Population genomics of domestic and wild yeasts

Population genomics of domestic and wild yeasts

David M. Carter

*, Gianni Liti

*, Alan M. Moses

, Leopold Parts

, Stephen A. James

Robert P. Davey

, Ian N. Roberts

, Anders Blomberg

, Jonas Warringer

, Austin Burt

Vassiliki Koufopanou

, Isheng J. Tsai

, Casey M. Bergman

, Douda Bensasson

Michael J. T. O'Kelly

, Alexander van Oudenaarden

, David B. H. Barton

, Elizabeth

Bailes

, Matthew Jones

, Michael A. Quail

, Ian Goodhead

, Sarah Sims

, Frances

Smith

, Richard Durbin

* & Edward J. Louis

Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton,

Cambridge, CB10 1HH, UK,

Institute of Genetics, Queen's Medical Centre, University

of Nottingham, Nottingham NG7 2UH, UK,

National Collection of Yeast Cultures,

Institute of Food Research, Norwich Research Park, Colney, Norwich NR4 7UA, UK,

Department of Cell and Molecular Biology, Lundberg Laboratory, University of

Gothenburg, Medicinaregatan 9c, 41390 Gothenburg, Sweden,

Division of Biology,

Imperial College London, Silwood Park, Ascot, Berks., SL5 7PY, UK,

Faculty of Life

Sciences, University of Manchester, Manchester M13 9PT, UK,

Department of

Physics, Massachusetts Institute of Technology, Cambridge MA 02139, USA

*These authors contributed equally to this work

+Present address: Department of Cell & Systems Biology, University of Toronto, Canada, M5S 2J4

Present address: School of Biological Sciences, University of Liverpool, Liverpool, LG9 3BX

The natural genetics of an organism is determined by the distribution of sequences

of its genome. Here we present one- to four-fold, with some deeper, coverage of the

genome sequences of over seventy isolates of the domesticated baker's yeast,

Saccharomyces cerevisiae, and its closest relative, the wild S. paradoxus, which has

never been associated with human activity. These were collected from numerous

geographic locations and sources (including wild, clinical, baking, wine, laboratory

and food spoilage). These sequences provide an unprecedented view of the

population structure, natural (and artificial) selection and genome evolution in

these species. Variation in gene content, SNPs, indels, copy numbers and

transposable elements provide insights into the evolution of different lineages.

Phenotypic variation broadly correlates with global genome-wide phylogenetic

relationships however there is no correlation with source. S. paradoxus populations

are well delineated along geographic boundaries while the variation among

worldwide S. cerevisiae isolates show less differentiation and is comparable to a

single S. paradoxus population. Rather than one or two domestication events

leading to the extant baker's yeasts, the population structure of S. cerevisiae shows

a few well defined geographically isolated lineages and many different mosaics of

these lineages, supporting the notion that human influence provided the

opportunity for outbreeding and production of new combinations of pre-existing

variation.

Since the completion of the genome sequence of Saccharomyces cerevisiae in 1996, the

first eukaryotic organism sequenced

1, 2

, there has been an exponential increase in

complete genome sequences accompanied by great advances in our understanding of

genome evolution. This has been particularly evident from comparative genomics both

of close relatives, e.g. for Saccharomyces yeasts

3, 4

and for Drosophila species

5, 6

, and of

more distant relatives, e.g. for hemiascomycetes yeast species

7, 8

. Less well understood

at the whole genome level are the evolutionary processes acting within populations and

species leading to adaptation to different environments, phenotypic differences and

reproductive isolation. There has been increased recognition of the importance of

understanding the genome sequence variation within a species in order to fully

understand the nature of a species and individual phenotypic variation, for example

single nucleotide or copy number variation in relation to human disease

. A single

genome sequence for a species is inadequate for answering many biological questions

which require large-scale population sampling. The tools for large-scale comparisons of

populations of whole genome sequences do not yet exist and require adequate and

realistic datasets for their development and evaluation. The Saccharomyces sensu stricto

yeasts, including the best understood model eukaryote S. cerevisiae, are ideal for

population genomics studies, as there are many isolates available from several species

from around the world, and many aspects of their biology are well characterised and

amenable to study. Although little is known about the natural and life histories of yeasts

in the wild, there are an increasing number of studies looking at ecological and

geographic distributions

10, 11

, population structure

12-15

, and sexual versus asexual

reproduction and gene flow

16, 17

. When combined with a population view of whole

genome sequence variation, these approaches will increase our understanding of the

ecology and evolutionary biology of these important organisms.

The baker's yeast, S. cerevisiae, has had a long association with human activity

leading to the idea that use in fermentation imposed selection on specific lineages

leading to domestication. A survey of five loci in many isolates supports the idea that

there were at least two independent domestication events arising from human activity,

one for sake strains and one for wine which were derived from wild populations

. In

contrast, the closest relative of S. cerevisiae, S. paradoxus, has never been associated

with human activity and is found globally and in many cases in the same locations as S.

cerevisiae

10, 11

. A preliminary comparison of a few genes from several isolates of each

species within the Saccharomyces sensu stricto group exhibited extensive variation

between three well defined geographic populations of S. paradoxus but limited variation

among S. cerevisiae isolates and no correlation with geographic location

. S. cerevisiae

is usually isolated from fermentation processes (beer, wine, sake, bread) but can also be

found in the wild (either associated with fermenting fruit or nectar or isolated from non-

fermenting substrates such as soil or bark), on humans (either as non-pathogenic

commensals or as disease causing clinical isolates), and in food spoilage situations.

Isolates from different sources have phenotypes associated with adaptation to that

source, so that baking strains, for example, are fundamentally different from wine

strains in that they exhibit vigorous maltose fermentation

, while clinical isolates

exhibit the ability to grow at high temperatures, which is associated with virulence

Here we present the results of the first large-scale multi-species population

genomics analysis in eukaryotes using the genome sequences of 36 strains S. cerevisiae

and 35 strains S. paradoxus isolated from several sources and locations. These

sequences not only provide a dataset to aid development of new population genomics

tools but will also help us understand the nature of Saccharomyces yeasts, the genetic

variation underlying important phenotypic differences, and the effect of human activity

on the evolution of one of the most important model and industrial organisms.

Strain selection

Strains of S. cerevisiae and S. paradoxus were chosen to maximize the potential

variation across geography as well as environments (Tables 1 and S1). Strain selection

was subject to certain constraints to avoid technical difficulties with sequencing as well

as maximize future utility. Except for the laboratory strains S288c, W303 and the

baking strains, the isolates selected were diploid. As nearly 10% of Saccharomyces

sensu stricto isolates appear to be hybrids between different species

, we selected

strains that behaved as diploid non-hybrids in meiosis and test crosses. We obtained a

single spore isolate from the strains in order to avoid the problem of heterozygosity

within the diploid genome, which would cause problems with SNP calling and

assemblies. Each single spore isolate sequenced, as well as the other three spores from

the same meiosis, have been stored and are available at NCYC

The S. cerevisiae isolates include strains from a large variety of sources:

laboratory, pathogenic, baking, wine, food spoilage, natural fermentation, sake,

probiotic, plant, soil as well as S288c, the progenitor of the reference strain sequence.

These strains were isolated from numerous locations around the world and were

obtained primarily from various culture collections or other researchers (Table S1). The

S. paradoxus isolates were mostly isolated from oak tree exudates and bark from

various locations around the world and include the three major populations now

recognised as European, Far Eastern and American

13, 15, 22

. The latter include two South

American isolates originally designated as a separate species, S. cariocanus

. The

European group also includes 18 strains isolated in the UK from the same geographic

area. Two new locations, Siberia, which is geographically located halfway between

European and Far East Asia, and the island of Hawaii, were also included. There is

overlap in the general geographic sources of isolates from both species.

