidence QTL on chromosome V. While it is challenging to establish a concrete LOD score threshold above which loci are statistically significant, a score of ~10 is comfortably above genome-wide significance thresholds of 3.1–6.3 established empirically in other studies (Treusch et al., 2015; Albert et al., 2014; Roberts et al., 2017). This locus likely encodes the strongest genetic determinant of plasmid instability in the Y9 genome.

Figure 4 with 4 supplements see all
QTL mapping identifies a plasmid instability locus on Y9 chromosome V.

(A) We plotted the mean Y9 SNP allele frequency in 20 kb windows for the ‘non-permissive’ (red) and ‘permissive’ (black) pools of meiotic haploid progeny from BY4742/Y9 heterozygous diploid parents. Associations with a plasmid instability locus would show an increased representation of Y9 alleles in the non-permissive pool and a decreased representation of the BY4742 haplotype in the permissive pool (dotted line indicates equal representation). The increased representation of Y9 alleles on chromosome XIV in both pools is a result of a segregating disomy in the Y9 parent that we show does not affect the plasmid instability phenotype (Figure 4—figure supplement 2). (B) Based on the allele frequencies of individual SNPs, we used MULTIPOOL to calculate LOD scores for association with the plasmid instability phenotype. We observe a highly significant LOD score (10.00) on chromosome V. The peak is fairly sharp and reaches maximal LOD score at chrV:122.3–122.7 kb (sacCer3 coordinates). All loci have allele frequencies skewed in the expected direction; the restrictive pool is enriched for Y9 alleles. (C) MULTIPOOL 90% (54 genes, 91.2 kb, chrV:92.4–183.6 kb) and 50% (16 genes, 23.1 kb, chrV:107.3–130.4 kb) credible intervals for the chromosome V QTL. Among the 16 genes in the 50% credible interval is MMS21, which encodes an essential SUMO E3-ligase.

A single variant of the essential SUMO ligase MMS21 contributes to 2μ mitotic instability in Y9

We focused our efforts on variants within the QTL on chromosome V to identify the genetic basis of Y9-encoded plasmid instability (Figure 4—figure supplement 3). The 90% confidence interval for this QTL is ~91 kb wide and contains 54 ORFs, with a 50% confidence interval 23 kb wide (16 ORFs) (Figure 4—figure supplement 3, Figure 4C). This region contains many polymorphisms between the Y9 and BY4742 genomes but very few structural variants (i.e., large insertions, deletions, translocations). One of these structural variants is the URA3 gene, which is present in Y9 and was specifically deleted in BY4742. Although the URA3 gene often falls within fitness-related QTLs in BY4742 crosses, detailed follow-up studies (Figure 4—figure supplement 4) allowed us to conclusively rule out a role for URA3 in the plasmid instability phenotype of the Y9 strain (Wilkening et al., 2014; Romano et al., 2010).

After excluding dubious ORFs, 44 bona fide protein-coding genes remained within the 90% confidence interval. 28 of these candidate genes contained a total of 94 missense changes between Y9 and BY4742, while the rest contained no non-synonymous differences between the parental strains for our QTL cross. We focused on 15 missense polymorphisms (in 11 genes) at which Y9 is identical to the phylogenetically close non-permissive strain, Y12, but different from the closely-related permissive strain, UC5 (Supplementary file 3). Our attention was drawn to MMS21, which contains a single Thr69Ile missense change common to Y9 and Y12, but distinct from the BY4742 laboratory strain and the permissive UC5 strain. This polymorphism is only 1.2 kb away from the apex of the LOD score peak in our bulk segregant analysis. Even though MMS21 has not previously been implicated in 2μ biology, it encodes one of the three mitotic SUMO E3 ligases in S. cerevisiae (Zhao and Blobel, 2005). The two other SUMO E3-ligases, encoded by SIZ1 and SIZ2, have known roles in SUMO-modification of the plasmid-encoded Rep and Flp1 proteins to cause instability or hyper-amplification phenotypes (Dobson et al., 2005; Pinder et al., 2013; Chen et al., 2005). Therefore, we evaluated the consequences of Y9’s MMS21 polymorphism on 2μ plasmid mitotic stability.

We first tested whether the Y9 MMS21 allele was sufficient to confer the plasmid loss trait. We integrated the Y9 MMS21 allele, with the flanking Y9 intergenic regulatory regions, into the ho locus of BY4742. These engineered BY4742 haploids thus express both the BY4742 and Y9 MMS21 alleles. Although the addition of Y9 MMS21 does lower plasmid stability, this difference is not statistically significant from BY4742 haploids that only express the BY4742 allele (Figure 5A). Thus, the Y9 MMS21 allele, by itself, does not appear to be sufficient to lower plasmid stability in the BY4742 genetic background.

Figure 5 with 3 supplements see all
A single SNP in Y9 MMS21 contributes to the plasmid instability phenotype.

(A) Introduction of the Y9 MMS21 allele into BY4742 haploid cells is not sufficient to significantly lower plasmid instability. (B) However, removal of the Y9 MMS21 allele but not the BY4742 MMS21 allele increases plasmid stability in BY4742/Y9 heterozygous diploids, showing that the Y9 allele of MMS21 plays an important role in the Y9 plasmid instability phenotype. **p<0.001, Kruskal-Wallis test, n.s. = not significant. (C) Plasmid prevalence (by plasmid class) for each MMS21 Thr9Ile genotype within 1011 sequenced S. cerevisiae strains. Plasmid data and genotypes from Peter et al., 2018. Strains with the Y9 MMS21 allele (I69) have a lower frequency of harboring 2μ plasmids in general, and A-type 2μ plasmids, in particular. However, this effect can be confounded by the phylogenetic relatedness of these strains.

Next we used reciprocal hemizygosity to test whether loss of Y9 MMS21 from heterozygous BY4742/Y9 diploids would lead to an increase in plasmid stability. Because MMS21 is an essential gene, we could not simultaneously delete both the Y9 and BY4742 MMS21 alleles in heterozygous diploids. Instead, we deleted either the Y9 or the BY4742 allele, yielding BY4742/Y9 diploids that are hemizygous for one or the other MMS21 allele. If the plasmid stability phenotype were affected by MMS21 haploinsufficiency, we would expect that deletion of either MMS21 allele would affect the stability phenotype. Contrary to this expectation, we found that deletion of the Y9 MMS21 allele, but not the BY4742 allele, results in a reproducible and statistically significant increase in plasmid stability of nearly 8% (Figure 5B). Indeed, otherwise identical heterozygous Y9/BY4742 strains that are hemizygous for either the Y9 or the BY4742 allele of MMS21 differ significantly in their plasmid instability phenotype (Figure 5B). Our results show that MMS21 allelic differences contribute significantly to the 2μ instability of the Y9 strain. These data are also consistent with Y9 plasmid instability being due to dominant plasmid restriction, rather than a haploinsufficient permissivity factor. However, MMS21 does not explain the entire Y9 phenotype, consistent with plasmid instability segregating as a multigenic trait through the cross. The remaining trait-determining loci in Y9 likely include some of the minor QTL peaks we found but could also include linked polymorphisms within the chromosome V region.

The phenotypic difference between hemizygous strains (Figure 5B) could be due to the Thr69Ile coding polymorphism found in Y9 and Y12 strains relative to BY4742, or due to regulatory differences, or both. We applied the same comparative genomics approach that initially identified Thr69Ile to the intergenic (candidate regulatory) regions flanking MMS21. In addition to the missense polymorphism, the Y9 strain differs from BY4742 at four synonymous sites within the MMS21 ORF and at a total of 11 sites in the two flanking regions (Figure 5—figure supplement 1). Y9 and Y12 are identical at all 16 of these sites. We next examined the closest outgroup strain, UC5, which still bears plasmids according to our PCR survey (Figure 2—figure supplement 1B). We found that UC5 differs from Y9 and Y12 at only three sites in the MMS21 locus: two synonymous SNPs and the single missense SNP. Thus, only these three sites strictly correlate with the plasmid instability phenotype, whereas the intergenic SNPs do not. Based on this finding, we chose to focus on the MMS21 Thr69Ile variant. However, regulatory differences between Y9 and BY4742 may still contribute to natural variation in plasmid stability.

MMS21 natural variation within S. cerevisiae and between sensu stricto species

The Thr69Ile change found in Y9 and Y12 strains is not found in the third non-permissive strain, oak YPS1009, suggesting that YPS1009 acquired 2μ plasmid instability through an independent evolutionary path. The Thr69 allele found in the BY4742 lab strain appears to be the ancestral allele, with Ile69 arising more recently in a subset (96) of 1,011 s. cerevisiae strains that were sequenced as part of a recent large-scale study (Peter et al., 2018). This study also reported which of the 1011 strains carry 2μ plasmids. Upon reanalyzing these data, we find that a smaller proportion of S. cerevisiae strains homozygous for the MMS21 Ile69 allele harbor 2μ plasmids compared to strains homozygous for the ancestral Thr69 allele (Figure 5C). In particular, the A-type 2μ plasmids, which we have tested using SCAMPR in this study, appear to be particularly depleted in strains with the Ile69 allele, suggesting that this allele might specifically restrict A-type 2μ plasmids (Figure 5C). While interesting, at present we cannot distinguish whether these observations are a result of a causal association or of shared evolutionary history, due to phylogenetic relatedness of the Ile69 allele-encoding strains.

