Himalayan Saccharomyces eubayanus Genome Sequences Reveal Genetic Markers Explaining Heterotic Maltotriose Consumption by Saccharomyces pastorianus Hybrids

S. pastorianus, an S. cerevisiae × S. eubayanus hybrid, is used for production of lager beer, the most produced alcoholic beverage worldwide. It emerged by spontaneous hybridization and colonized early lager brewing processes. Despite accumulation and analysis of genome sequencing data of S. pastorianus parental genomes, the genetic blueprint of industrially relevant phenotypes remains unresolved. Assimilation of maltotriose, an abundant sugar in wort, has been postulated to be inherited from the S. cerevisiae parent. Here, we demonstrate that although Asian S. eubayanus isolates harbor a functional maltotriose transporter SeAGT1 gene, they are unable to grow on α-oligoglucosides, but expression of S. cerevisiae regulator MAL13 (ScMAL13) was sufficient to restore growth on trisaccharides. We hypothesized that the S. pastorianus maltotriose phenotype results from regulatory interaction between S. cerevisiae maltose transcription activator and the promoter of SeAGT1. We experimentally confirmed the heterotic nature of the phenotype, and thus these results provide experimental evidence of the evolutionary origin of an essential phenotype of lager brewing strains.

Recently made S. cerevisiae ϫ S. eubayanus laboratory hybrids showed lager brewing performance similar to that of S. pastorianus strains, also with respect to maltotriose utilization (5,6,15,33). In these hybrids, maltotriose consumption depended on the presence of a functional ScAgt1 transporter encoded by the S. cerevisiae subgenome (34). However, in view of the nonfunctionality of ScAGT1 in current S. pastorianus strains, these laboratory hybrids did not fully recapitulate the genetic landscape of S. pastorianus with respect to maltotriose fermentation (2,6,33).
Studies on laboratory hybrids based on S. eubayanus strains whose genomes are more closely related to the S. eubayanus subgenome of S. pastorianus strains than that of the Patagonian type strain CBS 12357 might generate new insights into the evolution of maltotriose utilization in S. pastorianus. To date, Himalayan S. eubayanus isolates show the highest sequence identity with the S. eubayanus subgenome of S. pastorianus, with up to 99.82% identity, in contrast to 99.56% for S. eubayanus CBS 12357 T (35).
Here, we investigated if and how the genomes of Himalayan S. eubayanus strains could have contributed to maltotriose utilization in the earliest hybrid ancestors of current S. pastorianus strains. To this end, we generated chromosome-level genome assemblies of these strains by long-read DNA sequencing. Since the Himalayan strains were unable to utilize maltotriose, we functionally characterized the assembled MAL genes and identified genetic determinants that prevented maltotriose utilization. Subsequently, a laboratory hybrid of a representative Himalayan S. eubayanus strain with a maltotriose-deficient ale strain of S. cerevisiae was generated to investigate the genetics of maltotriose utilization in a hybrid context. We discuss the implications of the experimental results for the proposed role and origin of SeAGT1 in S. pastorianus and for the potential of hybridization to enable maltotriose consumption in novel Saccharomyces hybrids.

RESULTS
Sequencing of Himalayan S. eubayanus strains revealed variations of subtelomeric regions and the presence of novel putative maltose transporter genes. It has been proposed that the S. eubayanus genetic pool of S. pastorianus was inherited from an ancestor of the Asian S. eubayanus lineage (35). With 99.82% identity, the Himalayan S. eubayanus strains CDFM21L.1 and ABFM5L.1 that belong to the Holarctic lineage (36) present the closest characterized relatives of the S. eubayanus ancestor of lager brewing yeasts. However, this distance was based on a limited sequencing space (35), and the analysis did not investigate the presence of specific S. eubayanus genetic markers found in S. pastorianus hybrids. Therefore, we sequenced the genome of the Himalayan S. eubayanus strain CDFM21L.1 with a combination of long-read and short-read techniques (Oxford Nanopore MinION and Illumina technologies, respectively) to generate a nearly complete draft reference genome sequence. The resulting CDFM21L.1 genome assembly comprised 19 contigs, including the mitochondrial genome. All chromosomes were completely assembled from telomere to telomere, except for chromosome XII, which was fragmented into 3 contigs due to the repetitive ribosomal DNA (rDNA) region and manually assembled into a single scaffold. With a total size of 12,034,875 bp, this assembly represents the first nearly complete draft genome of an S. eubayanus strain of the Holarctic clade (36).
Chromosome-level assemblies were hitherto only available for the Patagonia B-clade strain CBS 12357 T (1,13). We identified three major structural differences in CDFM21L.1 relative to the structure of CBS 12357 T using Mauve (37): (i) a paracentric inversion in the subtelomeric region of chromosome VII involving approximately 8 kbp, (ii) a translocation of approximately 12 kbp from the left subtelomeric region of chromosome VIII to the right subtelomeric region of chromosome VI, and (iii) a reciprocal translocation between approximately 20 kbp from the right subtelomeric region of chromosome V and approximately 60 kbp from the center of chromosome XII (Fig. 1A). All structural variation involved subtelomeric regions, in accordance with their known relative instability (38)(39)(40).
