Deciphering the Hybridisation History Leading to the Lager Lineage Based on the Mosaic Genomes of Saccharomyces bayanus Strains NBRC1948 and CBS380T

Saccharomyces bayanus is a yeast species described as one of the two parents of the hybrid brewing yeast S. pastorianus. Strains CBS380T and NBRC1948 have been retained successively as pure-line representatives of S. bayanus. In the present study, sequence analyses confirmed and upgraded our previous finding: S. bayanus type strain CBS380T harbours a mosaic genome. The genome of strain NBRC1948 was also revealed to be mosaic. Both genomes were characterized by amplification and sequencing of different markers, including genes involved in maltotriose utilization or genes detected by array-CGH mapping. Sequence comparisons with public Saccharomyces spp. nucleotide sequences revealed that the CBS380T and NBRC1948 genomes are composed of: a predominant non-cerevisiae genetic background belonging to S. uvarum, a second unidentified species provisionally named S. lagerae, and several introgressed S. cerevisiae fragments. The largest cerevisiae-introgressed DNA common to both genomes totals 70kb in length and is distributed in three contigs, cA, cB and cC. These vary in terms of length and presence of MAL31 or MTY1 (maltotriose-transporter gene). In NBRC1948, two additional cerevisiae-contigs, cD and cE, totaling 12kb in length, as well as several smaller cerevisiae fragments were identified. All of these contigs were partially detected in the genomes of S. pastorianus lager strains CBS1503 (S. monacensis) and CBS1513 (S. carlsbergensis) explaining the noticeable common ability of S. bayanus and S. pastorianus to metabolize maltotriose. NBRC1948 was shown to be inter-fertile with S. uvarum CBS7001. The cross involving these two strains produced F1 segregants resembling the strains CBS380T or NRRLY-1551. This demonstrates that these S. bayanus strains were the offspring of a cross between S. uvarum and a strain similar to NBRC1948. Phylogenies established with selected cerevisiae and non-cerevisiae genes allowed us to decipher the complex hybridisation events linking S. lagerae/S. uvarum/S. cerevisiae with their hybrid species, S. bayanus/pastorianus.


Introduction
Beer and wine fermentations were acquired from the Middle East by Germanic and Celtic tribes and represent some of the most ancient fermentation technologies [1]. During the development of this technology, strains have been progressively selected for enhanced abilities according to standards for quality and production. Strains specific to beer fermentation have been isolated and studied for over one century and their diversity has led to the delineation of several species. During the 19th century ancient beers made with a mixture of yeasts were a source of yeast strains and species. In 1883, Emile Christian Hansen isolated the strain CBS1171 from beer. This strain was designated as the neo type of Saccharomyces cerevisiae previously described by Reess in 1870. The genus name Saccharomyces was first proposed by Meyen in 1838 [2]. Thereafter, many other species belonging to the genus Saccharomyces were isolated from beer including: S. bayanus (Saccardo, 1895) by Will in 1891, and S. carlsbergensis and S. monacensis by EC Hansen in 1908. Since 1888, progressively modern beers were made with pure cultures following the recommendation of EC. Hansen. Later, other sources have replaced beer as a reservoir of yeasts. For example S. uvarum (Beijerinck, 1898) was isolated from blackberry juice and S. paradoxus (Batschinskaia, 1914) was isolated from oak exudates. Forty one Saccharomyces species isolated between 1883 and 1965 were successively reclassified by many authors and were admitted to the genus Saccharomyces by van der Walt in 1970 [3]. In this new classification, S. carlsbergensis and S. monacensis described by EC Hansen, were reduced to synonyms of S. uvarum (Beijerinck).
However in 1985 [4], the Saccharomyces group known as Saccharomyces sensu stricto was restricted by DNA-DNA reassociation to four species: S. cerevisiae (neo type strain CBS1171 NT ), S. bayanus (CBS380 T ), S. paradoxus (CBS432 NT ), and S. pastorianus (CBS1538 NT ). These experiments also revealed that S. pastorianus, which includes two synonyms, S. carlsbergensis and S. monacensis, was a hybrid species with S. cerevisiae and S. bayanus as parents [4]. At the same time, S. uvarum (CBS395 T ) was reduced to a synonym of S. bayanus. Successive studies on other lager strains revealed the complexity of strains classified under the S. pastorianus and S. bayanus species. Some controversial issues appeared such as the hybrid nature of CBS380 T and its status of type strain. On the one hand, our previous studies revealed that S. bayanus CBS380 T was itself a chimer bearing Y' and SUC4 sequences identical to S. cerevisiae sequences localized on three chromosomes [5]. This hybrid nature has been identified in other strains such as: CBS378, CLIB271, and NRRLY-1551. The latter strain is the most similar to CBS380 T and has been confused with CBS 1538 NT , the current neo type strain of S. pastorianus. On the other hand, S. uvarum strains (classified as S. bayanus) examined in the same study lack S. cerevisiae SUC4 and Y' sequences [6].
Three sequencing projects of the genomes from S. uvarum representative strains have been developed, by Broad Institute (http://www.broadinstitute.org/, [7]) using CBS7001 ( = MCYC-623), by Génolevures (http://www.genolevures.org/, [8]) and the Washington University School of Medicine (http://genomeold. wustl.edu/projects/yeast/index.php, [9]) using the spore clone of CBS7001, strain 623-6c ura3-1 ( = CLIB533, CBS9787). The genome of S. uvarum CBS7001 has been recently upgraded by the Saccharomyces Sensu Stricto consortium (www.SaccharomycesSensuStricto.org) [10]. Sequence annotations confirmed that the S. uvarum genome is exempt of S. cerevisiae sequences. In 2005, based on sequence data, S. uvarum was proposed to be reinstated as a distinct species, thereby abolishing its synonym status with S. bayanus [5]. However, S. uvarum genomes are still listed as those of S. bayanus in the SGD, Génolevures, and NCBI databases to accommodate its synonymy status. For the same reason, sequences identical to those of S. uvarum have been labelled in databases as S. bayanus sequences because they were obtained from synonym strains of S. bayanus. Very recently, some authors have reminded that those genomes are S. uvarum and not S. bayanus as classified by the sequencing groups [11].
The taxonomic reinstatement of S. uvarum (Beijerinck) [5] as a real species has been admitted by some investigators [12,13], but is still contested by others [14,15] who argued that the presence of the subtelomeric Y' sequence is not necessarily indicative of a mosaic genome [16,17]. While at the same time, hybrids have been described with only one marker belonging to each parent species when found concomitantly in one strain [18,19].
Another controversial interpretation of the differences between S. bayanus and S. uvarum has been put forward by Rainieri et al. [20]: the S. bayanus taxon defined in [4] is a heterogeneous complex composed of pure and mixed genetic lines. Although the authors of this study do agree that the type strain of S. bayanus CBS380 T represents a mixed line, they designated three other strains to be representatives of the pure line: NBRC1948, NBRC539, and NBRC2031. Isolated from beer, these strains are not known as S. uvarum, even though the sequences of some markers display more than 99% nucleotide identity with S. uvarum strain CBS7001 (see for example the sequences Acc Nu: AB196324-AB196327, AB196329-AB196331).
The way S. bayanus and S. uvarum have been grouped has thus been the root of conflicting interpretations. Some propose that these species should be recognized as two varieties of the same species. This suggestion was put forth by G. Naumov [15] who relied upon the biological species concept, a taxonomy concept that was first proposed for plants, where varieties are inter-fertile [21]. The concept is partially derived for yeast based upon the notion that full fertility occurs when strains of the same species are crossed, whereas half fertility defines varieties within a species. Yeast genomic studies have led to new insights into the biological species concept since reduced fertility or absence of fertility between strains in a species may result from gross chromosomal polymorphism or translocations. This has been observed in a strain of S. paradoxus regarded as S. cariocanus, a distinct species [22].
Inter-fertility is thus an all-or-nothing phenomenon. Although appealing, the biological species concept is therefore difficult to apply in many cases. A further complication stems from the fact that interspecies gene transfers described between Saccharomyces yeasts [23] appear much more common than previously anticipated, thus blurring species boundaries. This has been largely confirmed by recent genome sequencing data now available for many Saccharomyces strains [22,24,25,26] indicating that up to 10% of the strains classified in collections as S. cerevisiae may be natural hybrids between S. cerevisiae and more or less closely related species [24]. This situation may also prevail for other species such as S. bayanus.
To try and better determine the delineation between S. bayanus mixed-line, pure-line, and S. uvarum, we decided to investigate the genomes of representative strains of the above species using, as references, the genomes of S. cerevisiae S288c, EC1118, and S. uvarum CBS7001 ( = MCYC623) available in public databases. In a first step, we amplified and sequenced 17 S. uvarum genes which revealed that S. bayanus strains present a mosaic genome including S. uvarum sequences and more divergent ''S. uvarum-like'' sequences. To differentiate hybrids from the genuine species, we initially relied upon physiological characterisation followed by PCR amplification-and-sequencing of specific markers from S. cerevisiae. Some of these markers were then localised on electrokaryotypes by chromosomal blotting and Southern hybridisation. In addition, whole genome scanning using Comparative Genomic Hybridization array (aCGH) was performed to detect all possible S. cerevisiae materials of suspected mosaic genomes. Beside the S. bayanus type strain CBS380 T , we also analysed the beer strains NBRC539 and NBRC1948 which have been described as nonhybrid, pure genetic lines of S. bayanus [20]. As S. bayanus, S. uvarum and S. cerevisiae have been regarded as contributors to the S. pastorianus genome [27], we included S. pastorianus CBS1538 NT , S. carlsbergensis CBS1513 and S. monacensis CBS1503 into our study. We also included the strain NRRLY-1551, since this strain has been confused with CBS1538 NT and been used many times instead of the S. pastorianus neo type strain. We revealed three S. cerevisiae contigs with as many as 27 genes in strain CBS380 T confirming that it indeed carries a mosaic genome. This was also the case for strains NBRC539 and NBRC1948 since two more chromosome contigs of S. cerevisiae origins were found in the genome of these strains. Obviously these transfers have resulted in the capacity to efficiently metabolise not only maltose but also maltotriose, two sugars which are abundant in beer wort. We then checked the fertility between S. bayanus and S. uvarum using strain S. bayanus NBRC1948 which sporulates and ensures high spore viability instead of strain CBS380 T which was infertile in our hands. The hybrid NBRC1948 and the pure line CBS7001 were crossed and were fully inter-fertile, resulting in F1 spores with chromosomal patterns similar to CBS380 T and NRRLY-1551. These strains were thus elements of tetrads issued from crosses involving either the S. uvarum and strain NBRC1948 or a similar strain.
We also compared the MEL1 gene amplified and sequenced from S. uvarum, S. bayanus CBS380 T and S. cerevisiae Mel + strains. This comparison revealed that S. carlsbergensis MEL1 gene is different from the SuMEL1 gene in S. uvarum. The Mel+ character has reduced S. carlsbergensis to a synonym of S. uvarum [3] making it possible to suggest that S. carlsbergensis is the same as S. uvarum [28,29].
Finally, the phylogenies obtained for many of these cerevisiae and non-cerevisiae markers, combined with microsatellite data on S. cerevisiae populations enable us to propose a new phylogeny of these beer lineages.