Sequencing and assembly

The majority of sequences were obtained using standard Sanger sequencing on

ABI3730 sequencing machines. For some strains short sequencing-by-synthesis (SBS)

reads were obtained using the Illumina Genetic Analyser. Most strains were covered to

a depth of 1-3X with a few covered extensively with both ABI and Ilumina reads. The

basic statistics are shown in Table 1 while details of number of reads, depth of coverage

and description of libraries are given in supplementary information (Table S2). High

levels of ABI coverage were obtained for the S288c version we sequenced as well as

laboratory strains W303, SK1, Y55 and of our single spore isolate of the S. paradoxus

type strain, CBS432. The sequence reads, assemblies, alignments, other data and a

genome browser are all publicly available

. Reads are also deposited at the

international trace archive

Reference-based genome assemblies were created for each strain in a series of

steps. First, each read was aligned to the reference genome for the relevant species

(S288c or CBS432). As this approach cannot deal with large indels or with sequences

not present in the reference genome, we developed an iterative parallel assembling tool,

PALAS (see Methods in SI), to introduce insertions that were allowed to share material

between related strains. Two versions of each strain sequence were produced, a partial

assembly derived just from data collected from that strain, and a more complete

assembly using an imputation process to infer the most likely sequence of the strain

taking into account data from related strains. In both cases confidence estimates are

given for each base call. Even using PALAS, some sequences such as the subtelomeric

regions, which structurally are highly polymorphic

, could not be reliably assembled.

Because of their extreme AT-richness, the mitochondrial genome sequences are also

incomplete. In all we identified 235127 high-quality SNPs and 14051 indels in the S.

cerevisiae nuclear genome, and 623287 SNPs and 25267 indels in S. paradoxus. Further

details are given in supplementary information (Table S2).

Nearly complete assemblies have been obtained for the four laboratory strains of

S. cerevisiae including our version of S288c. Our S288c varies from the reference

genome (downloaded from SGD 10-10-2007) by 659 high-quality SNPs, of which we

estimate at least 33 (5%) are due to accumulated divergence since the two versions

separated. Many of the differences found are in subtelomeric regions and there appears

to be a higher rate of discrepancy in chromosomes 1 and 2 than in the rest of the

genome (Fig. S1). Of the high-quality aligned positions, for 480 SNPs our S288c

sequence is supported by other strains sequenced whereas the reference version has no

support, while for 18 SNPs the reference genome sequence is supported by other strains

while our S288c has no support. Many of the former set are likely to represent errors in

the reference genome sequence, which still implies a low error rate of 1 in 24kb (better

than Q40). A list of likely errors is given in Table S3 and has been submitted to SGD.

The available sequence for the type strain of S. paradoxus

is not complete and so

we sequenced a single spore isolate from our version of CBS432 to 4.3X coverage with

ABI. Although the two sequences are closely related and are part of the same

population, there are numerous SNP differences between them. This could be due to the

fact that we used a single spore isolate which would not have any heterozygosity, while

the assembly created by Kellis et al

has the other allele at some SNP locations.

Alternatively, the difference may come from the strains actually not being the same as

they were obtained from separate culture collections (Northern Regional Research

Laboratory, NRRL, and Centraalbureau voor Schimmelcultures, CBS). Our version of

CBS432 is identical to that in the CBS collection over specific regions of SNP

differences confirmed by sequencing of an independent sample from CBS. In order to

circumvent this issue and to produce a S. paradoxus reference sequence with the

greatest contiguity we collected together all the reads from the UK isolates of S.

paradoxus, the reads from the Broad project

, and artificial reads created by shredding

the Broad contigs. We assembled these reads using Phusion

into 608 contigs, which

we then aligned to the S. cerevisiae reference to place them on chromosomes. We used

the resulting sequence as the starting point for the reference-based assemblies for the

individual S. paradoxus strains. We also aligned this sequence and the contigs for S.

bayanus, S. kudriavzevii and S. mikatae (downloaded from SGD

) to create cross-

species multiple alignments used in subsequent analyses. Further details of both

procedures are in the supplementary information.

SNPs and the Population structures of the two species

We generated neighbour-joining (NJ) phylogenetic trees based on pairwise SNP

differences in the strain-sequence alignments (Fig. 1 and Fig. S2) to obtain a global

picture of relationships among the strains and species. As has been previously

reported

13, 15

for a smaller gene sample, the S. paradoxus strains fall into three very well

separated populations with one lineage outside these (the Hawaiian isolate discussed in

more detail below). The S. cerevisiae isolates exhibit somewhat more variation species-

wide than a single S. paradoxus population but much less than the overall species-wide

variation seen in S. paradoxus. Most of the SNPs in S. paradoxus are private

polymorphisms within each population while most of the SNPs in the S. cerevisiae

samples are shared. There is some structure in the S. cerevisiae sequences, as will be

discussed below, but there is evidence for extensive recombination between lineages.

S. paradoxus exhibits a clear picture of population structure using the software

STRUCTURE

which indicates three populations plus the single Hawaiian isolate (Fig.

2A). Previous analysis with smaller segments sequenced resulted in the same picture of

three well defined populations

13, 15

. The European population was sampled extensively

which provided a picture of within population dynamics (Fig. 1B). The topology of the

NJ tree of the European population correlates with the West to East geographic

locations of isolation of the strains.

The S. cerevisiae population structure is more complicated. There are five lineages that

exhibit the same phylogenetic relationship across their entire genomes, which we

consider to be `clean' non-mosaic lineages (Fig. 1C). These are strain sets from

Malaysia, West Africa, sake and related fermentations, North American, and a large

cluster of mixed sources that contains many European and wine making strains

(Wine/European). The remaining strains are on long branches between the large

Wine/European cluster and the other four clean lineages. There are some lineages that

correspond to geographic origin, such as the wild isolates from North America and

Malaysia. However many of the closely related strains seem to be from widely

separated locations. The lab strains SK1 and Y55, isolated in North America and

Europe respectively, are on a branch with two West African strains; and one African

strain (Y12) is on a branch with sake strains while another (DBVPG1853) is on a long

branch unrelated to other strains. This mixed architecture could be due to human traffic

in yeast strains and subsequent recombination between the lineages. Analysis with

STRUCTURE is consistent with separate populations for the West African, Malaysian,

sake lineages and the Wine/European cluster at the far right of the NJ tree (Fig. 2B).

The wild North American isolates share some polymorphisms with all four separate

populations, while all of the rest share polymorphisms with the European lineage and at

least one other.

Trees constructed on a chromosome-by-chromosome basis or by smaller segments make

the mosaic nature of these genomes clear, as do segmental similarity comparisons (Fig.

2C and Fig. S3). The laboratory strains SK1 and Y55 appear to be the result of recent

crosses between the West African lineage and the European lineage, as can be clearly

seen in pairwise comparisons along the chromosomes (Fig. S3B). Similarly W303 is a

recent cross with the reference S288c lineage and one or more other lineages. Other

strains appear to be more ancient combinations of the different lineages with no long

segments (linkage disequilibrium blocks) retained from one or another lineage. Some

chromosomes or segments fall into different locations in the NJ tree (Fig. 2C) while

many stay on a long branch in roughly the same relationship to other strains. The

recently sequenced clinical derivative YJM789

, for example, is one of these long-

branch mosaic strains. The wild North American isolates represent a clean lineage as

most segments exhibit the same phylogenetic relationship with the other lineages across

the genome. The shared polymorphisms with the other four clean lineages indicate a

more ancient interaction. This complex population structure of S. cerevisiae is

consistent with previous analyses

and is similar to a recent study by Schacherer et al

(this issue). However, rather than concluding that there were two independent

domestication events in the history of this species

, the picture presented here is more

consistent with there being five well-delineated lineages, two of which contain multiple

isolates used in fermentation industries. In many cases, the nearest relative to a

fermentation strain over some portion of the genome is a wild or clinical strain. There is

the added complication that many extant strains are recombinants between lineages,

most but not all of which are also used for fermentation. Nevertheless, phenotypic

profiling and analysis of rDNA repeat unit variation both produce results consistent with

this overall picture of the S. cerevisiae population structure (see below).

