Saccharomyces arboricola and Its Hybrids’ Propensity for Sake Production: Interspeciﬁc Hybrids Reveal Increased Fermentation Abilities and a Mosaic Metabolic Proﬁle

: The use of interspeciﬁc hybrids during the industrial fermentation process has been well established, positioning the frontier of advancement in brewing to capitalize on the potential of Saccharomyces hybridization. Interspeciﬁc yeast hybrids used in modern monoculture inoculations beneﬁt from a wide range of volatile metabolites that broaden the organoleptic complexity. This is the ﬁrst report of sake brewing by Saccharomyces arboricola and its hybrids. S. arboricola x S. cerevisiae direct-mating generated cryotolerant interspeciﬁc hybrids which increased yields of ethanol and ethyl hexanoate compared to parental strains, important ﬂavor attributes of ﬁne Japanese ginjo sake rice wine. Hierarchical clustering heatmapping with principal component analysis for metabolic proﬁling was used in ﬁnding low levels of endogenous amino / organic acids clustered S. arboricola apart from the S. cerevisiae industrial strains. In sake fermentations, hybrid strains showed a mosaic proﬁle of parental strains, while metabolic analysis suggested S. arboricola had a lower amino acid net uptake than S. cerevisiae. Additionally, this research found an increase in ethanolic fermentation from pyruvate and increased sulfur metabolism. Together, these results suggest S. arboricola is poised for in-depth metabolomic exploration in sake fermentation. normalization by internal reference feature, cubed root transformation, pareto / range data scaling for endogenous / sake metabolites, clustering by Ward’s method, and Euclidean distance measurements. the


Introduction
As currently defined, the budding yeast clade known as Saccharomyces has eight members: S. cerevisiae, S. paradoxus (syn. S. cerevisiae var. tetraspora, S. cerevisiae var. terrestris), S. uvarum (syn. S. bayanus var. uvarum), S. mikatae, S. kudriavzevii, S. arboricola (syn. S. arboricolus), S. eubayanus, and S. jurei [1][2][3][4][5][6][7][8][9]. Additionally, S. pastorianus (syn. S. carlsbergensis, S. monacensis) and S. bayanus are two natural hybrids found in this group [10][11][12]. S. arboricola was first isolated from the tree bark of the Fagaceea family in the west and southwest of China [2] in 2008 and later also isolated in western Taiwan [5] and northern New Zealand [13]. The New Zealand isolate was reported to be highly divergent from the Chinese isolates used in this study, supporting the far East Asian origin theory for Saccharomyces yeast. The Chinese S. arboricola has a high quality annotated genomic sequence that when compared to S. cerevisiae shows a reciprocal translocation on chromosome (chr.) IV and XIII, inversion on chr.VI, high presence of non-syntenic genes near telomeric regions, lack of introgression, and four novel genes [14]. Between S. cerevisiae and other Saccharomyces members there is a 15-30% divergence, but the divergence between S. cerevisiae strains is 0.1-0.6% [15,16]. Additionally, subpopulations of S. paradoxus show a 1.5-4.5% divergence, while the divergence between S. eubayanus and S. uvarum is 6% [7,17]. The nucleotide divergence of yeast within Saccharomyces is near the man to bird comparison, but with the advantage of a physiological and structural similarity of man to chimp [18]. Favorable properties of different Saccharomyces yeast can be brought together by hybridization that often results in transgressive phenotypes. Typical environmentally isolated Saccharomyces strains are diploid while laboratory strains are often used in the haploid form to reduce variability and industrial yeast are frequently found to be polyploidy, potentially a result of favored fermentation environment. Although interspecies hybrids of the Saccharomyces genus are viable, the they do not produce viable gametes or ascospores in yeast resulting in reproductive isolation as a sterility barrier [9]. Hybridization of phenotypically diverse yeast has been exploited for the improvement of industrial strains for specific purposes.
Fermenting yeasts, specifically S. cerevisiae, has been unwittingly domesticated throughout humanity's existence via the production of various foods and beverages, including bread, beer, wine, and sake. One factor in S. cerevisiae selection is its unique metabolic capabilities as a Crabtree positive yeast; they amass ethanol in the presence of oxygen. The commonly preferred eukaryotic metabolic pathway in the presence of oxygen is oxidative respiration for its high ATP yield. But in the high concentrations of six-carbon carbohydrates, such as glucose, this pathway is repressed by the Crabtree effect. Under these conditions, the energy for growth comes from the glycolysis pathway until exhaustion of six-carbon molecules causes a diauxic shift to oxidation of the two-carbon ethanol molecules into CO 2 [19,20]. Crabtree positive microbes are beneficial in the production of alcoholic beverages which have been produced for at least 9000 years, and the earliest known recording of Japanese sake is mentioned ca. 285 AD in an ancient Chinese document called Gishiwajinden [21][22][23]. Sake is the alcoholic drink generated by fermenting rice and beer is that of barley, making their development specific to regions dominated by these grains. Rice and barley diverged 50 million years ago and rice was introduced to Japan from the Chinese Yangzi region around 400 BC, many differences exist between these two grains [24,25]. During the 15th century in Bavaria, lager-brewing arose [26] and is the most popular alcoholic beverage in the world today. Since the lager beer yeast, S. pastorianus was discovered to be a hybrid of S. eubayanus x S. cerevisiae, interest has sparked in the use of interspecific hybrids in fermentation [27]. Originally the S. eubayanus parent was identified in Patagonia and the question of lager yeast's origin has generated controversy [28][29][30][31]. Yeasts originated multiple times from various fungal lineages of which one gave rise to the budding yeasts [32]. The genomic diversity with geographical boundaries and many other lines of evidence support a Far East Asia origin for Saccharomyces [33][34][35].
Beer, wine, and sake are complex chemical matrixes resulting from many metabolic and chemical reactions in which raw ingredients determine initial sugar, assimilable nitrogen, and fatty acid profiles [25,36]. The volatile metabolic profile obtained during fermentation of different carbohydrates is reflective of the raw ingredients composition [37,38]. Wine is perhaps one of the oldest biotechnologies as yeast are often found on damaged grapes and crushing of the fruit facilitates the release of the fermentable sugars [39,40]. Obtaining fermentable sugars for beer and sake is quite different. Production of beer involves a saccharification of malted barley by the grain's own enzymes to convert their starches to maltose, glucose, and maltotriose found in descending concentrations. Boiling of the wort results in less than 3% of the malt lipids available to yeast during fermentation [41]. Fermentation of green beer is accomplished in 3 days to 10 days [42] and results in an average alcohol content of 5% (vol/vol) with some brews tipping the scales around 10% (vol/vol) or more.
Immature sake is generally produced over 41 days and involves several steamed rice additions at temperatures between 12-20 • C applying high osmotic pressure on the yeast from the rice mash ratios [43]. The highest alcohol yield in the world, 17-20% (vol/vol) without distillation, is obtained through this process [25]. Aspergillus oryzae, colloquially known as koji, is a keystone organism in the production of sake from polished rice [44]. By traditional Japanese sake mash methods, koji performs a continuous saccharification of highly milled rice starches to the fermentable sugars and the budding yeast, S. cerevisiae, ferments these sugars into ethanol [45]. Free sugars such as glucose and oligomers are produced by the koji enzymes glucoamylase and α-amylase during the sake mash [23]. Recent work has shown that koji secretes mannitol 2-dehydrogenase which converts fructose to mannitol in the sake mash. Interestingly, S. arboricola efficiently catabolizes mannitol, an abundant natural energy storage molecule that Aspergillus sp. produce in high concentrations [46,47]. This phenotype is superior to S. cerevisiae metabolism of mannitol [14]. Considering the temperature tolerance and growth capabilities on mannitol, it was hypothesized that S. arboricola and interspecific hybrids are uniquely fit for sake fermentation with A. oryzae because of their metabolic access to the underutilized sugar-alcohol and temperature preference which affects the flavor active metabolites.
The aim of this study was to characterize the metabolic capabilities of S. arboricola and its hybrids with S. cerevisiae in Japanese rice wine. For the first time, sake fermentation profiling of S. arboricola and newly generated S. arboricola x S. cerevisiae interspecific hybrids is reported. The wine and beer brewing industries have often employed interspecific hybrids for their fermentations. S. kudriavzevii x S. cerevisiae are known to ferment wines, S. mikatae x S. cerevisiae hybrids have been actively produced for wine, and the S. cerevisiae x S. eubayanus hybrid produces the most popular beer in the world [33,48,49]. Limited research has been conducted on the metabolic capabilities of S. arboricola and these results suggest S. arboricola and the hybrids' propensity for fermentation may be suited for sake production.

