Into the wild: new yeast genomes from natural environments and new tools for their analysis

ABSTRACT

Genomic studies of yeasts from the wild have increased considerably in the past few years. This revolution has been fueled by advances in high-throughput sequencing technologies and a better understanding of yeast ecology and phylogeography, especially for biotechnologically important species. The present review aims to first introduce new bioinformatic tools available for the generation and analysis of yeast genomes. We also assess the accumulated genomic data of wild isolates of industrially relevant species, such as Saccharomyces spp., which provide unique opportunities to further investigate the domestication processes associated with the fermentation industry and opportunistic pathogenesis. The availability of genome sequences of other less conventional yeasts obtained from the wild has also increased substantially, including representatives of the phyla Ascomycota (e.g. Hanseniaspora) and Basidiomycota (e.g. Phaffia). Here, we review salient examples of both fundamental and applied research that demonstrate the importance of continuing to sequence and analyze genomes of wild yeasts.

New genome sequencing technologies together with new bioinformatic tools provide unique insights into the fascinating stories of wild and non-conventional yeast.

INTRODUCTION

Recent advances in sequencing technologies, the availability of new bioinformatic tools and multiple genomic studies during the past five years have significantly improved our understanding of the evolution, phylogeography, ecology and biotechnology of yeasts. Although most studies have focused on the genus Saccharomyces, substantial progress has been achieved with other yeasts of the phylum Ascomycota and, to a lesser extent, with yeasts of the phylum Basidiomycota. Today, approximately one fifth of the 1500 described yeast species have had their genomes fully sequenced (Kurtzman, Fell and Boekhout 2011; Shen et al. 2018). In a few cases, sequences of multiple isolates are available for population genomic studies (Table S1, Supporting Information). Historically, most studies were performed on yeast strains isolated from anthropic environments. In recent years, the number of yeasts from natural environments (wild yeasts) whose genomes have been sequenced has increased rapidly, creating a new opportunity to more fully explore eukaryotic biological mechanisms. This review provides an update on recent advances in the bioinformatics tools available for assembling, annotating and mining yeast genomes from a broad evolutionary range of yeasts (Section 2). Besides the best-studied genus Saccharomyces (Section 3), we also include other examples of outstanding interest from the Ascomycota (Section 4.1) and Basidiomycota (Section 4.2), which are rising models of yeast evolution and are becoming important for specific industrial applications.

NEW BIOINFORMATIC TOOLS FOR de novo GENOME RECONSTRUCTION AND ANALYSIS OF YEASTS

Access to whole genome sequence data has significantly increased in the past few years. In particular, the number of species of yeasts of the subphylum Saccharomycotina whose genomes have been sequenced has increased at least three-fold (Hittinger et al. 2015; Shen et al. 2018). While these data are more accessible, their analysis can be challenging. Non-conventional yeasts can have ploidy variation, have high heterozygosity, or be natural hybrids. Although there are multiple tools available to explore a broad range of topics in yeast evolution, integrating these tools to answer biological questions can be daunting. Table 1 depicts a description of new bioinformatic tools useful for genomic data processing and their respective references.

Table 1.

List of bioinformatics tools.