Genome analysis
We investigated the genomes of the S. bayanus (Saccardo) strain group (Table 1) regarded as hybrid or pure lines such as CBS380 T and NBRC1948 by comparing them with the genomes of the related species S. uvarum (Beijerinck) strains CBS395 T or CBS7001, as well with the genome of S. cerevisiae and S. pastorianus lager strains CBS1538 NT , CBS1513 (ex S. carlsbergensis) and CBS1503 (ex S. monacensis).
Detection in S. bayanus CBS380 T of a 20 kb S. cerevisiae fragment extending from the MAL locus to the telomere of chromosome VII When testing the fermentation of maltose, the S. bayanus strain CBS380 T responded strongly and rapidly, whereas the S. uvarum strains responded more slowly ( Table 2). We hypothesized that the capacity to ferment maltose might reflect the presence of MAL genes originating from S. cerevisiae. Primers were then designed from the sequence of S288c at the SGD to amplify S. cerevisiae MAL33, MAL31 and MAL32 genes. PCR products were obtained from the genomic DNA of S. bayanus CBS380 T for MAL31 and MAL32 (designated SbMAL31 and SbMAL32) but not for MAL33. The latter gene was mutated in S288c, and was later amplified with primers designed from the sequence of the S. cerevisiae wine yeast EC1118 [24].
Sequencing revealed that SbMAL32 was identical to S. cerevisiae ScMAL32 whereas SbMAL31 shared only 90% nucleotide identity with ScMAL31 but 98% identity with MTY1, a gene encoding the maltotriose transporter described in S. carlsbergensis [30].
From the MAL32 gene in the direction of the telomere, we used chromosome walking to further amplify and sequence PAU24 and COS2 on the genomic DNA of CBS380 T which were found to be identical to S. cerevisiae sequences. In CBS380 T , all intergenic sequences could be amplified from the regions extending from MTY1 to COS2 as well as the COS2-telomere region. This last fragment that contain SUC4 and Y' of S. cerevisiae and that has been previously sequenced in CBS380 T [5,6], was assigned to the newly identified region. Finally a contig of 20kb, named SC20, was assembled spanning the following genes: MTY1-MAL32-PAU24-COS2-SUC4-SCY_1426-RTM1-Y' (Figure 1). RTM1 has been reported to be responsible for the resistance to molasses toxicity in industrial S. cerevisiae strains [31]. All intergenes of SC20 shared 99% sequence identity with their counterparts in S. cerevisiae. Thus, the entire region, except for the coding sequence of MTY1, has originated from S. cerevisiae genome.
As the two S. bayanus strains, NBRC1948 and NBRC539, have been admitted as genetically pure-lines [20], they should be devoid of the S. cerevisiae genes observed in S. bayanus CBS380 T . To verify the purity of these genomes, PCR amplifications with different combinations of primers for the contig SC20 were performed with genomic DNA from NBRC1948 and NBRC539. The fragments MTY1_MAL32, and SUC4_RTM1 from NBRC1948 and NBRC539 were identical with the corresponding fragments from CBS380 T . This suggests that SC20 is also present in NBRC1948 and NBRC539 and that these strains harbour the same composite genomes as S. bayanus CBS380 T . We therefore used both NBRC1948 and CBS380 T strains in further experiments.
Three chromosomes of S. bayanus CBS380 T and NBRC1948 bear the S. cerevisiae contig SC20 Localization of markers from contig SC20 on the chromosomes of S. bayanus CBS380 T and NBRC1948 was performed by karyotype comparison with the S. uvarum strain CBS7001 and the S. cerevisiae strain YNN 295 (Figure 2A). Chromosomal blotting and Southern hybridisation with MTY1 and RTM1 -two markers located at both extremities of the contig SC20-showed that the SC20 contig is repeated on three chromosomes of CBS380 T ( Figure 2B, 2D) as previously reported for SUC4 and Y' [6]. The same localisation was also observed for strain NBRC1948 although its karyotype is not identical to that of CBS380 T . As expected, the Y' probe was also localised in the same chromosomal bands as RTM1 in strains CBS380 T and NBRC1948. S. cerevisiae YNN295 does not contains RTM1 ( Figure 2D), whereas no signal of these S. cerevisiae markers was obtained on the S. uvarum CBS7001 chromosomes as expected. The MAL31 homologue from S. uvarum presents only 77% identity with S. cerevisiae MAL31. We designated the three S. cerevisiae contigs, cB, cA, and cC, located on the three composite chromosomes of CBS380 T and NBRC1948. These three chromosomes have respective sizes which are comparable to those of chromosomes 14, 11, and 8-9 in the strain S. uvarum CBS7001 ( Figure 2C). Chromosomes are   [4]. Previously, the S. cerevisiae MAL genes have been detected on the chromosomes of strain CBS380 T by Southern hybridization [17]. However, in the absence of sequencing, the authors did not recognize the MTY1 gene.
Detection of other S. cerevisiae fragments in strains CBS380 T and NBRC1948 We searched for further S. cerevisiae content in strains CBS380 T and NBRC1948 using array-CGH mapping. For this, we employed Affymetrix Yeast2 arrays which are specific for S. cerevisiae DNA [32]. The strain S. uvarum CBS7001 was used as a reference as it is known to be devoid of S. cerevisiae genes [7,9],