Yeast Isolation
Saccharomyces cerevisiae HT01 was the only strain isolated for this study and was isolated as described previously [54] with slight modification. Briefly, beer was harvested from a can sanitized on the outside with 70% ethanol into an enrichment media (3 g yeast extract, 3 g malt extract, 5 g peptone, 10 g sucrose, 76 mL EtOH, 1 mg chloramphenicol, and 1 mL of 1-M HCl per liter). Samples were transferred to sterile tubes and incubated until signs of yeast appeared. Cultures were then streaked onto YPD media and single colonies were grown and stocked. From these stocks, the yeast culture's genomic DNA was extracted, the ITS region was amplified via polymerase chain reaction (PCR) [55], Sanger sequenced at the West Virginia University Genomic Core, and the isolated yeast was confirmed to be S. cerevisiae via utilizing NCBI Blast.

Serial Spot Dilutions
Serial spot dilution growth assays were conducted as described previously [56] with some modification. Briefly, OD 600nm growth readings were taken from fresh, overnight cultures of each yeast grown at 23 • C. For each yeast assayed, a total of 3.2 × 10 5 yeast cells were transferred to a well in the first column of a 96 well plate. Each yeast assayed was serially diluted 4 times in a 1:10 ratio, using sterile distilled and MilliQ (MQ) filtered water, across the rows which resulted in 5 wells each. A handheld 48 pin plate was used to place the serial dilution spots onto solid YPD agar media. Plates were then incubated at 4 • C for ten days, 10 • C for seven days, and 15 • C for four days. Growth was assessed and photographed.

Sporulation and Spore Staining
Fresh liquid yeast cultures were grown in YPD (1% yeast extract, 2% peptone, 2% dextrose) overnight at 23 • C in a shaking water bath. Isolates were streaked out onto YEP(EG) buffered agar plates (1% succinic acid, 1% yeast extract, 2% peptone, 2% glycerol, 2% agar, potassium hydroxide to pH 5.5, 2.5% ethanol) for 2-3 days. Single colonies were thinly patched onto Snack-Spo Zn agar plates (1% potassium acetate, 0.1% yeast extract, 0.05% dextrose, 0.005% zinc acetate, 2% agar) and incubated for 6 days. To measure spore formation, a modified Schaeffer-Fulton endospore staining technique was used. Briefly, cultures from the sporulation plates were suspended in 1ml distilled water and recovered by pipette. The suspended culture was fixed to a glass microscope slide via flame. Malachite green (10%, 0.1 g/mL) was then used to stain the spore walls for 15 min (minute/s) before decoloring and rinsing with water. Safranin (0.25%, 2.5 mg/mL) was used as a counterstain for non-sporulated cells for 1 min before rinsing with water and mounting in glycerol. Malachite green cells were counted as spores and safarin red cells were counted as non-sporulated yeast cells [57]. Images were taken under a light microscope and sporulation efficiency for each strain was calculated, reported in the results section, to select the best possibility of a successful hybridization.

