Issatchenkia terricola 0621 is a non-Saccharomyces yeast strain isolated from Tannat grapes from Uruguayan vineyards; it stands out for its ability to produce high levels of β-glucosidase activity, which contributes to the aromatic complexity of wines. To delve into the potential oenological applications of this strain, its high-quality genome was obtained and explored, allowing the main central carbon and nitrogen metabolic pathways to be reconstructed. I. terricola is able to utilise glycerol as the sole carbon source in a way that has not previously been described for yeasts. The genes of the fermentome and those involved in stress resistance during winemaking were also identified, and differences were found when compared to S. cerevisiae, which may explain why I. terricola is unable to complete fermentation. The pathways responsible for natural aroma synthesis were also reconstructed, and the production of aromatic acids, alcohols, esters, acetates and lactones was verified experimentally. Finally, sequences encoding for β-glucosidases, in addition to the previously characterised one, were identified in the genome. The work presented here lays the groundwork for experimental research focused on the dissection of the metabolism of a native non-Saccharomyces strain and its application for oenological and biotechnological purposes.

Genome, non-Saccharomyces, Issatchenkia terricola, metabolism, wine

In wine production, yeast strains not only have a leading role in the alcoholic fermentation process but they are also responsible for the final characteristics of the beverage, such as colour, structure and aromatic composition (Lambrechts and Pretorius, 2000). Saccharomyces cerevisiae is considered the top fermenting yeast, but non-Saccharomyces (NS) yeasts that survive the early stages of alcoholic fermentation, and that represent 50-70 % of the total wine yeast population, are attracting increasing attention due to their different abilities to secrete enzymes, produce secondary metabolites, glycerol and ethanol and release mannoproteins, among others (de Ovalle et al., 2021; Jolly et al., 2006). Thus, an adequate selection and use of NS strains may contribute to the organoleptic and sensory properties of the final product when they are used in mixed fermentations with S. cerevisiae (Shi et al., 2019). More than fifteen NS yeast genera associated with wine fermentation have been identified so far, and, in the last decade, research on NS strains has advanced notably, with a focus on the improvement of specific parameters of wine quality and its positive role in the organoleptic characteristics of wine (de Ovalle et al., 2018; Padilla et al., 2016).

With a mild Atlantic climate comparable to that of European wine regions, Uruguay is the fourth largest wine producer in South America. In the last few years, Uruguay has focused research on native strains and their use in wine fermentations (de Ovalle et al., 2021; Lleixà et al., 2016). In previous research, native yeast producers of β-glucosidase (BGL) activity, Metschnikowia pulcherrima, Issatchenkia terricola and Issatchenkia orientalis were isolated from Uruguayan grapes and vineyards (de Ovalle et al., 2018, 2021; González-Pombo et al., 2008; González-Pombo et al., 2011). These enzymes play a crucial role in releasing volatile compounds from non-volatile glycosidic precursors present in must and wines. In the case of I. terricola, we found that BGL activity stands out and increases the level of free monoterpenes and norisoprenoids, which contribute to the aromatic complexity of Muscat and Tannat wines (de Ovalle et al., 2018; González-Pombo et al., 2011). The ability of strains of this species to produce high levels of BGL has also been reported by other authors, who also highlighted their potential for reducing the total acidity content of wines and producing acetaldehyde (Shi et al., 2019; Zhang et al., 2020).

Having knowledge of the genomes of NS yeasts is expected to contribute considerably to the understanding of their metabolic and ecological diversity, which is essential for their successful biotechnological application (Seixas et al., 2019). Nevertheless, sequenced and characterised genomes of native NS strains are still very scarce. Hence, to provide a basis for further research on the use of I. terricola in oenology, in this study we carried out the sequencing, annotation and functional analysis of the genome of strain 0621 isolated from Tannat grapes of a Uruguayan vineyard. In the study, there is an emphasis on the metabolic traits relevant to winemaking and the strain´s tolerance to the many stress factors that arise throughout the fermentation process.