Whether we have sampled the entire space of existing S. cerevisiae lineages is open to

debate. It is clear that segments from many of the strains, such as W303 and the mosaic

strains (Fig. S3), are not related to any of the lineages delimited here and are probably

derived from yet to be determined or no longer existing lineages. Many of these strains

are segmentally related to each other. Future work may be able to provide a partial

reconstruction of the other lineages from the sequences that now exist.

We have previously identified in the left arm of chromosome XIV a 23 Kbp region of

introgression from S. cerevisiae into the European population of S. paradoxus

. An

analysis of the sequence reads in this region of the genome is consistent with a single

event that spread through the population, as the end points of introgression are the same

in all strains sequenced over this region (Fig. S4). We searched for a possible S.

cerevisiae donor and found several smaller segments within the introgression that

appeared similar to different clean lineages indicating the donor was one of the mosaic

strains.

We also find a significant number of S. cerevisiae-like reads in the Hawaiian isolate of

S. paradoxus. In many cases there are S. paradoxus reads covering the

homologous/allelic locations. At present this is not easily explained as the strain

behaves as a diploid member of the S. paradoxus species by inherent fertility and

meiotic viability in test crosses. Both sets of sequences exist in the strain as confirmed

by PCR and independent isolation and DNA preparation. Further sequencing and

analysis of paired reads will help to resolve the potential hybrid origin of the Hawaiian

isolate of S. paradoxus.

Population genetics: recombination and selection

Sequence variability in populations is typically estimated in two ways, as the average

pairwise divergence of sequences in a population ( ) and from the proportion of

polymorphic or segregating sites (

)

. Here we estimate these parameters for (i) the

UK population of S. paradoxus (the closest we have to several isolates from a single

natural population); (ii) a global sample of S. cerevisiae, excluding all isolates that

appear to be clonemates; and (iii) the Wine/European cluster of S. cerevisiae (the best

sampled of any of the clean lineages, again excluding likely clonemates; Table S4).

Both

and

are about 0.001 in the UK population of S. paradoxus, indicating an

average of 1 difference per 1000bp between two strains. This is very similar to the value

previously found for chromosome III in the same population

. The Wine/European

cluster of S. cerevisiae has approximately the same level of diversity as the UK S.

paradoxus population, while the global sample of S. cerevisiae has >5 times more

diversity, but is less diverse than the global sample of S. paradoxus. The average

divergence of two S. cerevisiae strains is about 40% of that between European and Far

East S. paradoxus, and about 15% of that between the American S. paradoxus and

either of the other two populations

. In both the global and Wine/European samples of

S. cerevisiae Tajima's D

(a standardised measure of the difference between

and

)

is significantly negative, indicating an excess of singleton polymorphisms. This may be

a consequence of our sampling strategy for S. cerevisiae, which included single isolates

from many sources. By contrast, the UK sample of S. paradoxus is from a single

population, collected within a 10km

area, and Tajima's D is positive, indicating a

relative abundance of mid-frequency polymorphisms, though the deviation from random

expectation is small.

We have also analysed nucleotide diversity separately for each chromosome. In the

global sample of S. cerevisiae there is a significant negative correlation between the

length of the chromosome and the amount of variation, in particular for shorter

chromosomes (Kendall's

= -0.52, p=0.008). This appears to be because there is more

variation in the subtelomeric regions extending 30kb from each end of each

chromosome, and these regions make up a larger proportion of shorter chromosomes. If

these regions are excluded, the correlation is no longer significant.

The covariation of alleles at different sites, or linkage disequilibrium, also differs

between samples (Fig. 3A). For S. paradoxus linkage disequilibrium declines smoothly

as the distance between sites increases, decaying to half its maximum value at about

9kb, similar to that previously reported

. For both S. cerevisiae samples the linkage

disequilibrium decays much faster, with a half life of 3kb or less. This implies more

recombination in S. cerevisiae, perhaps due to selection for recombinants, or more

opportunity for strains to mate and recombine.

Our genome-scale population variation dataset, in conjunction with the high-quality

genome annotation available for S. cerevisiae

, allows us to compare systematically the

patterns of variation for mutations with respect to the functional genomic regions in

which they occur. Although the low coverage, variable completeness and complex

population structure in our data makes comparison of our observations to theoretical

expectations difficult to interpret, it is possible for us to compare our data between

functional classes to obtain a picture of the relative effects of selection on new

mutations.

As expected under a model of weakly deleterious effects of mutations, in coding regions

we find a shift in the derived allele frequencies (see methods) for amino acid

replacement polymorphism (Fig. 3B, a) towards lower frequencies compared to

synonymous polymorphism (Fig. 3B, s). To quantify this, we computed the a/s ratio at

different allele frequencies. For polymorphism with minor allele frequency (MAF) less

than 20%, there were 0.86 amino acid changing polymorphisms for each silent

polymorphism. In contrast, for those with MAF greater than 20% this ratio was 0.34,

indicating that at least 61% (1-0.31/0.86) of the 24418 amino acid changing

polymorphisms with minor allele frequency less than 20% are deleterious, and are

destined to be removed from the population by natural selection.

The preceding calculations make the assumption that synonymous polymorphisms are

neutral. In yeast, however, this is known not to be the case; genes with high expression

levels exhibit high codon bias

and this can be measured by the codon-adaptation index

(CAI)

. We computed the derived allele frequency spectrum for the silent

polymorphisms in genes with high levels of codon bias (average CAI>0.6) and found an

excess of polymorphism at both low and high frequency (Fig. 3B, s*). This suggests the

action of both positive and negative selection on polymorphism at these sites. To test

this we identified polymorphisms in which the derived allele was either a preferred or

un-preferred codon, as defined by the CAI, and compared their allele frequencies in

genes with high codon bias to those in the rest of the genome. Consistent with both

positive and negative selection acting on synonymous sites in highly expressed genes,

we observe both an excess of low-frequency (DAF<20%) SNPs that created un-

preferred codons (38% vs. 51%, p<10

-6

) and an excess of high-frequency (DAF>20%)

SNPs that create preferred codons (67% vs. 54%, p<0.01 Fisher's Exact Test). These

results indicate that codon bias in S. cerevisiae is maintained by both purifying and

positive selection, as suggested by the mutation-selection-drift model

An unexpected finding in previous genome-wide studies of population variation

was

the large number of seemingly highly deleterious alleles. Interestingly, we also found

large numbers of mutations that were predicted to introduce stop codons in our data

(Fig. 3B, `create stop'), including 5 stop codons in essential genes. These mutations

showed an extremely skewed allele frequency distribution, consistent with them being

deleterious and likely to be removed from the population. One possible explanation is

that the truncations they introduce are compatible with protein function. Indeed, we

found that these stop mutations were significantly enriched in the C-termini (the final

5% of protein length, Fig. 3B inset) of proteins, suggesting that in some cases the stop

codons cause relatively moderate consequences to protein function.