Array-CGH analysis of the CBS380 T and NBRC1948 genomes
After CGH scanning, the log-ratio calculated for each probe was plotted along each chromosome. This revealed the presence of several regions with hybridization signals stronger than the S. uvarum CBS7001 background, suggesting the presence of S. cerevisiae genes. Examples are shown for S. cerevisiae chromosomes A, B, and H ( Figure S1). Three regions larger than 1 kb with strong hybridization signals were detected in CBS380 T . These corresponded to YB (S. cerevisiae chromosome B) from nt 801475 to nt 809216, YG from nt 1063994 to nt 1067828, and a telomeric region carrying Y' (Table S1). These three regions represent a total of 17.3 kb of S. cerevisiae S288c genomic DNA. The first and the third regions include almost all of the genes amplified and previously assembled into contig SC20: MAL33, MAL31 or MTY1, MAL32, and PAU24, which are located in the subtelomeric regions of S. cerevisiae. The second region bears ZUO1, BIO2, and IMA1 encoding an isomaltase (a-glucosidase) activity [33].
Compared with CBS380 T , strain NBRC1948 presented a higher proportion of S. cerevisiae genes in its genome with 21 regions .1 kb totaling as much as 89 kb (Table S3). These included the S. cerevisiae regions already found in CBS380 T and many more S. cerevisiae genes -such as PHO12, IMD2, and FLO5 on YHR-as well as the YAL genes BDH2, BDH1, ECM1, CNE1, and GBP2. Thus in CBS380 T and NBRC1948, aCGH detected a S. uvarum genetic background interspersed with S. cerevisiae fragments or regions.
The classification of all of these regions into GO categories revealed that three categories of genes were overrepresented: genes belonging to the ribonucleoprotein complexes (20 genes), genes involved in key functions under anaerobiosis or high osmotic stress such as AUS1, PBS2, GPD1, TDH1 GPH1, and genes belonging to the maltose metabolic process (three genes of the MAL locus).
A S. cerevisiae contig cB of 30.8 kb in strains CBS380 T and NBRC1948 The aCGH analysis identified ZUO1, BIO2, and IMA1 which are contiguously located on the right arm of chromosome VII, YGR of S288c. Using S. cerevisiae primers on CBS380 T and NBRC1948, we could amplify and sequence a BIO2 block containing a truncated sequence of ZUO1, the entire sequence of BIO2, and a truncated IMA1. . Black arrows represent to scale genes oriented from the telomere to the centromer and grey arrows genes in the opposite orientation. The junction between S. uvarum and S. cerevisiae chromosomes was indicated by a colour change (white for S. uvarum, grey for S. cerevisiae) within the MAL33 gene of cA. The 59 ends of contigs A and C were deduced from segregants NBCB-6c and NBCB-6a issued from the cross between strain NBRC1948 and CBS7001. doi:10.1371/journal.pone.0025821.g001 Using Southern blotting, the BIO2 gene was detected on a single chromosome bearing the contig cB in strains CBS380 T and NBRC1948 ( Figure 2E, 2F). We hypothesized that cB may start in the BIO2 block and extend to SC20. Indeed, a 10 kb fragment spanning from BIO2 to MTY1 was successfully amplified in these two strains using the Expand Long Range PCR kit. Sequencing of the unknown parts of this 10 kb-fragment allowed us to complete and correct the MAL locus with MAL33-MAL31 (in place of MTY1) genes at the 39 end of the ZUO1-BIO2-IMA1 block. In the final assembly, contig cB is 30803 bp in length, spanning the following genes: truncated ZUO1, BIO2, truncated IMA1, and MAL33-MAL31-MAL32-PAU24-COS2-TEL14L_XC-SUC4-SCY_1426-RTM1-YPL283C (Accession number FN677930). The block SUC4, SCY_1426 and RTM1, has been recently described in two brewing yeasts [34].
In the genomes of the S. bayanus strains CBS380 T and NBRC1948, both MAL31 and MTY1 genes exist and respectively encode maltose and maltotriose transporters. Southern blotting showed that the MTY1 probe hybridized with three chromosomes on the karyotypes of CBS380 T and NBRC1948. As MTY1 and MAL31 share 90% nucleotide identity, cross hybridization of MTY1/MAL31 was observed. Later, segregants from the cross NBRC1948 x CBS7001, carrying single contig cB, cA, and cC, allowed us to show that cB carries MAL31, whereas cA and cC both carry MTY1 ( Figure 1). This was confirmed by sequencing the MAL locus in each segregant separately. In either segregants or parent strains, the presence of MAL31, MTY1, or both can be recognized by PCR using specific primers (Table S2).

Presence of cB contig in brewing yeasts
S. carlsbergensis CBS1513 can ferment maltotriose. Salema-Oom et al. [30] attribute this activity to MTY1 whereas Alves et al. [35] have implicated the AGT1 (or MAL11) gene. The BIO2_MAL31/ MTY1 fragment was amplified in the S. pastorianus strains CBS1513 (ex S. carlsbergensis) and CBS1503 (ex S. monacensis), but not in the S. pastorianus CBS 1538 NT (Table 3). In the S. carlsbergensis karyotype, the BIO2 probe hybridized with one chromosome, whereas the MTY1 probe hybridized well with four other chromosomes, and the Y' probe hybridized with many chromosomes ( Figure 2B, 2F, 2G). SCY_1426 and RTM1 are also present in CBS1513 and CBS1503, so we concluded that contig cB carrying MTY1 or MAL31 is also present in S. carlsbergensis and S. monacensis. Indeed, both genes could be amplified from strains of these latter species and sequenced using appropriate specific primers. Out of all the strains used in our study, AGT1 gene could only be amplified in CBS1513 (S. carlsbergensis) and CBS1503 (S. monacensis). Their sequences have the same T insertion at nucleotide 1183 which generates a stop codon giving a truncated protein of 394 aa whereas the normal protein has 616 residues (Acc Nu FR873106-07). This suggests that AGT1 may be inactive, thus implying that MTY1 is responsible for the fermentation of maltotriose in CBS1513 and CBS1503 [30,36,37].

Additional S. cerevisiae genes present in NBRC1948 and NBRC539
Array-CGH detected some additional S. cerevisiae genes in NBRC1948 other than those present in CBS380 T . We amplified and sequenced some of these genes in NBRC1948. A single 3913 bp fragment spanning the two entire PHO12 and IMD2 genes could thus be obtained, defining contig cD (Acc. Nu FR754543). Five other contiguous genes BDH2, BDH1, ECM1, CNE1, and GBP2 located on chromosome I (YAL genes) could be amplified in two PCR overlapping fragments: BDH1_ECM1 and ECM1_GBP2. Sequences of these two sub-fragments were assembled into a new 8487 bp contig named cE (Acc. Nu FR754541). As expected, no amplification with the primers used was obtained for the above genes in CBS380 T . Partial FLO5 (YHR211w) could be amplified and sequenced in strain NBRC1948 (Acc. Nu FR754545). Genes BDH2 and FLO5 were used as probes on chromosomal blots. BDH2 is localised on chromosome I of S. cerevisiae and on one of the two smallest chomosome bands in NBRC1948. In strains CBS380 T and S. uvarum CBS7001, BDH2 of S. cerevisiae also marked the smallest chromosome, though with less intensity. This is expected for a 20% diverging sequence ( Figure S2A, S2B). FLO5, which is a member of a multi-gene family, hybridised with two chromosomes in S. cerevisiae and with at least two chromosomes in S. uvarum CBS7001 and S. bayanus CBS380 T but with less intensity ( Figure  S2C, S2D).
Strain NBRC539, which has been described as another S. bayanus genetic pure-line [20] was submitted to several experiments carried out with NBRC1948. Results showed that NBRC539 was similar to NBRC1948. Indeed, both strains shared the same karyotype ( Figure S3) and several segments of contig cA, FLO5 (partial) were amplified and sequenced from NBRC539, as were the cD and cE. These sequences were identical to their counterparts in NBRC1948 (FR754544 and FR754542, Acc. numbers in Table S3). NBRC539 and NBRC1948 are thus genetically similar, although only the latter retained the capacity to sporulate.
Origin of the S. cerevisiae fragments characterized in CBS380 T and NBRC1948 The recent re-sequencing of many S. cerevisiae strains [16,24] enabled us to compare the S. cerevisiae sequences found in CBS380 T and NBRC1948 with sequences from strains of various origins including ale strains [34], or the lager strain Weihen-