Issatchenkia terricola 0621 isolated from Tannat grapes (González-Pombo et al., 2011) was precultured on a YPD medium (0.1 g/L yeast extract, 2 g/L peptone, 2 g/L glucose) and then cultured in simile must (Martin et al., 2016) or synthetic mediums (de Ovalle et al., 2018) at 28 °C and 100-180 rpm. For the growth assays, the required amount of preculture to reach an initial OD₆₀₀ of 0.2 was inoculated in 50 mL of synthetic medium. The latter was prepared with no carbon or nitrogen sources, which were added after sterilisation. The carbon sources comprised: glucose (10 g/L), fructose (10 g/L) or glycerol (8 mL/L). The nitrogen sources comprised: urea (4 g/L), ammonium (140 or 200 mg/L) or L-amino acids (Pro 20.5 mg/L, Gln 16.9 mg/L, Arg 1.25 mg/L, Trp 6.0 mg/L, Ala 4.9 mg/L, Glu 4.0 mg/L, Ser 2.6 mg/L, Thr 2.6 mg/L, Leu 1.6 mg/L, Asp 1.5 mg/L, Val 1.5 mg/L, Phe 1.3 mg/L, Ile 1.1 mg/L, His 1.1 mg/L, Met 1.1 mg/L, Tyr 0.6 mg/L, Gly 0.6 mg/L, Lys 0.6 mg/L, Cys 0.4 mg/L). The OD₆₀₀ of the culture was monitored every day until no more growth was observed.

Fermentations were assayed on the must from Viura grapes, which were pressed and then pasteurised with a short cycle of temperature increase to 105 °C and subsequent gradual decrease. Assays were performed in conical bottom tubes containing 40 mL of the must at 1.06 g/mL and containing 222 g/L of glucose and fructose. Each tube was inoculated with a preculture of I. terricola to an initial OD₆₀₀ of 0.2. Culture conditions were initially aerobic (180 rpm, 28 °C) for 3 days, and then they were anaerobic for 8 days (with airlock valves). Fermentation was monitored by weight loss.

The volatile fraction was obtained by headspace solid-phase micro extraction coupled with gas chromatography-mass spectrometry (HS-SPME/GCMS), as previously described (Pozo-Bayón et al., 2001). A Trace GC Ultra gas chromatograph with a Triplus Autosampler and ISQ mass detector was used (Thermo Scientific). In a 20 mL headspace-vial, 4 g of NaCl was added to 2 mL of wine sample, followed by 40 µL of internal standard solution (1000 ppm 4-methyl-2-pentanol, 20 ppm nonanol, and 1000 ppm n-heptanoic, in ethanol), which was filled up to a final volume of 9 mL with distilled water. The vial was tightly capped with a PTFE/Silicone cap and heated at 70 °C for 10 min. Then, a Supelco 100 µm PDMS fibre was exposed to the headspace of the sample vials for 30 min, after which volatiles were desorbed in the GC inlet for 4 min. The instrument was fitted with a 30 m × 0.25 mm TG-WAXMS fused-silica capillary column, 0.25 mm film thickness (Thermo, Fisher Scientific). The GC temperature programme was 40 °C (5 min hold), 3 °C/min up to 200 °C and 15 °C/min^-1 up to 240 °C (10 min hold), while the SPME liner (0.75 mm ID) was held at 180 °C. Helium was used as the carrier gas at a flow rate of 1 mL/min, operating in split mode (ratio 30). For the MS detector, the temperature of both the transfer line and ion source was 250 °C, the ionisation mode was electron impact at a voltage of 70 eV, and acquisitions were performed in SIM mode (dwell time 50 ms). Instrument control, data analysis and quantification results were carried out with the Xcalibur 2.1 software. Volatile compounds were identified and quantified as duplicates by comparison with standards and expressed as the value of the area ratio (sample/standard).