Our genome-wide variation data was derived from direct sequencing (rather than SNP

genotyping as in previous studies

) and this afforded the opportunity to consider

insertion and deletion mutations (indels) as well as single nucleotide polymorphisms

(Fig. 3C). We identified 3377 indels segregating in the coding regions of the S.

cerevisiae population (467 with minor allele frequency greater than 10%), including 622

(65 with minor allele frequency greater than 10%) in genes identified as essential in

S288c (methods in SI). Of the indels with minor allele frequency greater than 10%, 241

of 467 are predicted to cause frame shift mutations. Once again, we found an

enrichment for these mutations in the C-terminal 5% of the protein (Fig. 3C, inset),

consistent with the indels attaining high frequency in the population being those with

relatively mild effects. Nevertheless, we found that the proportion of frame-shift

mutations relative to in-frame indels decreases strongly as a function of minor allele

frequency, consistent with the action of purifying selection to remove these highly

deleterious mutations (Fig. 3D). For example, at minor allele frequency >15% there are

0.69 out-of-frame indels for every in-frame polymorphism, compared to 17.2 at minor

allele frequency <10%. Based on this, we estimate that 96% (1-0.69/17.2) of the 2750

out-of-frame indels segregating at minor allele frequency below 10% in the S.

cerevisiae population are deleterious and will be removed by natural selection.

Non-coding regions contain many functional sequences including regulatory sequences

and non-coding RNA genes. We computed the derived allele frequency spectrum for the

SNPs that we identified in non-coding regions (Fig. 3E). We found that there is strong

evidence for purifying selection in all classes of non-coding regions in yeast (compared

to synonymous sites), and that tRNA genes show a skew in their allele frequency

distribution comparable to amino acid altering mutations in protein coding regions.

One of the most exciting applications of population genomics is to systematically

identify mutations that may have been the targets of positive natural selection and hence

may underlie the adaptive differences between species. Comparisons of divergence to

diversity in different classes of sites, such as the McDonald-Kreitman test (M-K

)

provide a powerful means to do so, and are relatively robust to demographic

complexity

. As noted above, there was a large excess of amino-acid replacement

polymorphism at low allele frequencies (Fig. 3F inset). We therefore excluded SNPs

with minor allele frequency less than 20%

and performed directional Fisher's exact

tests on the 1105 genes for which there were at least 5 SNPs, to test the hypotheses that

there was an excess of fixed amino acid differences between S. cerevisiae and S.

paradoxus. Overall, we found a skew in the distribution of the M-K ratio towards values

less than one, indicating either pervasive purifying selection on the differences between

species, or reduction in the efficacy of selection within the current S. cerevisiae

population

. In the positive tail (M-K ratio>1) of this distribution lie genes enriched for

amino-acid changes between species, candidates for adaptive differences. However,

none of these were significant after a multiple testing correction.

Repeated sequence families: Ty elements, rDNA and CNVs

We estimated the overall level of Ty element abundance for each strain directly from

ABI sequencing reads, since a set of dispersed features like Ty elements is sampled

proportionally with light shotgun sequencing coverage and because large insertions like

Ty elements present a challenge for reference-based genome assembly. The proportion

of Ty sequences is typically less than 3% in all strains in both species, with the highest

abundance observed in the laboratory strain S288c (3.53%) (Fig. 4A). Globally, we find

that the proportion of Ty sequence per strain is higher among strains in S. cerevisiae

relative to S. paradoxus (Table S5; Wilcoxon Test, p = 0.02275). Interestingly, S.

paradoxus strains from South America (UFRJ50791 and UFRJ50816), which are

partially reproductively isolated from other S. paradoxus lineages and have been

previously known as a separate species (S. cariocanus)

15, 24

, have the highest Ty

abundance for this species. This correlates well with the increased rate of rearrangement

due to reciprocal translocations in the South America S. paradoxus lineage, and

supports the hypothesis that a burst of rearrangements occurred in this lineage due to

increased Ty activity

. We were able to map all 14 breakpoints of the large

translocations in the South American and three additional S. paradoxus strains most of

which involve Ty/LTRs (see Fig. S5). Levels of variation in Ty abundance are similar in

the Wine/European S. cerevisae population and in the UK S. paradoxus population, and

levels of variation in Ty abundance for the global S. cerevisae sample are substantially

higher than the Wine/European S. cerevisae population (Table S5). Finally, we find that

the mosaic strains of S. cerevisae have higher Ty abundance than the clean lineages

(Table S5; Wilcoxon Test, p=0.006896), suggesting that hybridization may have led to

an increase in Ty element activity in this species.

Sequence coverage was sufficient to ensure that each position in the 9.1kb rDNA repeat

was covered many times in each strain (average 140). A complete analysis of this

dataset is given elsewhere (O'Kelly et al, this issue) but we report here an interesting

correlation between intragenomic rDNA sequence variation and genome mosaicism

(Fig. 4B). By comparing the number of rDNA reads against total overall reads (Methods

in SI) we estimated rDNA copy number for each S. cerevisiae strain to range between

54 (K11) and 511 (YJM981). The estimates were generally in good agreement with

previous estimates

42, 43

. The exceptionally high count for strain YJM981 appears to be

linked to an unusual karyotype resulting in a much larger than average overall genome

size (Fig. S6). A link between rDNA copy number and genome size has been reported

previously

. We also investigated the possibility of a link between rDNA copy number

and genome mosaicism but found them to be unrelated. However, when polymorphic

sites in the rDNA repeats from individual strains were enumerated (Methods in SI),

mosaic genomes were found to have significantly more variable positions than the clean

lineages (p=1.3 x 10

-6

under Mann-Whitney U rank test). We further characterised this

variation in terms of number of substitutions per position and found the vast majority of

variant sites to be in the 0-10% minor allele frequency range (Fig. 4b). In contrast to

Ganley and Kobayashi

who found only four polymorphic sites in strain RM11-1A, we

identified 518 polymorphic sites of which 156 are resolved to a complete SNP (relative

to the S288c rDNA consensus sequence) in at least one strain. A possible reason for the

increased polymorphism within mosaic strains is that differences in the rDNA sequence

between the parents of the mosaic may not have yet been resolved by gene conversion

suggesting the mosaics are relatively recent.

This sequence survey provides the opportunity to address issues of CNV including

segmental duplications, as well as identification of sequences not found in reference

genomes. Table S2 shows the proportion of reads for each strain that were unplaced in

the assembly and hence potentially novel. Much of the unplaced material is likely to

originate from subtelomeric regions, which are repetitious and structurally variable and

hence hard to assemble. A genome-wide analysis of copy number based on the numbers

of reads of each strain aligning to each gene showed very little significant CNV outside

the rDNA region with the exception of three strains

. Because of the number of gene-

strain combinations, only copy numbers of 10 or more could be reliably detected, and

only six instances of this were found. The clearest such example, the CUP1 gene, is

discussed below.

Correlation of phenotypes to genotypes

Finding the genetic basis for natural as well as artificially generated variation in

phenotypes is a fundamental goal in biology, with implications for our understanding of

evolution, disease and biotechnological traits. We found surprisingly little correlation

between the phylogeny and the source environment (Fig. 1) among S. cerevisiae

isolates. The baking strains loosely cluster together globally but fall in different regions

of the tree when individual chromosomes and segments are analysed (not shown). Of

the six clinical isolates, three are well inside the European cluster while the other three

are well separated on individual long branches. Similarly, many of the wine and food

spoilage strains are on individual long branches. In order to provide a phenotypic map

of the sequenced strains, the entire set was subjected to high throughput phenotypic

analysis under a multitude of conditions (Fig. 5A). Growth curves were exhaustively

sampled (>250 time points) over several days and the physiologically relevant growth

variables growth lag, growth rate (slope) and growth efficiency (maximum density)

were extracted

to provide roughly 200 distinct phenotypic traits.

Overall there is enough variation in phenotype to allow clustering of strains into

phenotypic clusters. Remarkably, there is a high qualitative overlap between the

phenotypic clustering and the phylogenies based on SNPs (Fig. 5A and Fig. 1). Also on

a quantitative level the correlation between genotypic and phenotypic similarity within

S. cerevisiae is surprisingly good (linear correlation r = 0.37, sequence similarity vs.

Pearson correlation coefficients from phenotyping), given that conventional phenotypic

taxonomy generally fails even to resolve the Saccharomyces sensu stricto species. Here,

the S. paradoxus strains were well separated from the S. cerevisiae strains. The only

exception was the Hawaiian isolate, which was discussed above. In addition, the S.

cerevisiae isolates fell into two groups. The first contains most of the Wine/European

lineage and most of the long-branch recombinants, while the second consists of the

other lineages. The main phenotypic characteristic separating these two groups appears

to be rapid growth (short lag and steep slope in rate) for the Wine/European and

mosaics, which could be advantageous for the fermentation processes many of these

strains are used for.