To explore natural variation in MMS21 beyond S. cerevisiae, we aligned sequences from selected S. cerevisiae strains as well as other Saccharomyces sensu stricto species and two outgroups (N. castelli and N. dairenensis) (Figure 5—figure supplement 2). Interestingly, S. eubayanus and S. uvarum also seem to have independently acquired Ile69 but still harbor endogenous 2μ plasmids, whereas S. arboricola has yet another amino acid (alanine) at this position (Strope et al., 2015).

The location of the Y9/Y12 amino acid change in the Mms21 protein also provides important clues to its functional consequences. The Thr69Ile change occurs in the third of three alpha-helices in the Mms21 N-terminal domain, which makes contact with the Smc5/6 complex, and is essential for yeast viability (Duan et al., 2009Figure 5—figure supplement 3). Yeast cells deficient for MMS21 show gross chromosomal segregation defects and die as large, multi-budded cells (Bermúdez-López et al., 2010). However, the C-terminal zinc finger RING domain responsible for sumoylation of substrates is dispensable for Mms21’s essential function (Duan et al., 2009). We therefore speculate that the non-permissive MMS21 allele may act by directly affecting the Smc5/6 complex rather than through its sumoylation function. However, these possibilities may be hard to distinguish because the SUMO ligase function of Mms21 also depends on its docking with the Smc5/6 complex (Bermúdez-López et al., 2015). We also examined the Y9 strain for polymorphisms in other members of the Smc5/6 complex. While there are other polymorphisms in Smc5/6 complex members, we observed none in the regions of Smc5 that interact with any segment of Mms21 (Duan et al., 2009Supplementary file 4). Despite the Smc5/6 complex’s essential role in the removal of DNA-mediated linkages to prevent chromosome missegregation and aneuploidy, it has not been directly implicated in 2μ stability. Our finding that a single polymorphism at the Mms21-Smc5 interaction interface reduces 2μ stability thus reveals a novel facet of host control.


In this study, we leveraged natural variation to identify a gain-of-function variant that restricts 2μ plasmids in S. cerevisiae. Our approach is complementary to the traditional biochemical and genetic approaches that have previously used loss-of-function genetic analyses to study host regulation of 2μ plasmids. Natural variation studies can identify alleles of host genes that retain host function but still block SGEs like 2μ plasmids. Such studies can reveal novel mechanisms of host control, which may be otherwise challenging to discover via loss-of-function analysis.

Although 2μ-based vectors have long been used as an important tool in yeast genetics, studies of 2μ plasmids as natural SGEs have lagged behind considerably. Our new phenotyping assay, SCAMPR, makes the 2μ plasmid a more tractable system. SCAMPR captures single-cell data that facilitate studies of population heterogeneity, allowing inferences of the mechanisms by which plasmids may be controlled by their hosts. Although recent advances in single-cell genome sequencing make it possible to directly sequence and infer copy number of 2μ plasmids, this would be prohibitively expensive compared to the GFP-based flow cytometry profiling methodology we use. SCAMPR has potential for expanded use, for example to explore meiotic plasmid transmission dynamics. SCAMPR could also be paired with host lineage tracking to assess plasmid fitness burden alongside plasmid loss dynamics in competitive fitness assays. This paired strategy would provide a powerful approach for understanding the relative contribution of both plasmid fitness cost and host-plasmid incompatibility across hosts. In general, SCAMPR could be utilized to study high-copy number SGE plasmid dynamics, DNA replication and segregation, in any system where expression is well matched to copy number.

Our survey of 52 wild S. cerevisiae isolates identified three strains that naturally lack 2μ plasmids. Detailed studies of one of these strains, Y9, revealed that 2μ plasmid instability is heritable, dominant and likely the result of multiple contributing alleles. Through QTL mapping by bulk segregant analysis, we identified a significant locus on chromosome V associated with 2μ plasmid loss. We found that a single amino acid variant in Y9 MMS21, which encodes an essential SUMO E3 ligase in S. cerevisiae, contributes to 2μ plasmid instability. MMS21 does not fully account for the 2μ plasmid loss phenotype in heterozygous BY4742/Y9 strains. This result is unsurprising based on our tetrad analysis and QTL mapping, which both suggest that additional independently segregating loci affect plasmid stability. Although loss of Y9 MMS21 from heterozygous diploids leads to a relatively modest effect on 2μ plasmid instability, it may still account for all of the QTL signal we observe in chromosome V. Alternatively, the QTL on chromosome V could contain additional determinants of plasmid instability in close genetic linkage to MMS21, either in coding or regulatory sequences. CRISPR-Cas9 based approaches will be useful to test a large number of genomic changes rapidly and in parallel between Y9 and BY4742 to identify other determinants of plasmid instability in this QTL and in other candidate loci (Sadhu et al., 2016; Sharon et al., 2018).

The reciprocal hemizygosity experiment (Figure 5B) reveals that the Y9 variant of MMS21 acts dominantly to restrict mitotic stability of 2μ plasmids, ruling out the possibility that this allele is haploinsufficient. We considered two scenarios by which this allele may exert its dominant effect on plasmid stability. The first scenario is that the Y9 allele of MMS21 encodes a dominant-negative allele, which impairs the function of the Smc5/6 complex whether in a haploid or heterozygous state. Although this impairment does not negatively impact essential host functions when 2μ plasmid is absent (Y9 strains are viable and fit), this allele would have a fitness deficit in the presence of 2μ plasmids, as host functions become overburdened when hijacked by the parasite. Under this scenario, we would expect to see a greater fitness loss due to 2μ plasmid presence in Y9 compared to BY4742. However, both Y9 and BY4742 strains suffer an equal fitness loss in the presence of 2μ reporter plasmids (Figure 2—figure supplement 2). Moreover, plasmid-restrictive host alleles would only arise and propagate in natural populations if their fitness cost did not outweigh the modest 1–3% fitness cost imposed by the widespread 2μ plasmids in S. cerevisiae populations. We therefore favor a second scenario, in which the Y9 MMS21 allele represents a separation-of-function allele that is still capable of performing host functions but impairs 2μ mitotic stability.

The 2μ plasmids appear to have co-evolved with budding yeasts for millions of years and are prevalent in species such as S. cerevisiae. Long-term coevolution appears to have ‘optimized’ 2μ plasmids as tolerable parasites: not too great of a burden on host fitness, but still high enough plasmid copy numbers to ensure stable propagation. This copy number balance is achieved through both plasmid (e.g., Flp1 repression) and host (e.g., sumoylation) contributions. Nevertheless, there are hints that this truce between 2μ plasmids and yeast may be uneasy. 2μ plasmid stability is frequently compromised in heterospecific (other species) hosts, suggesting it is actively adapting to maintain stability within its native host species (Murray et al., 1988). Our discovery of a natural host variant of S. cerevisiae that impairs conspecific (same species) 2μ plasmid stability further supports the hypothesis that even the low fitness costs imposed by 2μ plasmids are sufficient to select for host evolutionary resistance.

Most studies of 2μ plasmid plasmids (including this one) have focused on the A-type variant that is most commonly found in laboratory strains. However, new sequencing studies have revealed that S. cerevisiae strains harbor a diverse set of 2μ plasmid plasmids (Peter et al., 2018; Strope et al., 2015). This diversity of 2μ plasmid plasmids might itself have arisen as a result of host defenses within S. cerevisiae, leading to plasmid diversification. For instance, although SCAMPR studies revealed the importance of the Y9 MMS21 variant against the stability of the A-type plasmid, it is possible that this variant is ineffective against the other 2μ genotypes. Thus, 2μ plasmids might exist in a frequency-dependent regime with their budding yeast hosts; A-type plasmids might thrive in certain host genetic backgrounds whereas B-type plasmids might thrive in others. The simultaneous presence of multiple 2μ plasmid types within species could explain the presence of standing variation in plasmid instability phenotypes in S. cerevisiae populations, including the low observed frequency of the Y9 MMS21 allele.

Testing the effects of MMS21 and other restrictive alleles on stability of different 2μ plasmids (e.g. B- or C-type) would provide a means to distinguish between universal versus plasmid-type-specific restriction. Future studies could employ SCAMPR to study the functional consequences of the natural diversity of 2μ plasmids in yeast. In particular, SCAMPR reporters from different types of S. cerevisiae 2μ plasmids and from divergent Saccharomyces species may reveal important biological determinants behind their co-evolution and long-term success in budding yeast species.

The identification of MMS21 led us to initially suspect that this locus might represent another connection between the SUMO-ligation machinery and 2μ plasmid stability (Dobson et al., 2005; Zhao et al., 2004; Pinder et al., 2013; Ma et al., 2019). However, the location of the Thr69Ile missense change at the binding interface between Mms21 and Smc5 (Figure 5—figure supplement 3) suggested a mechanism that relies on the Smc5/6 complex rather than the catalytic RING domain. Interestingly, a recent study demonstrated that even the SUMO ligase function of Mms21 depends on its docking with the Smc5/6 complex (Bermúdez-López et al., 2015). Thus, the polymorphism in the Mms21-Smc5 interaction site could either affect Mms21’s SUMO ligase function or other functions of the Smc5/6 complex. The Smc5/6 complex lies at the nuclear periphery, where it anchors dsDNA breaks to facilitate repair, resolves X-shaped DNA structures that arise during DNA replication and repair, and helps mediate sister chromatid cohesion (Bermúdez-López et al., 2010). All three of these cellular processes might directly impact stability of 2μ plasmids (Velmurugan et al., 2000). Alteration of Mms21 function, an essential component of the Smc5/6 complex, could thus directly affect both segregation of 2μ plasmids as well as interfere with their amplification via Flp1-induced recombination intermediates.