200 mL of YPD was inoculated with a preculture of I. terricola to an initial OD₆₀₀ of 0.2 and growth monitored for 6 days. DNA was extracted using Quick-DNA™ Fungal/Bacterial Miniprep Kit (Zymo Research, Catalogue N°: D6005). mRNA was extracted from yeast cultures grown on simile must (Martin et al., 2016) or on synthetic medium (de Ovalle et al., 2018). Total RNA was extracted using Quick-RNA™ Miniprep Kit (Zymo Research, Catalogue N°: R1054). To improve the quality and quantity of both DNA and mRNA, an additional step of cell disruption (2 min in a Mini-Beadbeater-16, Biospec products) was performed after the addition of Lysis Buffer.

DNA was sequenced with both Illumina and PacBio technologies. Libraries were constructed and sequenced in Macrogen (Korea). Illumina libraries were constructed using TruSeq DNA Sample Prep Kit (Illumina) and sequenced in Hiseq2500 equipment producing 100 bp paired end reads. A SMRTbell 20kb library was constructed and sequenced in one PacBio RSII SMRT cell. mRNA of I. terricola was sequenced using Illumina technology. Libraries were constructed using TruSeq mRNA library after polyA RNA purification and sequenced with Hiseq2500, producing 100 bp paired end reads.

The quality of Illumina reads was analysed with FastQC (v0.11.5, https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Bases with Phred score lower than 30, as well as the first ten bases of all reads, were trimmed using Trimmomatic (v0.36, Bolger et al., 2014). PacBio reads were corrected using Illumina reads with LoRDEC (v0.9, Salmela and Rivals, 2014) with parameters k: 19 s: 3. Genome assembly was done based on Illumina-corrected long reads using Flye (v2.3.6, Kolmogorov et al., 2019) with default parameters and genome size of 12Mb. For assembly polishing, short reads were mapped to the obtained genome assembly using Bowtie2 (v2.3.4.2, Langmead and Salzberg, 2012) with default parameters, and used as input to Pilon (v1.22, Walker et al., 2014a). The quality of the assembly was evaluated using Quast (v5.0.0, Gurevich et al., 2013), with the options “conserved genes finding” and “fungus”. Genome completeness was assessed using BUSCO (v3.0.2, Waterhouse et al., 2018) against ascomycete lineage downloaded from OrthoDB (v9.1, Zdobnov et al., 2017) using default parameters. Mitochondrial genome assembly was performed using Norgal (v0.1, Al-Nakeeb et al., 2017) with default parameters, using the short reads obtained with Illumina as input. BLASTn (v2.6.0+, Camacho et al., 2009) was used to determine homology between Norgal mitochondrial and Flye total genome assemblies with the parameter “e-value 1e-5”.

For gene annotation, Braker (v1.9, Hoff et al., 2016) was used, with the fungus option and all other parameters set in default. The mapping of RNAseq reads against the assembled genome was done with HISAT2 using default parameters (v2.0.0-beta, Kim et al., 2015). The mapping file was used to train GeneMark and AUGUSTUS gene predictors. Annotation quality was assessed with BUSCO (v3.0.2, Waterhouse et al., 2018) with annotated proteins as entry (-m prot parameter) against the ascomycete linage from the OrthoDB database (v9.1, Zdobnov et al., 2017), with default parameters. Mitochondrial annotation was performed using Mitos Web Server (Donath et al., 2019), using Yeast genetic code and RefSeq 63 Fungi as the reference.

InterProScan (v5.31.70.0, Jones et al., 2014) was used for protein domain assignment, with the option “goterm” to report Gene Ontology terms (GO) and the remaining parameters as default. A BLASTp search (v2.6.0+, Camacho et al., 2009) was run against the Swiss-Prot database (downloaded in February 2019, Poux et al., 2017) using the I. terricola annotated proteins as query, and “e-value 1e-5”, “max_target_seqs 1” and the other options set in default. KEGG terms were assigned with the online KAAS tool (Moriya et al., 2007), using the GHOSTX search program, Saccharomycetes organisms as Gene datasets, and BBH (bi-directional best hit) as assignment method. KEGG Pathways assignment and Brite classification were performed with the KEGG Reconstruct tool (Kanehisa and Sato, 2020). For comparison, the same analysis was performed for S. cerevisiae and Pichia kudriavzevii.