Some specific phenotypes were consistent with known genotypes or sequences

determined here. For example, S288c showed only limited growth on galactose due to a

known gal2 mutation, while the reference strain BY4741 is Gal

as this mutation was

corrected (all phenotypic data is displayed in relation to BY4741). Deleterious

mutations in GAL genes, notably a frameshift in the GAL2 gene in Y55 and

DBVPG6044, and a premature stop-codon in the GAL3 gene in Y55, DBVPG6044, and

NCYC110 (West African lineage), are likely to cause the severe galactose growth

defects observed in these strains (Fig. 5B, C). Complete absence of galactose growth

was only observed for S. cerevisiae strains 273614N and YS2 which have no GAL

genes despite good sequence coverage. Parallel inactivation of GAL genes has earlier

been reported as a means of ecological diversification

Another example of a stringent genotype - phenotype link is the ability to utilise

melibiose. Unperturbed growth on melibiose was only observed in five S. cerevisiae

strains: the three Malaysian, the Hawaian and the West African strain NCYC110, which

contain either one or two copies of the melibiose gene MEL1. Strains with two MEL1

copies tend to grow better on melibiose (Fig. 5D). The only exception to the strict link

between melibiose utilisation and MEL1 is DBVPG6044, which contains two MEL1

genes with no non-synonymous mutations but exhibited limited growth on melibiose.

However, expression of MEL1 is governed by the GAL regulatory network

suggesting that the above mentioned galactose negative phenotype and prevalent

mutations in Gal regulatory proteins may explain its melibiose phenotype. S. cerevisiae

isolates displayed clear phenotypes in relation to copper. Phenotypes can also be

correlated with copy number (Fig. 5E). The gene dosage of CUP1, which in S.

cerevisiae varies between 0 and up to 40 for DBVPG1373, is directly proportional to

copper resistance. There is a clear distinction between the two groups of S. cerevisiae

strains. The Wine/European and long branched mosaics strains tended to show higher

copper resistance. This may be partially explained by the use of copper as a fungicide or

copper containers in breweries, imposing a selective pressure for copper resistance and

thus for CUP1 copy number increase. The wild S. paradoxus strains contain only one

CUP1copy and, with the exception of the Far Eastern strains, show high sensitivity to

copper (Fig. S7). This could explain the lack of S. paradoxus in fermentation or the lack

of use has not imposed selective pressure for increase copy number.

Discussion

We have presented here the first full-sequence population genomic survey of two

closely related species, one associated with human activity and one a natural species

with no human association. Both are globally distributed and can be found in similar

niches, even from the same bark sample of an oak tree

10, 11

. Despite this ecological

similarity, we find extensive differences in the population genomic history of these two

species, mirrored by extensive phenotypic variation. This study offers an unprecedented

view of sequence diversity across the genome within closely related species, lays the

background for functional phenotypic analysis, and gives us further insight into the

processes underlying sequence and genome evolution in this important model organism.

With these powerful genomic resources, we can finally address the issue of what

selective influence human activity has had on baker's yeast, if any, and start to

determine the underlying genetics of important traits (for industrial purposes or for

biological studies). Is there evidence of domestication in S. cerevisiae as previously

debated

19, 50

? One could interpret the results of this analysis in two ways. One that there

was a domestication of at least one group, the Wine/European strains, or perhaps two

with the sake strains, with selection for improved fermentation properties. This

domesticated group then gave rise to feral and clinical derivatives as well as being

involved in the generation of outbred progeny found in all sources. Alternatively,

human activity simply utilised existing strains from populations that have appropriate

fermentation properties, and that all human activity had provided was the opportunity to

outbreed through movement of strains. Using domestication to imply "species bred in

captivity"

the strains that best fulfil this definition are the baking isolates as they have

clearly arise from crosses between lineages, however further studies are necessary.

Comparison to extant S. paradoxus should provide a picture of how S. cerevisiae looked

like prior human activity. Any lineages that were selected from captively bred strains

would be expected to have lower diversity than other lineages/populations not selected.

This is not the case for the Wine/European or sake lineages, which have similar or

greater levels of diversity compared to the other clean lineages or to S. paradoxus

populations. This view of human activity simply moving yeast strains around without

captive breeding is consistent with the results and interpretation of analysis of over 600

strains

In addition to providing a more complete picture of what the genome space of

the species S. cerevisiae actually looks like, our results raise a large number of questions

about genome evolution in general. For example, we have found a great deal of

variation within protein coding genes, even essential genes, raising questions about the

evolutionary mechanisms that maintain sequence variation in this species. Our strategy

to shotgun sequence genomes from many isolates allows us to determine additional

sequences as well as rearrangements, raising new questions that could not be found

using array CGH. Many of these rearrangements are likely to be subtelomeric, and may

explain adaptation to different environments including industrial fermentation

properties. The assembly and analysis of these regions will be important for

understanding genome evolution in these organisms. Tandem arrays of rDNAs can

contain many SNPs within the array that may be useful in developing an understanding

of tandem array dynamics and evolution in addition to their use in strain typing. Ty

distributions and copy numbers correlate with genome rearrangements over

evolutionary time scales, the analysis of which may add to our understanding of the

forces governing genome evolution.

A major use of the sequences will be to support genetic analysis between strains,

which given the high recombination rate in yeast will allow rapid fine mapping of the

genetic determinants of many phenotypic differences between the strains. One approach

that has already been taken is QTL analysis of crosses and backcrosses between pairs of

founder strains

53, 54

. The near-complete genome sequences and phenotypes identified

here provide a rich source of material for such studies. In the future, the recombinant

nature of the mosaic S. cerevisiae lineages, together with the variation identified here,

present an unprecedented opportunity to map to traits of phenotypic importance using

naturally occurring recombinant inbred lines by genome wide association studies. The

processes leading to these observations and their functional and evolutionary

References

Goffeau, A. et al. Life with 6000 genes. Science 274, 546, 563-7 (1996).

Mewes, H. W. et al. Overview of the yeast genome. Nature 387, 7-65 (1997).

Cliften, P. et al. Finding functional features in Saccharomyces genomes by

phylogenetic footprinting. Science 301, 71-6 (2003).

Kellis, M., Patterson, N., Endrizzi, M., Birren, B. & Lander, E. S. Sequencing

and comparison of yeast species to identify genes and regulatory elements. Nature 423,

241-54 (2003).

Begun, D. J. et al. Population genomics: whole-genome analysis of

polymorphism and divergence in Drosophila simulans. PLoS Biol 5, e310 (2007).

Clark, A. G. et al. Evolution of genes and genomes on the Drosophila

phylogeny. Nature 450, 203-18 (2007).

Dietrich, F. S. et al. The Ashbya gossypii genome as a tool for mapping the

ancient Saccharomyces cerevisiae genome. Science 304, 304-7 (2004).

Dujon, B. et al. Genome evolution in yeasts. Nature 430, 35-44 (2004).

McCarroll, S. A. & Altshuler, D. M. Copy-number variation and association

studies of human disease. Nat Genet 39, S37-42 (2007).

10.

Sampaio, J. P. & Goncalves, P. Natural populations of Saccharomyces

kudriavzevii in Portugal are associated with oak bark and are sympatric with S.

cerevisiae and S. paradoxus. Appl Environ Microbiol 74, 2144-52 (2008).

11.