Interspecific Hybridization
The ascus of the spore was digested over 1 hr. with Zymolase (25 mg/mL) at 37 • C. The digested spores were placed dropwise on YPD agar at an angle to facilitate a central streak of liquid across the plate. Micromanipulation of spores from S. arboricola strain 350 and S. cerevisiae M22 spores was performed by fiber optic needle under a light microscope using the SporePlay (Singer Instruments, Roadwater, England) dissection microscope by Singer Instruments and yielded 40 spore to spore mattings. Visual observations were conducted on the shmooing process during incubation at 25 • C. Single colonies were selected, streaked out for purification, and stocked at −80 • C with cryogenic media (1% yeast extract, 2% peptone, 15% glycerol).

Genomic DNA Extraction
Alkaline lysis genomic DNA extraction was performed on fresh, overnight cultures by harvesting (0.3 × 10 7 ) cells via cold centrifugation at 7500× g for 3 min. Cells were washed twice with phosphate-buffered solution (PBS) before NaOH (0.02 M) addition and subsequent incubation at 100 • C for 10 min. The cellular solution underwent cold centrifugation at 4 • C for 3 min before 250 µL of supernatant was mixed with 50 µL acetic acid (2.5 M) and then 300 µL of 2-propanol. Cold centrifugation was performed again for 15 min before the supernatant was discarded. Ethanol (80% v/v) was added to the DNA pellet to lightly mix the solution before a 5 min cold centrifugation. The DNA was dried at room temperature and resuspended in TE buffer.

Polymerase Chain Reaction
DNA concentrations were measured by OD 260nm via a glass black cell cuvette. PCR confirmed S. arboricola 350 and S. cerevisiae M22 (SAM hybrids) interspecific hybridization. Two multiplexed species-specific primer sets were used, detailed previously [58]. The oligonucleotide sequences used for primers were as follows; for S. arboricola sarbF1 (GGC ACG CCC TTA CAG CAG CAA TCG) and sarbR2 (TCG TCG TAC AGA TGC TGG TAG GGC GCT); for S. cerevisiae scerF2 (GCG CTT TAC ATT CAG ATC CCG AG) and scerR2 (TAA GTT GGT TGT CAG CAA GAT TG ATC). A master mix for the PCR reactions was made on ice and consisted of PCR reagents from kit: KOD-Plus-Neo (product code KOD-401) ordered from TOYOBO, Japan. The PCR Touchdown [59] thermocycle protocol was used. During gel electrophoresis [60], the S. cerevisiae band generated at 150bp (base pairs) and S. arboricola generated at 349 bp.

Sake Fermentations
Freshly grown cells (2.0 × 10 9 ) were measured by OD 600nm , harvested from YPD via centrifugation at 3200× g for 3 min, and washed in MilliQ water twice. The harvested yeast, rice, koji, lactate, and water were mixed together in mason jars covered with loose lids. Incubation occurred at 15 • C for 2 weeks with an initial remixing of the mash occurring after the first 24 h. Fermentation was monitored daily by mass reduction of the sake mash, which corresponded to CO 2 production. Fermentation trials were performed in triplicate. Post-fermentation, the sake mash was harvested into conical tubes by centrifuged at 4 • C and 3200× g. The supernatant was transferred to storage at 4 • C for further analysis. Statistical analysis of ethanol content was validated by one-way ANOVA with a two-tailed student's t-test post-hoc analysis and ordered difference report for each day of fermentation.

Gas Chromatography-Flame Ionization Detection
For gas chromatography-flame ionization detection (GC-FID) of the sake supernatant, ribitol (syn. adonitol) (0.2 mg/mL) was supplied as an internal standard. Samples were freeze-dried via liquid nitrogen bath followed by vacuum chamber drying for 6 h. The freeze-dried samples were mixed with 100 µL of methoxylamine diluted with pyridine (20 mg/mL] and incubated at 30 • C for 90 min with agitation (1500 rpm). The sample was mixed with 50 µL of N-methyl-N-TMS-trifluoroacetamide (MSTFA) incubated for 30 min at 37 • C with agitation and transferred into a glass vial for chromatography metabolome analysis. A Cp-sil8CB column from Agilent, model #CP8751 (30 mm, 0.25 mm, 0.25 µm) was used to assess the sake fermentation samples. The injection was at 230 • C with a 1 mL/min helium carrier gas flow rate. The column temperature was held at 80 • C for 2 min before rising 10 • C/min until 320 • C was achieved, the temperature was held for 12 min before reaching a 320 • C detector. The Shimadzu Corporation (Shimadzu Manufacturing Company, Kyoto, Japan) model GC-2014AFSC 100 V was used in performing the GC-FID analysis.

Head Space-Solid Phase Microextraction
For the head space-solid phase microextraction (HS-SPME), the sake supernatant samples were prepared with NaCl and methyl hexanoate (40 mg/100 mL) was supplied as an internal standard. In a glass vial, the solution was agitated under 80 • C heat with a stir bar (160 rpm) for 15 min. A syringe was inserted into the vial to expose a SPME fiber (100 µm PDMS) for 15 min. The fiber was retracted into the syringe before removal from the vial and inserted into the injector for a 5 min exposure. A DB wax column from Agilent, model #122-7032 (30 mm, 0.25 mm, 0.25 µm) was used. The injection was at 200 • C with a 0.95 mL/min helium carrier gas flow rate. The column temperature was held at 60 • C for 1 min before rising 3 • C/min until 100 • C was achieved, the rate increased to 15 • C/min until 200 • C was reached and held for 4 min before reaching a 320 • C detector. The Shimadzu Corporation (Shimadzu Manufacturing Company, Kyoto, Japan) model GC-2014AFSC 100 V was used in performing the HS-SPME analysis.