An alignment comparison of the CDFM21L.1 and CBS 12357 T genomes with MUMmer revealed that 557 kb were unique to CDFM21L.1, and, reciprocally, 428 kb were unique to CBS 12357 T . Sequences unique to CBS 12357 T (3.6% of its genome) and to CDFM21L.1 (4.6% of its genome) were located primarily in subtelomeric regions and in repetitive regions, such as rDNA on chromosome XII (Fig. 1B). Out of the 32 subtelomeric regions, 23 exhibited absence of synteny. Conserved synteny was observed for subtelomeric regions on ChrIII (left), ChrIV (left and right), ChrVI (left), ChrIX (right), ChrXI (right), ChrXII (right), ChrXIV (left), and ChrXV (right) (see File S1 and Table  S1 in the supplemental material).
The 428 kb of sequence that was absent in the Himalayan S. eubayanus strain included 99 annotated open reading frames (ORFs) (File S1). Of the 99 ORFs that were (partly) affected, 11 were completely absent in CDFM21L.1, involving mostly genes implicated in iron transport facilitation (File S1). The 557 kb of sequence that was not present in CBS 12357 T included 113 annotated ORFs (File S1). Of these 113 ORFs, 15 were completely absent in CBS 12357 T . These 15 ORFs showed an overrepresentation of genes involved in transmembrane transport (Fisher's exact test, P ϭ 4.8EϪ5) (Fig. 1C).
Of the 15 ORFs unique to CDFM21L.1, three were identical orthologs of S. cerevisiae MAL11 (AGT1) (Table S1). These three ORFs were found in the subtelomeric regions of chromosomes VII, XIV, and XV. Their sequence similarity values with the S. cerevisiae CEN.PK113.7D and S. pastorianus CBS 1483 MAL11 and AGT1 genes were 82.7% and 99.89%, respectively. In addition to these SeAGT1 genes, the CDFM21L.1 genome sequence harbored genes encoding three maltose transporters (SeMALTx), two maltases (SeMALSx), and three regulators (SeMALRx). In contrast to the situation in S. eubayanus CBS 12357 T , none of the SeMAL genes formed a canonical MAL locus in CDFM21L.1 (Fig. 2). A systematic sequence inspection of these CDFM21L.1 SeMAL genes revealed mutations that prematurely interrupted the reading frames of SeMALR1 ChrV ( 706 TGA 708 ), SeMALT2 ChrXII ( 694 TGA 696 ), and SeMALT3 ChrXIII ( 1045 TAA 1047 ). . Three copies of SeAGT1 were found close to the telomeres on chromosomes VII, XIV, and XV. Furthermore, there are two intact SeMALS genes on ChrII and ChrXII and three SeMALR genes on ChrV and ChrXIII whose copy on ChrV is also mutated (SeMALR1*). The gene and interval sizes are approximately to scale. Transporter genes SeAGT1, SeMALT1, SeMALT2, and SeMALT3 are denoted with blue arrows, the hydrolase genes SeMALS1 and SeMALS2 are denoted with red arrows, and the regulator genes SeMALR1, SeMALR2, and SeMALR3 are denoted with green arrows. Any other genes are shown with black arrows. (B) Phylogeny of Saccharomyces SeAGT1 genes described in S. cerevisiae, S. eubayanus, and the lager brewing hybrid S. pastorianus. (C) Nucleotide percentage identities between AGT1 orthologs from S. cerevisiae, S. eubayanus, and the lager brewing hybrid S. pastorianus. Green indicates highest similarity between SeAGT1 and SeAGT1 genes from S. pastorianus strains CBS 1483 and WS3470. Red indicates similarity between SeAGT1 from North American strains with SeAGT1 genes from Asian S. eubayanus and S. pastorianus strains CBS 1483 and WS3470.
In addition to the S. eubayanus CDFM21L.1 strain, a second Himalayan S. eubayanus isolate (ABFM5L.1) was sequenced. These two strains were 99.97% genetically identical at the nucleotide level, their MAL genes were syntenic, and the premature stop codons in SeMALR1 (ChrV), SeMALT2 (ChrXII), and SeMALT3 (ChrXIII) were conserved. Two additional mutations were identified in one of the three SeAGT1 genes. A nucleotide variation at positions 53 and 939 (T instead of an A and A instead of a G) resulted in a glycine-to-valine and arginine-to-lysine change, respectively.
Paradoxically, Himalayan S. eubayanus strains do not utilize maltose and maltotriose. Identification of SeAGT1 in the two Himalayan S. eubayanus strains suggests an ability to grow not only on maltose but also on maltotriose. Strains from the Holarctic clade have previously been hypothesized to be the donor of the S. eubayanus subgenome in S. pastorianus hybrids (35,36). However, no physiological data regarding their ability to grow on the sugars present in wort are available. To assess their growth characteristics, the Asian S. eubayanus strains CDFM21L.1 and ABFM5L.1, the Patagonian S. eubayanus type strain CBS 12357 T , and the S. pastorianus strain CBS 1483 were grown on diluted industrial brewer's wort at 12°C. As reported previously, S. pastorianus strain CBS 1483 could utilize all three sugars but did not fully consume maltotriose (Fig. 3) (11). Also in accordance with previous observations (6), CBS 12357 T consumed glucose and maltose completely but left maltotriose untouched. However, in marked contrast to S. eubayanus CBS 12357 T , neither CDFM21L.1 nor ABFM5L.1 consumed maltose after growth on glucose. Moreover, like CBS 12357 T , maltotriose was not metabolized by these two S. eubayanus strains. While in CBS 12357 T an ability to grow on maltose and an inability to grow on maltotriose could be readily attributed to its MAL gene complement, CDFM21L.1 and ABFM5L.1 failed to grow on maltose even though they appeared to contain complete genes encoding maltose (SeMALT1 and SeAGT1) and maltotriose (SeAGT1) transporters.