Sniegowski, P. D., Dombrowski, P. G. & Fingerman, E. Saccharomyces

cerevisiae and Saccharomyces paradoxus coexist in a natural woodland site in North

America and display different levels of reproductive isolation from European

conspecifics. FEMS Yeast Res 1, 299-306 (2002).

12.

Aa, E., Townsend, J. P., Adams, R. I., Nielsen, K. M. & Taylor, J. W.

Population structure and gene evolution in Saccharomyces cerevisiae. FEMS Yeast Res

6, 702-15 (2006).

13.

Koufopanou, V., Hughes, J., Bell, G. & Burt, A. The spatial scale of genetic

differentiation in a model organism: the wild yeast Saccharomyces paradoxus. Philos

Trans R Soc Lond B Biol Sci (2006).

14.

Kuehne, H. A., Murphy, H. A., Francis, C. A. & Sniegowski, P. D. Allopatric

divergence, secondary contact, and genetic isolation in wild yeast populations. Curr Biol

17, 407-11 (2007).

15.

Liti, G., Barton, D. B. & Louis, E. J. Sequence diversity, reproductive isolation

and species concepts in Saccharomyces. Genetics 174, 839-50 (2006).

16.

Ruderfer, D. M., Pratt, S. C., Seidel, H. S. & Kruglyak, L. Population genomic

analysis of outcrossing and recombination in yeast. Nat Genet 38, 1077-81 (2006).

17.

Tsai, I. J., Bensasson, D., Burt, A. & Koufopanou, V. Population genomics of

the wild yeast Saccharomyces paradoxus: Quantifying the life cycle. Proc Natl Acad Sci

U S A 105, 4957-62 (2008).

18.

Pretorius, I. S. Tailoring wine yeast for the new millennium: novel approaches to

the ancient art of winemaking. Yeast 16, 675-729 (2000).

19.

Fay, J. C. & Benavides, J. A. Evidence for domesticated and wild populations of

Saccharomyces cerevisiae. PLoS Genet 1, 66-71 (2005).

20.

Bell, P. J., Higgins, V. J. & Attfield, P. V. Comparison of fermentative

capacities of industrial baking and wild-type yeasts of the species Saccharomyces

cerevisiae in different sugar media. Lett Appl Microbiol 32, 224-9 (2001).

21.

McCusker, J. H., Clemons, K. V., Stevens, D. A. & Davis, R. W. Genetic

characterization of pathogenic Saccharomyces cerevisiae isolates. Genetics 136, 1261-9

(1994).

22.

Liti, G., Peruffo, A., James, S. A., Roberts, I. N. & Louis, E. J. Inferences of

evolutionary relationships from a population survey of LTR-retrotransposons and

telomeric-associated sequences in the Saccharomyces sensu stricto complex. Yeast 22,

177-92 (2005).

23.

Pope, G. Saccharomyces Genome Resequencing Project Strains. (2008).

http://www.ncyc.co.uk/sgrp.php

24.

Naumov, G. I., James, S. A., Naumova, E. S., Louis, E. J. & Roberts, I. N. Three

new species in the Saccharomyces sensu stricto complex: Saccharomyces cariocanus,

Saccharomyces kudriavzevii and Saccharomyces mikatae. Int J Syst Evol Microbiol 50

Pt 5, 1931-42 (2000).

25.

Carter, D. M. Saccharomyces Genome Resequencing Project. (2005).

http://www.sanger.ac.uk/Teams/Team71/durbin/sgrp/index.shtml

26.

The Sanger Institute/EMBL. Ensembl Trace Server. http://trace.ensembl.org/

27.

Liti, G. & Louis, E. J. Yeast evolution and comparative genomics. Annu Rev

Microbiol 59, 135-53 (2005).

28.

Mullikin, J. C. & Ning, Z. The phusion assembler. Genome Res 13, 81-90

(2003).

29.

Hong, E. L. et al. Gene Ontology annotations at SGD: new data sources and

annotation methods. Nucleic Acids Res 36, D577-81 (2008).

30.

Pritchard, J. K., Stephens, M. & Donnelly, P. Inference of population structure

using multilocus genotype data. Genetics 155, 945-59 (2000).

31.

Wei, W. et al. Genome sequencing and comparative analysis of Saccharomyces

cerevisiae strain YJM789. Proc Natl Acad Sci U S A 104, 12825-30 (2007).

32.

Bensasson, D., Zarowiecki, M., Burt, A. & Koufopanou, V. Rapid evolution of

yeast centromeres in the absence of drive. Genetics 178, 2161-7 (2008).

33.

Tajima, F. Statistical method for testing the neutral mutation hypothesis by DNA

polymorphism. Genetics 123, 585-95 (1989).

34.

Coghlan, A. & Wolfe, K. H. Relationship of codon bias to mRNA concentration

and protein length in Saccharomyces cerevisiae. Yeast 16, 1131-45 (2000).

35.

Sharp, P. M. & Li, W. H. The codon Adaptation Index--a measure of directional

synonymous codon usage bias, and its potential applications. Nucleic Acids Res 15,

1281-95 (1987).

36.

Bulmer, M. The selection-mutation-drift theory of synonymous codon usage.

Genetics 129, 897-907 (1991).

37.

Clark, R. M. et al. Common sequence polymorphisms shaping genetic diversity

in Arabidopsis thaliana. Science 317, 338-42 (2007).

38.

McDonald, J. H. & Kreitman, M. Adaptive protein evolution at the Adh locus in

Drosophila. Nature 351, 652-4 (1991).

39.

Fay, J. C. & Wu, C. I. Sequence divergence, functional constraint, and selection

in protein evolution. Annu Rev Genomics Hum Genet 4, 213-35 (2003).

40.

Fay, J. C., Wyckoff, G. J. & Wu, C. I. Positive and negative selection on the

human genome. Genetics 158, 1227-34 (2001).

41.

Fischer, G., James, S. A., Roberts, I. N., Oliver, S. G. & Louis, E. J.

Chromosomal evolution in Saccharomyces. Nature 405, 451-4 (2000).

42.

Kobayashi, T., Heck, D. J., Nomura, M. & Horiuchi, T. Expansion and

contraction of ribosomal DNA repeats in Saccharomyces cerevisiae: requirement of

replication fork blocking (Fob1) protein and the role of RNA polymerase I. Genes Dev

12, 3821-30 (1998).

43.

Petes, T. D. Yeast ribosomal DNA genes are located on chromosome XII. Proc

Natl Acad Sci U S A 76, 410-4 (1979).

44.

Prokopowich, C. D., Gregory, T. R. & Crease, T. J. The correlation between

rDNA copy number and genome size in eukaryotes. Genome 46, 48-50 (2003).

45.

Ganley, A. R. & Kobayashi, T. Highly efficient concerted evolution in the

ribosomal DNA repeats: total rDNA repeat variation revealed by whole-genome

shotgun sequence data. Genome Res 17, 184-91 (2007).

46.

Ohta, T. Some models of gene conversion for treating the evolution of multigene

families. Genetics 106, 517-28 (1984).

47.

Warringer, J. & Blomberg, A. Automated screening in environmental arrays

allows analysis of quantitative phenotypic profiles in Saccharomyces cerevisiae. Yeast

20, 53-67 (2003).

48.

Hittinger, C. T., Rokas, A. & Carroll, S. B. Parallel inactivation of multiple

GAL pathway genes and ecological diversification in yeasts. Proc Natl Acad Sci U S A

101, 14144-9 (2004).

49.

Johnston, M. A model fungal gene regulatory mechanism: the GAL genes of

Saccharomyces cerevisiae. Microbiol Rev 51, 458-76 (1987).

50.

Vaughan-Martini, A. & Martini, A. Facts, myths and legends on the prime

industrial microorganism. J Ind Microbiol 14, 514-22 (1995).

51.

Diamond, J. Evolution, consequences and future of plant and animal

domestication. Nature 418, 700-7 (2002).