Although Smc5/6 has not been previously implicated in 2μ stability, this complex is involved in the stability of viral episomes, as human Smc5/6 acts as a restriction factor against hepadnaviruses such as human Hepatitis B virus (Decorsière et al., 2016; Murphy et al., 2016). To counteract this restriction function, diverse hepadnaviruses encode antagonist HBx proteins that degrade mammalian Smc5/6 and restore viral fitness (Decorsière et al., 2016; Murphy et al., 2016; Abdul et al., 2018). Our findings that components of the yeast Smc5/6 complex affect 2μ stability suggest that the Smc5/6 complex might provide a general mechanism to protect host genomes from the deleterious consequences of multicopy genetic parasites.

Materials and methods

Strain growth and construction

Request a detailed protocol

Yeast strains were grown in standard yeast media at 30°C unless otherwise noted (Amberg et al., 2005). Transformations were carried out using a high-efficiency lithium acetate method (Amberg et al., 2005). The GFP-2μ plasmid was created by Gibson assembly directly into otherwise plasmid-less yeast strains cir0 BY4741 (MATa haploid) and BY4742 (MATα haploid), which had been cured of their endogenous plasmids by previously published methods (Tsalik and Gartenberg, 1998; Gibson, 2011). To avoid disruption of the plasmid’s endogenous replication and segregation machinery, a cassette containing both markers was integrated into the A-type 2μ sequence found in the S. cerevisiae laboratory strain BY4741 at a restriction site reported to tolerate insertions of up to 3.9 kb without impacting copy number or stability (Ludwig and Bruschi, 1991). We did not use any bacterial cloning vector sequences to minimize unnecessary or destabilizing changes to the 2μ reporter plasmid, so the reporter plasmid was directly assembled in yeast. Assembling in cir0 strain backgrounds avoided multiple plasmid genotypes within a strain background that could have led to plasmid competition or recombination.

We used the NEBuilder HiFi DNA Assembly Master Mix (product E2621) for Gibson assemblies. Yeast plasmids were recovered using Zymoresearch Zymoprep Yeast miniprep kits (D2004). The assembled plasmids were then retransformed to the same cir0 yeast backgrounds to ensure plasmid clonality. Genetic crosses were carried out on a Singer Sporeplay dissection scope, for both tetrad dissection and selection of unique zygotes for mating strains. Strain mating type was confirmed by halo formation in the presence of known mating type tester strains. Strains used in this work are listed in Supplementary file 1.

Natural isolates were obtained as homothallic diploids (capable of mating type switching and self-diploidization). We made stable heterothallic haploid strains (no longer capable of mating type switching) by first knocking out ho endonuclease prior to sporulation (hoΔ::HphNT1). We found that the natural isolates required significantly longer homology arms for proper DNA targeting when making integrated genomic changes (e.g. gene deletions) via homologous recombination. Where BY4742 lab strains utilized ~50 bp homology arms for high efficiency recombination, Y9 required ~1 kb flanking homology. Even with longer homology, a substantial number of clones in any transformation did not contain the desired edit. These hurdles made editing the Y9 genome challenging.

Growth rate measurements

Request a detailed protocol

Growth rate measurements were performed in 96-well plates in Biotek Powerwave incubating plate readers with Gen5 software at 30°C with shaking. Logarithmically dividing cells were seeded at approximately 2000 cells in 100 ul of defined medium per well, with or without selection as indicated. Synthetic complete media was prepared with monosodium glutamate as the nitrogen source to facilitate G418 selection. Twelve replicate wells were run for each strain background and OD660 measurements were taken every 10 min until each well reached saturation. Data were trimmed to time points during log phase growth, and replicate data points are plotted as the mean value of replicates, with error bars indicating SD. The outside wells of the plate were incubated as media blanks to measure baseline OD660 and to help monitor and minimize evaporation during the runs.

Colony sectoring

Request a detailed protocol

Confirmed transformants were cultured under G418 selection, then plated to YPD medium where colonies were allowed to form without selective pressure to maintain the reporter plasmid. After 2 days growth at 30°C, colonies were imaged under white light and GFP excitation to assess qualitative plasmid loss in the different strains using a Leica M165 FC dissection scope with a GFP filter and Leica DFC7000 T camera. Colony sectoring was visually assessed. We then performed image processing using ImageJ to split channels and recolor the GFP channel.

MCM assay

Request a detailed protocol

MCM assays were performed as previously published (Maine et al., 1984). However, samples were taken at only two time points. Therefore, we reported changes in frequency of 2μ plasmid rather than an estimated rate of loss per generation. This two-timepoint measurement also provided a more direct comparison to the SCAMPR assay. At time = 0 hr and 24 hr, cells were plated on both selective and non-selective media to determine what fraction of the population maintains the plasmid by virtue of encoding the selectable marker. Samples were plated at multiple dilutions to ensure between 30–300 CFU per plate. All strains containing the reporter plasmid were grown under G418 selection to ensure 2μ plasmid presence prior to the start of the assay. At time = 0 hr, cultures were transferred into liquid media with shaking, but without drug selection for 24 hr. After 24 hr, cultures were diluted in PBS and plated on YPD either with or without G418 selection at multiple dilutions, targeting 30–300 CFU per plate. Plates were incubated for 2 days, then colonies were manually counted to determine what fraction of the population were G418 positive. Calculations were based on whichever dilution gave a countable (30–300 CFU) plate. Multiple replicates (at least 8) were performed for each strain to measure variability in plasmid retention. A subpopulation of GFP-negative, G418-negative cells can be found even under selection. This ‘phenotypic lag’ occurs because of protein persistence following DNA loss; while cells that lose the plasmid die under selection, new plasmid-free cells are constantly generated as well. We therefore normalize data to account for different starting frequencies of plasmid-negative cells by comparing cells grown with or without G418 selection for 24 hr relative to their starting frequency (Figure 1B).


Request a detailed protocol

SCAMPR samples were prepared as for MCM assays. When grown in 96-well format at 30°C, cultures were shaken using a Union Scientific VibraTranslator to ensure aeration. Fluorescence was directly measured by flow cytometry at 0 and 24 hr timepoints. A BD Canto-2 cytometer was used to collect cell data. FlowJo software was used for subsequent data analysis: samples were gated for single cells, omitting doublets/multiple cell clumps and any cell debris. Single cells were gated for GFP-positive and -negative populations, using GFP-negative strains and single-copy integrated GFP-positive strains as gating controls. Summary statistics (frequency of GFP-positive and -negative cells, GFP intensity) were exported from FloJo. Each strain was measured in at least triplicate per assay and means are reported here.

Statistical analyses

Request a detailed protocol

For SCAMPR and MCM assay results, we determined significance by non-parametric tests. To compare two strains we used two-tailed Mann-Whitney tests, and to compare three or more strains, we used Kruskal-Wallis with Dunn’s multiple comparison tests. Graphs were prepared and statistical analysis done using GraphPad Prism seven software.

Screening for endogenous 2μ plasmid in natural isolates of S. cerevisiae

Request a detailed protocol

Natural isolates (Supplementary file 1) were generously shared by Dr. Justin Fay. DNA from these strains was isolated using a standard Hoffman and Winston preparation method, then probed by PCR and Southern blot (Amberg et al., 2005). Two pairs of primers were designed to amplify either REP1 or FLP1 (FLP1_F: CCACAATTTGGTATATTATG, FLP1_R: CTTTCACCCTCACTTAG, REP1_F: AATGGCGAGAGACT, REP1_R: CGTGAGAATGAATTTAGTA), the two best conserved coding regions of the plasmid, as previously described (Xiao et al., 1991a). Only strains that showed negative PCR results for both sets of primers were further validated by chemiluminescent Southern blot using the Thermo North2South kit (17097). Briefly, whole genome DNA was digested, run on an agarose gel in TAE, transferred to membrane and probed with chemiluminescent probes created from digested endogenous 2μ plasmid collected from BY4741 by Zymoresearch yeast plasmid miniprep kit (D2004). Gels and blots were imaged on a Bio-Rad ChemiDoc.

Illumina sequencing, library preparation, and QTL mapping via bulk segregant analysis

Request a detailed protocol

We prepared high quality genomic DNA for sequencing using Zymoresearch Yeastar kits according to the manufacturer’s instructions (D2002 - using chloroform method). Sequencing libraries were prepared using the TruSeq method for genomic DNA (Illumina), multiplexed and run on an Illumina HiSeq by the Fred Hutchinson Sequencing core facility to generate 50 bp paired-end sequences (SRA accession PRJNA637093). 100 bp paired-end reads for the lab strain, BY4742, were downloaded from the SRA database (accession SRR1569895). Reads that failed Illumina’s ‘chastity filter’ were removed using a custom R script, and adapters and low-quality regions were trimmed using cutadapt with parameters -q 10 --minimum-length 20 (Martin, 2011). Trimmed read pairs were aligned to the sacCer3 reference genome assembly using BWA-backtrack (Li and Durbin, 2009). Mean coverage in non-overlapping 20 kb windows across the genome was calculated and plotted using R and Bioconductor.