Protein sequences from related yeasts with available annotated genomes were downloaded (see Supplementary Table 1). Orthologs were determined using Orthofinder (v2.4.0, Emms and Kelly, 2019) with options M: msa and S: diamond. S. cerevisiae genes of interest were used to retrieve I. terricola orthologous genes from orthologous cluster analysis. When no clear ortholog gene could be determined, Enzyme Commission numbers or protein domains (Pfam or InterPro) of S. cerevisiae proteins were obtained from UniProt and searched in I. terricola annotation. S. cerevisiae proteins were also used as queries for local BLASTp or BLASTx (v2.6.0+, Camacho et al., 2009) using the I. terricola proteome and genome as databases respectively. For these, an e-value cut-off of 1e-5 was used and the BLAST results were manually filtered for each protein.

Different methods were used for the identification of β-glucosidases in I. terricola. First, enzymes with the assigned term “beta-glucosidase” from InterProScan analysis and homology to Swiss-Prot β-glucosidases were retrieved from genome functional annotation. Second, BLAST homology search of annotated proteins with those from existing databases was performed. The BLASTp programme (v2.6.0+, Camacho et al., 2009) was used with “e-value 1e-5”, “max_target_seq 1” and the other parameters by default. Yeast β-glucosidases were retrieved from NCBI Refseq protein database (research performed 26th February 2022), searching by the terms “beta glucosidase” in all fields, with the filter “Organism = fungi and budding yeasts". These protein sequences were used for BLASTp search against I. terricola annotated proteins. The dbCAN2 meta server (Zhang et al., 2018) was used to classify the identified glucosidases using the three available tools DIAMOND, HMMER y Hotpep.

To obtain the genome of the I. terricola 0621 reference strain, we followed a hybrid strategy combining PacBio long reads and Illumina short reads. Genome sequencing produced 138.868 long reads and 22.509.739 short, paired end reads. The total genome assembly size was 12,5 Mb, consisting of 15 scaffolds, 5 of them longer than 1Mb (Figure 1A), with an N₅₀ of 3,4 Mb and a GC content of 36.4 %, similar to those of related yeasts (Douglass et al., 2018). The mitochondrial genome assembly, obtained exclusively from short reads, produced a circular scaffold of 32.088 bp, showing high homology to the shorter contigs of the assembly obtained from the long reads (contigs 7, 12, 14-21, Figure 1A). These results suggest that the shorter contigs are fragments of the mitochondrial genome, while longer scaffolds correspond to nuclear chromosomes. Genome polishing resulted in the correction of 137 single nucleotide polymorphisms (SNPs), 1793 small insertions (making up a total of 6,911 bases), and 119 small deletions (making up a total of 195 bases). No ambiguous bases were found. The polishing result did not give major differences in the quality metrics with respect to the originally assembled genome (Supplementary Table 2). The coverage was 76X and 25729X for the nuclear and mitochondrial genomes, respectively. Raw reads and genome assembly were deposited in NCBI under Bioproject PRJNA833333.

Table

(A): Genome assembly graph visualized with Bandage (v0.8.1, Wick et al., 2015). Contigs and scaffold were assembled using Flye and polished with Pilon. Coloured segments correspond to chromosomes. Three short segments (light brown squares) corresponding to rRNA sequences were included in Scaffold 2, Contigs 5 and 6. Putative telomeres are indicated by asterisks (*). Segment lengths and the number of annotated genes are shown. The grey segments enclosed in the small inset correspond to mitochondrial fragments assembled by Flye. The order of the different segments is shown in the big inset, as obtained by comparison with the mitochondrial genome. (B): Annotation of the I. terricola nuclear genome. Coding genes (yellow), tRNAs (orange), rRNA clusters (purple) and transposable elements (TE) (aquamarine) are shown in each nuclear contig. (C): Mitochondrial genome annotation. Coding genes for NADH:ubiquinone oxidoreductase subunits (nad), the 3 subunits of ATP synthase (atp), the 3 subunits of the cytochrome c oxidase (cox), apocytochrome b (cob-a), and ribosomal protein S3 (rps3) are shown in aquamarine, tRNAs are shown in purple, rRNAs are shown in yellow and mobile introns (LAGLI, GIY) in genes cox1 and cob are indicated in orange. Plots in B and C were obtained using the karyoploteR package (v1.16.0, Gel and Serra, 2017).