Gas Chromatography-Mass Spectroscopy
For gas chromatography-mass spectroscopy (GC-MS) of endogenous metabolites, stationary phase yeast cultures were grown in YPD liquid media. Stationary phase was defined at an OD 600nm of 2.0. Cells (4 × 10 7 ) were harvested, washed, and freeze-dried. An extraction solution (chloroform:methanol:MQ [2:5:2]) with ribitol (0.2 mg/mL) as the internal standard was added to the freeze-dried samples and incubated for 30 min at 37 • C under agitation (1500 rpm). The samples were centrifuged, and the supernatant was harvested. MQ water was added to the supernatant and centrifuged. The supernatant was concentrated under vacuum centrifugation (−100 kPa) at 37 • C. MQ water was added, and the water-soluble metabolites were freeze-dried. Methoxyamine (20 mg/mL) was added to the samples and incubated for 90 min at 30 • C under agitation (1500 rpm). MSTFA (N-methyl-N-trimethylsilyl trifluoroacetoamide) was added to the samples and incubated for 30 min at 37 • C with agitation (1500 rpm). Samples were supplied in glass vials for GC-MS analysis using an Agilent DB-5MS+DG (30 m, 0.25 mm, 0.25 um) column, splitless injection, and helium carrier gas (1 mL/min). The column temperature rose from 80 • C at a rate of 13 • C/min until reaching a 320 • C detector. The Agilent Technologies, Inc. (Agilent Technologies, Santa Clara, California, United States of America) model 7890 B was used for the GC system, model 5977 A was used for the MS system, and the National Institute for Standards and Technology (NIST) mass spectral library (United States Department of Commerce, Gaithersburg, MD, USA) was used.

Data Analysis
The peak area of identified compounds was divided by the peak area of the internal standard for normalization of the data. Normalized peak areas of the compounds and ethanol concentrations measured during sake fermentations were auto-scaled and used as independent variables in the loading plots for construction of the principal component analysis (PCA). Statistical analysis was performed by soft independent modeling by class analogy (SIMCA) methodology by SIMCA version 13.0.3.0 (Umetrics, Umeå, Sweden) computer software. Analysis of variance (ANOVA), post-hoc students t-test, and post hoc Tukey-Kramer honest significance test (Tukey HSD), p < 0.05 was used for statistical significance. Deviations from this statistical analysis procedure, if occurred, was reported in individual results or methods sections. Heatmap was generated through MetaboAnalyst 4.0 [61]. Heatmaps were based on data normalization by internal reference feature, cubed root transformation, pareto/range data scaling for endogenous/sake metabolites, clustering by Ward's method, and Euclidean distance measurements.