For bulk segregant analysis, we first identified a conservative set of 47,173 high quality SNPs that distinguish the cross parents (Y9 and BY4742) as follows. Before SNP-calling, BWA output files were processed using Picard’s MarkDuplicates tool and indels were realigned using GATK’s RealignerTargetCreator and IndelRealigner tools (Picard, 2020; DePristo et al., 2011). We then called SNPs using samtools mpileup (parameters --skip-indels -t DP -uBg -d 6660) and bcftools call (parameters -vmO z -o), and counted reads matching each allele using GATK’s VariantAnnotator DepthPerAlleleBySample module (with --downsampling_type NONE option) (Li, 2011). We used R and Bioconductor to further filter SNPs to obtain the final set of 47,173 SNPs, removing any that overlapped repetitive elements, SNPs with QUAL score <200, SNPs with unusual coverage in any sample, and SNPs with an apparent mix of alleles in either of the haploid parental strains. We then ran MULTIPOOL in ‘contrast’ mode on allele frequencies at each SNP in the permissive and non-permissive pools to generate LOD scores across all chromosomes (Edwards and Gifford, 2012).

To identify candidate functional polymorphisms in each sequenced strain, we took two approaches: (a) we performed more sensitive SNP-calling, including small insertions and deletions; (b) to detect larger insertion/deletion events, we generated de novo assemblies from each strain, aligned them to the reference genome assembly, and identified locations where assemblies differed. In more detail, the first approach used processed alignments (see above) as input to GATK’s HaplotypeCaller (parameters -stand_call_conf 30.0 -stand_emit_conf 10.0) (DePristo et al., 2011). Functional consequences of each variant were annotated using Ensembl’s Variant Effect Predictor (McLaren et al., 2016). For the second approach (de novo assemblies), we performed error correction on the adapter-trimmed reads using musket (parameters -k 28 536870912) and then used SOAPdenovo2 across a range of k-mer sizes and fragment sizes, choosing the combination for each sample that yielded the assembly with highest N50 length as determined using QUAST (Liu et al., 2013; Luo et al., 2012; Gurevich et al., 2013). These Whole Genome Shotgun projects have been deposited at DDBJ/ENA/GenBank under the accessions JABVXK000000000, JABVXL000000000, JABVXM000000000, JABVXN000000000, JABVXO000000000 and JABVXP000000000. The versions described in this paper are versions JABVXK010000000, etc. We obtained tiling path alignments of each assembly to the sacCer3 reference genome assembly using MUMMER (nucmer parameters -maxmatch -l 100 c 500, delta-filter options -m) (Kurtz et al., 2004). Structural variants were determined from genome alignments using Assemblytics (variant size range 1 bp-100kb) (Nattestad and Schatz, 2016).

We identified all missense polymorphisms in the chrV 90% credible interval where genotype was shared between the non-permissive Y9 and Y12 strains, but distinct from the closely-related plasmid-permissive strain, UC5, and the permissive laboratory strain, BY4742 (Supplementary file 3). In addition, we expanded our analysis to include all candidate regulatory and synonymous polymorphisms at the MMS21 locus (Figure 4—figure supplement 4).

Structure visualization

Request a detailed protocol

The Cn3D viewer was used to visualize Thr69Ile on a crystal structure of MMS21 with SMC5 made available by Duan et al., 2009; Wang et al., 2000.

Analysis of MMS21 natural variation

Request a detailed protocol

To examine natural variation in MMS21 across S. cerevisiae strains and in other fungal species, we first extracted the MMS21 (YEL019C) open reading frame from the reference assembly (sacCer3, chrV:120,498–121,301, - strand) and translated that sequence. We then used this MMS21 protein sequence as the query in tblastn searches against various databases (Altschul et al., 1997). Searching the NR database, using taxonomic restrictions as needed, yielded MMS21 sequences from S. paradoxus (XM_033909904.1), S. eubayanus (XM_018364578.1), S. jurei (LT986468.1, bases 125,344–126,147, - strand), S. kudriavzevii (LR215939.1, bases 100238–101041, - strand), N. castellii (XM_003677586.1) and N. dairenensis (XM_003671024.2). For additional orthologs, we downloaded individual genome assemblies from NCBI and performed local tblastn searches for S. arboricola (GCA_000292725.1; MMS21 at CM001567.1:97,893–98,699, - strand), S. mikatae (GCA_000166975.1; MMS21 at AABZ01000034.1:31,665–32,468, - strand) and S. uvarum (GCA_002242645, MMS21 at NOWY01000012.1:107,550–108,353, + strand). Additional S. cerevisiae strain sequences come from our own de novo assemblies, where we used blastn to identify MMS21.

MMS21 genotypes in the 1011 isolates previously sequenced (Peter et al., 2018) were accessed via that publication’s supplementary data file 1011Matrix.gvcf.gz. Plasmid status was obtained from another supplementary file (Supplementary file 1) and cross-referenced with genotype in R.

Data availability

Raw sequencing data have been deposited to the SRA database, accession PRJNA637093. De novo assemblies are in GenBank with accessions JABVXK000000000, JABVXL000000000, JABVXM000000000, JABVXN000000000, JABVXO000000000 and JABVXP000000000.

The following data sets were generated
    1. Hays M
    2. Young JM
    3. Levan PF
    4. Malik HS
    (2020) NCBI BioProject
    ID PRJNA637093. Natural variation among Saccharomyces cerevisiae strains in resistance to 2-micron plasmid.
    1. Hays M
    2. Young JM
    3. Levan PF
    4. Malik HS
    (2020) NCBI Nucleotide
    ID JABVXK000000000. Saccharomyces cerevisiae strain Y9_Hap1, whole genome shotgun sequencing project.
    1. Hays M
    2. Young JM
    3. Levan PF
    4. Malik HS
    (2020) NCBI Nucleotide
    ID JABVXL000000000. Saccharomyces cerevisiae strain Y12, whole genome shotgun sequencing project.
    1. Hays M
    2. Young JM
    3. Levan PF
    4. Malik HS
    (2020) NCBI Nucleotide
    ID JABVXM000000000. Saccharomyces cerevisiae strain NC-02, whole genome shotgun sequencing project.
    1. Hays M
    2. Young JM
    3. Levan PF
    4. Malik HS
    (2020) NCBI Nucleotide
    ID JABVXN000000000. Saccharomyces cerevisiae strain YPS1009, whole genome shotgun sequencing project.
    1. Hays M
    2. Young JM
    3. Levan PF
    4. Malik HS
    (2020) NCBI Nucleotide
    ID JABVXO000000000. Saccharomyces cerevisiae strain PW5, whole genome shotgun sequencing project.
    1. Hays M
    2. Young JM
    3. Levan PF
    4. Malik HS
    (2020) NCBI Nucleotide
    ID JABVXP000000000. Saccharomyces cerevisiae strain UC5, whole genome shotgun sequencing project.
The following previously published data sets were used
    1. Song G
    2. Stanford University
    (2014) NCBI Sequence Read Archive
    ID SRR1569895. Saccharomyces cerevisiae strain genomes commonly used in laboratories.


  1. Book
    1. Amberg DC
    2. Burke DJ
    3. Strathern JN
    Methods in Yeast Genetics: A Cold Spring Harbor Laboratory Course Manual
    Cold Spring Harbor Laboratory Press.
  2. Book
    1. Burt A
    2. Trivers R
    Genes in Conflict
    Harvard University Press.
  3. Book
    1. Dawkins R
    The Selfish Gene
    Oxford University Press.
    1. Lande R
    The minimum number of genes contributing to quantitative variation between and within populations
    Genetics 99:541–553.
    1. Lander ES
    2. Botstein D
    Mapping mendelian factors underlying quantitative traits using RFLP linkage maps
    Genetics 121:185–199.
    1. Maine GT
    2. Sinha P
    3. Tye BK
    Mutants of S. cerevisiae defective in the maintenance of minichromosomes
    Genetics 106:365–385.
    1. Scholes DT
    2. Banerjee M
    3. Bowen B
    4. Curcio MJ
    Multiple regulators of Ty1 transposition in Saccharomyces cerevisiae have conserved roles in genome maintenance
    Genetics 159:1449–1465.

Decision letter

  1. Christian R Landry
    Reviewing Editor; Université Laval, Canada
  2. Detlef Weigel
    Senior Editor; Max Planck Institute for Developmental Biology, Germany

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

[Editors' note: this paper was reviewed by Review Commons.]

Acceptance summary:

Parasitic genetic elements offer a fascinating framework for the study of evolution as they battle with their hosts. In this paper, the authors identify mutations in yeast that may tilt the balance in favour of the host by increasing the loss rate of a parasitic selfish plasmid. This study lays the foundation for a new study system to better understand the evolutionary arms race between genetic parasites and their hosts at the molecular level.