In S. cerevisiae, the use of different high-throughput phenotypic screenings has allowed determinants of the resistance to the numerous stressors that occur during wine fermentation to be identified (Gonzalez et al., 2016; Teixeira et al., 2009; Teixeira et al., 2010). Based on these data, we analysed I. terricola for homologs of the genes involved in ethanol and high-sugar toxicity, which is of major importance (Supplementary Table 9). In the case of ethanol resistance, of the 641 identified genes in the budding yeast, 477 were found to be present in I. terricola. As in the case of S. cerevisiae, these genes are functionally related to vacuolar protein sorting and peroxisomal function, V-ATPase complex assembly, membrane and cell wall composition and the control of carbohydrate and amino acid metabolism. Regarding high-glucose tolerance, of the 278 genes identified as important or relevant in S. cerevisiae, 194 are present in I. terricola (Supplementary Table 9). Those genes are related to vacuolar organisation, late endosome to vacuole transport and the regulation of transcription, mainly at the level of nutrient metabolism control, as in the budding yeast (Gonzalez et al., 2016; Teixeira et al., 2010).

During alcoholic fermentation, adaptations in terms of lipid content are also important (Tesnière, 2019). S. cerevisiae imports unsaturated fatty acids (UFAs) and sterols through specific transporters or by endocytosis, to cope with the inability to synthesise these compounds in anaerobiosis (Tesnière, 2019). I. terricola have orthologous genes to FAT1 and YPK1 (g4763.t1 and g115.t1, respectively), possible regulators of fatty acid uptake in the budding yeast (Zou et al., 2002). Ethanol stress affects yeast membrane integrity, which is counteracted by an increase in fatty acid unsaturation (oleic acid), sterol and phosphatidylcholine content. In I. terricola we identified the genes and pathways involved in the synthesis of phospholipids, TAGs, sphingolipids and sterols as ergosterol, the most abundant lipids in yeast (Supplementary Table 7). In addition, during adaptation to the variable temperatures employed in winemaking, changes in membrane lipid content takes place (Henderson et al., 2013), eliciting the induction at the transcriptional level of genes involved in lipid biosynthetic pathways (López-Malo et al., 2014). It was possible to identify the I. terricola orthologs of these genes, such as DPL1 and LCB3 of the sphingolipid synthesis pathway (g512.t1 and g3137.t1 respectively), OLE1, which encodes for a desaturase (g4642.t1 and g5162.t1), or PSD1, which is involved in the synthesis of phosphatidylethanolamine (g4129.t1).

In the course of fermentation, maturation and aging, secondary metabolites present in musts are modified by yeasts and other microorganisms, resulting in an increase in the aroma complexity of wines. The compounds responsible for wine aroma are mainly alcohols, acids, esters, ketones and aldehydes, and their production depends on the set of microorganisms present in musts and the conditions applied during winemaking and aging (Ruiz et al., 2019).

Given our previous promising results in relation to the BGL activity of I. terricola 0621 strain on the glycosylated aroma precursors of musts and wines, we decided to continue the search for other β-glucosidases for their further characterisation and oenological application. Ten proteins were identified as putative β-glucosidases. The classification of these ten enzymes according to specific families was confirmed (Table 3): four of them belong to the GH17 family (g989.t1, g1025.t1, g2300.t1, g4410.t1), three to the GH132 (945, 3296, 3698), two to the GH5 (g1395.t1, g.4180.t1) and the last to the GH3 family (4598). The g4180 gene codes for the previously characterised β-glucosidase (González-Pombo et al., 2011).