Results
Endogenous metabolome characterization of S. arboricola was put in context of industrially relevant S. cerevisiae yeast. Two yeast strains were chosen to represent either beer, wine, or sake fermentation industries. WLP001 and HT01 represented beer, M22 and DBVPG6044 represented wine, K7 and TCR7 represented sake production. SA350, SA351, and SA352 were used to represent wild S. arboricola (Table 1). Endogenous metabolite profiling was visualized in a hierarchical clustering heatmap that showed clustering of yeast by industry and species ( Figure 1A). Ward's linkage clustering algorithm created two main clusters separated by species with the exception of West African bili wine strain DBVPG6044 that grouped within S. arboricola. The yeast in this cluster, located in the right half of the heatmap, all showed a low concentration of endogenous amino acids. Beer strains, sake strains, and wine strain M22 composed the other cluster on the left half. Beer strains paired together, and sake strains paired together with Italian wine strain M22. These industrial strains had similar but varying metabolic profiles displaying high endogenous levels of many amino and organic acids. paired together, and sake strains paired together with Italian wine strain M22. These industrial strains had similar but varying metabolic profiles displaying high endogenous levels of many amino and organic acids. A score plot generated through principal component analysis separated the strains into general industry-specific groups ( Figure 1B). A total of 19 metabolites identified by GC-MS were used as variables in the construction of the loading plot ( Figure 1C). Sugar utilization played a major role in separation of the strains. S. arboricola strains were grouped to the left while beer and sake strains grouped opposite to the right on principal component (PC1) explaining 36.2% of the sample variance and separating the groups by species. On the axis of PC2, sake yeast grouped high, wine yeasts were neutral/variable, and beer yeasts were low explaining 18.0% of sample variance and separating the  A score plot generated through principal component analysis separated the strains into general industry-specific groups ( Figure 1B). A total of 19 metabolites identified by GC-MS were used as variables in the construction of the loading plot ( Figure 1C). Sugar utilization played a major role in separation of the strains. S. arboricola strains were grouped to the left while beer and sake strains grouped opposite to the right on principal component (PC1) explaining 36.2% of the sample variance and separating the groups by species. On the axis of PC2, sake yeast grouped high, wine yeasts were neutral/variable, and beer yeasts were low explaining 18.0% of sample variance and separating the yeast by their industrial niche domestication. Glucose, sucrose, and trehalose negatively contributed to PC1, which exemplified sugar uptake differences and was inferred to translate to fermentation aptitude. Beer brewing strains were found to be the most inclined yeasts in this regard and S. arboricola were found to be the least adept. A third component, explaining 12.7% was used to create a 3D PCA plot of the endogenous metabolites ( Figure S1). Citric acid, malic acid, fumaric acid, glutamic acid, nicotinic acid, succinic acid, aspartic acid, threonine, serine, glycine, phenylalanine, pyroglutamine, methionine, uracil, inositol, and adenine positively contributed to PC1. A loading matrix was constructed for the variables (Table S1). Glycine, threonine, malic acid, fumaric acid, nicotinic acid, succinic acid, sucrose, trehalose, inositol, and adenine positively contributed to PC2. Phenylalanine, pyroglutamine, methionine, uracil, aspartic acid, and glucose contributed negatively to PC2, which distinguished the beer brewing strains from the sake strains. These two groups are represented in the heatmap's industrial cluster. Factors from the heatmap show the difference between sake and beer yeast include the endogenous inositol levels; which were high in sake yeast and affect membrane fluidity during osmotic stress such as an industrial sake mash.
The potential of Saccharomyces hybridization to broaden metabolic capabilities and sensory complexity from volatile metabolites was explored using S. arboricola and S. cerevisiae. Sporulation efficiency was used in selection of strains for direct-mating hybridization as industrial strains are typically poor sporulators. S. arboricola 350 and S. cerevisiae M22 yielded the highest sporulation efficiency at 69% and 91% respectively ( Figure S2A). Selection of the S. cerevisiae industrial strain relied on sporulation results to optimize spore-spore mattings and given M22 grouped similar to the sake strain metabolism in Figure 1B, M22 was choosen. Other S. arboricola strains sporulated at 68% for 351 and 20% for 352. Interspecific SAM (S. arboricola 350 and S. cerevisiae M22) hybrids, SAM8c and SAM9a, generated double bands representing genomic input from both species. Two spore-to-spore mating colonies were advanced to lab-scale sake fermentation after confirmation of hybrid state by species-specific multiplexed PCR ( Figure S2B). Cryotolerance of S. arboricola and the hybrids was shown as substantial growth at 4 and 10 • C after ten and seven days, respectively (Figure 2). At 15 • C, there was no perceivable difference between yeast growth. Thermosensitivity was not detected using the serial spotting dilution assay in temperatures up to 30 • C, slow growth of S. arboricola cultures during extreme temperature highs in the summer months was observed. nicotinic acid, succinic acid, sucrose, trehalose, inositol, and adenine positively contributed to PC2. Phenylalanine, pyroglutamine, methionine, uracil, aspartic acid, and glucose contributed negatively to PC2, which distinguished the beer brewing strains from the sake strains. These two groups are represented in the heatmap's industrial cluster. Factors from the heatmap show the difference between sake and beer yeast include the endogenous inositol levels; which were high in sake yeast and affect membrane fluidity during osmotic stress such as an industrial sake mash. The potential of Saccharomyces hybridization to broaden metabolic capabilities and sensory complexity from volatile metabolites was explored using S. arboricola and S. cerevisiae. Sporulation efficiency was used in selection of strains for direct-mating hybridization as industrial strains are typically poor sporulators. S. arboricola 350 and S. cerevisiae M22 yielded the highest sporulation efficiency at 69% and 91% respectively ( Figure S2A). Selection of the S. cerevisiae industrial strain relied on sporulation results to optimize spore-spore mattings and given M22 grouped similar to the sake strain metabolism in Figure 1B, M22 was choosen. Other S. arboricola strains sporulated at 68% for 351 and 20% for 352. Interspecific SAM (S. arboricola 350 and S. cerevisiae M22) hybrids, SAM8c