Dispensable essential gene datasets

We compiled a comprehensive list of DEGs in S. cerevisiae from two large-scale studies (Liu et al, 2015; van Leeuwen et al, 2020) and from individual cases described in the literature (van Leeuwen et al, 2020) (Fig 1A). We only considered studies in which gene essentiality was bypassed in a laboratory yeast strain, as these often involve a single causal bypass suppressor gene, and disregarded studies that focused on essentiality changes across natural yeast strains, which are frequently driven by complex combinations of genetic variants (Dowell et al, 2010; Chen et al, 2022; Wang et al, 2022). In total, 205 DEGs had been identified, representing ∼20% of all tested essential genes (Fig 1B). Cases of bypass suppression were identified by looking for rare survivors in populations of 100–150 million cells deleted for an essential gene (van Leeuwen et al, 2020; experimental dataset), by following germination of single deletion mutant spores (Liu et al, 2015), or by a mixture of methods, including directly testing the effect of a mutation on essential gene deletion mutant viability (van Leeuwen et al, 2020; literature dataset). These methodological differences could possibly affect the detected DEGs.

To determine whether the datasets could be merged, we compared various properties of the DEGs described in each dataset. The DEGs identified in the three datasets overlapped significantly (P < 0.001, randomization test). In all datasets, DEGs showed similar functional enrichments (Fig S1A) and were depleted for fundamental cellular processes like RNA processing or translation and enriched for more peripheral functions related to signaling or transport (P < 0.05, Fisher’s exact tests, and false discovery rate [FDR] < 10%). Furthermore, protein complexes tended to be either completely dispensable or indispensable across datasets (P < 0.05 in the combined dataset, randomization test, Fig S1B). For instance, the combined dataset contained 14 protein complexes with only dispensable essential subunits (Fig S1C), significantly more than expected by chance (P < 0.002, randomization test, Fig S1B). DEGs were more likely than core essential genes to be non-essential in the closely related S. cerevisiae strain Sigma1278b (P < 0.0005 in the combined dataset, Fisher’s exact test, Fig S1D), and to be absent in the S. cerevisiae core pangenome (P < 0.05 in the combined dataset, Fisher’s exact tests, Fig S1E). Because the properties of the combined and individual datasets were similar, we used the combined dataset in the following analyses.

Properties of dispensable essential genes

By querying an extensive panel of 21 gene features (see the Materials and Methods section, Fig S1F), we compared the properties of dispensable and core essential genes and found several significant differences (P < 0.05 using Mann–Whitney U tests for numeric features and Fisher’s exact tests for binary features, and FDR < 10%). DEGs tended to exhibit more stable gene expression levels and lower transcript counts, to be less conserved across species, to have more gene duplicates and higher evolutionary rates, and to be coexpressed with fewer genes than core essential genes. The proteins encoded by DEGs tended to be more multifunctional, to lack structural domains, to localize to a membrane, to be absent from protein complexes, and to have fewer protein–protein interactions, lower abundances, and shorter half-lives compared with those encoded by core essential genes (Fig 1C and Table S1). Interestingly, the observed differences between dispensable and core essential genes resembled the differences between non-essential and essential genes (Fig 1C and Table S1). Thus, we asked whether dispensable essential and non-essential genes shared the same properties and found that they comprised two different classes of genes with clearly distinct features (Fig 1C and Table S1). Broadly, features of DEGs fell between those of core essential and non-essential genes, consistent with and extending previous findings in a smaller dataset (Liu et al, 2015).