Table

titre du tableau
Gene	Family*	Protein Domains**	BLAST Swiss-Prot***
		Pfam	InterPro	Description	Organism
g4598.t1	GH3	Glycosyl hydrolases family 3 N terminal domain	Glycoside hydrolase superfamily	Beta-hexosaminidase	Photobacterium profundum
g1395.t1	GH5	Cellulase (glycosyl hydrolase family 5)	Glycosyl hydrolase family 5	Glucan 1,3-beta-glucosidase	Saccharomyces cerevisiae
g4180.t1		Cellulase (glycosyl hydrolase family 5)	Glycoside hydrolase, family 5, conserved site	Glucan 1,3-beta-glucosidase	Saccharomyces cerevisiae
g989.t1	GH17	Glycosyl hydrolases family 17	Glycosyl hydrolase family 17	Glucan 1,3-beta-glucosidase	Saccharomyces cerevisiae
g1025.t1		Glycosyl hydrolases family 17	Glycosyl hydrolase family 17	Probable family 17 glucosidase	Saccharomyces cerevisiae
g2300.t1		-	Glycoside hydrolase superfamily	Probable family 17 glucosidase	Saccharomyces cerevisiae
g4410.t1		-	Glycoside hydrolase superfamily	Probable family 17 glucosidase	Pichia angusta
g945.t1	GH132	Beta-glucosidase (SUN family)	SUN family	Uncharacterized protein	Saccharomyces cerevisiae
g3296.t1		Beta-glucosidase (SUN family)	SUN family	Probable secreted beta-glucosidase	Saccharomyces cerevisiae
g3698.t1		Beta-glucosidase (SUN family)	SUN family	Probable secreted beta-glucosidase	Saccharomyces cerevisiae

*Classification of BGLs based on the CAZy database. **Domains of I. terricola annotated proteins in the Pfam and InterPro databases. **Identification by BLAST search of I. terricola annotated proteins in the Swiss-Prot database. For details see Methods section.

Knowledge of non-conventional yeasts and their biotechnological use has increased considerably over the last 10 years. However, there is still a considerable lack of data on these yeasts compared to the conventional wine yeast S. cerevisiae. In this paper, we present the whole-genome sequencing of the 0621 strain of I. terricola. The assembly revealed a 12.5 Mb genome with a GC content of 36.4 %, which encodes 5377 proteins. Those data are in accordance with genomic data for other yeasts (Douglass et al., 2018; Giorello et al., 2018; Seixas et al., 2019). When assigned to Brite categories, the number of proteins in each category is comparable to those of related organisms, which supports the quality of the genome annotation. The analysis of the genome allowed us to determine the strain´s metabolic traits relevant for winemaking, as well as for its ability to grow on the different nutrients available in musts, adapt to stressful conditions, and generate substances that can confer desirable properties to the wine.

We were able to identify the catabolic pathways and the transporters for an ample range of carbon sources, which suggests that I. terricola adapts by utilising different available nutrients. It is worth mentioning that, as in other yeasts and fungi, we found homologs of the key regulators of carbon catabolite repression (Gancedo, 1998; Kayikci and Nielsen, 2015): g4645 is orthologous to S. cerevisiae MIG1 and g1591.t1 to SNF1. The enzymes involved in the catabolism of hexoses (except galactose), pentoses, monocarboxylic acids, lipids and glycerol were identified in the genome of I. terricola. As already mentioned, most of the genes identified in S. cerevisiae as being part of the fermentation process are present in I. terricola, but some of those that were found to be essential or important for a successful and well-timed fermentation are absent in the latter. Ourexperimental data supports this observation, since, in contrast to S. cerevisiae, I. terricola was not able to perform a complete fermentation, consuming only 25 %of total fermentable sugars (Supplementary Figure 2). Alternatively, or additionally, the lack of homologs of many of the genes involved in stress tolerance could explain this metabolic trait.

The putative ability of I. terricola to metabolise pentoses contrasts with S. cerevisiae, which is unable to grow on ribose, xylose and L-arabinose (Ruchala and Sibirny, 2021). However, we were not able to identify the permeases responsible for the transport of these sugars, although it has been shown that they can enter the cell through transporters of the HXT hexose transporter family and through the galactose permease GAL2 (Nijland and Driessen, 2020). I. terricola has a gene that has homology with GAL2, despite the enzymes for the catabolism of galactose being absent. I. terricola also has five homologs of HXT genes, which is a low number compared to the 20 HXT genes present in S. cerevisiae (Perez et al., 2005). Thus, the potential ability of I. terricola to metabolise pentoses makes it possible to study its capacity to synthesise chemicals and biomass from pentoses, a topic that has also attracted much attention in recent years (Ruchala and Sibirny, 2021).