Growth defects on maltose and maltotriose are caused by deficiency of the regulatory SeMalR proteins in S. eubayanus CDFM21L.1. The recent characterization of maltose metabolism in CBS 12357 T showed that the coding regions of transcriptionally silent maltose transporter genes in S. eubayanus can potentially encode functional proteins (13). The inability of the Himalayan S. eubayanus isolates to grow on ␣-oligosaccharides precluded direct testing of transporter gene functionality by deletion studies. Instead, these genes were expressed in S. cerevisiae IMZ616, which is devoid of all native maltose metabolism genes (41). The CDFM21L.1 transporter gene SeMALT1, SeMALT2, SeMALT3, or SeAGT1 was integrated at the ScSGA1 locus in IMZ616 along with the S. cerevisiae maltase gene ScMAL12 (13), yielding a series of strains overexpressing a single transporter [(IMX1702 (SeMALT1), IMX1704 (SeMALT2), IMX1706 (SeMALT3), and IMX1708 (SeAGT1)]. These strains, as well as the negative-and positivecontrol strains IMZ616 and IMX1365 (IMZ616 expressing ScAGT1 and ScMAL12), were grown on synthetic medium (SM) supplemented with either maltose (SMM) or maltotriose (SMMt). On maltose, not only the positive-control strain IMX1365 but also IMX1702 (SeMALT1) and IMX1708 (SeAGT1) were able to grow on maltose, consuming 30 and 60%, respectively, of the initially present maltose after 100 h (Fig. 4A). As anticipated, the SeMALT2 and SeMALT3 alleles with premature stop codons did not support growth on maltose. Of the two strains that grew on maltose, only IMX1708 (SeAGT1) also grew on maltotriose. These results demonstrate that SeAGT1 from a Holarctic S. eubayanus encoded a functional maltotriose transporter and, consequently, that the inability of Holarctic strains to grow on maltose and maltotriose was not caused by transporter dysfunctionality. In addition to transport, metabolism of ␣-oligoglucosides requires maltase activity. Functionality of the putative SeMALS1 and SeMALS2 maltase genes was tested by constitutive expression in strain IMZ616, together with a functional ScMAL31 transporter gene, yielding strains IMZ752 and IMZ753, respectively. The maltase-negative strain IMX1313 was used as negative control. In SM with maltose, both IMZ752 (SeMALS1) and IMZ753 (SeMALS2) grew and completely consumed maltose within 65 h, demonstrating functionality of both hydrolase genes (Fig. 4C).
In S. cerevisiae transcriptional regulation of MALx2 and MALx1 genes is tightly controlled by a transcription factor encoded by MALx3 genes. Malx3 binds an activating site located in the bidirectional promoters that control expression of MALx2 and MALx1 genes (42,43). To test whether absence of maltose consumption in Himalayan S. eubayanus strains was caused by a lack of transcriptional upregulation of SeMALT and SeMALS, the S. cerevisiae ScMAL13 gene was integrated at the SeSGA1 locus in S. eubayanus CDFM21L.1, under the control of a constitutive ScPGK1 promoter and ScTEF2  (2), IMX1706 overexpressing SeMALT3 (), and IMX1708 overexpressing SeAGT1 (e) were grown on SM with 2% maltose or maltotriose at 20°C. It is worth mentioning that the symbols corresponding to the strains IMZ616, IMX1704, and IMX1708 are highly overlapping and therefore might be difficult to visualize. Growth on maltose (A) and on maltotriose (B) was monitored based on optical density (OD 660 ), and concentrations of maltose and maltotriose in culture supernatants were measured by HPLC. Data are presented as averages and standard deviations of two biological replicates. (C) IMX1313 overexpressing only ScMAL31 (OE), IMZ752 overexpressing ScMAL31 and SeMALS1 (o), and IMZ753 overexpressing ScMAL31 and SeMALS2 (ٌ) grown on SM with 2% maltose. Growth was monitored based on optical density measurement at 660 nm (OD 660 ), and maltose in culture supernatants was measured by HPLC. Data represent averages and standard deviations of two biological replicates.
terminator. ScMAL13 expression in CDFM21L.1 enabled growth on maltose and maltotriose (Fig. 5A), indicating that a lack of transcriptional upregulation was indeed the cause of the parental strain's inability to grow on these oligoglucosides. However, consumption of maltose and maltotriose was incomplete, and consumed sugars were almost exclusively respired, as no ethanol was measured after 60 h of cultivation.