Phylogenetic analysis of dispensable essential genes

We further explored the differences in gene conservation between dispensable and core essential genes using the phylogeny of S. cerevisiae, starting with a large panel of sequenced S. cerevisiae strains (Peter et al, 2018). DEGs were more likely than core essential genes, but less than non-essential genes, to harbor deleterious mutations disrupting protein sequences (P < 0.0005, Fisher’s exact test, Fig S2A), to present higher non-synonymous mutation rates (P < 0.0005, Mann–Whitney U test, Fig S2B), and to show copy number loss (CNL) events in other S. cerevisiae strains (P < 0.0005, Fisher’s exact test, Fig S2C). To further investigate differences in the evolutionary pressure on dispensable essential and core essential genes, we analyzed essentiality data and orthology relationships in Candida albicans, S. pombe, and human cell lines (Figs 2A and S2D and E and Table S2). Genes that were dispensable essential in S. cerevisiae were more often absent than core essential genes in each of the analyzed species (P < 0.0005, Fisher’s exact tests, Fig 2B). We hypothesized that this bias could be caused by: (i) genes specific to the S. cerevisiae phylogenetic branch and, thus, not present in their common ancestor or (ii) genes present in their common ancestor but lost in the phylogenetic branch of the analyzed species. To determine the contribution of each factor, we calculated the age of each S. cerevisiae gene by identifying the furthest species with an orthologous gene. DEGs were enriched for younger genes with respect to core essential genes (P < 0.0005, Mann–Whitney U test, Fig 2C), particularly for genes with no ortholog in any other species (i.e., specific to S. cerevisiae; P < 0.005, Fisher’s exact test, Fig 2D). Next, for each species, we defined lost genes as those absent in that species but present in its common ancestor with S. cerevisiae. We found DEGs were more often lost in other species than core essential genes (P < 0.0005, Mann–Whitney U test, Fig 2E). Thus, the absence of DEGs in other species can be explained both by genes specific to S. cerevisiae and by gene loss events in those species.

Furthermore, DEGs present in other species were more frequently duplicated and had more N:1 orthology relationships (P < 0.05, Fisher’s exact test, Fig 2B) than core essential genes. For genes with a 1:1 ortholog in other species, DEG orthologs were more often non-essential than orthologs of core essential genes (P < 0.0005, Fisher’s exact test, Fig 2B), also in the closely related Saccharomyces uvarum species (P < 0.05, Fisher’s exact test, Fig S2F). Similarly, fitness data from a panel of 1,070 cancer cell lines (Meyers et al, 2017) revealed that knockout of DEG orthologs led to less severe proliferation defects than knockout of core essential gene orthologs (P < 0.0005, Mann–Whitney U test, Fig 2F). Thus, genes that can be bypassed by genetic mutations in S. cerevisiae tend to be non-essential in other species. We show the comparison between essential and non-essential genes and dispensable essential and non-essential genes to contextualize the observed differences (Fig S2G–I and Table S2).

Finally, we compared sequences of S. cerevisiae proteins and their 1:1 orthologs in S. pombe and C. albicans. Gene products of DEGs had lower sequence identity and differed more in sequence length than core essential proteins (P < 0.05, Mann–Whitney U tests, Figs 2G and S2J–L), in line with the dN/dS data (Figs 1C and S2B). Overall, orthology relationships, phenotypic changes, and sequence divergence reflect that the evolutionary pressure on DEGs is more lenient than on core essential genes but more strict than on non-essential genes.