and SAM9a, generated double bands representing genomic input from both species. Two spore-tospore mating colonies were advanced to lab-scale sake fermentation after confirmation of hybrid state by species-specific multiplexed PCR ( Figure S2B). Cryotolerance of S. arboricola and the hybrids was shown as substantial growth at 4 and 10 °C after ten and seven days, respectively (Figure 2). At 15 °C, there was no perceivable difference between yeast growth. Thermosensitivity was not detected using the serial spotting dilution assay in temperatures up to 30 °C, slow growth of S. arboricola cultures during extreme temperature highs in the summer months was observed. To determine if important sake flavor molecules changed with hybridization, ethyl hexanoate, isoamyl acetate, and ethanol fermentation was measured. In fact, it was found that S arboricola, S. cerevisiae, and the hybrids metabolite profiles after lab-scale sake mash fermentation trials were different. The hybrids had a more vigorous fermentation, yielding a higher alcohol content than S. cerevisiae by day seven (p = 0.031) of the 14 day trial fermentation ( Figure 3A). A one-way ANOVA with a two-tailed student's t-test post-hoc analysis and ordered difference report for each day of  To determine if important sake flavor molecules changed with hybridization, ethyl hexanoate, isoamyl acetate, and ethanol fermentation was measured. In fact, it was found that S arboricola, S. cerevisiae, and the hybrids metabolite profiles after lab-scale sake mash fermentation trials were different. The hybrids had a more vigorous fermentation, yielding a higher alcohol content than S. cerevisiae by day seven (p = 0.031) of the 14 day trial fermentation ( Figure 3A). A one-way ANOVA with a two-tailed student's t-test post-hoc analysis and ordered difference report for each day of fermentation was generated in determine ethanol content significance (Table S2). The gap between fermentation rates declined over time trending towards non-significance between all samples if more time was allotted. The fermentations were stopped as the fermentation rate slowed down and a fair amount of ethanol 10%+ was produced. The fermentation trials were designed with time constraints as this is often a major factor in production settings and attenuation was secondary. The final alcohol yield at day 14 was significant between S. cerevisiae and the hybrids (p = 0.038), while S. arboricola was not significantly different than either hybrid nor S. cerevisiae. Hybrid SAM9a had the most productive fermentation rate, followed by hybrid SAM8c, S. arboricola 350, and S. cerevisiae M22. The mean final alcohol content of sake produced was 12.92% for the hybrids, 12.15% for S. arboricola 350, and 11.46% for S. cerevisiae M22.
Detection of volatile aromatic compounds allows insight into flavor and sensory attributes of the final sake produced and HS-SPME technology in the gas chromatography system was utilized. Sake fermented with S. arboricola contained significantly less ethyl hexanoate (syn. ethyl caproate) (0.207 mg mL −1 , st error 0.016) than S. cerevisiae (0.316 mg mL −1 , st error 0.019, p = 0.005), or the hybrids (0.308 mg mL −1 , st error 0.014, p = 0.003) ( Figure 3B). Sake fermented by S. arboricola contained a concentration of isoamyl acetate (0.368 mg mL −1 , st error 0.064) that was high compared to S. cerevisiae (0.259 mg mL −1 , st error 0.064, p = 0.25) and the hybrids (0.252 mg mL −1 , st error 0.045, p = 0.17), but fell short of a significant p-value ( Figure 3C). Ethyl hexanoate imparts apple or aniseed and isoamyl acetate imparts a banana-like attribute to human sensory receptors. Whole metabolite heatmap by sample was created showing a separation of S. arboricola from the hybrids and M22 individual samples when volatile compounds were used ( Figure S3A) or a separation of all three when non-volatile metabolites were used ( Figure S3B).
Sake fermentation profiles were visualized in a hierarchical clustering heatmap with hybrids clustering in closer proximity to the S. arboricola yeast ( Figure 4A). Parental strain S. cerevisiae M22 s sake metabolite profile contained high amounts of glucose, malic acid, threonine, and valine while parental S. arboricola's profile contained high amounts of citric acid, aspartic acid, glutamic acid, leucine, isoleucine, alanine, phenylalanine, inositol, and glycerol. Parental profiles almost completely opposed one another with only serine, succinic acid, and urea showing similar parental concentrations. M22 was the most distant parent to the hybrids' profile. The heatmap of the hybrid strains showed a mosaic blending of the two parental species. Metabolomic analysis using principal components clearly separated interspecific hybrids from each parent based in a PCA score plot ( Figure 4B). A total of 20 metabolites, including 17 non-volatile compounds, were identified by GC-FID and 3 volatile compounds identified by HS-SPME These were used as variables in construction of the loading plot ( Figure 4C). Principal component (PC1) separated samples by their species and explained 43.2% of the variance between the samples. Fermentation profiles showed a separation of the two Saccharomyces species on opposite ends of PC1 and the hybrid yeast grouped to the middle. PC2 separated the hybrid from the parental samples by the novel hybrid fermentation features that explained 23.1% of the sample variance. Threonine, valine, malic acid, glucose, ethyl hexanoate, and dimethyl trisulfur (DMTS) positively contributed to PC1. Phenylalanine, isoleucine, leucine, alanine, aspartic acid, citric acid, glutamic acid, succinic acid, inositol, glycerol, ethanol, and isoamyl acetate were amino acids, organic acids, and other molecules that negatively contributed to PC1. Ethanol, DMTS, phenylalanine, ethyl hexanoate, glycerol, isoleucine, glutamic acid, valine, and threonine negatively contributed to PC2. Serine, succinic acid, malic acid, glucose, urea, isoamyl acetate, aspartic acid, inositol, and alanine positively contributed to PC2. A loading matrix was constructed for the variables (Table S3).
acetate imparts a banana-like attribute to human sensory receptors. Whole metabolite heatmap by sample was created showing a separation of S. arboricola from the hybrids and M22 individual samples when volatile compounds were used ( Figure S3A) or a separation of all three when nonvolatile metabolites were used ( Figure S3B).  aspartic acid, citric acid, glutamic acid, succinic acid, inositol, glycerol, ethanol, and isoamyl acetate were amino acids, organic acids, and other molecules that negatively contributed to PC1. Ethanol, DMTS, phenylalanine, ethyl hexanoate, glycerol, isoleucine, glutamic acid, valine, and threonine negatively contributed to PC2. Serine, succinic acid, malic acid, glucose, urea, isoamyl acetate, aspartic acid, inositol, and alanine positively contributed to PC2. A loading matrix was constructed for the variables (Table S3).