Software	Description	Input	Output	Notes	Pros	Cons	Reference
AAF	Assembly and alignment-free phylogenetic approach.	Raw-sequencing reads.	Phylogenetic tree.		Works with raw short read data. Alignment- and assembly-free. Can work with low coverage data and has low computational demands.	Does not work with hybrid genomes. Cannot analyze deep nodes. Does not report informative sites.	Fan et al. 2015
BUSCO	Assessment of genome quality and completion. Gene annotation of universal orthologous genes.	Assembled genome.	Summary output of gene counts and locations. Amino acid and protein sequences for complete genes in the genome.	Requires a set of genes to search for, which are available through their website.	Easy to use. Provides a set of genes that can be used in downstream analyses (e.g. phylogenomics).	Cannot find novel genes and does not allow for customizable gene sets. Gene sets are always single copy orthologs.	Waterhouse et al. 2018
GenomeTools	Collection of bioinformatic tools. For example, genomediff can be used to calculate pairwise differences between genomes (not gene based).	Varies depending on tool being used.	Varies depending on tool being used.	This is a collection of tools.	Has multiple tools, which can be used for whole genome analyses. Easy to use.	Not all tools are useful for the approaches discussed here.	Gremme, Steinbiss and Kurtz 2013
HybPiper	Assembles genes of interest from short reads.	Raw-sequencing reads and sequence of gene of interest.	Assembled contig with gene of interest (e.g. FASTA).	Output requires manual assessment to interpret the results.	Works with raw short read data. Assembles genes in regions that tend to be difficult for de novo assemblers. It can potentially retrieve paralogs.	Negative results do not necessarily mean the gene does not exist. Does not give information about functionality. Downstream steps required for in silico gene functionality assessment.	Johnson et al. 2016
iWGS	Wrapper which integrates multiple de novo genome assemblers. It allows for customization of which assemblers to include in analyses. It has upstreaming trimming and assembly quality assessments downstream.	Raw-sequencing reads (e.g. FASTQ).	Multiple assembled genomes.	Output requires the manual selection of appropriate assembly for your question. A subset of assemblers can be used on long-read data. See reference for list of assemblers.	Parallelizes genome assembly, with a range of assemblers, including assemblers that are ploidy-aware. Upstream steps, such as trimming, are included. Can simulate genome sequencing experiments	Computationally intensive and can be difficult to set-up initially due to multiple dependencies.	Zhou et al. 2016
jEMBOSS	A package with the EMBOSS software.	An assembled genome.	Depends on your analysis. For example GC content and length of scaffolds/chromosomes.		Useful to detect the mitochondrial scaffold.		Carver and Bleasby 2003
LRSDAY	Genome assembly for Nanopore and PacBio data.	Raw-sequencing reads.	Chromosome-level scaffolds.		Assembler for yeast long read data. There is a well-detailed extensive step-by-step workflow.	Initial set-up can be difficult due to multiple dependencies. Requires a reference genome for assembly.	Yue and Liti 2018
MAKER2	de novo gene annotation.	Assembled genome.	Integrated gene annotations across multiple platforms.	MAKER2 uses SNAP, Augustus, GenMark. MAKER2 can also predict genes using evidence-based approaches. Prior to final annotation, multiple training rounds must occur.	Runs multiple gene predictors. Integrates multiple tools and lines of evidence for accurate predictions.	Requires multiple dependencies and training for annotations. Difficult to run without prior experience.	Campbell et al. 2014
MITObim	Mitochondrial baiting and iterative mapping.	A reference mitochondrial genome and NGS reads.	A mitochondrial genome assembly.		It targets mitochondrial reads to increase the chances of recovering a mitochondrial genome assembly.	When reference mitochondrial genome is quite different from your strain, the performance of the pipeline is poor.	Hahn, Bachmann and Chevreux 2013
MUMmer4	Ultra-fast alignment of long DNA and protein sequences.	Two assembled genomes.	PNG or PDF comparing the structure of both genomes.		Visualization of large structural variants. Helps during ultrascaffolding of closely related strains	Small structural variants are much difficult to visualize.	Marçais et al. 2018
nQuire	Estimation of ploidy.	Short reads mapped to a reference genome.	Estimates of ploidy up to tetraploid, including plots to visualize the data.		Works with raw short-read data, with some upstream steps. Can detect aneuploidies.	Needs a reference genome. Can only detect up to tetraploids.	Weiß et al. 2018
RAxML	Phylogenetic placement.	Aligned sequences.	Phylogenetic tree.		Statistically robust. Is able to recover deeper nodes. Can generate larger phylogenies. Allows for parallelization.	Computationally intensive and maybe difficult for first-time users. It takes multiple upstream steps, including sequence alignments.	Stamatakis 2014
RepeatMasker	Screens DNA sequences for interspersed repeats and low complexity DNA sequences.	An assembled genome.	A detailed annotation of the repeats that are present in the query sequence.
A modified version of the query sequence in which all the annotated repeats have been masked.		Facilitates the annotation of transposable elements.	Some libraries, such as RepBase, require payment.	http://www.repeatmasker.org/
SISRS	Assembly- and alignment-free phylogenetic approach.	Raw sequencing reads.	Phylogenetic tree.		Works with raw short-read data. It is alignment- and assembly-free.	Does not work with hybrid genomes. It does not report informative sites. Deeper nodes are harder to recover.	Schwartz et al.2015
splitsTree	Phylogenetic network.	Aligned sequences.	Phylogenetic network.		Works to visualize admixture and hybrids. Allows thresholding and gives confidence intervals.	Networks can be difficult to interpret and are not regularly used in phylogenomics.	Huson and Bryant 2006
sppIDer	Quick inference of genome composition, hybrid and admixture detection.	Raw-sequencing reads.	Multiple assessment (tables and visualizations) of genome composition.	Output requires manual assessment to interpret the results.	Works with raw short read data. Detects hybrids. Keeps intermediate steps, which can be used in other analyses. Includes statistical analyses.	Needs multiple closely related reference genomes.	Langdon et al.2018
YGAP	Gene annotation based on synteny.	Assembled genome.	Multiple gene annotation outputs.		Trained on multiple yeast species. It can support pre- and post-whole genome duplication species. Web-based interface easy to use.	Web-based approach limits high throughput analyses. Does not work well on species highly diverged from S. cerevisiae.	Proux-Wera et al.2012

Several tools have been developed to quantify ploidy levels and detect hybrids from short-read sequencing data. Both nQuire and sppIDer are alignment-based approaches developed for detecting ploidy variation and hybridization events, respectively. They are useful to run on raw data prior to genome assembly since these factors create challenges for de novo genome assembly programs that affect performance and increase the frequency of assembly errors. Multiple de novo genome assembly programs are available that can use short-reads, many of which are available in the wrapper iWGS, including the ploidy-aware genome assemble programs PLATANUS and dipSPAdes, which perform well on highly heterozygous sequences. Additionally, genome assemblies with long-reads can be performed with the wrapper LRSDAY (Yue and Liti 2018). Prior to these phylogenetic analyzes, the bioinformatic tool BUSCO can be used to assess genome quality and completeness, as well as to curate a robust set of orthologous genes to build phylogenies in programs, such as RAxML (Shen et al. 2018). As an alternative to traditional phylogenetic approaches that require aligned sequences, phylogenetic analyzes can be performed prior to genome assembly using AAF and SISRS. Genome annotations can be performed using MAKER2 and YGAP. MAKER2 is a wrapper that calls multiple gene annotation tools and makes for multiple sets of gene predictions simultaneously, while YGAP is a web-based tool built specifically for yeast genome annotation, especially genomes that are syntenic with the model yeast Saccharomyces cerevisiae. Additionally, HybPiper can be used to detect candidate genes that are located in hard-to-assemble regions of the genome and does not require genome assembly.