S. monacensis CBS1503
BIO2 Sequences with more than 99% identity with S288c, RTM11-1a, YJM789 and EC1118 are considered as S. cerevisiae. For RTM1 the sequence Acc. NuU02618 was used for comparison. Most of the sequences are deposited (see Table S3). ND: Not determined. stephan 34/70 [25]. A phylogeny constructed for the BIO2 nucleotide sequence clearly indicates that CBS380 T and NBRC1948 have a wine/European origin, and that the sequence is different from the one encountered in ale yeast strains ( Figure  S4). Similarly, the ScBDH2 gene from NBRC1948 also indicates a wine/European origin which is clearly different from the fragments sequenced in the two ale strains as well as in the lager yeast, Weihenstephan 34/70.
Differentiation of S. bayanus, S. uvarum and S. pastorianus strains by fermentation of maltotriose and melibiose S. bayanus CBS380 T can ferment maltose more quickly and strongly than S. uvarum strains. However, the difference between these species was even more conspicuous when maltotriose was used in fermentation tests; only S. bayanus can ferment this trisaccharide, S. uvarum strains cannot. Strains CBS380 T , NBRC539 and NBRC1948 all fermented maltotriose. This activity in S. bayanus is likely to be mediated by MTY1 as demonstrated for S. carlsbergensis [30,36]. Fermentation of maltotriose is thus a common character between S. bayanus and the group of brewing yeasts.
Melibiose utilisation is a characteristic common to S. carlsbergensis and S. uvarum. This latter is known as Mel+ and carries the MEL1 gene, which is found in the S. uvarum genome (contig AACA01000043). Primers were selected to amplify the MEL1 in the S. uvarum strains CBS395 T and CBS2946, S. bayanus CBS380 T , and strain NRRLY-1551. They all share an identical sequence designated SuMEL1 (Acc. Numbers FR750556-59). Sequence alignment showed that SuMEL1 displays 79% and 94% nucleotide identities with MEL1 genes from S. cerevisiae (ScMEL1, Acc Nu M10604) and from S. carlsbergensis, respectively (Acc Nu M58484) [38]. The ScMEL1 gene was also amplified and sequenced in two S. cerevisiae strains, ATCC42367 and CBS2354 (Acc. Nu FR75054-55). On chromosomal blotting, the SuMEL1 probe hybridised to chromosome 3 of S. uvarum strains CBS395 T and CBS7001 as well as to its isomorphic chromosome in S. bayanus CBS380 T ( Figure  S5B, S5D). However, in S. cerevisiae Mel+ strains, S. carlsbergensis CBS1513, and S. monacensis CBS1503, the ScMEL1 probe hybridised with a larger chromosome ( Figure S5B, S5D). MEL1 could not be amplified in strains NBRC539 and NBRC1948 with either S. uvarum or S. cerevisiae primers and could not be detected on the karyotype of NBRC1948 ( Figure S5). If the gene is indeed absent, this may explain why these two strains cannot ferment melibiose.
Inter-fertility of S. bayanus NBRC1948 and S. uvarum CBS7001 To determine whether S. bayanus and S. uvarum are conspecific according to the biological species concept, we tested the fertility of a hybrid diploid strain.
Strain CBS380 T was able to sporulate albeit poorly giving no viable spores from 20 dissected tetrads under normal growth conditions on rich medium YPD. This strain is therefore practically infertile in our hands. However, Ryu et al. [39] have obtained a sporal clone from CBS380 T , strain B19-3c, presenting a karyotype with three missing bands compared with CBS380 T [6].
In contrast, strain NBRC1948 sporulated efficiently. After selfsporulation, 19 asci were dissected giving a spore viability of 54%. Karyotypes of segregants from two complete tetrads were indistinguishable from the karyotype of the parental strain NBRC1948 (data not shown). Since the fertility of NBRC1948 was sufficiently high, we crossed this strain with S. uvarum CBS7001, which is homothallic and fertile, as self-sporulation and dissection of 24 asci gave 98% of viable spores.
Two hybrids were constructed by crossing strain NBRC1948 with CBS7001 using the spore-to-spore mating technique [40,41]. One hybrid named NBCB-10D was confirmed by karyotyping and retained for further study. Its karyotype united two sets of chromosomes, each derived from one parent ( Figure 3A). As expected, the hybrid cumulated the phenotypes of the two parental strains since it can ferment both maltotriose and melibiose. Thirteen asci were dissected from the sporulated NBCB-10D: 36 spores have germinated giving 69% of viable spore clones. This viability is in the range observed for many intraspecific crosses within S. uvarum strains, i.e. 45 to 85% [40]. Four complete tetrads, five with three, one with two, and three with one viable spore were obtained ( Figure 3B). The four complete tetrads named NBCB-2, NBCB-6, NBCB-9, and NBCB-13 were further analysed; chromosomal patterns of two parent strains and segregants of the tetrads NBCB-6 are shown ( Figure 3C). This proved that the cross S. bayanus NBRC1948 x S. uvarum CBS7001 produced fertile offspring. As a control, an interspecies hybrid was constructed by crossing NBRC1948 with CLIB219 (ade2), a S. cerevisiae tester strain [40]. The hybrid S. bayanus x S. cerevisiae was confirmed by the white colour of the colony and its capacity to grow at 37uC was inherited from S. cerevisiae; S. bayanus cannot grow at this temperature. This hybrid sporulated well and 26 asci were dissected, but no viable spore was obtained on rich medium. Consequently S. bayanus is conspecific with S. uvarum but not with S. cerevisiae according to the biological species concept [42].

Segregation of S. cerevisiae contigs in the NBRC 1948 x CBS 7001 offspring
The offspring of the cross S. bayanus/S. uvarum allowed us to follow the transmission of the cerevisiae contigs and several other markers in the segregants of the tetrads. Karyotypes of the four complete tetrads from the cross NBRC1948 x CBS7001 showed segregation of the two chromosome sets from both parental strains ( Figures 3C, 4A, 4C). In some tetrads (NBCB-2, NBCB-9), none of the karyotypes are completely similar to one parental karyotype and the chromosomes have been reassorted. Chromosomal recombination was clearly observed in segregants NBCB-2a, NBCB-2d, NBCB-9a, and NBCB-9b changing the size of the shortest chromosome which was isomorphic with S. uvarum chromosome 1 ( Figure 4A, 4C).
Chromosomal blotting following Southern hybridization was first performed with RTM1 which hybridized with one, two or three chromosomes in the segregants ( Figure 4B, 4D). Probing with MTY1 showed that it also hybridised with the chromosomes bearing RTM1 in each segregant, whereas the BIO2 probe hybridised with two segregants bearing the contig cB (data not shown). Thus, the three S. cerevisiae contig cA, cB and cC segregated as three independent markers during meiosis. Tetrads NBCB-2 and NBCB-6 were analysed more extensively and we detected segregants bearing single contigs: contig cB in NBCB-2d and NBCB-6b, contig cA in NBCB-6c, and contig cC in NBCB-6a. The segregant NBCB-6d harbours the three contigs cB, cA, and cC. The continuity of the three contigs was confirmed by PCR of segregants NBCB-2d and NBCB-6b (cB), NBCB-6c (cA), and NBCB-6a (cC). Seven overlapping fragments were successfully amplified for contig cB and six overlapping fragments were amplified for cA and cC using a set of suitable primers (Table S2). Sequencing of the MAL genes in NBCB-6b, NBCB-6c, and NBCB-6a confirmed the presence of the usual S. cerevisiae MAL locus composed of MAL33, MAL31, and MAL32, in contigs cB and the modified MAL locus composed of MAL33, MTY1 and MAL32 in contigs cC and cA (Figure 1). MTY1 encoding a maltotriose transporter has been first cloned and sequenced in S. carlsbergensis but the same gene named MTT1, has been later identified in many S. pastorianus strains [36]. The sequence of MTT1 or MTY1 is divergent but to a lesser degree with S. cerevisiae MAL31 (90% nucleotide identity) than with the S. uvarum MAL31 homologue (77% nucleotide identity).
PCR amplification of ScIMA1-ScMAL33, yielded a shorter fragment than expected for contig cA in segregant NBCB-6c. Sequencing showed that the MAL33 gene is a chimer; at the 59end, 397 nucleotides have 91% identity with S. uvarum MAL33 (78% with S. cerevisiae) whereas at the 39-end, there are 1010 nucleotides which are identical to the S. cerevisiae strain RM11-1a. Thus, the junction between the S. uvarum-like chromosome part and the S. cerevisiae contig cA occurred in the MAL33 gene. The size difference in the segment ScIMA1-ScMAL33 in cA compared with cB or cC is due to the nucleotide polymorphism of the intergenic region between S. uvarum-like IMA1 and MAL33 (992 bp in cA sequence Acc. number FR845777 versus 2098 bp in contig cB). Both amplicons of ScIMA1-ScMAL33 were obtained from NBRC1948, CBS380 T , NRRLY-1551 and from segregants carrying cA and cB ( Figure S6A).
PCR amplification using specific primers for MAL31 and MTY1 also confirmed the presence of MAL31 in cB and MTY1 in cA and cC ( Figure S6B). Furthermore, PCR/RFLP of MAL31/MTY1 with HinfI can be used to recognize strains carrying S. cerevisiae MAL31 or MTY1 or both ( Figure S6C).
When checked for the capacity to ferment maltose and maltotriose, segregants NBCB-2b, NBCB-2c, and NBCB-6c containing MTY1 but not MAL31 fermented maltotriose more rapidly than maltose. NBCB-6a which also possesses MTY1 but not MAL31 did not ferment maltose at all. This could be explained by other deficiencies because NBCB-6a grows slowly even on rich medium. Strains NBCB-2d and NBCB-6b carrying MAL31 but not MTY1 could only ferment maltose, as expected. Contig cE segregated 2:2 in the tetrads NBCB-2 and NBCB-6 as confirmed by PCR of the two overlapping regions BDH2_ECM1 and ECM1_GBP2 (Table 2). It was localised on one of the smallest chromosome in strain NBRC1948 using the BDH2 probe ( Figure  S2B). Other markers non-cerevisiae SuMEL1 and HO segregated 2:2 as expected for a single gene ( Table 2). Segregation of HO and SuMEL1 in the tetrads NBCB-2 and NBCB-6 implies a cross involving two haploid parent strains.
For all of the cerevisiae and non-cerevisiae makers analysed, the genetic exchanges between S. bayanus and S. uvarum confirmed that they are fully inter-fertile and the transmission of cerevisiae characters -such as the capacity to ferment maltose/maltotriose-to future generations can be assumed by many segregants of this cross.