The possibility to grow an engineered variant of S. eubayanus CDFM21L.1 on ␣-oligoglucosides offered an opportunity to study transporter function in its native context. Complementary functional characterization by gene deletion of SeMALT1 and Overview of constructed knockout strains. Knockouts of SeMALT1 (IMK820) and SeAGT1 (IMK823) were made with CRISPR-Cas9. Subsequently the SeSGA1 locus was replaced by ScPGK1 p -ScMAL13-ScTEF2 t using CRISPR-Cas9 in both strains, resulting in IMX1939 and IMX1940, respectively. (C) S. eubayanus strains IMK820 (y), IMK823 (OE), IMX1939 (), and IMX1940 (ࡗ) were characterized on SM with maltose or maltotriose at 20°C. The OD 660 was measured (black), and sugar (black) and ethanol (red) concentrations were determined from the supernatant by HPLC. It is worth mentioning that the symbols corresponding to the strains IMK820, IMK823, and IMX1940 are highly overlapping and therefore might be difficult to visualize. All data represent averages and standard deviations of biological duplicates.
SeAGT1 was performed using CRISPR-Cas9 genome editing method (13,44). Deletion of SeMALT1 and SeAGT1 in CDFM21L.1 resulted in strains IMK820 and IMK823, respectively. Complete deletion of SeAGT1 required disruption of six alleles. To confirm the complete removal of all copies, the genome of IMK823 was sequenced. Mapping reads onto the reference S. eubayanus CDFM21L.1 genome assembly confirmed that all six alleles were removed simultaneously. Subsequently, the regulator expression cassette (ScPGK1 p -ScMAL13-ScTEF2 t ) was integrated in IMK820 and IMK823 at the SeSGA1 locus, yielding strains IMX1939 and IMX1940, respectively (Fig. 5B). The four deletion strains IMK820 (SemalT1Δ), IMK823 (Seagt1Δ), IMX1939 (SemalT1Δ Sesga1Δ::ScMAL13), and IMX1940 (Seagt1Δ Sesga1Δ::ScMAL13) were characterized on SM with glucose (SMG), SMM, or SMMt. All four strains were able to grow on glucose (Fig. S1). While strains IMK820, IMK823 and IMX1940 were unable to grow on maltose or maltotriose (Fig. 5C), strain IMX1939 (SemalT1Δ Sesga1Δ::SeMAL13), which harbored functional SeAGT1 copies, grew on maltose as well as on maltotriose. However, after 64 h of growth, these sugars were only partially consumed. Only 1.2 g liter Ϫ1 ethanol was produced from maltose, and no ethanol formation was observed during growth on maltotriose. The low ethanol concentration and the relatively high optical density at 600 nm (OD 660 ) suggest that, under the experimental conditions, strain IMX1939 exhibited a Crabtree-negative phenotype and exclusively respired maltotriose. S. eubayanus IMX1940 (Seagt1Δ Sesga1Δ:: SeMAL13) did not consume maltotriose after 84 h of incubation. Moreover, despite the presence of SeMALT1, which encoded a functional maltose transporter upon expression in S. cerevisiae IMZ616, strain IMX1940 was also unable to consume maltose.
In addition to a functional Malx3 transcription factor, transcriptional activation of MAL genes also requires presence of a cis-regulatory motif in the promoter of regulated genes. Transcriptome analysis of S. eubayanus CBS 12357 T recently showed that absence of a canonical cis-regulatory motif in SeMALT1 and SeMALT3 of S. eubayanus CBS 12357 T caused a deficiency in their expression (13). To further explore regulation of SeMAL and SeAGT1 genes, we investigated the impact of carbon sources on the genome-wide transcriptome and, specifically, on transcriptional activation of genes involved in maltose metabolism. Duplicate cultures of S. eubayanus strain IMX1765 (ScPGK1 p -ScMAL13-ScTEF2 t ) were grown on SMG, SMM, and SMMt at 20°C and sampled in mid-exponential phase. After mRNA isolation and processing, sequencing reads were mapped onto the newly annotated S. eubayanus CDFM21L.1 genome to calculate the number of fragments per kilobase per million reads mapped for the gene of interest (FPKM). The heterologous regulator ScMAL13, expressed from the constitutive ScPGK1 promoter (ScPGK1 p ) displayed the same expression levels in glucose-and maltosegrown cultures. Although ScMAL13 was efficiently expressed on glucose, none of the nine S. eubayanus maltose genes (the three identical SeAGT1 copies being undistinguishable) were transcriptionally induced under these conditions ( Fig. 6 and Table S2), confirming that the hierarchical regulatory role of glucose catabolite repression (42,45) also takes place in S. eubayanus. During growth on maltose, all nine genes were significantly upregulated relative to levels in glucose-grown cultures, but large variations in expression levels were observed. The maltase genes SeMALS1 and SeMALS2 and the transporter gene SeAGT1 showed the highest upregulation, with fold changes of 148, 161, and 2,355 respectively. Although upregulated SeMALT1 displayed a fold change of 13, its normalized expression in maltose-grown cultures was 886-fold lower than that of SeAGT1. This weaker upregulation might explain why, despite the ability of its coding region to support synthesis of a functional maltose transporter, SeMALT1 alone could not restore growth on maltose. The transcriptome data also revealed that the absence of maltose induction in CDFM21L.1 was not associated with defective cis-regulatory elements in SeMALR promoter sequences since the regulator genes were properly activated; instead, these results would suggest that the SeMalR regulators are not functional.