In recent years, DNA reassociation also referred as DNA–DNA hybridization (DDH) has been gradually replaced by high-throughput sequencing, which allows the in silico calculation of overall genome related indices (OGRI) (Chun and Rainey 2014). OGRI include any measurements indicating how similar two genome sequences are, but they are only useful for differentiating closely related species (Chun et al. 2018). Examples of OGRI include average nucleotide identity (ANI) and digital DDH (dDDH), which are widely used, and relevant software tools are readily available as web-services and as standalone tools (for a detailed list see Chun et al. 2018; Libkind et al. 2020). Other approaches include the calculation of pairwise similarities (Kr, with the tool genomediff of Genometools) and genome-wide alignments (MUMmer, Marçais et al. 2018). The resulting alignments can be used to obtain syntenic regions, study conservation and assist in ultra-scaffolding.

There are many bioinformatic tools and pipelines available that are not listed here. For example, approaches have been developed to explore gene functions (Pellegrini et al. 1999; Jones et al. 2014), horizontal gene transfers (HGT) (Alexander et al. 2016), species phylogenetic tree inference (Shen et al. 2016) and copy number variation (Steenwyk and Rokas 2017, 2018). Furthermore, new tools are being developed regularly. The availability of these bioinformatic tools, coupled with access to hundreds of genomes, allows us to address a broad range of questions in yeast genomics, evolution and genetics.

GENOMICS IN THE MODEL GENUS Saccharomyces

The understanding of the S. cerevisiae genome has been driven by the advent of novel-sequencing technologies. Indeed, S. cerevisiae was the first eukaryote to be completely sequenced (Goffeau et al. 1996) (Fig. 1). Furthermore, the development of next-generation sequencing (NGS or 2nd generation) technologies and long-read sequencing (3rd generation) technologies, together with bioinformatic tools (see Section 2) (Fig. 2), have enhanced our understanding of yeast genome evolution and led to nearly complete assemblies of the nuclear genomes of four of the eight known Saccharomyces species. The new combined data sets allowed the annotation of most eukaryotic genetic elements: centromeres, protein-coding genes, tRNAs, Ty retrotransposable elements, core X’ elements, Y’ elements and ribosomal RNA genes. The study of the population-scale dynamics of repetitive genomic regions has been relatively underexplored due to the emphasis on short-read (< 300 bp) technologies, such as Illumina sequencing. Regardless, in combination with short-read data sets, new long-read technologies are beginning to unravel the differences in Ty and other repeat content between different Saccharomyces strains and species (Istace et al. 2017; Yue et al. 2017; Czaja et al. 2019), including their contribution to differences in genome size between S. cerevisiae and S. paradoxus genomes (Yue et al. 2017; Czaja et al. 2019). Assembly of subtelomeric regions has also benefited from combining short-read and long-read data. A recent study comparing the evolutionary dynamics of subtelomeric genes found that the length of subtelomeric regions to vary greatly (0.13–76 Kb with 0–19 genes) and demonstrated an accelerated rate of evolution in domesticated S. cerevisiae strains compared to wild S. paradoxus isolates (Yue et al. 2017). Important traits for environmental adaptations and phenotypic diversification can now be detected among subtelomeric structural variants (which can also be important in speciation), the copy number variants can now be quantified and localized (e.g. those observed in the CUP1 gene and ARR cluster) and the presence and absence of metabolic genes can be accurately assessed (McIlwain et al. 2016; Yue et al. 2017; Naseeb et al. 2018; Steenwyk and Rokas 2018).

Figure 1.

The genomes of more than three thousand Saccharomyces strains have been sequenced. At least 3077 unique Saccharomyces strains have had their genomes sequenced using various sequencing technologies in the past 23 years (Table S1, Supporting Information). About 71.5% of the sequenced Saccharomyces strains belong to S. cerevisiae, 11.0% are S. paradoxus, 8.5% are S. eubayanus, 5.95% are interspecies hybrids, and 2.0% are Saccharomyces uvarum. At least 105 Saccharomyces strains have been sequenced by more than two studies (Table S1, Supporting Information). Colored circles highlight the total genome sequences published per year per technology (symbol shape) for each Saccharomyces species or for interspecies hybrids. Bar plots represent the total number of sequenced strains from each Saccharomyces species or interspecies hybrids, including (panel B) and excluding (panel C) S. cerevisiae strains. Bar plots are colored according to species.

Figure 2.

Pros and cons of the genome sequencing methods that are currently used most widely. Pros and cons of technologies used for for de novo genome assembly and population genomics (Goodwin et al. 2015; Chen et al. 2017; Giordano et al. 2017; Istace et al. 2017). CLR: continuous long read; CC, circular consensus; 4mC, N4-methylcytosine; 5mC, 5-methylcytosine; 6 mA, N6-methyladenine; Kb, kilobase.

Mitochondrial genome sequence assemblies have also been missing from most Illumina sequencing studies, except in a handful studies (Baker et al. 2015; Wu, Buljic and Hao 2015; Sulo et al. 2017). In contrast, long-read technologies better capture and facilitate the assembly of mitochondrial genome sequences (Wolters, Chiu and Fiumera 2015; Giordano et al. 2017; Yue et al. 2017). Similarly, despite the fitness disadvantages of possessing the 2µ plasmids (1.5%–3% growth rate disadvantage compared to cured cells) (Mead, Gardner and Oliver 1986), few genome sequencing studies explicitly comment about the recovery of 2µ plasmid sequences (Baker et al. 2015; Strope et al. 2015; McIlwain et al. 2016; Peter et al. 2018).