Discussion
This is the first report of sake brewing by Saccharomyces arboricola and its hybrids. The exploratory metabolic analysis of S. arboricola, a neglected Saccharomyces yeast lacking metabolic evaluation, adds to the novelty of this study. Through direct mating of S. arboricola to S. cerevisiae, interspecific hybrids gained a higher final ethanol content over S. cerevisiae and a higher ethyl hexanoate over S. arboricola. It was found that the low levels of endogenous amino/organic acids clustered S. arboricola apart from the industrial strains by principal component analysis and hierarchical clustering heatmapping. In sake fermentations, hybrid strains showed a mosaic profile of parental strains except where parental profiles were similar. This metabolic analysis suggested S. arboricola had higher transportation activity of pyruvate to the mitochondria and a lower amino acid net uptake than S. cerevisiae. Additionally, hybrid metabolic analysis suggested an increase in ethanolic fermentation from pyruvate and increased sulfur metabolism. Together, these results suggest S. arboricola is poised for in-depth and focused metabolomic exploration.
Endogenous metabolite comparison of nine industrially relevant and diverse yeast by GC-MS provided a separation of metabolomic profiles where industry and species appeared to be drivers of distinction in both heat map clustering and PCA. The internal composition is characteristic of their metabolic tendencies and influences the environment in which many endogenous metabolic reactions occur. Of the industrial yeast, sake yeast separated from beer yeast by PC2 of the PCA ( Figure 1B) which is characterized by the glucose to fructose ratio, inositol level, as well as the organic and amino acid profile. The endogenous inositol levels were high in sake yeast and affect membrane fluidity during osmotic stress such as an industrial sake mash. Sugar utilization played a major role in separation of the strains as evident in the loading plot which highlighted the importance of sugar metabolism. Metabolic distinctions between the glucose to fructose ratio of beer to sake or wine yeasts can be inferred to resemble the kinetic differences of hexokinase-mediated sugar phosphorylation. The glucose to fructose ratio between wine strains was correlated with the kinetics of hexokinase-mediated sugar phosphorylation and the inherent properties of each unique strain [62,63]. High amounts of endogenous inositol and trehalose were characteristic of S. arboricola. Furthermore, serine and glycine were found to be low in the S. arboricola cluster and are important glycolytic metabolites biosynthesized from 3-phosphoglycerate or alternatively through the gluconeogenic pathway utilizing glyoxylate [64]. Organic and amino acid metabolism shift between alternative biosynthesis pathways and their byproducts likely play a role in the difference seen between strains. Additional genetic and environmental factors are also suspected to play a role in strain metabolomic profiles as seen by increased copy number of MAL (maltose) genes in sake and beer yeast, yet wine strains tend to lose this attribute because their typical industrial fermentation sugar availability lacks maltose [65]. Copy number variations (CNV) are concentrated in the sub telomeric regions, with chr. I, VII, VIII, IX, X, XII, XV, and XVI being more susceptible to variation [65]. Gene ontology related to carbon and nitrogen metabolism, ion transport, and flocculation are prone to CNV [66,67] which often results in deletions rather than amplifications of genetic material [65]. The CNVs are thought to underline niche adaptation and domestication [65,66].
Flavor-active elements formed during yeast fermentation can be divided into five key groups: sulfur-containing molecules, organic acids, higher alcohols, carbonyl compounds, and volatile esters [68][69][70][71]. Of these classifications, volatile esters are the largest and are responsible for the highly desired fruity character in several alcoholic beverages [72]. Production of ethyl hexanoate and isoamyl alcohol is maximized during fermentations at 15 • C [73,74]. Ethyl hexanoate imparts apple or aniseed and isoamyl acetate imparts a banana-like attribute to human sensory receptors. Ethyl hexanoate, a MCFA (medium chain fatty acids) ester with a 0.17-0.21 mg mL −1 flavor threshold, is an important flavor molecule for ginjo sake [68,75]. The fruity ester, formed from hexanoic acid or hexanoyl CoA with ethanol, imparts a light, sweet, apple complexity to the delicate and fragrant beverage. Ginjo sake is produced from rice with the outer 40% of the grain polished away to remove the hull, bran, and part of the starchy endosperm, leaving 60% of the rice to create the sake mash. During the sake fermentation trials, S. arboricola produced significantly less ethyl hexanoate than either the hybrid or S. cerevisiae wine strain M22. Increased inositol concentrations have been shown to decrease production of ethyl hexanoate during fermentation [76]. This relationship was seen in PC1 of the hybrids and suggests S. arboricola's inositol biosynthesis pathway is more active, in turn producing low ethyl hexanoate levels seen in this study. Additionally, S. arboricola's production of isoamyl acetate showed a trend towards a significant increase over the hybrid or S. cerevisiae M22 ( Figure 3C). Isoamyl acetate, an acetate ester with a 0.6-1.2 mg mL −1 flavor threshold, imparts a distinct banana character typical of hefeweizen yeast [68]. Hybrid strains produced a pleasant aromatic profile containing both ethyl hexanoate and isoamyl acetate. Ethanol hexanoyl transferase, Eht1, is believed to catalyze the formation of ethyl hexanoate [68]. Alcohol acetyltransferase genes Atf1 and Atf2 are responsible for the total biosynthesis of isoamyl alcohol and divergence in the Atf proteins has been suggested to influence the whole ester aroma profile of yeast strains [72]. Protein identity between S. arboricola and S. cerevisiae for Atf1 was 81%, Atf2 was 36%, and Eht1 was 84%. Together the aromatic ester profile of S. arboricola was unique in the combination of low ethyl hexanoate and isoamyl alcohol concentrations are speculated to be owing from inositol lipid activity and variation in factors affecting Atf activity. This area of research is merits further investigation.