The bypass suppressor interaction network

Identification of the relevant genetic changes (i.e., suppressors) required to tolerate the deletion of an essential gene is key to interpreting the presence of deleterious genetic variants in natural populations. To improve our knowledge on the mechanisms of genetic suppression, we built an interaction network between DEGs and their bypass suppressors by combining data from our recent systematic study (van Leeuwen et al, 2020) and the literature (van Leeuwen et al, 2020). The two individual suppression interaction networks overlapped significantly (P < 0.001, randomization test, Fig 3A) and were similarly enriched in functional associations (P < 0.0005, Fisher’s exact tests, Fig S3A). The combined network included a total of 319 unique bypass suppression gene pairs, corresponding to 243 suppressors and 137 DEGs out of the 205 known DEGs. For the remaining DEGs (33% of the dataset), the suppressor variants have not been identified. Dispensable essential and suppressor genes tended to be functionally related (P < 0.05, randomization test, and FDR < 10%, Fig S3B), particularly for close functional relationships like cocomplex or copathway membership (P < 0.0005, Fisher’s exact tests, Fig S3A), and suppressors related to nuclear-cytoplasmic transport and transcription processes were more frequent than expected by chance (P < 0.05, Fisher’s exact test, and FDR < 10%, Fig S3B). For a subset of bypass suppressors, we and others have previously determined experimentally whether a suppressor mutation had a loss-of-function (LOF) or gain-of-function (GOF) effect, by testing the effect of suppressor gene mutation or overexpression on the viability of the corresponding DEG deletion mutant (van Leeuwen et al, 2020). Here, we found that for 50% and 26% of the dispensable genes, only LOF and GOF suppressors had been identified, respectively, and in 15% of the cases, both types of suppressors had been described (Fig S3C). For the remaining cases, the nature of the suppressor had not been determined.

Genetic interactions identify combinations of mutants that result in unexpected phenotypes given the phenotypes of the individual mutants. In negative genetic interactions, the resulting phenotype is more severe than expected, whereas in positive genetic interactions, the phenotype is healthier than expected. In a bypass suppression interaction, a secondary mutation recovers the lethal phenotype caused by an essential gene deletion, therefore representing an extreme form of positive genetic interaction. We grouped interacting pairs in the bypass suppression network by their suppression mode (i.e., LOF or GOF) and evaluated their overlap with a global genetic interaction network (Costanzo et al, 2016), generated using hypomorphic alleles of essential genes and deletion alleles of non-essential genes (see the Materials and Methods section). We first analyzed bypass suppression gene pairs with LOF suppressors and found that LOF alleles of these gene pairs often had a positive genetic interaction with each other in the global network (P < 0.0005, Fisher’s exact test, Fig 3B). In spite of the different experimental protocols, this overlap is expected because both bypass suppression and genetic interactions were identified using LOF alleles. Conversely, when analyzing bypass suppression gene pairs with GOF suppressors, we found that the corresponding LOF alleles mainly showed negative genetic interactions (P < 0.05, Fisher’s exact test, Fig 3B). Thus, GOF and LOF alleles of the suppressor gene have opposite effects when combined with a LOF allele of the corresponding DEG, being beneficial or detrimental as shown by the bypass suppression and genetic interaction networks, respectively.

Structure of the bypass suppression interaction network

Interaction density (i.e., the percentage of gene pairs with an interaction) of the bypass suppression network ranged from 0.007% to 0.96% depending on whether we considered all possible gene pairs or only pairs between the identified dispensable essential and suppressor genes, respectively. In spite of the sparsity of this network, several patterns emerge showing its structure and modularity. For instance, all DEGs in the same protein complex tended to interact with either GOF or LOF suppressors. These monochromatic interactions affected 13 out of 17 non-redundant protein complexes with at least two dispensable essential subunits in our dataset (P < 0.05, randomization test, Fig 3C), suggesting similar suppression types apply for functionally related genes. Importantly, both individual suppression networks contributed to this result (Fig S3D), discarding the potential bias from specific hypothesis-driven experiments in the literature dataset. We analyzed the topology of the network and found that for 45% of the DEGs, multiple suppressors had been described (Fig S3E). This set of genes exhibited specific features compared with DEGs for which only a single suppressor had been described (Fig S3F). For instance, DEGs with multiple identified suppressors tended to have higher multifunctionality and an increased number of structural domains (P < 0.05, Fisher’s exact test, and FDR < 10%), which suggest multiple different molecular mechanisms of suppression may exist for these DEGs. Suppressors were more specific than DEGs, and only 23% of them interacted with multiple genes (Fig S3E). Next, we explored the relationship between functional similarity and connectivity patterns. We found that genes in the same protein complex tended to have the same interactors: 52% of the DEGs encoding members of the same complex shared suppressor genes, and 70% of the suppressor genes encoding members of the same complex shared DEGs (Fig 3D), more than expected by chance (P < 0.0005, Fisher’s exact test).