Hybridization of two maltotriose-deficient S. eubayanus and S. cerevisiae lineages results in heterosis through regulatory cross talk. The genetic makeup of S. pastorianus lager brewing yeasts strongly indicates that they originate from hybridiza-tion of S. cerevisiae and S. eubayanus parental lineages that were both unable to metabolize maltotriose (2). This hypothesis is consistent with the recurrent mutation in the S. cerevisiae AGT1 allele of S. pastorianus strains as well as with the inability of Himalayan strains of S. eubayanus to grow on these oligoglucosides.
Spores of the Himalayan S. eubayanus CDFM21L.1 were hybridized with S. cerevisiae CBC-1. This top-fermenting S. cerevisiae is recommended for cask and bottle conditioning and unable to consume maltotriose (Lallemand, Montreal, Canada). Analysis of the CBC-1 assembly, obtained by a combination of long-and short-read sequencing, linked its maltotriose-negative phenotype to a total absence of the MAL11 (AGT1) gene. The resulting laboratory interspecific hybrid HTSH020 was characterized at 12°C on synthetic wort, a defined medium whose composition resembles that of brewer's wort. While S. eubayanus CDFM21L.1 consumed only glucose and S. cerevisiae CBC-1 consumed glucose and maltose after 103 h (Fig. 7A), the interspecific hybrid HTSH020 completely consumed glucose and maltose and partially consumed maltotriose after 105 h, thus resembling characteristics of S. pastorianus strains (e.g., CBS 1483) (11). In addition to this gain of function, the hybrid HTSH020 outperformed both of its parents in maltose consumption since it depleted this sugar in 70 h instead of the 95 h for strain CBC-1. Since S. cerevisiae grows generally more slowly at 12°C, the experiments were also performed at 20°C, at which temperature HTSH020 consumed all maltose 16 h earlier than CBC-1 (Fig. S2).
Transcriptome analysis of the hybrid strain HTSH020 grown on SM with different carbon sources showed that SeAGT1 expression was repressed during growth on glucose, with a normalized expression level of 7 FPKM (Fig. 7B). When grown on SM with maltose, SeAGT1, SeMALS1, and SeMALS2 were significantly induced, with fold increases of 816, 109, and 116, respectively ( Fig. 7B and Table S3). Although SeMALT1 and SeMALT2 were induced, these transporters do not contribute to maltose metabolism due to truncation of their ORFs. These transcriptome data demonstrated that SeAGT1 and SeMALS genes are induced by regulatory cross talk between regulators encoded from the CBC-1 S. cerevisiae subgenome and maltotriose transporter genes harbored by the S. eubayanus genome. This laboratory hybridization experiment may be the closest reproduction yet of how, centuries ago, maltotriose fermentation capacity arose in the first hybrid ancestor of S. pastorianus.

DISCUSSION
The ability to consume maltose and maltotriose represents a key performance indicator of S. pastorianus lager brewing strains (10). This study demonstrates how mating of S. cerevisiae and S. eubayanus strains that cannot themselves ferment maltotriose can yield maltotriose-fermenting hybrids. This laboratory study illustrates how, centuries ago, maltotriose fermentation capacity may have arisen in the first hybrid ancestor of S. pastorianus.
While the origin of the S. eubayanus parent of S. pastorianus strains is still under debate (46)(47)(48), phylogenetic analysis suggested a Far East Asian origin (35). However, this interpretation was based on a limited sequencing space and was constrained by the quality of available sequence assemblies. Since an ortholog of SeAGT1 had previously been found only in the S. eubayanus subgenome of S. pastorianus strains, this finding revived the discussion on the geographical origin of the ancestral S. eubayanus parent (14). The high-quality, annotated genome assemblies of the Himalayan S. eubayanus strains CDFM21L.1 and ABFM5L.1 presented in the present study revealed several copies of SeAGT1, whose very high sequence identity with S. pastorianus SeAGT1 is consistent with the previously proposed Asian origin of the S. eubayanus subgenome of S. pastorianus (14,35,36). Next, genome sequence comparison of the Patagonian B subclade S. eubayanus strain CBS 12357 T and the Holarctic subclade strain CDFM21L.1 revealed homoplasy of SeAGT1, probably reflecting that these subclades evolved in different ecological niches. The variation in genes encoding iron transport facilitators between the Patagonian and Himalayan S. eubayanus lineages further supports this idea of a difference of the natural habitat of these yeasts. The abundance of these transporters in CBS 12357 T could indicate a lower iron concentration in the environment of Patagonian S. eubayanus, therefore requiring a higher transport capacity to sustain sufficient intake. It could also indicate the presence of other organisms competing for this essential trace element in its ecosystem. The S. eubayanus wild stock whose genome sequence most closely corresponds to the S. eubayanus subgenome of S. pastorianus originates from the Tibetan plateau of the Himalayas (35). However, the first S. cerevisiae ϫ S. eubayanus hybrid, from which current lager yeasts evolved by centuries of domestication, likely originates from a region between Bavaria and Bohemia in Central Europe. So far, European S. eubayanus isolates have not been reported. This may indicate that the original hybridization event occurred elsewhere or that the ancestral European lineage became extinct. The recent detection, in a metagenomics analysis of samples from the Italian Alps, of internal transcribed spacer 1 (ITS1) sequences corresponding to S. eubayanus could indicate that a wild European lineage exists after all (49).