Distribution of S. uvarum genes and lager-type sequences in S. bayanus strains
Our previous results [6] and array CGH data obtained in this study demonstrate that several strains classified as S. bayanus carry S. cerevisiae-like sequences in a CBS7001-like genomic background (hereafter referred to as S. uvarum). Other studies suggest the presence of a third ancestral parent in S. bayanus [20,43,44,45]. To confirm the presence of different backgrounds in S. bayanus and S. uvarum strains, we compared several non-cerevisiae sequences of CBS380 T with those from different S. bayanus and S. uvarum strains using primers designed from the CBS7001 genome (Table 4).
Among the 17 protein-coding genes from S. bayanus CBS380 T , 12 are identical to those of CBS7001 whereas five share around 93% nucleotide identity with their S. uvarum counterparts ( Table 4). Most of the nucleotide substitutions are neutral so that the deduced protein sequences have over 99% similarity. Four of these five genes share between 99.9 and 100% nucleotide identity with their homologs in S. carlsbergensis CBS1513 or S. monacensis CBS1503 (Acc. numbers in Table S3; [5]). This type of sequence, which represented 30% (5/17) of the non-cerevisiae sequences found in the S. bayanus genome, has already been identified in S. bayanus CBS380 T and NBRC1948 and published as lager-type sequences (Lg sequences [43,44,45]). A small set of random fragments of genomic DNA from CBS 380 T have been previously compared with S. cerevisiae proteins [46]. We revisited these sequences by Blast against contigs of S. uvarum CBS 7001 deposited by the MIT and the Washington University (accession numbers AACA00000000 and AACG00000000). Sequences containing protein-coding genes with over 98% nucleotide identity with CBS 7001 genes were considered as being derived from S. uvarum. However, this ancestor represented only 40% of the sequences (34 GSSs out of 86, Table S4). Given the quality of the CBS380 T sequences (single-strand sequencing) these GSSs cannot be as representative as the sequences obtained in our study. As we found in CBS380 T and NBRC1948, non-cerevisiae gene sequences are either identical to S. uvarum genes (ADH1, MET2, NAM2, and ERG10) or to lager-type genes (BAP2, GDH1, GPI13, HO, and PMA1); we concluded that CBS380 T and NBRC1948 carry three types of sequences originating from S. cerevisiae, S. uvarum, and a third parent common to S. carlsbergensis and other S. pastorianus or lager strains.
Phylogenetic trees based on MET2, PMA1, and MAL31/MTY1 exemplified the variability in strain clustering in uvarum, lager, and cerevisiae depending on the sequences considered. For instance, strain CBS380 T clusters with uvarum type for both MET2 and PMA1 whereas strain NBRC1948 belongs to the uvarum cluster for MET2 but to the lager cluster for PMA1 ( Figure 5).
The ERG10, GDH1 and HO sequences were also considered. The trees obtained from these depicted another set of relatedness features: CBS380 T is clustered in lager whereas NBRC1948 is clustered in uvarum for ERG10 and in lager for GDH1 ( Figure 6). The HO phylogenetic tree showed two clusters: bayanus/lager and uvarum ( Figure 6). As expected, CBS380 T and NBRC1948 clustered with both lager and cerevisiae regarding MTY1 and MAL31 markers (Figure 7). S. carlsbergensis CBS1513 with all lager- type sequences always belongs to the lager cluster reflecting a more lager-type than uvarum-type background as observed in Table 4. Additional strains were included in the trees and show variable proportions of each background. Based on the three markers MET2, PMA1, and MAL31, NBRC539 and NBRC2031 appear as mixed-line (Figures 5, 7).

Origin of the S. cerevisiae moiety of the genome of lager hybrids
To characterize the origin of the S. cerevisiae partner of beer hybrids, the 12 microsatellite loci formerly used to characterize S. cerevisiae diversity [47] were amplified in seven beer hybrids: CBS1513, CBS1503, CBS1538 NT , CLIB276, CLIB277, CLIB278, and CLIB279. Amplification was successful for 5 to 11 loci depending on the strain. Comparison of these allelic profiles with those obtained for different strains by Liti et al. [16] and with samples from our database revealed some original features ( Figure S7). The different beer isolates were split into three clusters. The first and largest cluster included ale beer strains and lager beer strains of German origin, as previously described [47]. A second group of strains contained S. monacensis CBS1503, the neotype strain of S. pastorianus CBS1538 NT , and two other beer strains CLIB276 and CLIB277. Strains of this group exhibit a lower S. cerevisiae content in their genomes as only five or six loci could have been amplified. This confirmed the separation of lager yeast into ''Frohberg-type'' strain and ''Saaz-type'' strains in agreement with the results of Dunn and Sherlock [45].
Strain S. carlsbergensis CBS1513 clustered away from these two former groups and was more related to some rum and distillery isolates (identical alleles at five loci over 10). These different clusterings indicate that several hybridization events have given rise to these different brewing strains.