Genomic differences among wild, pathogenic and domesticated Saccharomyces strains

The importance of S. cerevisiae for a multitude of industrial processes, such as making wine, ale beers, biofuels, sake and bread, has greatly influenced genome sequencing efforts, including in other Saccharomyces species. Indeed, more 2500 S. cerevisiae strains have been sequenced, including many that were independently sequenced by different labs (Fig. 1). These efforts have helped differentiate the genome characteristics of wild, pathogenic/clinical and domesticated S. cerevisiae strains (Fig. 3). However, it has also been necessary to increase isolation efforts of other Saccharomyces species to generalize the genomic traits found in wild S. cerevisiae strains to other species where all or most known strains are wild. In contrast to wild strains, pathogenic/clinical strains and domesticated strains are both associated with humans. Wild and human-associated strains differ for several genomic characteristics: (i) low heterozygosity in wild isolates, suggesting high-inbreeding rates (Magwene et al. 2011; Wohlbach et al. 2014; Leducq et al. 2016; Peris et al. 2016; Duan et al. 2018; Naseeb et al. 2018; Peter et al. 2018; Nespolo et al. 2019; Langdon et al. 2019b); (ii) fewer admixed strains from the wild, supporting low levels of outcrossing (Liti et al. 2009; Almeida et al. 2014; Leducq et al. 2016; Peris et al. 2016; Eberlein et al. 2019); (iii) the rarity of wild interspecies hybrids [currently only one is known (Barbosa et al. 2016)], suggesting limited opportunities or low fitness for interspecies hybrids in wild environments (Fig. 3); (iv) strong geographic structure of wild Saccharomyces populations (Hittinger et al. 2010; Almeida et al. 2014; Gayevskiy 2015; Leducq et al. 2016; Peris et al. 2016; Duan et al. 2018; Peter et al. 2018), which highlights the limited influence of humans on the expansion of wild strains and (v) more copy number variants (CNVs), especially in subtelomeric genes and more aneuploidies in domesticated lineages (Gallone et al. 2016; Gonçalves et al. 2016; Steenwyk and Rokas 2017). In addition, wild strains have evolved mainly by accumulating SNPs, whereas domesticated and clinical samples are more prone to Ty element and gene family expansions (Peter et al. 2018). However, there are common genomic characteristics among wild and human-associated strains: (i) 75% genes not found in a reference genome are located in subtelomeric regions and are often related to flocculation, nitrogen metabolism, carbon metabolism and stress (Bergström et al. 2014; Steenwyk and Rokas 2017); (ii) subtelomeric regions are hotspots of gene diversity, which influences traits (McIlwain et al. 2016; Yue et al. 2017) and (iii) loss-of-function (LOF) mutations usually occur in non-essential genes and are more frequently found in regions closer to the 3’ end of protein-coding sequences (Bergström et al. 2014).

Figure 3.

Genomic traits of wild, pathogenic, and domesticated Saccharomyces yeasts. Main genomic trait differences inferred from whole genome sequencing studies between wild Saccharomyces, domesticated, and clinical S. cerevisiae strains (Table S1, Supporting Information). Heterozygosity is represented as the percentage of heterozygous sites in the genome. Arrows (→) indicate the introgression/HGT direction inferred. Arrow (↑) indicates an increase in copy number. ADY, active dry yeast; CNVs, copy number variants; HGT, horizontal gene transfer; LOF, loss of function; POF, phenolic off-flavor; SNPs, single nucleotide polymorphisms; Sacc, Saccharomyces; Scer: S. cerevisiae; Spar, S. paradoxus; Sjur, S. jurei; Suva, S. uvarum; Seub, S. eubayanus; Efae, Enterococcus faecium; Tmic, Torulospora microellipsoides; Zbai, Zygosaccharomyces bailii; Lthe, Lachance thermotolerans; PB, Patagonia B; PA, Patagonia A; NA, North America; HOL, Holarctic.

Several HGT events have been described in Saccharomyces, including several specific examples that are well supported (Fig. 3) (Hall and Dietrich 2007; Novo et al. 2009; Galeote et al. 2010; League, Slot and Rokas 2012; Marsit et al. 2015; Peter et al. 2018). Nonetheless, caution is warranted for cases built solely using automated BLAST analysis, which can lead to premature conclusions for two main reasons. First, the absence of published genome sequences for most species make the unequivocal identification of donor and recipient species or clades challenging. Second, gene presence and absence variation of a horizontally acquired gene within or between species can mislead the inference of the history of a gene if population or species sampling is insufficient or biased. For example, large gene families found in subtelomeric regions are particularly prone to being identified as involved in HGT events using simple BLAST criteria due to cryptic paralogy. In these cases, the fact that a gene's best BLAST hit is to a distant species may just be due to missing data. For these reasons, we recommend using BLAST-based statistics, such as Alien Index (Alexander et al. 2016; Wisecaver et al. 2016), to identify interesting candidates, followed by explicit gene tree-species tree reconciliation and phylogenetic topology testing to evaluate candidate HGT events (Alexander et al. 2016; Wisecaver et al. 2016; Shen et al. 2018). Furthermore, the identification of HGT events, as well as more accurate identification of donors and recipients, will greatly benefit from the completion of comprehensive whole genome sequencing projects from diverse species, such as the Y1000 + Project (Hittinger et al. 2015; Shen et al. 2018). In summary, a combination of improved genome sampling and formal phylogenetic approaches together provides the best path forward to generating robust inferences about which genes have been horizontally acquired.