Functional characterization by heterologous complementation of an S. cerevisiae mutant strain established that the SeAgt1 transporters from the Himalayan S. eubayanus strains CDFM21L.1 and ABFM5L supported uptake of maltose and maltotriose. After it was shown that these strains also encoded a functional maltase gene, their inability to grow on maltose and maltotriose was attributed to an inability to transcriptionally upregulate maltose metabolism genes likely caused by regulator loss of function. In S. cerevisiae and, to some extent, in S. eubayanus strains of the Patagonian B subclade such as CBS 12357 T (13,24,50), MAL loci exhibit a specific organization in which a transporter (MALT) and a hydrolase (MALS) gene are expressed from the same bidirectional promoter and are located adjacent to a regulator gene (MALR) (19). In contrast, of the seven genomic regions harboring MAL genes in the two Asian S. eubayanus strains, none showed this canonical organization (Fig. 2), and the subtelomeric regions carrying SeAGT1 did not harbor sequences similar to hydrolase or regulator genes. Subtelomeric regions harboring the other MAL genes indicated intensive reorganization as a result of recombination. In particular, subtelomeric regions on ChrII, ChrV, and ChrXII provide clear indications for recombination events that scattered genes from ancestral MAL1 and MAL2 loci over several chromosomes. A similar interpretation could explain the reorganization MAL3 on ChrXIII (Fig. 2). Similar events may have contributed to loss of function of the MAL regulators (MalR), as exemplified by the occurrence of a nonsynonymous mutation in SeMALR1 resulting in loss of function. These rearrangements did not, however, inactivate the cis-regulatory sequences of the MAL genes since complementation with a functional ScMAL13 allele caused induction of most SeMAL genes ( Fig. 6 and 7B) and, thereby, the heterotic maltotriose-positive phenotype of the hybrid strain HTSH020. Together with the high copy number of SeAGT1, this heterotic complementation may have been the main driver for colonization of lowtemperature brewing processes by the early hybrid ancestors of current S. pastorianus strains.
Recent work on adaptation to brewing environments of laboratory S. cerevisiae ϫ S. eubayanus hybrids showed loss of maltotriose utilization during serial transfer in wort (34). A similar loss of maltotriose utilization is frequently encountered in S. cerevisiae ale strains (52), as well as in some Saaz-type S. pastorianus strains (53). This is thus in contrast with retention of a maltotriose assimilation phenotype by Frohberg-type S. pastorianus strains. This may have been facilitated by the occurrence of multiple copies of the SeAGT1 gene in the S. eubayanus ancestor, which could act as a sequence buffer to counteracting adverse effects of gene copy loss. The recent release of the first long-read sequencing assembly of S. pastorianus enabled a precise chromosomal mapping of the maltose metabolism genes (54) and showed that the Frohberg-type S. pastorianus strain CBS 1483 harbored one copy of SeAGT1 on the S. eubayanus ChrXV section (as in CDFM21L.1) of the chimeric chromosome formed from S. eubayanus ChrXV (SeChrXV) and SeChrVIII (54).
Differential retention and loss of maltotriose consumption in S. pastorianus lineages may reflect different brewing process conditions during domestication. In modern brewing processes based on high-gravity wort, cell division is largely constrained to the glucose and maltose phases, which occur before depletion of nitrogen sources (55). It may be envisaged that, in early lager brewing processes, nonstandardized mashing processes generated wort with a higher maltotriose content, which would have allowed for continued yeast growth during the maltotriose consumption phase. During serial transfer on sugar mixtures, the selective advantage of consuming a specific sugar from a mixture correlates with the number of generations on that sugar during each cycle (56,57). Such conditions would therefore have conferred a significant selective advantage to a maltotriose-assimilating S. cerevisiae ϫ S. eubayanus hybrid, especially if, similar to current ale yeasts, the S. cerevisiae parent was unable to ferment maltotriose.
The heterotic phenotype that was reconstructed in the interspecies S. cerevisiae ϫ S. eubayanus hybrid HTSH020 resulted from combination of dominant and recessive genetic variations from both parental genomes. S. eubayanus contributed the SeAGT1 gene and its functional cis-regulatory sequences but also harbored recessive mutations in MALR genes that allowed full expression of the heterotic phenotype. These mutations were complemented with a set of S. cerevisiae genes including a functional MALR and an absence of the ScAGT1 gene to match the mutations found in S. pastorianus (2). Although some S. pastorianus strains harbor an additional maltotriose transporter encoded by SpMTT1 (28), this gene was recently proposed to have emerged after the original hybridization event as a result of repeated recombination between MALT genes from both subgenomes (32). It is worthwhile mentioning that this hypothesis would also stand if the parental S. cerevisiae was carrying a functional AGT1 gene, as do about 60% of S. cerevisiae (ale) strains (52). The gene could have been mutagenized and lost its function through domestication (34).