Discussion
Our results present a new picture of beer strain lineages. We show that despite the use of classical or molecular approaches, the hybrid nature of many strains of the S. bayanus taxon has until now remained unidentified. Here, we have decisively confirmed that the type strain of S. bayanus, CBS380 T , shows a hybrid nature. Unexpectedly, other strains regarded hitherto as pure genetic lines [20] were also revealed to be mixed genetic-lines: NBRC1948, NBRC539, and NBRC2031.
The hybrid nature of CBS380 T has been proposed based on the presence of S. cerevisiae SUC4, and Y' in its genome [5,6]. However, with few cerevisiae-genes detected, our proposal has not been generally accepted. New investigations revealing the capacity of CBS380 T to strongly ferment maltose have paved the way to detecting more S. cerevisiae genes by PCR. Further investigation by aCGH mapping of the CBS380 T genome allowed us to amplify and assemble three contigs cA, cB and cC giving a total of 70 kb of S. cerevisiae sequences. Concerning strain NBRC1948, its hybrid nature has never been suspected even when using PCR/RFLP of three markers per chromosome [20]. This is probably due to the sporadic distribution of S. cerevisiae genes in NBRC1948. Here, it was revealed to be a hybrid carrying over 82 kb of S. cerevisiae sequences composed of all the S. cerevisiae genes found in CBS380 T together with contigs cD and cE. This shows that chromosome blotting and Southern hybridization, as well as aCGH, when used separately, may lead to mis-identification of related sequences. For instance, aCGH did not distinguish MAL31 and MTY1 in S. carlsbergensis [45] because they share 90% identity [30].
A second species related to S. uvarum in the genetic background of S. bayanus CBS380 T and NBRC1948 As previously reported, our study confirmed the presence of two ancestral partners in the genomes of S. bayanus and lager brewing strains [20,45]. Analysis of the non-cerevisiae background in S. bayanus NBRC1948, CBS380 T enabled us to detect more sequences sharing around 93% identity with their S. uvarum counterparts. Some of these alleles have already been described for S. pastorianus CBS1503 and CBS1513 and were qualified as lagertype [43,44]. In the present study, additional lager-type markers, such as BGL2, ERG10, and PMA1 were amplified with primers designed from S. uvarum homologues and sequenced ( Table 4). None of these lager-sequences were found in the S. uvarum genome. In our analyses, the proportion of lager/uvarum sequences is 1/3 in S. bayanus CBS380 T but it seems to be higher in strain NBRC1948 (Table 4) and in S. pastorianus.
Strikingly, these lager-sequences present a very low nucleotide divergence compared with S. uvarum sequences. The SNPs are almost all neutral, which indicates a common origin for both kinds of sequences.
The species carrying these lager-type sequences should be the partner of S. uvarum forming the genetic background of the S. bayanus mosaic genomes. The genome of this S. uvarum-like species should present approximately 93% nucleotide identity with S. uvarum and be inter-fertile with it. We propose to tentatively name this S. uvarum lineage Saccharomyces lagerae. Traces of this lineage can be found in strains of the S. bayanus taxon such as CBS424 (e.g. S. globosus) and CBS2946. These strains carry the S. uvarum proteincoding sequences analysed but with a CARB type IGS typical of lager strains (Acc. NuAJ243214) as evidenced by the AluI pattern of the NTS2 ( Figure S8A) described in [6]. The differential distribution of lager-type markers in the genome of CBS380 T and NBRC1948 sustains the possibility of independent crosses between each ancestor of these strains. Strains CBS424 and CBS2946 may be considered as mosaic genomes S. uvarum/S. lagerae. In addition, strains CBS378, CLIB271, and NBRC2031 without any lagerae markers detected might have arisen from a parent strain similar to NBCB-2d or NBCB-6b which was backcrossed with the S. uvarum strain. They carry the S. uvarum NTS2 and HO ( Figure S8B) with the usual MAL locus carrying MAL31 but not MTY1 in two chromosomes of CBS378 [17] and NBRC2031 (data not shown). Another explanation for strains CBS424, CBS2946, CBS378, CLIB271, and NBRC2031 was that they were produced from backcrosses between segregants of NBRC1948 x CBS7001 with S. uvarum. This hypothesis is schematically presented in Figure 8.
In traditional brewing conditions, back crossing of NBRC1948 or a similar strain with a S. uvarum or with a S. lagerae strain might have generated many hybrids which persisted in brewing processes until pure cultures came into practice following the recommendation of E.C. Hansen. Since then, only hybrids capable of fermenting at low temperatures with improved biotechnological characteristics have been selected and maintained. Therefore, the pure S. lagerae line endowed with a weak MAL system may have been lost in the brewing environment.
S. bayanus consists of two lineages resulting from separated ancestral crosses between S. uvarum and S. cerevisiae Our combined molecular analyses revealed that the genome of NBRC1948 was also mosaic with five chromosome contigs of S. cerevisiae origin; three of them had been transferred to S. bayanus CBS380 T . In the mosaic genome of NBRC1948, the presence of long S. cerevisiae contigs seemed to be the result of transfer events involving the subtelomeric region of three chromosomes. The circulating mechanism that has led to the integration of Zygosaccharomyces bailii material into wine yeast genomes [34,48] is inadequate to explain the presence of contigs cB, cA and cC. This is because these contigs do not show a circular permutation of gene order of the inserted fragment. In addition, the 59-end junction is different in the three contigs. This suggests that they propagated by unequal crossing over between homologous sequences at subtelomers leading to reciprocal unbalanced chromosomal ends in meiotic segregants. In contig cA, the junction point occurred within the MAL33 gene which became a chimer with one third S. lagerae MAL33 and two thirds S. cerevisiae MAL33. This corroborates the results reported for strains CBS378 which could be re-interpreted as follows: this strain bears three copies of the MAL locus with different truncations in the 59 ends, one downstream of MAL33 and another one downstream of MAL32 [17].
In strain NBRC1948, additional S. cerevisiae genes were detected, which constituted a high proportion and greater diversity of genes transferred from S. cerevisiae. These findings suggest that NBRC1948 may have arisen from a hybrid S. uvarum or S. uvarum-like strain which received S. cerevisiae DNA fragments in many successive horizontal transfer events. Recurrent backcrosses between this ancestral hybrid and S. uvarum strains may have Figure 6. Evolutionary relationships of GDH1, ERG10 and HO in mixed and pure lines. The evolutionary history was inferred as in Figure 5 but with limited number of strains analysed for GDH1, ERG10 and HO. Sequences GDH1 have been obtained in this study and from [5], ERG10 and HO are sequenced in this study except for bayanus-HO and Lg-HO. The optimal tree with the sum of branch length = 0.06421927 is shown. The percentages of replicate trees in which the associated taxa clustered together in the bootstrap test (100 replicates) are shown next to the branches. The tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. The evolutionary distances were computed using the Kimura 2-parameter method [61] and are in the units of the number of base substitutions per site. The analysis involved 44 nucleotide sequences. All ambiguous positions were removed for each sequence pair. There were a total of 4382 positions in the final dataset. Strain CBS 380 T clustered in lager group based on ERG10 and GDH1 phylogenetic trees. doi:10.1371/journal.pone.0025821.g006 occurred, explaining the progressive loss of S. cerevisiae material in some strains isolated from ancient beer. This is exemplified by the results obtained for the NBRC1948/CBS7001 cross. Indeed, karyotypes of tetrads from the hybrid NBCB-10D showed that segregants harbour parental and recombinant chromosomes. However, none of the karyotypes is completely similar to one parental karyotype, as the chromosomes have been reassorted. The segregant NBCB-6d, retains the three S. cerevisiae contigs cB, cA and cC, and is very similar to both CBS380 T and NRRL Y-1551 except that the latter carries the contig cE whereas CBS380 T and NBCB-6d have not inherited it. Based on these features, we propose that strain NBRC1948 is an initial hybrid while CBS380 T and NRRLY-1551 are segregants of tetrads deriving from a backcross between NBRC1948 -or a similar strain-with S. uvarum.

Acquisition of new functions by interspecies genes transfer
The S. cerevisiae moiety of the mosaic genome of NBRC1948 includes genes with key functions in anaerobiosis (AUS1), high osmotic stress (PBS2, GPD1, TDH1 and GPH1), maltose and maltotriose fermentations (MAL loci), biotin synthesis, sucrose degradation (SUC4), and resistance to inhibitory substance in molasses (RTM1). These genes were transferred from a S. cerevisiae wine strain, as shown by the phylogenetic trees established with BIO2 and BDH2 coding sequences. Thus, the presence of contigs cB, cA, and cC conferred great advantages under brewing conditions with one authentic MAL locus (contig cB) and two MAL-like loci with MTY1 (contigs cA and cC) leading to the fermentation of maltotriose. The presence of many genes involved in maltose or maltotriose metabolism in a brewing strain is rather common. Despite the lack of conclusive arguments, the most plausible hypothesis concerning the presence of both MTY1 and MAL31 in S. bayanus is that two successive events occurred, firstly a transfer of the S. cerevisiae MAL locus containing MAL31 and then the replacement of MAL31 by MTY1 by conversion. The biological origin of this latter gene is yet to be determined [37].
Analysis of the offspring of the cross NBRC1948 x CBS7001 clearly showed the independent segregation of the three S. cerevisiae chromosomal contigs bearing the MAL locus. However, segregants bearing more than one MAL locus harbouring two copies of MTY1 and one copy of MAL31 would have been more frequently retained. This is because these genes allow these strains to maintain and even increase their capacity to ferment maltose and maltotriose, two saccharides abundant in beer wort. This may explain why different offspring of NBRC1948 carrying the three contigs have persisted until now: CBS380 T , and NRRLY-1551. These strains resulted from multiple hybridization events to acquire the ability to ferment malt wort at low temperatures as required for lager beer production.
The three genes SUC4, SCY_1426, and RTM1 present in S. bayanus, S. carlsbergensis, and S. monacensis lager yeasts have also been found in the recently sequenced genomes of ale strains [34]. By relying on the genome of S. cerevisiae strain S288c, genes present only in wild, or industrial strains such as SCY_1426 and RTM1 have not been detected in S. carlsbergensis and S. monacensis [45].