Decision letter after peer review:

Thank you for submitting your article "A natural variant of the essential host gene MMS21 restricts the parasitic 2-micron plasmid in Saccharomyces cerevisiae" for consideration by eLife. Your article has been reviewed by three peer reviewers at Review Commons, and the evaluation has been overseen by a Reviewing Editor and Detlef Weigel as the Senior Editor.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

In this manuscript, the authors examine the genetic basis for parasitic plasmid resistance in budding yeast. They develop a new method that allows to measure plasmid copy number and retention at the single cell level. They leverage this tool to map QTLs underlying variation in plasmid retention or stability among strains of yeast. They identify a locus that explains a significant fraction of the variation found among strains. The candidate locus is involved in biological processes that have been associated with plasmid transmission in previous works. The study is of interest from a technical and conceptual standpoint and will be of broad interest.

From reading the comments, responses and the manuscript in details, it seems that two comments remain to be resolved.

First, you mention that you could include in the manuscript the results shown in response to reviewer #1, comment 4. Since a similar comment was made by another reviewer, it would indeed be important to include this analysis in a revised submission. In addition, the figure as it is now in the rebuttal is difficult to read.

Second, reviewer #1 points out in their comment #6 that allele replacement would be important to confirm the role of the candidate amino acid substitution. I understand that the paper itself goes beyond the identification of a precise variant and that the current COVID-19 conditions limit your access to lab facilities. However, since this claim appears to be key for some of the downstream analyses on the 1001 genomes and on closely related species, and for the discussion about the mechanisms by which the plasmid is unstable in Y9, this appears to be a key experiment. One of the sub-section title states for instance that this single amino acid variant in MMS21 contributes to mitotic instability of the plasmid in Y9.

Here are some observations that make the contribution of this amino acid substitution uncertain. First, the Y9 instability appears to be dominant as measured in the heterozygote whereas the addition of the Y9 allele in BY does not affect its phenotype. Second, on Figure 5C, I appreciate that the phenotype of Het-Y9 MMS21 is different from the Het, but a relevant comparison here would be Het-Y9 MMS21 versus Het-4742 MMS21. These two strains are the ones for which the only difference is the nature of the allele at the locus and not the number of alleles. The statistical test for this difference is not reported but it is likely not significant given what is shown on the figure. The results shown could for instance (and one could think of other dosage effects) be consistent with a difference in expression of the two alleles acting in cis, which would also alter the phenotypes depending on which allele is deleted. Since the analysis is focusing on missense variants, regulatory variants have not been considered so this cannot be ruled out. I therefore believe that allele replacement would be needed to confirm the identity of the causal variant.


Author response

We would like start by thanking all three reviewers for their time, effort and extremely helpful comments. This has been a very constructive review process and we are grateful.

Reviewer #1 (Evidence, reproducibility and clarity):

The paper by Hays et al. addresses the evolution of host genomes to eliminate parasitic selfish genomes that coevolve with them. They use the multi-copy 2-micron yeast plasmid as a model selfish genome to address this issue. Using a series of cleverly designed experiments, the authors identify a high confidence quantitative trait locus (QTL) on chromosome V in the Y9 yeast strain associated with mitotic plasmid instability. Furthermore, they demonstrate Thr69Ile substitution in MMS21 (a SUMO E3 ligase) to be a heritable dominant determinant for plasmid exclusion. The work is significant for its conceptual and methodological advances.

Although the multigenic nature of the commensal relationship between the plasmid and host would make the task rather arduous, the present study exemplifies the utility of natural variations in understanding the mechanisms of genetic conflicts and their resolution during the coevolution of selfish genetic elements and their hosts.

We thank the reviewer for their kind comments and constructive suggestions and corrections, which we have now incorporated into our revision as detailed below.

1) SCAMPR; Figure 1, Relevant text. The fluorescence based single cell high throughput analysis of plasmid retention, SCAMPR, is an important advancement. However, the use of reporter plasmids containing insertion of markers that minimally disturb native plasmid organization, and containing no bacterial vector sequences, has been described earlier.

We thank the reviewer for pointing out this oversight. We have now modified the main text to specify that our strategy was not the first to mark the 2μ plasmid without bacterial sequences.

2) Figure 1C and the relevant text. For total missegregation of the GFP-plasmids, the mother cell should contain 2 x n plasmid copies, and therefore twice the GFP intensity. The schematic diagram in Figure 1C does not seem to depict this increase in copy number in plasmid retaining cells. The legend suggests that “no change” in the median GFP intensity in plasmid containing cells is expected. And according to the text, the copy number is not too high to saturate the signal detector. Perhaps some clarification here might help. Is there potential suppression of plasmid replication at high copy numbers? Or do such cells have a growth disadvantage and are selected against?

The reviewer raises an important question and one that we tried to address in our main text but have revised for clarity: we do not see a population of super-green cells. We speculate that this is because these cells are not fit. Lack of plasmid segregation is likely to result in cumulatively higher levels of plasmid retention in “mother” cells as the reviewer points out, and others have shown that high copy numbers of the plasmid are detrimental to the host. However, we agree with the reviewer that the SCAMPR schematic can make clear that this may be why we do not see the super-green cells that we would expect to see under a segregation defect. We have edited both the main text and amended the figure to reflect this important point, which was also raised by reviewer #2. Thank you for helping us to improve the clarity of this point.

3) Related to point 2 above. One wonders whether the GFP-plasmid is capable of Flp mediated amplification and copy number correction? This aspect has not received scrutiny, understandably because of the emphasis of this study on “plasmid resistance/exclusion”. In principle, there is nothing to preclude plasmid amplification by the generally accepted Futcher-Volkert-Broach mechanism. Copy number correction in the case of unequal segregation might be responsible for the lack of shift in median GFP intensity. One is curious to know whether a plasmid harboring inactive FLP or a nonfunctional FRT target site would exhibit a different behavior.

This is an excellent point. We believe that the modified reporter plasmid should be fully capable of FLP copy number repair because of the insertion location (we now make this point more explicitly when we cite the previous reference about our choice of insertion site). Although we do not test this possibility directly, we expect a FLP minus plasmid would be less stable in both the lab strain (as others have observed) and Y9 cells. Since we wished our reporter to be as similar as possible to the native 2μ plasmid (to focus on host interference with normal plasmid function), we did not study mutations in Flp or FRT sites, but this is an exciting research avenue to pursue in the future.

4) Figure 2. From panel A of the figure, plasmid-free colonies appear to be larger than plasmid containing colonies – more so for the Y9, Y12 and YPS1009 strains than BY4742. The phenotype would be consistent with a growth disadvantage of plasmid containing cells in the first three strains. The effect does not appear to be caused by GFP expression/toxicity as the size difference is less apparent for BY4742.

We thank the reviewer for this observation, which was also commented on by reviewer #2. Indeed, as the reviewer suggests, this could suggest an increase in plasmid fitness cost in the wild strains’ backgrounds. Since it can be difficult to decouple 2μ plasmid instability from a growth advantage of plasmid-lacking strains, we performed growth curve assays of Y9 and BY4742 haploid strains either lacking or containing plasmids

(with G418 selection to maintain plasmids). We found that the growth curves from plasmid-lacking BY4742 strains were indistinguishable from Y9 strains. Moreover, both strains had lower fitness in the presence of 2m plasmids (and G418 selection) but this fitness drop was also indistinguishable during log phase growth. This result argues against a strong Y9-specific plasmid carriage cost. If recommended by the editor, we would be happy to include this result in our revision.

Of course, it is possible that there is a more subtle relative fitness cost that can only be measured under competitive co-culture conditions, but this would require significant re-engineering of strains. This is why we leave open this possibility in the revised Discussion.

5) Text related to Figure 3—figure supplement 1. If plasmid segregation is strongly coupled to chromosome segregation, in accordance with the hitchhiking model, it would be quite unlikely that plasmid exclusion would be a monogenic effect. I think the present results are consistent with the interaction of MMS21 with the Smc5/6 complex, which appears to have pleiotropic effects on chromosome dynamics and function.

We agree with the reviewer; there are many possible genes that could pleiotropically affect both host chromosome stability and plasmid stability.

6) Figure 5A and related text. Is it possible to replace the MMS21(Thr69) in BY4742 with MMS21(Ile69) from Y9? Is that substitution compatible with normal chromosome functions? The expectation based on the hemizygous diploid experiments is that the Y9 allele would destabilize the GFP-plasmid. Based on the results from Figure 5B, either MMS21(Thr69) or MMS21(Ile69) is functional in the diploid background.

The reviewer is right on several points. Our hemizygous MMS21 experiment in Y9/BY4742 heterozygous diploids demonstrates that either MMS21 (Thr69 or Ile69) is functional in the het. diploid background, at least in terms of providing MMS21 essential function. Although we previously showed that addition of the Y9 MMS21 allele into an intact BY4742 background does not appear sufficient to significantly lower plasmid stability (Figure 5A), it is possible that the presence of the BY4742 MMS21 allele masked the Y9 allele’s plasmid instability phenotype. We can create a Y9 MMS21-only version of a BY4742 strain to assess viability differences.

We would also like to measure plasmid stability in this allele swap strain. However, due to COVID-related restrictions placed on our FACS facilities, SCAMPR is not possible at this time. The MCM assay is possible; however, the plasmid stability phenotype difference in the hemizygous mutants is subtle, even by SCAMPR (which samples tens of thousands of cells per assay rather than tens of colonies per plate). We therefore anticipate that this phenotypic difference would be overwhelmed by noise in the MCM assay. However, we would undertake these experiments if they are required by the editors.