Genomic insights into the fascinating phylogeography of the wild lager-brewing yeast ancestor, Saccharomyces eubayanus

The yeast species S. eubayanus has been isolated exclusively from wild environments; yet, hybridizations between S. cerevisiae and S. eubayanus were key innovations that enabled cold fermentation and lager brewing (Libkind et al. 2011; Gibson and Liti 2015; Hittinger, Steele and Ryder 2018; Baker and Hittinger 2019; Mertens et al. 2019; Langdon et al. 2019a). Industrial isolates of S. uvarum, the sister species of S. eubayanus, with genomic contributions from S. eubayanus have also been frequently obtained from wine and cider (Almeida et al. 2014; Nguyen and Boekhout 2017; Langdon et al. 2019a), indicating that this species has long been playing a role in shaping many fermented products. Even so, pure strains of S. eubayanus have only ever been isolated from the wild. This association with both wild and domesticated environments makes S. eubayanus an excellent model where both wild diversity and domestication can be investigated.

Saccharomyces eubayanus was initially discovered in 2011 in Patagonia (Argentina) from locally endemic tree species of the genus Nothofagus (Libkind et al. 2011). Since then, it has received much attention for brewing applications and as a model for understanding the evolution, ecology and population genomics of the genus Saccharomyces (Sampaio 2018). Many new globally distributed isolates have been found in different parts of the world since its discovery (Bing et al. 2014; Peris et al. 2014; Rodríguez et al. 2014; Gayevskiy and Goddard 2016; Peris et al. 2016; Eizaguirre et al. 2018) but the abundance and genetic diversity measured by multilocus genetic data is still by far highest in Patagonia (Eizaguirre et al. 2018). Recently, two independent investigations significantly increased the number of S. eubayanus American isolates, mainly from Patagonia (Chile and Argentina), and together provide the largest genomic data set for this species with a total of 256 new draft genome sequences (Nespolo et al. 2019; Langdon et al. 2019b). This data set confirms the previously proposed population structure (Peris et al. 2014, 2016; Eizaguirre et al. 2018), where two major populations were detected (Patagonia A/Population A/PA and Patagonia B/Population B/PB), which has been further divided into five subpopulations (PA-1, PA-2, PB-1, PB-2 and PB-3) (Eizaguirre et al. 2018). Other isolates from outside Patagonia belong to PB, either the PB-1 subpopulation that is also found in Patagonia (Gayevskiy and Goddard 2016; Peris et al. 2016), or a Holarctic-specific subpopulation (PB-Holarctic) that includes isolates from Tibet and from North Carolina, USA (Bing et al. 2014; Peris et al. 2016; Brouwers et al. 2019), which represents the closest known wild relatives of the S. eubayanus subgenomes of lager-brewing yeasts (Bing et al. 2014; Peris et al. 2016). Furthermore, heterosis was recently demonstrated in a S. cerevisiae x Tibetan S. eubayanus hybrid, which showed that regulatory cross talk between the two subgenomes is partly responsible for maltotriose and maltose consumption (Brouwers et al. 2019). Multilocus data suggested that two more lineages from China, West China and Sichuan, diverged very early from all other known S. eubayanus strains, while Holarctic isolates from China had unusually low sequence diversity (Bing et al. 2014). In this way, S. eubayanus can be subdivided into a total of eight non-admixed subpopulations (six likely Patagonian–2 PA, 3 PB and 1 PB-Holarctic and 2 Asian–1 West China and 1 Sichuan) and two admixed lineages (one North American lineage with a broad distribution and South American strain sympatric to the Patagonian lineages) (Langdon et al. 2019b). The global distribution and geographically well-differentiated population structure of S. eubayanus is similar to what has been observed for Saccharomyces species, such as S. paradoxus (Leducq et al. 2014, 2016) and S. uvarum (Almeida et al. 2014).

While this species has been easily and repeatedly isolated from South American Nothofagus trees (Libkind et al. 2011; Eizaguirre et al. 2018; Nespolo et al. 2019), only a handful of isolates have been recovered from trees in China, New Zealand and North America (Bing et al. 2014; Peris et al. 2014; Gayevskiy and Goddard 2016; 2016). These data suggest that S. eubayanus is abundant in Patagonia but sparsely found in North America, Asia and Australasia. Most subpopulations display isolation by distance with genetic diversity that mostly scales with the geographic range of a subpopulation. In Patagonia, one sampling location can harbor more genetic diversity than is found in all of North America (Langdon et al. 2019b). The levels of diversity found within Patagonia is further underscored by the restriction of four subpopulations to this region, suggesting that Patagonia is the origin of S. eubayanus diversity or at least the last common ancestor of the PA and PB-Holarctic populations, the latter of which gave rise to lager-brewing hybrids. Different hypotheses and scenarios are discussed in more depth by Langdon et al. (2019b) and Nespolo et al. (2019).