As already shown, I. terricola is able to grow on glycerol as the sole carbon source in aerobic conditions, and its genome harbours genes encoding the enzymes of the DHA and the GA glycerol catabolic pathways (Klein et al., 2017). The latter was suggested by Tom et al. (1978) as a way of carrying out glycerol dissimilation in the filamentous fungus Neurospora crassa; however, it has not yet been reported for any yeast species. It would be interesting to evaluate in vitro the activities of the enzymes involved in these pathways in extracts of I. terricola on different carbon sources (Tom et al., 1978). As previously mentioned, the functionality of the DHA pathway relies on a putative NAD⁺-dependent activity of the g3745.1 gene, or on the existence of another enzyme with such an activity, which remains to be determined. Three genes predicted to code for glycerol transporters were identified.

Regarding the use of nitrogen sources, our findings suggest that I. terricola has adapted so as to transport and catabolise the variable and broad spectrum of assimilable nitrogenous compounds of which wine musts are composed. The key regulators of nitrogen catabolite repression (Magasanik and Kaiser, 2002) were identified in its genome: g5180.t1 is orthologous to URE2, g1046.t1 to GLN3 and g3408.t1 to GAT1 DAL80 and DEH1. However, those which are the preferred sources in pure cultures or in combined cultures with S. cerevisiae still need to be determined. Among the identified transporters, the high number of GABA transporters present in the genome of I. terricola is noteworthy. This compound may account for up to 20 %of the assimilable nitrogen in grape juices. Moreover, in S. cerevisiae the assimilation of GABA has favourable effects on yeast growth, fermentation rate and glycerol production under scarce nitrogen conditions, and is a source of succinate, which may contribute to pH reduction (Bach et al., 2009).

During wine production, the participating microorganisms must generate appropriate responses to cope with the variety of environmental stresses to which they are exposed, like elevated sugar and ethanol concentration, or different temperatures from those which are optimal for growth. According to our results, only 70-75 %of the S. cerevisiae resistance genes involved in high sugar and high ethanol concentrations, have homologs with I. terricola. What is more surprising is that even genes identified as relevant or essential determinants for survival are missing. In the case of high ethanol resistance, no homologs of CWC25 and SPP381 could be found. These two genes are linked to the splicing of mRNAs, which in S. cerevisiae is thought to be a way of eliciting rapid and specific changes in gene expression in response to environmental stress (Bergkessel et al., 2011). Another relevant resistance gene missing in I. terricola is FPS1, which encodes an aqua glycerol protein involved in the control of intracellular levels of glycerol, as well as the maintenance of ethanol intracellular levels during fermentation (Teixeira et al., 2009; Yoshikawa et al., 2009). Orthologs of MSN2/MSN4, which are crucially regulated in response to high-glucose tolerance (Brion et al., 2016; Gasch, 2007), are missing in I. terricola. MSN2 and MSN4 bind to stress response elements (STRE) present in the promoters of a number of the genes involved in the response to environmental stress (Martínez-Pastor et al., 1996). MSN2 and/or MSN4 are also missing or little conserved in other NS-yeasts, like members of the Hanseniaspora genus, or Lachancea kluyveri (Brion et al., 2016; Seixas et al., 2019). Nevertheless, we were able to identify the orthologs of HOG1 and PBS2 (g3226.t1 and g4853.t1 respectively), which lay upstream in the HOG hyperosmotic stress signalling pathway and are among the S. cerevisiae genes involved in high-glucose resistance. Taken together, these results reinforce the idea that there are important differences in the way yeast species respond to environmental stresses during winemaking (Brion et al., 2016; Seixas et al., 2019). However, as previously stated, I. terricola stops growing before fermentation is complete. It cannot be ruled out that the lack of the aforementioned genes renders this yeast more susceptible to environmental stress.