New proposal for the affiliation of S. pastorianus
As proposed by Vaughan-Martini and Kurtzman [4] based on the determination of DNA relatedness between type strains, S. pastorianus is regarded as a S. bayanus/S. cerevisiae hybrid. Since then, studies based on single-locus techniques have produced divergent conclusions. Using the MET2 gene sequence comparison, S. monacensis has been proposed to be the progenitor of S. carlsbergensis [49]. However, comparative analysis of the proteomes of CBS380 T , NRRLY-1551, and S. pastorianus obtained by 2D-gel electrophoresis suggested that the proteomes of S. monacensis (CBS1503) and S. carlsbergensis (CBS1513) result from the superimposition of two patterns. One of these corresponded to S. cerevisiae and the other corresponded to strain NRRLY-1551 [50]. In the present study, MTY1 and the S. cerevisiae sequences obtained for S. pastorianus CBS1513 and CBS1503 as well as for NBRC1948 and NRRLY-1551 confirmed the relatedness depict-ed by proteome analysis. Strain NRRLY-1551 is very similar to the F1 segregant NBCB-6d, suggesting that this strain is derived from a backcross between a strain closely related to NBRC1948 and a S. uvarum strain. We have indeed shown that some segregants issued from such crosses are fertile (Table 2), and such offspring may then have subsequently mated with S. cerevisiae strains to give S. carlsbergensis and S. monacensis. By analysing the genome of a group of S. pastorianus strains using aCGH, Dunn and Sherlock [45] confirmed the S. uvarum and S. uvarum-like genetic background of lager strains and defined one ale strain as the S. cerevisiae partner. Our findings corroborate the triple hybrid nature of these lager yeasts.
Overall phylogeny of the S. bayanus/pastorianus group: diversity of lager yeasts and clarification of the neo type status of CBS1538 We propose a scheme (Figure 8) to resume the generation of strains carrying mosaic genomes from NBRC1948 to CBS380 T , NRRLY-1551 and other strains in the S. bayanus taxon as well as S. pastorianus lineages. The two main groups of lager brewing strains are composed of the Froherg-type lager strains including strain Weihenstephan 34/70 and the Saaz-type lager strains including S. monacensis CBS1503. These two groups are clearly related to lineages with different S. cerevisiae, S. uvarum, and S. lagerae contents, Figure 8. General scheme deciphering the history of crosses between S. uvarum, S. lagerae and S. cerevisiae. Current S. bayanus strains are segregants of tetrads issued from a cross S. uvarum x NBRC1948 or a similar strain. Strain NBRC1948 represents a generation hybrid anterior to CBS380 T . S. lagerae is the missing species contributor to the genome of strain NBRC1948 or one similar strain. S. lagerae materials reflected by Lg sequences that have been transmitted to S. bayanus and S. pastorianus (Lg presenting 2 to 6% of divergence with S. uvarum sequences). doi:10.1371/journal.pone.0025821.g008 generating the diversity of strains found in the S. pastorianus taxon defined by Vaughan-Martini and Kurtzman [4]. For example many genes of the contig cB could be amplified and sequenced in S. carlsbergensis CBS1513 and S. monacensis CBS1503 but not in the S. pastorianus CBS1538 NT . Physiologically, the latter cannot utilise maltotriose and melibiose (Mel-), whereas S. carbergensis can better utilise these two sugars than S. monacensis (data not shown).
Some discrepancies found in the literature on S. pastorianus may be attributed to the misuse of two of its neotype strains: CBS1538 NT (from the CBS) and NRRLY-1551 (from the ARS in the past). In the neotype strain of S. pastorianus, Dunn and Sherlock [45] found three S. cerevisiae chromosomes to be lost instead of the eight found by Kodama et al., [44] and Rainieri et al., [20]. Proteome profiles obtained by Joubert et al. [50], in addition to karyotypes and melibiose degradation in the present study, demonstrate that NRRLY-1551 has been misidentified and should be reclassified as S. bayanus. In the ARS collection the S. pastorianus neotype is currently strain Y-27171 ( = CBS1538).

Conclusion
Beer was made by a mixture of yeasts in the past and the development of this technology has led to the formation of several lines of hybrids. Lager beer hybrids have been formerly characterized, and the present study showed that many more beer strains including NBRC1948 or CBS380 T are actually mixed lines bearing mosaic genomes related to S. cerevisiae, S. uvarum and a pure line closely related to S. uvarum which was previously unidentified. We have termed this pure line, Saccharomyces lagerae to avoid any confusion with formerly named species. Hybridizations between probable homoploid strains and horizontal transfer(s) might have generated a mosaic genome bearing by S. bayanus and the hybrid species S. pastorianus. The origins of S. bayanus, S. pastorianus and their relatedness were summarized as follows. A hybrid lagerare/uvarum received genetic material from S. cerevisiae probably by horizontal transfer(s). This strain, similar to NBRC1948 (NBRC1948-like), in a backcross with S. uvarum, has generated two segregants in each tetrad which were more or less similar to CBS380 T and NRRLY-1551. Fertile segregants of NBRC1948-like/S. uvarum might have crossed with S. cerevisiae wine strains to produce different hybrids constituting the S. pastorianus lager strain group. These events have permitted the transfer of S. cerevisiae loci -e.g. genes of the MAL and MAL-like containing MTT1into strains of the S. uvarum group allowing a better ability to metabolize maltose and maltotriose. In contrast, the cryophilic ability has been very likely gained from the S. uvarum/lagerae background.
The most striking fact is how the genomes of these strains have gradually gained their respective contents from the genetic materials of the different species, Saccharomyces cerevisiae, Saccharomyces uvarum, and Saccharomyces lagerae. The genome sizes measured for S. bayanus and S. pastorianus were 1.15 and 1.46, respectively [4].
Many cases of interspecific hybrids involving sibling or no sibling species of S. cerevisiae were recently described and appear to be rather common [12]. In addition to S. bayanus and S. pastorianus, hybrids of S. cerevisiae/S. kudriavzevii [26] and S. cerevisiae/S. paradoxus [51] have been reported.
Finally, beer strains appear as a population of triple or double hybrids descended from three pure lines: S. uvarum, S. lagerae, and S. cerevisiae. Each strain harbours a set of genes tracing a particular evolutionary route. Hence, studies carried out with different strains using few markers have led to contradictory conclusions. The type strain of the species S. bayanus CBS380 T is actually a segregant of the back-cross from a first mosaic strain, most likely strain NBRC1948, with S. uvarum. This finding leads to a new classification: S. uvarum and S. bayanus strain CBS380 T which are actually a parent/offspring pair. Our proposal to reinstate S. uvarum (Beijerinck) as a real species [5] and abolish its synonym or varietal status with S. bayanus is fully supported by the data obtained in this study.
The genome of strain CBS7001 and its spore clone 623-6c labelled as S. bayanus in data bases should be relabelled as the S. uvarum genome [11].

Additional note
During the revision of our manuscript, the article cited below appeared in PNAS Early Edition in which the authors described a new species named Saccharomyces eubayanus whose genome corresponds to the one we named Saccharomyces lagerae. Strains S. bayanus CBS380 T and NBRC1948 are triple hybrids containing sequences from S. uvarum, S. eubayanus and S. cerevisiae.
Microbe domestication and the identification of the wild genetic stock of lager-brewing yeast.

Yeast strains
Strains listed in Table 1

Media and fermentation tests
YPD medium, growth at 37uC, Mc Clary sporulation medium and fermentation tests were carried out as in [16]. Maltose and maltotriose at final concentration of 1% were used in fermentation tests.

DNA extraction and molecular techniques
All techniques of DNA extraction, PCR amplification, karyotyping and Southern hybridisation, were performed as previously [5].