NON-CONVENTIONAL YEASTS WITH NON-CONVENTIONAL GENOMES

Besides the well-studied genus Saccharomyces, more than 1500 recognized yeast species are known, which belong either to the Ascomycota or Basidiomycota (Kurtzman, Fell and Boekhout 2011). In this section, we review the interesting stories recently revealed through the use of genome data of two representative genera of both respective phyla, Hanseniaspora and Phaffia.

The yeasts with the least; the reductive genome evolution of Hanseniaspora

A hallmark of evolution in the budding yeast subphylum Saccharomycotina is the loss of traits and their underlying genes (Shen et al. 2018). Arguably, the most dramatic example of reductive evolution observed is the Hanseniaspora (Steenwyk et al. 2019), a genus of bipolar budding, apiculate yeasts in the family Saccharomycodaceae. Hanseniaspora yeasts can be assigned to two lineages, a faster-evolving one and a slower-evolving one (FEL and SEL, respectively), which differ dramatically in their rates of genome sequence evolution as well as in the extent and types of genes that they have lost (Fig. 4). The types of genes lost can be broadly ascribed to three categories: metabolism, DNA repair, and cell-cycle.

Figure 4.

The evolutionary trajectories of Hanseniaspora lineages are marked by differential rates of sequence evolution and rates of loss of metabolism, DNA repair and cell-cycle genes. (A), There are two lineages in the budding yeast genus Hanseniaspora: the faster-evolving and slower-evolving lineage (FEL and SEL, respectively). The FEL has a long and thicker stem branch indicative of higher rates of sequence evolution or higher mutation rates, whereas the SEL has a much shorter and thinner stem branch indicative of lower rates of sequence evolution or lower mutation rates. (B), Each lineage has lost many genes associated with metabolism, DNA repair and cell-cycle processes; squares with colors toward the red end of the spectrum correspond to greater rates of gene loss, whereas squares on the white end of the spectrum correspond to lower rates of gene loss.

Metabolism-related genes have been lost in both FEL and SEL. Analysis of 45 growth traits across 332 Saccharomycotina yeasts revealed that Hanseniaspora species can assimilate fewer carbon substrates compared to most of their relatives (Opulente et al. 2018; Shen et al. 2018) and have lost many of the associated genes and pathways (Steenwyk et al. 2019). Although less pronounced, similar gene and trait losses have been observed in wine strains of S. cerevisiae (Gallone et al. 2016; Steenwyk and Rokas 2017) and are thought to be signatures of adaptation to the wine must environment (Steenwyk and Rokas 2018). These gene losses may play a similar role in the ecology of Hanseniaspora species, considering their frequent isolation from fruit juices and fermenting musts (Cadez 2006; Kurtzman, Fell and Boekhout 2011), which likely reflects the specialization of Hanseniaspora species to sugar-rich environments.

Hanseniaspora species, especially those in the FEL, have lost numerous DNA repair genes spanning multiple pathways and processes (Steenwyk et al. 2019). For example, yeasts in both lineages have lost 14 DNA repair genes, including PHR1, which encodes a photolyase (Sebastian, Kraus and Sancar 1990) and MAG1, which encodes a DNA glycosylase that is part of the base excision repair pathway (Xiao et al. 2001). However, FEL yeasts have lost 33 additional DNA repair genes, which include polymerases (i.e. POL4 and POL32) and numerous telomere-associated genes, such as CDC13 (Lustig 2001). Inactivation or loss of DNA repair genes can cause hypermutator phenotypes, such as those observed in microbial pathogens and in human cancers (Jolivet-Gougeon et al. 2011; Billmyre, Clancey and Heitman 2017; Campbell et al. 2017). In the short-term, hypermutation can facilitate adaptation in maladapted populations by increasing the chance of occurrence of beneficial mutations (e.g. conferring drug resistance); in the long-term, however, hypermutation is not a viable strategy due to the increased accumulation of deleterious mutations (Ram and Hadany 2012). Molecular evolutionary analyzes suggest that the stem lineages of FEL and SEL yeasts were hypermutators; interestingly, the increased mutation rates in the two stem lineages reflect the degree of observed DNA repair gene loss in the two lineages. The larger number of gene losses in FEL stem branch is consistent with its higher mutation rate and the smaller number of gene losses in the SEL stem branch is consistent with a lower increase in its mutation rate (Steenwyk et al. 2019). However, the mutation rates of both FEL and SEL crown groups (i.e. every branch after the stem) are similar to those of other yeast lineages, consistent with evolutionary theory's predictions that long-term hypermutation is maladaptive (Ram and Hadany 2012; Steenwyk et al. 2019). Altogether, Hanseniaspora have lost DNA repair genes, undergone punctuated sequence evolution, and slowed down their overall mutation rate, despite having a reduced DNA repair gene repertoire. Finally, Hanseniaspora yeasts have lost genes associated with key features of the cell cycle, including cell size control, the mitotic spindle checkpoint and DNA-damage-response checkpoint processes, but these losses are more pronounced in the FEL. For example, both lineages have lost WHI5, a negative regulator of the G1/S phase transition in the cell cycle that is critical for cell size control (Jorgensen et al. 2002). Other gene losses are exclusive to the FEL, such as the loss of MAD1 and MAD2, which bind to unattached kinetochores and are required for a functional mitotic spindle checkpoint (Heinrich et al. 2014), as well as RAD9 and MEC3, which function in the DNA-damage-checkpoint pathway and arrest the cell cycle in G2 (Weinert, Kiser and Hartwell 1994). The loss of checkpoint genes is thought to contribute to bipolar budding in both lineages and greater variance in ploidy, as well as strong signatures of mutational burden due to aberrant checkpoint processes in FEL compared to SEL (Steenwyk et al. 2019). These observations suggest landmark features of cell cycle processes are absent in Hanseniaspora and warrant future investigations into the functional consequences of these losses.