7) The final paragraph of Results before Discussion. Based on the location of Thr/Ileu in MMS21, the proposed functional relevance of the MMS21 variants via Smc5/6 interaction would seem reasonable. Are there variations in the Smc5/6 complexes of BY4742 and Y9? And are the MMS21 variants able to functionally interact with both Smc5/6 complexes? This issue would be related to the point raised under 6 above.

In response to the reviewer query, we have now carried out a comprehensive analysis of variants in all members of the Smc5/6 complex which we now report in our revision (Supplementary file 4). However, we detect no polymorphism in the Smc5 sites that would likely interact with residue 69 of Mms21. Beyond these inferences, we are not able to make stronger statements about any epistatic interactions between MMS21 and any other components of the Smc5/6 complex in the absence of much more detailed biochemical analyses.

8) The SUMO E3 ligase activity of MMS21. It is implied in the text that the ligase activity of MMS21 is not involved in plasmid plasmid exclusion/retention. Has this been tested by inactivating mutations. To my knowledge, most of the published effects of MMS21 have been interpreted in terms of its E3 ligase activity in the context of the Smc5/6 complex. An E3 ligase independent function for MMS21 in commensalism would be quite interesting.

We think the reviewer is referring to our hypothesis that the Mms21 polymorphism is more likely to affect its interaction with Smc5/6 rather than the SUMO ring domain. We are excited by the prospect of follow-up studies to understand the mechanism of this polymorphism on Smc5/6 function. We feel these mechanistic experiments are outside of the scope of this current paper. We do more explicitly point out this possibility in our revised Discussion: “Interestingly, a recent study demonstrated that even the SUMO ligase function of Mms21 depends on its docking with the Smc5/6 complex. Thus, the polymorphism in the Mms21-Smc5 interaction site could either affect Mms21’s SUMO ligase functions or other functions of the Smc5/6 complex.”

If the editors require, we could explore plasmid stability in other MMS21 mutants, known to be deficient in sumoylation in others’ hands. However, there are several technical concerns we have about undertaking this experiment. MMS21 is an essential gene, so deleterious mutations in this gene are likely to cause substantial host fitness defects, which would confound any observed changes in plasmid stability. We again note that we are currently unable to perform SCAMPR due to COVID restrictions, so while we could follow this experiment up by MCM analysis, the sensitivity and sample size limitations of that assay (in conjunction with host viability concerns) are likely to make the results of this experiment challenging to interpret.

9) Is MMS21 function always coupled to Smc5/6? Although the location of the Thr69Ile location is consistent with MMS21 interaction with Smc5/6, there is no direct evidence to suggest that this interaction is involved in the plasmid stability/instability phenotype. Is there any known instance where MMS21 functions independently of Smc5/6?

We are not aware of any studies that have discovered non-SMC5/6 roles for Mms21. In contrast, a recent paper quite conclusively demonstrated that even the SUMO ligase functions of Mms21 are tied to its membership in the Smc5/6 complex (we now cite this paper more explicitly to make this point). Thus, the location of the polymorphism at the Smc5 interaction site is most consistent with Mms21's interaction with Smc5/6 function being affected.

Reviewer #1 (Significance):

The paper addresses the genetic determinants that influence the coevolution of host genomes and associated selfish DNA genomes by exploiting natural variations occurring in the host species. The problem and the approaches are of broad significance and interest.

We are grateful for the positive appraisal of the reviewer and their constructive suggestions to improve our paper.

Referees cross commenting :

The points raised in the other reviews are valid.

The issue that I am struggling with is the schematics in Figure 2 and the lack of change in median fluorescent intensity. The only way to reconcile the apparent discrepancy is to assume that cells with the high plasmid copy are eliminated from the population (selected against?). Nevertheless, if plasmid replication is not affected, in a daughter pair with one receiving all plasmid copies should be 2 n in copy number, which is not what the figure shows.

If I am missing something, could one of the other two reviewers correct me. please?

As we highlight in point 2 (above), the reviewer has raised an important issue that we previously only addressed in the text. We now expand on this possibility in the main text, have included this inference in Figure 1C, and explicitly state our hypothesis that cells with high plasmid copy number (super-green) are most likely not observed due to proliferation defects.

Reviewer #2 (Evidence, reproducibility and clarity):


The 2-micron plasmid is a selfish DNA element, typically found at high copy number in most lab strains of the yeast Saccharomyces cerevisiae. In this paper, a PCR survey of 52 natural isolates identified 3 strains lacking 2-micron plasmid. A novel single-cell high-throughput plasmid retention assay (SCAMPR) was developed and in combination with traditional yeast genetics, used to demonstrate plasmid instability in one of these was due to an inherited, dominant, multigenic trait. Next generation sequencing of pooled haploid segregants (identified as either permissive or non-permissive for plasmid maintenance using SCAMPR) allowed quantitative trait locus (QTL) mapping. A focus on candidate genes in the major QTL identified a single amino acid change (Thr69Ile) in a conserved essential SUMO E3 ligase, Mms21, as being associated with reduced 2-micron plasmid stability. Variation in Mms21 in species closely related to S. cerevisiae and correlation of the Ile69 allele with lower likelihood of possessing A-type 2-micron plasmids in S. cerevisiae was also reported. Mms21 is a subunit of the conserved Smc5/6 complex which plays an essential, but incompletely understood role in chromosome maintenance. Prior crystal structure data was used to identify the position of Thr69Ile in Mms21 as being in contact with the Smc5 subunit rather than in the SUMO ligase domain.

The paper is clearly written, and the key conclusions are well supported by the data provided. The data and methods are generally clearly presented and represent a useful template for others undertaking similar studies. Replicates and statistical analysis are adequate.

We are very grateful for the positive summary and constructive feedback from the reviewer.


1) On page 9, last line in 2nd paragraph in this section – the consequences of plasmid missegregation here are correctly stated as "some cells inheriting no plasmid, while others maintain or even increase plasmid copy number (Figure 1C)". However, in Figure 1C, the accompanying diagram is misleading as it shows copy number staying uniform in the cells that have plasmid when it would be expected to increase in these due to plasmid replication having doubled the copy number and lack of delivery to daughter cells due to missegregation. Similarly, the legend for this part of Figure 1 states that missegregation would "cause a rapid increase in GFP-negative cells but no change in the median expression of GFP-positive cells". The prediction should be no change or an increase in the median expression for the GFP-positive cells as stated. What is striking in Figure 1—figure supplement 1B is that the predicted increase it not seen. This suggests that for the Y9 strain, cells that received the higher than normal number of plasmids are very rapidly lost from the population. This would be consistent with the high number being toxic. Figure 1C and the legend should be amended to reflect the predictions and also to make it clearer that the plasmids are not lost from the cells, rather they are being lost from the cell population, due to failure to be delivered to daughter cells and presumably because plasmid-free cells rapidly out-compete plasmid-containing cells during the 24 hr when they are cultured in the absence of selection.

We completely agree with the reviewer, and appreciate the assistance in clarifying this point. This concern was also raised by reviewer #1 (see point 2) and we have revised the text and figure to incorporate these comments. We thank both the reviewers for pointing this inconsistency to us and suggesting remedies to improve its clarity and accuracy.

2) The deletion of the URA3 gene in BY4742 is referred to as a large indel. I appreciate that indel is the term used for genomic comparisons where the process leading to the difference is unknown but since we know that the URA3 gene was actively deleted in the construction of BY4742, this seems like a misleading term here.

We agree with the reviewer and have edited the language in our revision to reflect this.

3a) One part that would benefit from a bit more clarity is on page 17, where the assumption is made that since nibbled colonies were not observed when their GFP-tagged 2-micron plasmid was introduced into the Y9 cells, then any change to Mms 21function due to the Y9 allele is not synthetically lethal with high levels of the plasmid. I assume they mean the abnormally elevated plasmid copy number that occurs in sectors of ulp1 mutant colonies causing cells in those sectors to cease proliferating. A reader would be more apt to interpret the term "synthetic lethality" as being immediate inviability, or at least that the cells would not be able to form a viable colony, so this might not be the most helpful term here or at least requires clarification. For the Y9 strain, the combination of MMS21 allele and 2-micron can be a negative genetic interaction (synthetically sick) without giving rise to nibbled colonies.

We agree with this comment from the reviewer and have amended the text to be more precise with our description.

3b) There does seem to be a suggestion of this in Figure 2A where the Y9 colonies that express GFP seem on average to be smaller than those that do not. This contrasts with BY4742 where GFP-positive colonies similar in size to large GFP-negative ones are seen. It might be useful to compare the relative growth rates of Y9 versus BY4742, transformed with a 2-micron plasmid versus transformed with a CEN plasmid and versus both untransformed to see if there is a significant synthetic sick phenotype only when a 2-micron plasmid is present.

We thank the reviewer for this comment and observation. We note that reviewer #1 also made this observation (reviewer #1 point 4), which we now have addressed in our response to reviewer #1.

4) 2-micron plasmids are stated to "physically localize to the nuclear periphery" but the cited paper does not conclude this and other reports seem to be somewhat at odds as to localization, concluding that the plasmid is either spindle pole body-associated in mitotic cells (Mehta et al., 2005), spindle-associated (Velmurugan et al., 2000), explicitly not at the nuclear periphery (ScottDrew et al., 2002), and at the nuclear periphery, but only in meiosis (Sau et al., 2014). Under the circumstances, this statement needs to be removed or qualified.