52.

Legras, J. L., Merdinoglu, D., Cornuet, J. M. & Karst, F. Bread, beer and wine:

Saccharomyces cerevisiae diversity reflects human history. Mol Ecol 16, 2091-102

(2007).

53.

Brem, R. B., Yvert, G., Clinton, R. & Kruglyak, L. Genetic dissection of

transcriptional regulation in budding yeast. Science 296, 752-5 (2002).

54.

Steinmetz, L. M. et al. Dissecting the architecture of a quantitative trait locus in

yeast. Nature 416, 326-30 (2002).

55.

Kim, J. M., Vanguri, S., Boeke, J. D., Gabriel, A. & Voytas, D. F. Transposable

elements and genome organization: a comprehensive survey of retrotransposons

revealed by the complete Saccharomyces cerevisiae genome sequence. Genome Res 8,

464-78 (1998).

56.

Fingerman, E. G., Dombrowski, P. G., Francis, C. A. & Sniegowski, P. D.

Distribution and sequence analysis of a novel Ty3-like element in natural

Saccharomyces paradoxus isolates. Yeast 20, 761-70 (2003).

`Supplementary Information

Acknowledgements. We thank all the members of the Sanger sequencing production teams for

generating the sequence data. We thank members of the Durbin and Louis laboratories for comments and

suggestions and to L. Kruglyak and J. Schacherer for sharing their unpublished manuscript. We also

thank the British Council and Chinese Academy of Sciences for providing the opportunity to conceive

and develop this project. Research at The Sanger Institute (D.M.C., A.M.M., L.P, M.J., M.A.Q., I.G.,

S.S., F.S. and R.D.) is supported by The Wellcome Trust. G.L., D.B.H.B., E.B. and E.J.L. were supported

by The Wellcome Trust and the BBSRC. S.A.J., R.P.D. and I.N.R. were supported by the BBSRC. A.B.

and J.W. were supported by the Swedish Research Council and the Swedish Foundation for Strategic

is available.

Research. A.B., V.K. were supported by NERC and I.J.T. by The Wellcome Trust. D.B. was supported by

NERC. M.J.T.O. and A.V. were supported by NSF, NIH and a Hertz fellowship.

Author Contributions R.D. and E.J.L. conceived and designed the project. G.L. selected and

manipulated yeast strains and extracted DNA samples. M.J., M.A.Q., I.G., S.S., F.S. performed the

subcloning and sequencing. D.M.C. did the reference comparison and assembly of the sequences. D.M.C.

and G.L. coordinated the collection of data. D.M.C. and R.D. performed much of the global analysis,

which was the basis for specific analyses performed by the rest. A.M.M. did the selection studies. E.J.L.,

G.L. D.M.C., L.B. did the population structure analysis. C.M.B. and D.B. performed the analysis of Ty

elements abundance. S.A.J., R.P.D., M.J.T.O., A.V. and I.N.R. analysed the rDNA. A.B., V.K. and I.J.T.

did the recombination analysis. J.W. and A.B. generated the phenomics data.. E.J.L. with G.L. wrote the

paper, coordinating everyone's contributions.

Author Information Correspondence and requests for materials should be addressed to R. D.

(

rd@sanger.ac.uk

) or E. J. L. (

ed.louis@nottingham.ac.uk

Fig.1. Saccharomyces phylogenomics

NJ trees based on SNP differences of a, S. cerevisiae and S. paradoxus strains

sequenced in this project using S. mikatae, S. kudriavzevii and S. bayanus as

outgroups; b, Close-up of the European S. paradoxus, with UK isolates

highlighted in violet; c, S. cerevisiae strains with clean lineages highlighted in

grey, colour name refers to source and colour dots refer to geographic origin.

Fig. 2. Saccharomyces population structure

a, Inference of population structure using STRUCTURE on S. paradoxus,

(markers: 7544 SNPs with >30 strains passing neighbourhood quality standard,

NQS), assuming K=6 subpopulations and correlated allele frequencies, linkage

model based on marker distances in basepairs, 15000 iteration burn in, and

5000 iterations of sampling. Each mark on the x axis represents one strain, and

the blocks of colour represent the fraction of the genetic material in each strain

assigned to each cluster.

b, As a, but for S. cerevisiae (markers: 3413 SNPs with >30 strains passing

NQS)

c, Changing topology of the NJ trees along chromosome VIII. The clean

lineages exhibit the same NJ topology across the genome whereas mosaic

strains exhibit different topologies for different segments. For example in strain

UWOPS83.787.3 (green) the leftmost 80kb of chromosome VIII groups with the

Wine/European cluster. In the interval from 240-320 kb the strain groups with

the North American strains. For the same intervals the lab strain S288c (violet)

is in a long branch followed by the Wine/European cluster.

Fig. 3. Population genomics: variation and selection

a, Linkage disequilibrium between pairs of sites as a function of the distance

between them. Each point is the average for a 1kb bin. Insets show the decline

in linkage disequilibrium over the first 10kb. Points in the S. cerevisiae

Wine/European plot are more scattered due to the smaller number of strains

analysed. Numbers of strains and sites shown in Table S4.

b, Derived allele frequencies in single nucleotide polymorphisms (SNPs) in

coding regions. Amino acid changing SNPs (blue bars, `a') show an excess of

low frequencies compared to synonymous SNPs (unfilled bars, `s').

Synonymous SNPs in genes with strong codon bias (`s*', defined as CAI>0.6)

show an excess of SNPs at both low and high frequency, suggesting the action

of both positive and negative selection. SNPs that create stop codons (red bars,

`create stop') show a very strong skew to low frequencies, suggesting that they

are on average more deleterious than amino acid changing polymorphisms.

Inset is the number of mutations occurring over the length of the protein, which

exceeds three standard deviations (dotted trace) from the mean (solid trace) in

the extreme C-terminus (95% of length) indicating that stop codons that remain

in the population tend to be those causing relatively mild effects on protein

function.

c, Distribution of sizes of insertion/deletion polymorphisms (indels) in coding

regions. High frequency indels (minor allele frequency greater than 10%, red

bars) show a greater tendency to occur in multiples of 3 than low frequency

indels (grey bars), indicating selection against indels that disrupt the open

reading frame. Inset is the number of mutations occurring over the length of the

protein, which exceeds three standard deviations (dotted trace) from the mean

(solid trace) in the extreme C-terminus (95% of length) indicating that

segregating high frequency indels tend to be those causing relatively mild

effects on protein function.

d, Minor allele frequency distribution of indels in coding regions. Out of frame

indels (not multiples of 3, grey bars) show great excess at low frequencies when

compared to in frame indels (unfilled bars). The proportion of out of frame indels

(solid trace) decreases as minor allele frequency increases indicating the action

of purifying selection to remove most out of frame indels. Error bars represent

the standard error of the proportion.

e, Derived allele frequencies in single nucleotide polymorphisms (SNPs) in non-

coding regions compared to those in synonymous sites (unfilled symbols). All

categories of non-coding SNPs show an excess of low frequency alleles

indicating the action of purifying selection, with SNPs in tRNAs (red bars)

showing the strongest effect.

f, Distribution of McDonald-Kreitman (M-K) ratios for yeast genes. Dotted trace

indicates the expectation in the absence of selection or changes in effective

population size. Inset is the ration of amino acid changing to synonymous

polymorphisms as a function of allele frequency. Because of the excess of

amino acid polymorphism at low frequency, only SNPs with minor allele

frequency >20% were used for M-K tests.