Tetrad analysis and hybrids construction
Tetrads dissections were made under microscope using the micromanipulator de Fonbrune. For each dissection, asci formed on Mc Clary medium were suspended in 25 ml of 120 mg/ml Zymolyase 20T (ICN Biochemicals, Aurora Ohio USA). After 10 min of incubation at 37uC, 50 ml of sterilized water were added to stop the enzyme action. Tetrads were dissected on YPD thin layer which was afterward placed on YPD plate and incubated at 28uC; germinating spores formed visible colonies after three days.
Crosses between NBRC1948 and S. cerevisiae tester strain CLIB219 (ade-, red phenotype) or S. uvarum CBS7001 were carried out as follows: each strain was allowed to sporulate on Mc Clary at 28uC for 3-4 days, asci were digested with zymolyase as above. One ascus of each parent strain was dissected and two spores, one from each parent were put together by micromanipulation and incubated. After growing at 28uC for 3 days, putative hybrids formed large colonies compared to sporal clones and were purified on YPD. Hybrids between NBRC1948 and S. cerevisiae CLIB219 formed white colony, the ademutation being recessive. Hybrids were confirmed by electrokaryotyping and subsequently allowed to sporulate. Dissection of about 20 tetrads of each cross was carried out to assess fertility. Autodiploidized segregants from F1 tetrads were sporulated to determine the viability of the F2 generation; similar tests were done for the F3 generation.
Primer selection, PCR amplification and Sequencing of protein-coding genes For S. cerevisiae genes, primers were selected using the GCG Wisconsin package (Genetics Computer Group, Madison, USA) based on the S. cerevisiae sequences available for strains S288c, YJM789 and RM11-1a on SGD (http://db.yeastgenome.org) except for the MAL33 gene: the sequence was from the wine yeast genome EC1118 (Acc Nu FN393070). For S. uvarum genes the sequences used are from strain CBS7001 ( = MCYC 623), labelled S. bayanus in SGD.
For POL3 of S. uvarum strain CBS7001, the gene was fully sequenced (Acc Number FR822194) starting from two nonoverlapping contigs AACA01000496 and AACA01000623 in SGD.
PCR amplifications on S. cerevisiae and S. uvarum genes were carried out as in [16] with a temperature of 50uC for primer hybridization (Tm) except for the PMA1, POL1, and POL3 genes of S. uvarum for which the Tm was 54uC. Primers used are listed in Table S2. Sequences obtained are deposited in EMBL Nucleotide Sequence Data base, Accession numbers are in Supplementary Table S3.

Sequence analyses
Sequences were analysed with the Staden package [52] and the GCG Wisconsin package (Genetics Computer Group, Madison, USA). Nucleotide sequence alignments were performed with Clustal [53] included in MEGA5.0 [54]. The evolutionary distances were computed by the neighbor-joining algorithm [55] using the Kimura 2-parameter method also included in MEGA5.0. Phylogenetic trees were visualized with NJ-Plot [56].

Identification of Saccharomyces lineages by PCR/RFLP of the NTS2 region
The NTS2 of the IGS (rDNA) was amplified and sequenced from Saccharomyces strains using primer pair NTSU/ETSL (Table  S2); RFLP profiles were determined after digestion with AluI enzyme as previously described [6].

CGH analyses
Total genomic DNA of CBS7001, CBS380 T and NBRC1948 was prepared from cultures grown on YPD. The genomic DNA was labeled and hybridized against GeneChipH yeast genome 2.0 array from Affymetrix (Santa Clara, CA) which covers all S. cerevisiae S288C genes, according to Winzeler et al. [57]. Labeled fragments were prepared from 200 to 500 ng of genomic DNA using the BioPrime labeling kit (Invitrogen). Hybridization and detection steps were performed at the 'IGBMC Microarray and Sequencing Platform' (Illkirch, France). Two arrays were used for strains CBS380 T and CBS7001 and one for NBRC1948. Data were analyzed using the apt 1.12 software: first background was subtracted according to MAS5.0 with apt-cel-Transformer and then intensities for S. cerevisiae pm probes extracted with apt-celextract. For each pair of arrays, the slope between the two hybridization signals was calculated in order to normalize the signals of the two chips before averaging the hybridization signals of two probes. For each perfect match probe the hybridization ratio between CBS380 T (mean of the two arrays) or NBRC1948 and CBS7001 (mean of the two arrays) were calculated with Microsoft Excel which enabled us to score the hybridization intensity all along the genome map. The variations of the hybridization signal were analyzed with CGHScan [58] in order to detect regions with hybridization signals stronger than the background. In some cases CGHScan detected amplified regions separated by less than 1-kb gaps. In such cases, these gaps were interpreted as false negative hybridization, and thus, these regions were considered as contiguous. Telomeric genes were only counted once in the analysis. In the graphs, the log ratios were averaged in an 11 probes sliding window corresponding approximately to one gene.
The S. cerevisiae genes encountered in these regions were analyzed with Funspec (http://funspec.med.utoronto.ca/) [59], and categories were filtered with a P value cut-off of 0.01 after Bonferroni correction. All data is MIAME compliant and the raw and transformed data are available at GEO, a MIAME compliant database, under Accession number GSE28225 (http://www.ncbi.nlm.nih.gov/geo/query/acc. cgi?token = tvondwequyqwols&acc = GSE28225).

S. cerevisiae microsatellite analysis
Amplifications and PCR products were analyzed as described before for S. cerevisiae strains [47]. The data obtained for the beer strains were compared to a subset of our former data [47], however, in order to cope with missing data, genetic distances between each strains were computed with MicrosatAnalyzer [60] and in case of aneuploidy one of the supplementary allele was discarded randomly in order to restore diploidy. This did not change the overall topology of the tree in comparison to our former data. Figure S1 CGH scan of CBS380 T and NBRC1948 genomes detecting S. cerevisiae YA,YB and YH fragment. Graphical representation of the log ratio of the hybridization intensity values for Saccharomyces bayanus CBS380 T and NBRC1948 in comparison to Saccharomyces uvarum CBS7001 of which the DNA was hybridized against GeneChipH yeast genome 2.0 (Affymetrix). The graphs cover different regions of chromosome where S. cerevisiae introgressions A, B and H are revealed. (TIF) Figure S2 Localisation of BDH2 and FLO5 on chromosomes of S. uvarum and S. bayanus. A and C. Electrophoretic karyotypes of yeast strains stained with ethidium bromide. B. Probing with S. cerevisiae BDH2 showing its localisation on the chromosome 1 and its segregation in the tetrad NBCB-2. D. Probing with S. cerevisiae FLO5 showing its localization on one chromosome of strain NBRC 1948, and on two chromosomes in S. bayanus, S. uvarum, S. carlsbergensis and S. cerevisiae. Arrow heads indicate chromosome 1 of S. uvarum, S. bayanus and S. cerevisiae hybridized with S. cerevisiae BDH2 and FLO5 gene probes. S. uvarum chromosomes are numbered according to [6]. (TIF) Figure S3 Comparative karyrotypes of strains S. bayanus, S. uvarum, S. pastorianus and S. cerevisiae S288c. Note 1: Similarity between NBRC539 and NBRC1948 and their difference with NBRC2031. Note 2: Heterogeneity of S. pastorianus group: S. monacensis CBS1503, S. carlsbergensis CBS1513, S. pastorianus CBS1538 NT , NBRC2003. NRRLY-1551 and CBS1538 NT exhibit two clearly different chromosomal patterns. (TIF) Figure   S4 Evolutionary relationships between NBRC1948 and other S. cerevisiae strains depicted by BIO2 and BDH2 genes. The BIO2 and BDH2 genes of S. cerevisiae strains from various origins were compared. The sequences used are originating from the data published in [16,24,25,34]. The evolutionary history was inferred as for figure 5.  Figure S7 Clustering of S. pastorianus in S. cerevisiae according to microsatellite markers analysis. Neighborjoining tree showing the clustering of beer isolates among a subset of 140 yeast strains isolated from different sources [24,47] including the set of sequenced strains of Liti et al. [16]. The tree was constructed from the chord distance between strains based on the polymorphism at 12 loci and is rooted according to the midpoint method. Branches are coloured according to the substrate from which strains have been isolated. NColor code: Wine -Europe dark green; Bread yellow; Beer orange; Sake -Japan dark blue; sorghum beer or palm wine -Africa brown; Oak tree -America blue-green; distillery from South America and rum from French Indies purple; Laboratory strains red, Bertram palm -Malaysia blue. Clinical isolates black. (TIF) Figure S8 CARB and UVAR profiles of the NTS2 differentiating Saccharomyces yeasts. A. NTS2 AluI patterns of S. bayanus, S. uvarum, S. carlsbergensis and S. cerevisiae. CARB type pattern of S. carlsbergensis is common for S. bayanus strain group. NBRC2031 exhibits the UVAR type pattern, while the lager strain NBRC2003 exhibits the S. cerevisiae SACE pattern. B. NTS2 AluI patterns of S. uvarum CBS7001, S. bayanus NBRC1948 and of the hybrid NBCB-10D. Segregation 2:2 of CARB/UVAR patterns in the tetrads NBCB-6 and NBCB-13. M: marker 1kb plus Invitrogen. (TIF)