Maltotriose fermentation is likely not the only heterotic phenotype of S. pastorianus strains. Flocculation and formation of complex aroma profiles (26,58) are phenotypes that are not fully understood and difficult to reproduce and also might result from heterosis (34).
Laboratory-made S. cerevisiae ϫ S. eubayanus hybrids hold great potential for brewing process intensification and for increasing product diversity. In addition to increasing our understanding of the evolutionary history of lager yeast genomes, this study has implications for the design of new hybrids. Hitherto, laboratory crosses of S. cerevisiae ϫ S. eubayanus strains were designed based on a combination of dominant traits of the parental strains. Our results show that recessive traits can be just as important as contributors to the genetic diversity of such hybrids.

MATERIALS AND METHODS
Strains and maintenance. All strains used in this study are listed in Table 1. Stock cultures of S. eubayanus and S. cerevisiae strains were grown in YPD medium (10 g liter Ϫ1 yeast extract, 20 g liter Ϫ1 peptone, and 20 g liter Ϫ1 glucose) until late exponential phase, complemented with sterile glycerol to a final concentration of 30% (vol/vol) and stored at -80°C as 1-ml aliquots until further use.
Media and cultivation. S. eubayanus batch cultures were grown on synthetic medium (SM) containing 3.0 g liter Ϫ1 KH 2 PO 4 , 5.0 g liter Ϫ1 (NH 4 ) 2 SO 4 , 0.5 g liter Ϫ1 MgSO 4 ·7H 2 O, 1 ml liter Ϫ1 trace element solution, and 1 ml liter Ϫ1 vitamin solution (59). The pH was set to 6.0 with 2 M KOH prior to autoclaving at 120°C for 20 min. Vitamin solutions were sterilized by filtration and added to the sterile medium. Concentrated sugar solutions were autoclaved at 110°C for 20 min or filter sterilized and added to the sterile flasks to give a final concentration of 20 g liter Ϫ1 glucose (SMG), maltose (SMM), or maltotriose (SMMt). With the exception of IMZ752 and IMZ753, S. cerevisiae batch cultures were grown on SM supplemented with 150 mg liter Ϫ1 uracil (60) to compensate for loss of plasmid pUDC156 ( Table 2)  S. eubayanus strains transformed with plasmid pUDP052 carrying a guide RNA targeting SeSGA1 [(gRNA SeSGA1 )], pUDP091 (gRNA SeMALT1 ), and pUDP090 (gRNA SeAGT1 ) were selected on modified SMG medium in which (NH 4 ) 2 SO 4 was replaced by 6.6 g liter Ϫ1 K 2 SO 4 and 10 mM acetamide (SM Ace G) (61). SM-based solid medium contained 2% Bacto agar (BD Biosciences, Franklin Lakes, NJ). S. cerevisiae strains expressing either SeMALT, SeMALS, or ScMALR were selected on SM Ace G. For plasmid propagation, Escherichia coli XL1 Blue-derived strains (Agilent Technologies, Santa Clara, CA) were grown in lysogeny broth (LB) medium (10 g liter Ϫ1 tryptone, 5 g liter Ϫ1 yeast extract, 5 g liter Ϫ1 NaCl) supplied with 100 mg liter Ϫ1 ampicillin. Synthetic wort medium (SWM) for growth studies contained 14.4 g liter Ϫ1 glucose, 2.3 g liter Ϫ1 fructose, 85.9 g liter Ϫ1 maltose, 26.8 g liter Ϫ1 maltotriose, 5 g liter Ϫ1 (NH 4 ) 2 SO 4 , 3 g liter Ϫ1 KH 2 PO 4 , 0.5 g liter Ϫ1 MgSO 4 ·7H 2 O, 1 ml liter Ϫ1 trace element solution, and 1 ml liter Ϫ1 vitamin solution, supplemented with the anaerobic growth factors ergosterol and Tween 80 (0.01 g liter Ϫ1 and 0.42 g liter Ϫ1 respectively), as previously described (59).  Bottle caps were equipped with a 0.5-by 16-mm Microlance needle (BD Biosciences) and sealed with cotton to prevent pressure buildup. Sampling was performed aseptically with 3.5-ml syringes using an 0.8-by 50-mm Microlance needle (BD Biosciences). Microaerobic cultures were inoculated at an OD 660 of 0.1 from stationary-phase precultures in 50 ml of Bio-One Cellstar Cellreactor tubes (Sigma-Aldrich) containing 30 ml of the same medium and grown for 4 days at 12°C. Bottles were incubated at 12°C and shaken at 200 rpm in a New Brunswick Innova 43/43R shaker (Eppendorf Nederland B.V.). At regular intervals, 3.5-ml samples were collected in deep 24-well plates (EnzyScreen BV, Heemstede, The Netherlands) using a LiHa liquid handler (Tecan, Männedorf, Switzerland) to measure the OD 660 and external metabolites. Thirty microliters of each sample was diluted 5-fold in demineralized water in a 96-well plate, and the OD 660 was measured with a Magellan Infinite 200 Pro spectrophotometer (Tecan). From the remaining sample, 150 l was vacuum filter sterilized using 0.2-m-pore-size multiscreen filter plates (Merck, Darmstadt, Germany) for high-pressure liquid chromatography (HPLC) measurements.