Fig. 4. Transposon abundance and rDNA variability

a, Abundance of Ty transposable element sequences across Saccharomyces

strains. The proportion of Ty sequences in ABI shotgun sequencing reads

identified by RepeatMasker is shown for each strain of S. cerevisae (red) and S.

paradoxus (blue). Population are defined as in Fig. 1 and abbreviated as

follows: WE - Wine/European, CL - Clean Lineage, MO - Mosaic, UK - United

Kingdom, RU - Russia, FE - Far East, DK - Denmark, HA - Hawaii, NA - North

America, SA - South America. The 6 strains identified as potential clonemates

in Table S5 are also excluded from this analysis. Comparison of overall TE

abundance in shotgun reads (3.53%) from S. cerevisiae strain S288c with the

reference genome sequence from the same strain (3.35%) reveals that

estimates of overall TE content from reads are similar to finished genome

assemblies. Our estimate of Ty abundance for the finished S288c assembly is

slightly higher than previous estimates (3.1%)

since we used an expanded

RepeatMasker library with newly reported Saccharomyces Ty variants

b, Ribosomal DNA polymorphism associated with genome mosaicism. Variable

nucleotide positions (substitutions only) found in the 9.1 kb rDNA repeat are

mapped for each S. cerevisiae strain. The proportion of sequencing reads

showing base substitution at a given variable position is indicated by the colour

of the bar. Fully resolved polymorphisms (i.e. SNPs) are excluded. Low

frequency polymorphisms, indicated by red bars, predominate, with increasing

within-strain variation particularly evident in the non-coding IGS1 and IGS2

regions. Strain names are given on the left, with structured genome strains

shown in blue and mosaic strains in red. Ribosomal DNA copy number estimate

(in green) and total number of polymorphic positions (in black) are shown for

each strain on the right. Strains are sorted top to bottom by increasing number

of polymorphic positions. Strains with mosaic genomes possess significantly

greater numbers of polymorphisms. The average for mosaic strains was 98

polymorphic sites compared to the structured strain average of 34 sites (p=1.3 x

-6

under Mann-Whitney U rank test). L-1528 was excluded from this analysis

due to a high number of unaligned reads resulting in unreliable data.

Fig. 5. Saccharomyces phenotype variation

a, Phenotype variation among sequenced S. cerevisiae and S. paradoxus

strains. Strain growth phenotypes in 67 environments were quantified using

high resolution micro-cultivation, automated measurements of population

density (OD) and calculation of strain doubling times, lags and maximum

densities. See SI for details. A) Strain (n=2) doubling time phenotypes in

relation to the S288c derivative BY4741 (n=20) are displayed (Logarithmic

strain coefficients, LSC = LN (strain/BY4741)). Green = poor growth, red = good

growth. Hierarchical clustering of strain doubling time, lag and maximum density

phenotypes was performed using a centered pearson correlation metric and

average linkage mapping. Blue lines = S. paradoxus, pink lines = S. cerevisiae,

grey line = S. bayanus isolate CBS7001. The complete set of phenotypes,

including lag and maximum density data, is displayed in Fig. S7.

b-c, Example of large effect polymorphisms (stop codons and indels) that

correlated with a phenotype. b, shows 15 residues window around SNPs that

create a stop codon or out of frame deletion (red boxes) in Gal2 and Gal3

respectively. c, strains harbouring large effect polymorphisms in genes in the

gal pathway tend to show slow growth when galactose is the only carbon

source. Red = West African lineage, Blue = reference BY4741.

d-e, Linking gene copy number to growth phenotypes. Best estimates of gene

copy number were derived by dividing the number of BLAST matches of

Table 1. Origin of strains and sequence coverage

Source or location

Strains

ABI

AGI

S. cerevisiae

43.3

37.3

Fermentation

11.4

Clinical

5.6

Wild

10.0

Laboratory

10.6

37.3

Baking

2.6

Unknown

3.1

S. paradoxus

38.5

71.8

England

10.7

71.8

Continental Europe/Siberia

12.6

Far East Russia/Japan

6.8

North & South America

6.8

Hawaii

1.6

Geographic origin of Saccharomyces strains and more detailed information are given in

Supplementary table S1. Number of

ABI and

Illumina GA (Solexa) nucleotides successfully

aligned and divided by the genome size estimate to give mean coverage depth.

Y55

SK1

DBVPG6044

NCYC110

UWOPS

05.217.3

UWOPS

05.227.2

UWOPS

03.461.4

UWOPS

83.787.3

K11

Y12

YPS128

YPS606

UWOPS87.2421

NCYC361

DBVPG6040

378604X

W303

SGD

S288c

YJM789

YS4

YS2

YS9

273614N

DBVPG1853

YIIc17_E5

322134S

BC187

0.001

Europe

Americas

Asia

Oceania

Africa

Unknown

Laboratory

Wild

Fermentation

Clinical

Baking

Unknown

Z1.1

Q62.5

Q31.4

Q95.3

S36.7

Q69.8

Q32.3

Q74.4

Y8.5

W7 Y9.6

Y6.5

T21.4

Y8.1

Q59.1

Q89.8

NRRLY

DBVPG4650

N-17

KPN3829

KPN3828

CBS432

CBS5829

0.0001

Siberia

Russia

Italy

Denmark

Silwood Park

Windsor Park

Wine/European

North American

Sake

Malaysian

West African

S. cerevisiae

S. bayanus

S. mikatae

S. paradoxus

0.02

European

Far Eastern

American

Hawaiian

Figure 1

RM11_1A

YJM981

YJM978

YJM975

DBVPG1373

DBVPG1106

DBVPG6765

DBVPG1788

L-1374

L-1528

S. kudriavzevii

S. paradoxus

100

80
70
60
50

40
30
20
10

S. cerevisiae

European Far Eastern American Hw

100

90
80

70
60
50
40

30
20
10

Malaysian Sake NA WA Mosaics Wine/European

Figure 2

240-320kbp

K11

Y12

YPS128

YPS606

UWOPS83.787.3

NCYC110

DBVPG6044

BC187

UWOPS05.217.3

UWOPS05.227.2

UWOPS03.461.4

A=YJM978

B=YJM981

C=YJM975

D=DBVPG1373

E=L-1374

F=DBVPG1788

G=L-1528

H=DBVPG6765

I=RM11-1A

J=DBVPG1106

S288c

0-80kbp

VIII

UWOPS05.217.3

UWOPS05.227.2

UWOPS03.461.4

K11

Y12

YPS128 YPS606

S288c

NCYC110

DBVPG6044

250

200

mber)

0.4

0.6

0.8

a/s

rati

100

150

enes

(nu

0.2

0.4

0.6

0.8

Fixed

Derived allele frequency

M K ratio

M-K ratio

0.5

0.4

0.45

0 25

0.3

0.35

s
intron
upstream

NPs

0 15

0.2

0.25

upstream
downstream
tRNA

ction

0.05

0.1

0.15

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Deri ed Allele Freq enc

Derived Allele Frequency

0.2

0.4

0.6

0.8

1.0

200

400

600

800

1000

0.2

0.4

0.6

0.8

1.0

0.2

0.4

0.6

0.8

1.0

0.2

0.4

0.6

0.8

1.0

8 10

0.2

0.4

0.6

0.8

1.0

0.2

0.4

0.6

0.8

1.0

Linkage disequilibrium (r

)

Physical distance (kb)

S. paradoxus

S. cerevsiae

Wine/European

S. cerevisiae

global

0 7

0.8

0.9

MAF<0.1

MAF>0.1

f

mutatio

! + 3

0.5

0.6

0.7

Numbe

SNPs

0.2

0.3

0.4

% of protein length

raction

0.1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Size of indel (basepairs)

0.8

0.9

out of frame

in frame

0 5

0.6

0.7

dels

proportion of
indels out of frame

0.3

0.4

0.5

ction

i
n

0.1

0.2

0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45

Minor Allele Frequency

Figure 3

0.9

0.7

0.8

f

mutatio

! + 3!

0.5

0.6

a
s

SNPs

Number

o 10

! 3!

0.3

0.4

s*
create stop

raction

% of protein length

0.1

0.2

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Derived Allele Frequency