To illustrate the underlying modular structure of the bypass suppression interaction network, we explored the connectivity of NCB2 and BUR6, both DEGs with known suppressors and the only two members of the negative cofactor 2 transcription regulator complex (ID CPX-1662 in the Complex Portal [Meldal et al, 2021]). NCB2 and BUR6 have seven and 10 identified bypass suppressor genes, respectively, six of which are in common, again showing that functionally related DEGs tend to share suppressors (Fig S3G). Two of these common suppressors belong to the core Mediator complex that plays a role in the regulation of transcription (CPX-3226), showing that interactors of the same dispensable gene tend to be functionally related both to each other and to the DEG they are suppressing. The other four shared suppressor genes also affect transcription and encode subunits of the transcription factor TFIIA complex (CPX-1633), the general transcription factor complex TFIIH (CPX-1659), and the DNA-directed RNA polymerase II complex (CPX-2662). Interestingly, the NCB2-specific suppressor, TOA2, also encodes a member of TFIIA, and three of the four BUR6-specific suppressors members of RNA pol II or Mediator, further illustrating the modularity of the network. In another example (Fig 3E), members of the RPD3L histone deacetylase complex (CPX-1852) suppress two different protein complexes. DEP1, SAP30, and SDS3 suppress the two essential subunits of piccolo NuA4 histone acetyltransferase complex (CPX-3185), whereas RPD3, SIN3, and SDS3 interact with the Rer2 subunit of the dehydrodolichyl diphosphate synthase complex (CPX-162). This modularity in the suppression interaction pattern of RPD3L subunits is also observed in genome-wide genetic interaction patterns, which are more similar for RPD3L subunits that suppress the same query gene than for RPD3L subunits that suppress functionally diverse query genes (Fig 3F). These patterns suggest a functional modularity within the complex which is supported by its modeled structure (Sardiu et al, 2009).

Mutational landscape of S. cerevisiae strains reflects bypass suppression relationships

We wondered if the genetic dependencies described in the suppression interaction network were reflected in the genomic variation present in natural populations. Because homozygous deletions of essential genes are extremely rare across S. cerevisiae strains (median of one per strain), we first focused on DEGs with CNL events. Hemizygosity is associated with a decrease in gene expression levels and can impact cell growth (Pavelka et al, 2010), particularly in essential genes. For instance, even if only ∼10% of essential genes were haploinsufficient under rich media conditions (Deutschbauer et al, 2005), this increased to 30–50% when more conditions and phenotypes were tested (Delneri et al, 2008; Ohnuki & Ohya, 2018). In strains in which a copy of a DEG was lost, we evaluated if the corresponding suppressor gene had a simultaneous copy number change. Interestingly, bypass suppression gene pairs with LOF and GOF suppressor mutations showed different preferences for co-occurring copy number changes, in agreement with their LOF or GOF phenotype. Bypass suppression gene pairs that involved a LOF suppressor mutation were enriched for co-loss of both dispensable essential and suppressor genes (P < 0.0005, Fisher’s exact tests, Figs 4A and S4A). In contrast, cases with GOF suppressor mutations were enriched for events in which CNL of the DEG was accompanied by a copy number gain of the suppressor gene (P < 0.005, Fisher’s exact tests, Figs 4A and S4A). Thus, when the DEG has a CNL in a natural strain, the functional effect of the bypass suppressor mutation (GOF or LOF) identifies the most likely copy number change of the suppressor gene in that same strain. Next, we asked whether deleterious coding mutations in DEGs and in identified bypass suppressor genes co-occurred in S. cerevisiae isolates. We only considered haploid strains so the deleterious effects of mutations would not be masked by other alleles. When considering only bypass suppression gene pairs in which the suppressor carried a LOF mutation, we found 18 cases in which both the DEG and the suppressor gene carried deleterious mutations in at least one of the haploid strains, significantly more than in randomized gene pairs (P < 0.05, randomization test, Fig 4B). As expected, we did not observe a similar enrichment in diploid strains (P > 0.05, randomization test, Fig S4B) or for gene pairs involving GOF suppressor mutations (P > 0.05, randomization test, Fig S4C). Thus, the bypass suppression network mapped in a laboratory environment reflects evolutionary outcomes in natural S. cerevisiae strains.