The amount and length of lipids have a strong impact on membrane thickness, dynamics and fluidity characteristics, which is relevant in stressful conditions (Henderson et al., 2013). In the case of I. terricola, we found all the central pathways of membrane and reserve lipid synthesis, such as the synthesis of phosphatidic acid, the precursor in the synthesis of membrane phospholipids and neutral lipids. Under certain nutritional conditions, these lipids are stored in cytoplasmic lipid droplets, which indicates that there is a crossroads between membrane biogenesis and lipid storage (Ejsing et al., 2009). In fact, lipid droplets which store neutral lipids, such as triglycerides and steryl esters, are important in the cell as they supply of energy and membrane lipids and also regulate cell longevity (Goldberg et al., 2009; Welte and Gould, 2017).

Our experimental results show that I. terricola produces several active flavour compounds, like aliphatic and aromatic alcohols, acetate and ethyl esters and γ-nonalactone. We were able to identify the genes and pathways involved in the synthesis of these compounds. In some cases, orthologs of the S. cerevisiae genes involved in the pathways were identified; in others, the involvement of genes different to those which catalyse a certain step in the budding yeast, but that have the same activity, is likely. For example, it is possible that the alcohol acetyl-transferase activity encoded by S. cerevisiaeATF1 and ATF2 genes, which are involved in the production of acetate esters, can be fulfilled by the products of g20.t1 and g3217.t1, which possess an alcohol acetyltransferase/N-acetyltransferase domain. However, these genes are phylogenetically closer to SLI1, a gene shown to be involved in resistance to the sphingolipid biosynthesis inhibitor drug myriocin in the budding yeasts (Momoi et al., 2004). Hence, further research must be conducted in order to confirm these possible pathways.

I. terricola 0621was selected to add aromatic complexity to Muscat and Tannat wines because of its outstanding BGL activity, which increases the level of free monoterpenes and norisoprenoids (de Ovalle et al., 2018; González-Pombo et al., 2011). However, BGLs are inactivated in acidic pH and fermentation conditions, which limits the use of the I. terricola 0621 strain in mixed cultures to increase the varietal aroma during must fermentation. As an alternative, it is possible to consider the heterologous production of these BGLs. Indeed, our previous findings related to I. terricola g4180.t1 and the remaining BGL genes identified in the present study motivated the characterisation of the oenological properties of the encoded enzymes. This will eventually allow us to design new application strategies for aroma production in vinification, such as the cloning and expression of S. cerevisiae.

To sum up, this study constitutes an example of how the availability and exploration of the genomes of non-conventional native strains can provide tools for the study of metabolic mechanisms not yet fully understood, and of the reservoir of genes that could eventually be used to generate recombinant strains or for the biotechnological production of the enzymes.

We would like to thank Ramón Gonzalez and Pilar Morales from the Instituto de Ciencias de la Vid y del Vino (ICVV), CSIC, Universidad de La Rioja, Spain for hosting a student's internship, and Francisco Carrau from Área Enología y Biotecnología de Fermentaciones (Facultad de Química, UdelaR) for kindly providing us with I. terricola strain. This study was funded by the Agencia Nacional de Investigación e Innovación (Grant N° FMV_1_2017_1_136574) and Programa de Desarrollo de las Ciencias Básicas (PEDECIBA, Uruguay). A.C., A.R., A.V., P.G-P. and S.O. are members of Sistema Nacional de Investigadores (Uruguay). J.D. was recipient of a fellowship from Comisión Académica de Posgrado (CAP, UdelaR) and a fellowship from Agencia Nacional de Investigación e Innovación (Nº POS_FMV_2018_1_1007776). S.O. was recipient of fellowships from Agencia Nacional de Investigación e Innovación (N° POS_NAC_2014_1_102369) and Comisión Académica de Posgrado (CAP, UdelaR). SO internship at ICVV was funded by Comisión Sectorial de Investigación Científica (UdelaR).