Analytical methods. Optical densities of yeast cultures were measured with a Libra S11 spectrophotometer (Biochrom, Cambridge, United Kingdom) at a wavelength of 660 nm. Biomass dry weight was measured by filtering 10-ml culture samples over preweighed nitrocellulose filters with a pore size of 0.45 m. Filters were washed with 10 ml of water, dried in a microwave oven (20 min at 350 W), and reweighed. Sugars were measured using an Agilent Infinity 1260 series high-pressure liquid chromatograph (Agilent Technologies) using a Bio-Rad Aminex HPX-87H column at 65°C with 5 mM sulfuric acid at a flow rate of 0.8 ml min Ϫ1 . Compounds were measured using a refractive index detector (RID) at 35°C. Samples were centrifuged at 13,000 ϫ g for 5 min to collect supernatant or filter sterilized with a 0.2-m-pore-size filter before analysis.
Strain construction. S. cerevisiae IMZ616, which cannot grow on ␣-glucosides (41), was used as a host to test functionality of individual S. eubayanus (putative) maltose transporter genes (13). S. cerevisiae IMX1702 was constructed by integrating ScTDH3 p -ScMAL12-ScADH1 t and ScTEF1 p -SeMALT1-ScCYC1 t at the ScSGA1 locus of strain IMZ616. A fragment containing the ScTDH3 p -ScMAL12-ScADH1 t transcriptional unit was PCR amplified using Phusion High-Fidelity DNA polymerase (Thermo Scientific) from pUDE044 with the primer pair 9596/9355, which included a 5= extension homologous to the upstream region of the ScSGA1 locus and an extension homologous to the cotransformed transporter fragment, respectively. The DNA fragment carrying the S. eubayanus SeMALT1 maltose symporter (ScTEF1 p -SeMALT1-ScCYC1 t ) was PCR amplified from pUD794 using the primer pair 9036/9039, which included a 5= extension homologous to the cotransformed transporter fragment and an extension homologous to the downstream region of the ScSGA1 locus, respectively. To facilitate integration in strain IMZ616, the two PCR fragments were cotransformed with plasmid pUDR119 (amdS), which expressed a gRNA targeting ScSGA1 (spacer sequence: 5=-ATTGACCACTGGAATTCTTC-3=) (68). The plasmid and repair fragments were transformed using the LiAc yeast transformation protocol (69), and transformed cells were plated on SM Ace G. Correct integration was verified by diagnostic PCR with the primers pair 4226/4224. Strains S. cerevisiae IMX1704, IMX1706, and IMX1708 were constructed following the same principle, but instead of using pUD794 to generate the transporter fragment, pUD795, pUD796, and pUD797 were used to PCR amplify ScTEF1 p -SeMALT2-ScCYC1 t , ScTEF1 p -SeMALT3-ScCYC1 t , and ScTEF1 p -SeAGT1-ScCYC1 t respectively. IMX1313 was constructed in a similar way using only ScTEF1 p -ScMAL31-ScCYC1 t amplified with the primer pair 9036/ 11018 which contain 5= and 3= extensions homologous to the upstream and downstream regions of the ScSGA1 locus. Correct integration was verified by diagnostic PCR with the primer pair 4226/4224 (see Fig.  S3 in the supplemental material). All PCR-amplified genes were Sanger sequenced (BaseClear). IMX1313 was grown on YPD medium to lose pUDR119 (URA3) and pUDC156 (amdS). An isolate unable to grown on SMG without uracil and with acetamide was selected and named IMX1313Δ. This strain was able to grow on SMG supplemented with 150 mg liter Ϫ1 uracil.
S. eubayanus IMK820 (SemalT1⌬) was constructed by transforming CDFM21L.1 with 200 ng of pUDP091 and 1 g of a 120-bp repair fragment obtained by mixing an equimolar amount of primers 12442/12443, as previously described (44). As a control, the same transformation was performed without including the repair DNA fragment. Transformants were selected on SM Ace G plates. S. eubayanus IMK823 (Seagt1⌬) was constructed similarly, using pUDP090 and the primer pair 11320/11321. Deletion of SemalT1 was verified by PCR with the primer pair 11671/11672 and Sanger sequencing. The Seagt1 deletion was verified by PCR using the primer pair 12273/12274 and by Illumina whole-genome sequencing and read alignment to the reference genome of CDFM21L.1.
Strains IMX1765, IMX1939, and IMX1940 were constructed by inserting ScPGK1 p -ScMAL13-ScTEF2 t at the SeSGA1 locus of CDFM21L.1, IMK820, and IMK823, respectively. A repair fragment containing ScPGK1 p -ScMAL13-ScTEF2 t was amplified from pUDE780 with the primer pair 12917/12918. Strains CDFM21L.1, IMK820, and IMK823 were transformed by electroporation by addition of 350 ng of repair fragment and 560 ng of pUDP052 (amdS) into the cells as previously described (44). Transformants were plated on SM Ace G and incubated at 20°C. IMX1762 was constructed similarly using a repair fragment with ScTDH3 p -ScMAL12-ScADH1 t amplified from pUDE044 with the primer pair 12319/12320. Strains were verified by PCR using the primer pair 12635/12636 and Sanger sequencing.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.