We thank the reviewer for pointing out our erroneous phrasing, which we now removed and edited this section of the manuscript accordingly.

5) For Figure 3B, the legend and should be rewritten to make it clear whether these are ~600 random spores from each of the homozygous Y9 and BY4742 diploids and therefore directly comparable to the ~600 random spores from the heterozygotes in Figure 3C or if these are the distributions from the respective parental haploids.

We thank the reviewer for pointing out the lack of clarity here. We now clarify that these represent parental haploid distributions from across experiments (red and blue), and not homozygous progeny distributions.

Reviewer #2 (Significance):

This study is one of the first to leverage yeast genetics and the extensive genomic resources now available for yeast to identify natural variation in yeast populations that contribute to a trait of interest. Specifically, here the authors identify a variant of the SUMO ligase Mms21 that is associated with reduced mitotic stability of the 2-micron plasmid, a selfish genetic element, that despite conferring a growth fitness defect, is found in many yeast strains. The results are significant due to:

1) The conserved and essential role of the Smc5/6 complex, of which Mms21 is a subunit. Unlike the related cohesin and condensin complexes, the function of Smc5/6 complex is much less well understood.

2) The value of identifying non-conditional alleles, such as the Mms21 Thr69Ile, by this approach that can be used as tools for dissecting function of essential genes where the lethality of knock outs precludes many conventional analyses.

3) This is the first report to suggest a role for the Smc5/6 complex in 2-micron plasmid maintenance and is particularly intriguing in light of prior work implicating human Smc5/6 in limiting Hepatitis B virus.

4) The SCAMPR assay developed for this study will facilitate future high-throughput analyses of plasmid maintenance.

5) Taken together, the results represent a valuable addition to our understanding of 2-micron plasmid distribution and host interactions.

This paper will be of interest to those who use yeast as a model organism for molecular genetics and cell biology studies, those in the biotechnology and industrial sector who work with vector systems for model and non-model yeasts, and researchers with an interest in evolutionary biology, population genetics, eukaryotic chromosome biology, DNA repair and cell cycle regulation.

We are highly grateful for the accurate and positive summary of our article by the reviewer and for their comments towards improving and correcting our paper.

Referees cross commenting:

The other reviews seem quite reasonable to me. Clearly all three of us found this to be an interesting paper.

You are not missing anything, the cartoon in Figure 1C is incorrect as I also indicated in my review. In the text, it does say that plasmid copy number could increase in some cells if the plasmid missegregates (they should have stated that it would increase if replication is not affected). If this is total lack of delivery to daughters, copy number would double in mother cells and if GFP intensity is not saturated, one would have expected an increase in fluorescence for those cells (the cartoon should show a shift to the right in in the median GFP intensity for the GFP-positive cells. They did not see this (Figure 1—figure supplement 1) which could be due to the higher copy number being toxic resulting in loss of those cells from the population during the 24 hr culture under non-selective conditions as plasmid-free cells out-compete plasmid-plus ones. It did look like most of the GFP signal was gone after 24 hrs for the Y9 strain. This would not be unexpected. The authors do need to fix the cartoon and at least speculate as to why they did not see the expected increase.

We agree with these comments and suggestions by both reviewers #1 and #2 and have modified the figure and main text accordingly in our revision.

I agree that they should also have mentioned Flp-mediated correction of copy number in this context as it likely contributes to maintaining the median level of GFP intensity in the population of GFP-positive cells (if at least some plasmid copies are delivered to daughters in the absence of effective equipartitioning), although Flp cannot reduce copy number in cells that have already received higher than normal numbers.

As reviewer #2 states, Flp-mediated correction is likely to increase, not decrease copy number, which is why we favor the hypothesis that the lack of observed super-green cells with increased plasmid copy number is most likely the result of loss of proliferative fitness of these cells. Nevertheless, we agree that the issue of Flp-mediated amplification needs to be brought up in the main text as well as our motivation for working with a Flp positive plasmid: because the endogenous plasmid is capable of copy number recovery in the lab strain, but is not natively found in Y9, we wondered if Y9 was capable of restricting endogenous plasmid propagation (even with this amplification system intact). Our reporter plasmid is therefore as similar to an endogenous plasmid as we possible while still facilitating plasmid selection and screening. We have tried to better address these Flp ideas in our manuscript revision, and greatly appreciate the reviewers’ suggestion.

Reviewer #3 (Evidence, reproducibility and clarity):

In this manuscript, the authors examine diversity in 2-micron plasmid stability across Saccharomyces cerevisiae strains. The 2-micron plasmid is a selfish genetic element that has co-evolved with the Saccharomyces genus. Notably, they find that some Saccharomyces cerevisiae strains lack 2-micron plasmids. These strains also show low 2-micron plasmid stability after plasmid introduction. The authors map a major locus involved in this phenotype and localize its effect to MMS21, which had not been previously connected to 2-micron plasmid stability.

This manuscript is written in a way that is easy to read. The authors do a good job of conveying their results in a manner that is interesting, but not hyperbolic. Technically, I didn’t have any qualms with the work; it is rigorous.

We thank the reviewer for their constructive comments and positive appraisal.

More specific comments follow:

The first two paragraphs of the results are focused on assays that have been historically used to measure plasmid stability. This information is technical background. Should it be elsewhere, such as in the introduction or a supplementary note?

We appreciate the reviewer’s point. We did consider putting details about the earlier assays in the Introduction but felt that discussing them together with the SCAMPR assay in the Results made clear where our reporter plasmids and strategies differed and where it built upon previous assays. We would therefore prefer to leave it as it is in the beginning of the Results section if possible.

How much larger is the GFP-2 micron reporter plasmid than the endogenous 2 micron plasmid? Could difference in size between be another possible explanation for difference in stability?

Previous reports have shown that the 2μ plasmid can tolerate an insertion of ~3.9 kb. The cassette introduced in is 2703 bp. We now specify these additional details in the revision and emphasize the relevant citation.

I am not sure that the segregation data provide much clarity on polygenicity. The data don’t rule out the possibility that a good amount of the variability is just stochastic and the Castle-Wright estimator has limited utility. The authors might want to be more equivocal in this paragraph.

We agree with this comment. We have edited the text accordingly (we do, however, retain mention of the Castle-Wright estimate, as the expected complexity of the trait contributed to our choice of using a bulk segregant approach, and we are often asked for this estimate).

During discussion of confidence intervals, it might be worthwhile to note the position of the peak marker(s). Often those are pretty close to the causal gene in these bulk segregant studies.

We greatly appreciate this suggestion. Indeed, the T69I SNP is only 1.2kb away from the peak markers! We now include this information in the text, with details of peak locations and confidence interval coordinates in the figure and supplementary figure legends.

I found the emphasis on SCAMPR at some points distracting. I recognize the assay is an improvement upon other techniques, but the generalist reader might not find the assay that compelling.

Although we think this is a fair comment by the reviewer, we note that the SCAMPR assay allowed us the throughput, sensitivity and reproducibility to do the genetic mapping studies. Since this method was critical for our approach and is not published separately at this time, we feel most comfortable leaving SCAMPR as one focus of this paper.

If any additional work were to be done, I would make sure MMS21 is the sole quantitative trait gene at the detected locus. The locus was not resolved very precisely and, as the authors themselves note, other quantitative trait genes affecting the same trait might be present. I do not regard this work as essential.

We appreciate the comment that the mapping of additional resistance genes is a definite avenue for future work.

Reviewer #3 (Significance):

Whether genetic differences among hosts impact the propagation of selfish genetic elements is an interesting question. This paper not only demonstrates that host genetics affect the stability of selfish genetic elements, but also provides insights into molecular mechanism. A number of additional factors likely also contribute, but this does not diminish the MMS21 finding. I thought this was a nice story.

We appreciate the kind comments by the reviewer, and their helpful suggestions to streamline and improve the paper.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

First, you mention that you could include in the manuscript the results shown in response to reviewer #1, comment 4. Since a similar comment was made by another reviewer, it would indeed be important to include this analysis in a revised submission. In addition, the figure as it is now in the rebuttal is difficult to read.

We agree, and now provide growth curves as an additional Figure 2—figure supplement 2, along with detailed methods.

Second, reviewer #1 points out in their comment 6 that allele replacement would be important to confirm the role of the candidate amino acid substitution. I understand that the paper itself goes beyond the identification of a precise variant and that the current COVID-19 conditions limit your access to lab facilities. However, since this claim appears to be key for some of the downstream analyses on the 1001 genomes and on closely related species, and for the discussion about the mechanisms by which the plasmid is unstable in Y9, this appears to be a key experiment. One of the sub-section title states for instance that this single amino acid variant in MMS21 contributes to mitotic instability of the plasmid in Y9.

Here are some observations that make the contribution of this amino acid substitution uncertain. First, the Y9 instability appears to be dominant as measured in the heterozygote whereas the addition of the Y9 allele in BY does not affect its phenotype. Second, on Figure 5C, I appreciate that the phenotype of Het-Y9 MMS21 is different from the Het, but a relevant comparison here would be Het-Y9 MMS21 versus Het-4742 MMS21. These two strains are the ones for which the only difference is the nature of the allele at the locus and not the number of alleles. The statistical test for this difference is not reported but it is likely not significant given what is shown on the figure. The results shown could for instance (and one could think of other dosage effects) be consistent with a difference in expression of the two alleles acting in cis, which would also alter the phenotypes depending on which allele is deleted. Since the analysis is focusing on missense variants, regulatory variants have not been considered so this cannot be ruled out. I therefore believe that allele replacement would be needed to confirm the identity of the causal variant.