Phaffia rhodozyma: A colorful genome from the Basidiomycota

The orange-colored yeast Phaffia rhodozyma ( = Xanthophyllomyces dendrorhous), an early diverging Agaricomycotina (Basidiomycota), possesses multiple exceptional traits of fundamental and applied interest. The most relevant is the ability to synthesize astaxanthin, a carotenoid pigment with potent antioxidant activity and of great value for the aquaculture and pharmaceutical industries. Hyperpigmented mutants of P. rhodozyma are currently being exploited biotechnologically as a natural source of astaxanthin in aquaculture feed (Rodríguez-Sáiz, De La Fuente and Barredo 2010). These mutants were derived from an initial collection from 1976 from bark exudates of specific tree species (e.g. Betula sp.) from the Northern Hemisphere. Today, P. rhodozyma is known to have specific niches in association with trees of mountainous regions and a worldwide distribution comprising at least seven different genetic lineages (David-Palma, Libkind and Sampaio 2014). One of these lineages was obtained from Andean Patagonia (Argentina) on Nothofagus trees, the same substrates as S. eubayanus and S. uvarum (Section 3.2) (Libkind et al. 2011), and based on genomic analyzes, Patagonian wild strains were recently proposed as a potential novel variety of P. rhodozyma (Bellora et al. 2016). The 19-Mb genome of P. rhodozyma CRUB 1149 wild Patagonian isolate was sequenced and assembled, achieving a coverage of 57x. Analysis of its gene structure revealed that the proportion of intron-containing genes and the density of introns per gene in P. rhodozyma are the highest hitherto known for fungi, having values more similar to those found in humans than among Saccharomycotina where intronless genes predominate. An extended analysis suggested that this trait might be shared with other members of the order Cystofilobasidiales (Bellora et al. 2016).

Genome mining revealed important photoprotection and antioxidant-related genes, as well as genes involved in sexual reproduction. New genomic insight into fungal homothallism was obtained, including a particular arrangement of the mating-type genes that might explain the self-fertile sexual behavior. All known genes related to the synthesis of astaxanthin were annotated. Interestingly, a hitherto unknown gene cluster potentially responsible for the synthesis of an important UV protective and antioxidant compound (mycosporine-glutaminol-glucoside) (Moliné et al. 2011) was found in the newly sequenced and mycosporinogenic strain. However, this gene cluster was absent in a strain (CBS 6938) that does not to accumulate this secondary metabolite, which has potential applications in cosmetics (Colabella and Libkind 2016). Genome mining also revealed an unexpected diversity of catalases and the loss of H₂O₂-sensitive superoxide dismutases in P. rhodozyma. Altogether, the P. rhodozyma genome is enriched in antioxidant mechanisms, in particular those most effective at coping with H₂O₂, suggesting that the environmental interaction with this reactive species has definitely contributed to shaping the peculiar genome of P. rhodozyma.

YEAST BIOTECHNOLOGY GETS WILD WITH GENOMICS

The identification of new yeast strains and novel species could offer valuable innovative opportunities for applied research by taking advantage of traits found by bioprospecting in extreme environments (Pretscher et al. 2018; Cubillos et al. 2019). Newly isolated yeasts are expanding the repertoire of phenotypic diversity, and therefore, the current known variation in physiological and metabolic traits. These yeasts from extreme environments are of considerable interest in biotechnology, owing to diverse advantages, such as: rapid growth rates at extreme temperatures, (Choi, Park and Kim 2017; Yuivar et al. 2017; Cai, Gao and Zhou 2019), extraordinary capacity of fermentation in large-scale cultures (Choi, Park and Kim 2017; Krogerus et al. 2017) and the production of cold-active hydrolytic enzymes (such as lipases, proteases, cellulases and amylases) (Martorell et al. 2019). For example, the cryotolerant yeast S. eubayanus exhibits a wide set of relevant traits appropriate for brewing, including efficient biomass production at low temperature and production of high levels of esters and preferred aroma compounds in beer (Libkind et al. 2011; Hebly et al. 2015; Mertens et al. 2015; Alonso-del-Real et al. 2017; Gibson et al. 2017; Krogerus et al. 2017). Similarly, an Antarctic isolate of Wickerhamomyces anomalus has been indicated as a high producer and secretor of glucose oxidases, invertases and alkaline phosphatases enzymes at lower temperatures, decreasing the temperature requirement for their production (Schlander et al. 2017; Yuivar et al. 2017). In this context, the availability of new yeasts as biological and genetic resources from the wild immediately opens new avenues, not only for their direct utilization in industrial processes, but also to gather and obtain new genomic data so that their genes can be integrated into complex industrial systems already in use. However, the use and manipulation of these genetic resources are restricted by the limited knowledge in terms of the molecular basis underlying metabolic traits of industrial interest. Mining this genomic and phenotypic diversity provides a great opportunity to pinpoint unique pathways of biotechnological importance, which can then be exported to other systems or improved within the same genetic backgrounds. Recent advances in bioinformatics, quantitative genetics, systems biology and integrative biology, together with the large number of new genome sequencing projects are providing the means to address these challenges (Liti 2015; Peter et al. 2018; Viigand et al. 2018; Cai, Gao and Zhou 2019; Nespolo et al. 2019; Langdon et al. 2019b). Thus, leveraging wild yeast genomes, together with other ‘multi-omic’ approaches can generate possible targets for biotechnological applications.