Co-occurrence of viability changes and fixed bypass suppressor mutations

We have shown that genes that are dispensable essential in S. cerevisiae are often non-essential in other species (Fig 2B). Differences in the genetic background in those species may be responsible for these changes in essentiality. Here, we hypothesized that the genetic changes that bypass the essentiality of a gene in S. cerevisiae should be reflected in the genome of species in which the gene is also dispensable (i.e., non-essential or absent). To test this, we evaluated whether changes in essentiality for DEGs in a given target species co-occurred with bypass suppressor mutations that were fixed in the target genome. Briefly, we considered as equivalent bypass mutations those that could reduce or increase the gene activity in the target species, for LOF and GOF suppressors, respectively (see the Materials and Methods section). Given that genome-scale essentiality data are scarce, we focused our analysis on S. pombe, for which high-quality essentiality data are available for most genes (Harris et al, 2022).

We found that 67% (18/27) of the S. cerevisiae DEGs that are non-essential in S. pombe co-occurred with bypass suppressor mutations in that species, whereas this happened for only 26% (12/47) of the DEGs that were essential in S. pombe (P < 0.005, Fisher’s exact test, Fig 5A and B). A similar trend (48%) was observed for S. cerevisiae DEGs that were absent (i.e., without an ortholog) in S. pombe, although this difference was not significant compared with the set of essential orthologs (P > 0.05, Fisher’s exact test, Fig 5B). To increase the statistical power of our analyses, we combined the non-essential and absent genes in S. pombe into a single set and observed a clear difference with the essential orthologs (2.3-fold enrichment, P < 0.005, Fisher’s exact test).

We controlled for potential biases to ensure the robustness of our observation (Fig 5B). We evaluated the effect of interaction degree by generating 1,000 randomized bypass suppression networks while respecting the original topology (Fig S5A) and by considering only DEGs with a single known bypass suppressor (Fig S5B). In addition, we removed bypass suppression interactions from the literature which may have been identified because of phylogenetic properties (Fig S5C), functionally related bypass suppression pairs which may be prone to present similar evolutionary patterns (Fig S5D), and every node in the network to discard dependence on a single gene (Fig S5E). We only considered suppressors with 1:1 orthologs or absent in S. pombe to account for the potential expression divergence of duplicated genes (Fig S5F) and calculated the genes with large expression changes between both species to identify gene activity changes (Fig S5G). Also, we applied three alternative orthology mappings (Fig S5H) and used essentiality annotations and orthology mappings from C. albicans (Fig S5I). In all these analyses, DEGs without orthologs or with non-essential orthologs more often co-occurred with bypass suppressor mutations than DEGs with essential orthologs (P < 0.05, Fisher’s exact tests). Conversely, switching LOF and GOF annotations resulted in a non-significant difference, as expected (P > 0.05, Fisher’s exact test, Fig S5J).

Finally, we selected DEGs with 1:1 orthologs in both S. pombe and C. albicans and found that DEGs with non-essential orthologs in both species were more likely to have bypass suppressor mutations in those species than DEGs with essential orthologs (P < 0.05, Fisher’s exact test, Fig 5C). In all, these analyses reveal that the relationship between DEGs and their bypass suppressor genes identified in S. cerevisiae is reflected in the gene essentiality and mutational space of other species.