There are several excellent points made here, including some that were not addressed by us and the reviewers. We note each of them here and our proposed remedy, which we hope will prove satisfactory.

1) In Figure 5B, as you note, the phenotype of Het minus Y9 MMS21 is different from the Het, and you are right that we did not compare the Het minus Y9 MMS21 to the Het minus BY4742 MMS21, which would also correct for the allelic dosage comparison. Surprisingly, while we had previously tested the significance of this difference, somehow it did not make it into our manuscript. We gratefully acknowledge your attention to this detail and now include this comparison. Although this looks like a modest difference in the figure, it is highly significant (p=0.0013, Kruskal-Wallis test). Indeed, one of SCAMPR’s advantages is large sample numbers with relatively tight variance. We have amended Figure 5 accordingly. As you can see, this analysis demonstrates that with copy number held the same, removal of BY4742 versus Y9 MMS21 has significantly different effects on plasmid stability. We apologize for the omission of this important statistical comparison in our earlier version.

2) Your second point is that we did not adequately address whether it is regulatory sequence changes or the missense change in Y9 MMS21 that is responsible for the instability phenotype. We have now subjected MMS21’s regulatory regions to the same comparative genomics approach we used in our original submission to identify candidate coding SNPs. Our whole genome sequencing shows that in addition to the MMS21 missense SNP, Y9 differs from BY4742 at 4 synonymous sites within the ORF and a total of 11 sites in the two flanking (candidate regulatory) regions. For all 16 of these SNPs, the Y9 allele is shared with the closely-related plasmid-lacking sister strain Y12. We next examined the close outgroup UC5, which still bears plasmids according to our PCR survey: while the MMS21 missense SNP and two synonymous SNPs differ between Y9 and UC5, and thus strictly correlate with the plasmid instability phenotype, the candidate regulatory SNPs do not. We have added this analysis (new Figure 5—figure supplement 1) to our revision to explain why we favor the missense mutation rather than regulatory sequences as causal for plasmid instability. Despite this reasoning, we also acknowledge in our Discussion the possibility that regulatory sequences might play a role.

The ideal disambiguation experiment between regulatory and coding SNPs would be allele swaps in the Y9/BY Het background. However, this experiment is not trivial, because the Y9 genome is much less amenable to engineering than standard lab strain genomes. Engineering allelic loss of MMS21 in the Het background took us nearly two years under ideal (pre-COVID) circumstances, so we are unable to do this experiment in a reasonable time frame. We hope that your concerns will be sufficiently addressed by the addition of our analysis of regulatory and coding differences (above) and the statistical significance of plasmid instability differences between reciprocal deletions in the Het background (above).

3) We would also like to clarify why we do not expect the Y9 MMS21 allele introduced into BY4742 (Figure 5A) to be sufficient to drive plasmid instability in the BY background. Our tetrad analysis (Figure 3—figure supplement 1), the phenotype distribution among F1 progeny (Figure 3C) and our QTL analysis (Figure 4B) strongly suggest plasmid instability is a multilocus trait. We believe that the combination of Y9 MMS21 together with other genetic variants in the Y9 genome is required to explain the full plasmid instability phenotype. All of our MMS21 genetic data are consistent with a multilocus dominant trait: first, while we do observe loss of instability upon removal of Y9 MMS21 but not BY4742 MMS21 from the heterozygote, this loss only partially recapitulates the parental phenotypes; second, addition of Y9 MMS21 (with the Y9 native regulatory sequence) to the BY background is not sufficient to confer plasmid instability alone.


Article and author information

Author details

  1. Michelle Hays

    1. Molecular and Cellular Biology program, University of Washington, Seattle, United States
    2. Division of Basic Sciences & Fred Hutchinson Cancer Research Center, Seattle, United States
    Present address
    Department of Genetics, Stanford University, Palo Alto, United States
    Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8540-3516
  2. Janet M Young

    Division of Basic Sciences & Fred Hutchinson Cancer Research Center, Seattle, United States
    Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8220-8427
  3. Paula F Levan

    Division of Basic Sciences & Fred Hutchinson Cancer Research Center, Seattle, United States
    Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Harmit S Malik

    1. Division of Basic Sciences & Fred Hutchinson Cancer Research Center, Seattle, United States
    2. Howard Hughes Medical Institute, Fred Hutchinson Cancer Research Center, Seattle, United States
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Visualization, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6005-0016


National Institute of General Medical Sciences (R01 GM074108)

  • Harmit S Malik

National Science Foundation (DGE-1256082)

  • Michelle Hays

National Human Genome Research Institute (5T32HG000035-20)

  • Michelle Hays

Howard Hughes Medical Institute (Investigator award)

  • Harmit S Malik

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.


We thank past and present members of the Malik, Dudley and Raghuraman/Brewer labs for helpful discussions throughout this project. We especially thank Aimée Dudley and Mosur Raghuraman for their constant support throughout this project. We also thank Yu-Ying (Phoebe) Hsieh, Tera Levin, Antoine Molaro, Courtney Schroeder, and Gavin Sherlock for their comments on the manuscript, and Howard Chang for reminding us of the Smc5/6 restriction of Hepatitis B viruses. We are grateful to Justin Fay for the generous gift of natural yeast isolates. We appreciate all the generous assistance and advice from Flow Cytometry and Genomics shared resource facilities at the Fred Hutchinson Cancer Research Center. This work was supported by an NSF graduate research fellowship (Grant No. DGE-1256082 to M.H.), NIH/NHGRI Genome Training Grant at the University of Washington (5T32HG000035-20 to M.H.), NIH R01 grant GM074108 (to H.S.M.) and an Investigator award from HHMI (to H.S.M.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Senior Editor

  1. Detlef Weigel, Max Planck Institute for Developmental Biology, Germany

Reviewing Editor

  1. Christian R Landry, Université Laval, Canada

Version history

  1. Received: August 23, 2020
  2. Accepted: October 15, 2020
  3. Accepted Manuscript published: October 16, 2020 (version 1)
  4. Version of Record published: November 9, 2020 (version 2)


© 2020, Hays et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.


  • 2,289
    Page views
  • 271
  • 4

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Michelle Hays
  2. Janet M Young
  3. Paula F Levan
  4. Harmit S Malik
A natural variant of the essential host gene MMS21 restricts the parasitic 2-micron plasmid in Saccharomyces cerevisiae
eLife 9:e62337.

Share this article


Further reading

    1. Evolutionary Biology
    2. Genetics and Genomics
    Thomas A Sasani, Aaron R Quinlan, Kelley Harris
    Research Article

    Maintaining germline genome integrity is essential and enormously complex. Although many proteins are involved in DNA replication, proofreading, and repair, mutator alleles have largely eluded detection in mammals. DNA replication and repair proteins often recognize sequence motifs or excise lesions at specific nucleotides. Thus, we might expect that the spectrum of de novo mutations – the frequencies of C>T, A>G, etc. – will differ between genomes that harbor either a mutator or wild-type allele. Previously, we used quantitative trait locus mapping to discover candidate mutator alleles in the DNA repair gene Mutyh that increased the C>A germline mutation rate in a family of inbred mice known as the BXDs (Sasani et al., 2022, Ashbrook et al., 2021). In this study we developed a new method to detect alleles associated with mutation spectrum variation and applied it to mutation data from the BXDs. We discovered an additional C>A mutator locus on chromosome 6 that overlaps Ogg1, a DNA glycosylase involved in the same base-excision repair network as Mutyh (David et al., 2007). Its effect depends on the presence of a mutator allele near Mutyh, and BXDs with mutator alleles at both loci have greater numbers of C>A mutations than those with mutator alleles at either locus alone. Our new methods for analyzing mutation spectra reveal evidence of epistasis between germline mutator alleles and may be applicable to mutation data from humans and other model organisms.

    1. Chromosomes and Gene Expression
    2. Evolutionary Biology
    Katherine Rickelton, Trisha M Zintel ... Courtney C Babbitt
    Research Article Updated

    Primate evolution has led to a remarkable diversity of behavioral specializations and pronounced brain size variation among species (Barton, 2012; DeCasien and Higham, 2019; Powell et al., 2017). Gene expression provides a promising opportunity for studying the molecular basis of brain evolution, but it has been explored in very few primate species to date (e.g. Khaitovich et al., 2005; Khrameeva et al., 2020; Ma et al., 2022; Somel et al., 2009). To understand the landscape of gene expression evolution across the primate lineage, we generated and analyzed RNA-seq data from four brain regions in an unprecedented eighteen species. Here, we show a remarkable level of variation in gene expression among hominid species, including humans and chimpanzees, despite their relatively recent divergence time from other primates. We found that individual genes display a wide range of expression dynamics across evolutionary time reflective of the diverse selection pressures acting on genes within primate brain tissue. Using our samples that represent a 190-fold difference in primate brain size, we identified genes with variation in expression most correlated with brain size. Our study extensively broadens the phylogenetic context of what is known about the molecular evolution of the brain across primates and identifies novel candidate genes for the study of genetic regulation of brain evolution.