Genomics can support predicting biochemical traits in organisms with biotechnological potential, where the combination of comparative genomic and physiological studies can allow key genomic features to be inferred in non-conventional organisms (Riley et al. 2016). Furthermore, efforts to unravel the complexity of yeast genomes have proven successful in providing genome-scale models that can determine their potential metabolic profiles (Loira et al. 2012; Lopes and Rocha 2017). These models can be applied to new yeast genomes to predict an organism's chemical repertoire by reconstructing metabolic pathways and elucidating their biotechnological potential (Wang et al. 2017). Thus far, these approaches have been successfully applied to a subset of strains in model yeasts, such as Yarrowia lipolytica (Loira et al. 2012), S. cerevisiae (Heavner and Price 2015; Mülleder et al. 2016), and Komagataella phaffii (formerly known as Pichia pastoris) (Saitua et al. 2017). Their utilization in novel organisms is still in its infancy, but the integration of transcriptional regulatory networks and metabolic networks could guide novel metabolic engineering applications (Shen et al. 2019) to convert new yeasts (strains or species) into potential resources for the production of biofuels and biochemicals.

Biotechnological applications in non-conventional organisms are poised to be enhanced by recent advances in genome-editing techniques, such as CRISPR-Cas9 (Donohoue, Barrangou and May 2018). The utilization of CRISPR-Cas9 requires whole genome sequences so that gRNAs can be designed to specifically target genes of interest. This system is highly effective in S. cerevisiae and other Saccharomyces species, mostly due to their efficient homology-directed DNA repair machinery (Akhmetov et al. 2018; Kuang et al. 2018; Mertens et al. 2019). For example, novel S. eubayanus strains recently isolated from Patagonia (Rodríguez et al. 2014) were successfully engineered for the lower production of phenolic off-flavors (Mertens et al. 2019). Interestingly, high success rates have also been reported in other non-conventional yeasts, demonstrating the large spectrum of genomes that can be modified using the CRISPR-Cas9 system (Wang et al. 2017; Juergens et al. 2018; Kuang et al. 2018; Cai, Gao and Zhou 2019; Lombardi, Oliveira-Pacheco and Butler 2019; Maroc and Fairhead 2019). For example, CRISPR–Cas9-assisted multiplex genome editing (CMGE) in the thermotolerant methylotrophic yeast Ogataea polymorpha allowed for the introduction of all the genes necessary for the biosynthesis of resveratrol, along with the biosynthesis of human serum albumin and cadaverine (Wang et al. 2017). The seemingly universal capacity of the CRISPR-Cas9 genome-editing technique means that many, if not all, yeasts will ultimately be susceptible to being modified using this system. Thus, even newly isolated yeasts and novel species could be used as microbial cell factories, allowing the spectrum of applications and products to be expanded.

CONCLUSIONS

The power of genomics in the study of yeast biology, evolution and biotechnology is highly dependent on the number of genome sequences available, and this factor is currently the main limitation for comprehensive studies. So far, studies have focused mostly on model species or taxa of specific fundamental or applied interest, mainly for ascomycetous yeasts. In contrast, few projects have dealt with basidiomycetous yeast genomes, many of which also likely harbor interesting characteristics. The description of novel species based on complete genome sequences is still not a trend among yeast taxonomists, probably due in part to cost and due in part to the lack of general guidelines for this practice. A review included in this issue represents the first attempt to establish minimal advice for taxonomic descriptions using whole genome sequence data for the formal descriptions of novel yeast species (Libkind et al. 2020 submitted). As this practice becomes more widespread and the genomic database for non-conventional yeasts grows, our ability to answer different biological questions about their history, ecological adaptations and dynamics will increase. Even so, new bioinformatic tools that are more user-friendly and automated will make the power of genomics more accessible to researchers without bioinformatic training. On the technological side, the gradual increase in the use of long-read sequencing technologies will enable the exploration of complete or near-complete genome assemblies, including repeats and telomeres, of non-conventional yeasts.

Here, we provided clear examples of how our understanding of many biological and evolutionary processes has been improved by widening the spectrum of yeasts studied, especially by including non-conventional yeasts from the wild. Emblematic cases from the anthropogenically affected genus Saccharomyces were addressed as an example of how genomics helped to cast light into complex microbial domestication processes and to detect genomic signatures of pathogenicity and domestication. This insight would not have been possible if large genomic data sets from wild isolates of S. cerevisiae were not available. Similarly, the previously missing wild ancestor of lager-brewing yeasts would have not been found if yeast explorations into pristine and remote environments had not been carried out. Studies in the less known genus Hanseniaspora, including both domesticated and wild strains, revealed unexpected evolutionary histories, with surprising and interesting modes of genome evolution. The basidiomycetous yeast Phaffia rhodozyma provided an illustrative example of the unique genomic traits that can be found within this understudied phylum. In the future, the large number of new yeast genomes, along with transcriptomic, proteomic and other multi-omic studies, will rapidly improve our understanding of non-conventional and indeed all organisms at the systems level.

Conflict of interest . None declared.