Interspecific hybrids are commonly employed in agricultural and horticultural settings where hybrid vigor provides an increased or transgressive phenotype [77]. In oenology Saccharomyces hybrids, such as S. cerevisiae x S. kudrazevii, are often employed [78,79]. Many times Saccharomyces hybrids are able to yield more ethanol during fermentation under lab controlled environments [80], and these results support previous findings. The use of S. arboricola in sake production is novel and no other study had S. arboricola's metabolism as the primary focus of the manuscript. To date several labs have investigated Saccharomyces hybrids, one study included S. arboricola in a beer brewing context while one other included S. arboricola among their hybrids in a biofuel context [80,81]. This research showed that in a sake mash, SAM interspecific hybrids produced a more vigorous fermentation in a short amount of time which may be beneficial if used or further developed in a production context. Of the range of temperatures for Saccharomyces growth, S. arboricola is cryotolerant and thermosensitive [2,80,82]. Cryotolerance is a common attribute of yeast hybrids as seen in the lager yeast Saccharomyces pastorianus, which is a hybrid between S. cerevisiae x S. eubayanus [7,[83][84][85]. This phenotype has proven beneficial for the cold fermentations typical of lagering beer or brewing sake. The ability to ferment at colder temperatures allows for more volatile compounds to be retained and a decrease in bitter or acidic taste [86]. The mechanism by which thermal phenotypes are determined has not been fully elucidated, but mitotype, ploidy, and lipid composition are factors [80,[86][87][88][89][90][91]. Lipid profiles with a higher number of MCFA attributes to cryotolerance of S. cerevisiae x S. kudriavzevii hybrids [92]. At colder temperatures, increasing the MCFA composition and reducing saturated and unsaturated fatty acid composition allows increased fluidity of the membranes.
Previous reports show that wine yeast produce lower amino acidity and less succinic acid, but more malic acid than sake yeast [93]. Yeast strain has a major impact on metabolic profiles as seen in this study's two wine strains. During fermentation, organic acid production is attributed to the metabolism of pyruvate in the mitochondria and the reverse tricarboxylic acid cycle (TCA) cycle in the cytoplasm [94,95]. Most succinic acid and malic acid are produced through mitochondrial TCA cycle, however reductive TCA can supplement their production during sake fermentation [95]. These amino and organic acidity profiles are metabolically unique to individual strains and these may contribute a favorable fresh acidity such as malate while succinate contributes an unfavorable acidity to sake. During sake metabolic analysis, PC1 was inferred to represent the amino acid net uptake and mitochondrial transportation of pyruvate which contributes to the amino and organic acidity profiles ( Figure 4C). Of the sake metabolites detected, 75% of the compounds were produced in high abundance in the sake by S. arboricola as compared to S. cerevisiae which produced 44% of the compounds in high amounts. S. arboricola was considered to have a lower amino acid net uptake than S. cerevisiae strains in this study. The abundance of residual amino acids detected in the sake medium and the low abundance of endogenous metabolites were considered in these findings. Moreover, membrane fluidity and amino acid and organic acid transport of S. arboricola has not been investigated and merits further research.
The generated interspecific hybrids were considered to have vigorous glycolysis and fermentation of pyruvate to ethanol compared to parental strains. The reduced malic acid, succinic acid, and glucose ( Figure 4A) compared to the increased ethanol yield ( Figure 3A) supported these claims. The hybrid condition was distinctly exemplified in principal component 2 of the sake metabolite score plot. In this study, where parental metabolic profiles were similar, the hybrids revealed an inversion of these compounds in the heatmap as seen in DMTS, serine, urea, and succinic acid. DMTS is associated with the stale aroma of Japanese sake called hineka. Recycling of sulfur via the methionine salvage pathway produces DMTS-P1, a precursor of this aromatic compound [96]. Mercaptans (thiols) typically have a low sensory threshold and the hybrids were considered to have an active sulfur metabolism. Additionally, urea is a yeast waste product from arginase activity and can be catabolized to form succinate in the cell [97]. Mechanistic elucidation of these metabolic variances is beyond the scope of this initial report of sake fermentation by S. arboricola and the hybrids, further research is required.
Research for S. arboricola yeast is very limited and one other report [80], to the author's knowledge, has explored their use and their interspecific hybrids' use in beverage fermentation. During their wort fermentation trials, S. arboricola produced high levels of ethyl hexanoate at 0.35 mg mL -1 , while this study revealed low production at 0.207 mg mL −1 by substituting sake mash for wort. Yeast strain and raw ingredients certainly play a role in the final fermentation product qualities. The chemistry of sake and beer have some similarities, but there are also many differences including raw materials such as rice or barley and the presence of A. oryzae in the sake mash. These results indicate koji was an integral part of the sake mash converting rice carbohydrate into fermentable sugars. Aspergillus spp. produce mannitol at high levels [46,47] and S. arboricola has great growth utilizing mannitol [14], but the bioavailability of this sugar for fermentation by S. arboricola remains unexplored. This metabolic capability likely facilitated the increase in fermentation performance seen in this study. S. arboricola and their hybrids may be uniquely fit for use in sake fermentations.
In two recent large phylogenetic analysis of 157 and 1011 Saccharomyces cerevisiae strains, an Asian cluster was found basal to the root of the trees [34,65] suggesting little divergence from their wild ancestors in comparison to the highly diverged beer1 clade. Although 1011 genome study was a great step in unmasking the lost ecology and domestication of natural Saccharomyces yeast, it had an overrepresentation of beer brewing strains. In order to dive deeper into the Out of Asia Origin Hypothesis, many more Asian and diverse strains will need to be sequenced and compared phylogenetically. The investigation into S. arboricola's biotechnological value yielded promising results, but the dynamics between A. oryzae's mannitol production and S. arboricola's mannitol utilization merits further research. Exploration of the biotechnical potential of the Saccharomyces clade is only now beginning and understanding the diversity of yeast facilitates humanity's exploitation of Saccharomyces.

Conclusions
This study explored the novel use of S. arboricola in Japanese rice wine known as sake, traditionally produced from polished rice with A. oryzae during winter months. The metabolites in the sake produced from S. arboricola, S. cerevisiae, and their interspecific hybrids were distinct by PCA analysis. S. arboricola had a lower amino acid net uptake than S. cerevisiae and the hybrids showed a mosaic profile of parental strains. Hybrid strains had a more vigorous ethanolic fermentation in the sake mash than S. cerevisiae but had a higher ethyl hexanoate yield than S. arboricola. In the ginjo style of Japanese sake, ethanol and ethyl hexanoate are important sensory metabolic compounds. S. arboricola differed from S. cerevisiae industrial strains because of lower levels of endogenous amino/organic acids, suggesting a lower amino acid net uptake. The endogenous metabolite study highlighted the impact of strain selection which industry niche has unwittingly had an influence in each strains metabolism. The diversity of the Saccharomyces clade of budding yeast is expansive and future studies focused metabolic phenotypes of new strains and novel applications will most likely surface in the following years to come.