Transcriptional Analysis of l-Methionine Catabolism in the Cheese-Ripening Yeast Yarrowia lipolytica in Relation to Volatile Sulfur Compound Biosynthesis

Abstract

Yarrowia lipolytica is one of the yeasts most frequently isolated from the surface of ripened cheeses. In previous work, it has been shown that this yeast is able to convert l-methionine into various volatile sulfur compounds (VSCs) that may contribute to the typical flavors of several cheeses. In the present study, we show that Y. lipolytica does not assimilate lactate in the presence of l-methionine in a cheeselike medium. Nineteen presumptive genes associated with l-methionine catabolism or pyruvate metabolism in Y. lipolytica were transcriptionally studied in relation to l-methionine degradation. The expression levels of the YlARO8 (YALI0E20977g), YlBAT1 (YALI0D01265g), and YlBAT2 (YALI0F19910g) genes (confirmed by real-time PCR experiments) were found to be strongly up-regulated by l-methionine, and a greater variety and larger amounts of VSCs, such as methanethiol and its autooxidation products (dimethyl disulfide and dimethyl trisulfide), were released in the medium when Y. lipolytica was grown in the presence of a high concentration of l-methionine. In contrast, other genes related to pyruvate metabolism were found to be down-regulated in the presence of l-methionine; two exceptions were the YlPDB1 (YALI0E27005g) and YlPDC6 (YALI0D06930g) genes, which encode a pyruvate dehydrogenase and a pyruvate decarboxylase, respectively. Both transcriptional and biochemical results corroborate the view that transamination is the first step of the enzymatic conversion of l-methionine to VSCs in Y. lipolytica and that the YlARO8, YlBAT1, and YlBAT2 genes could play a key role in this process.

The enzymatic degradation of l-methionine and the subsequent formation of volatile sulfur compounds (VSCs) are essential for the development of the characteristic aromatic notes of several cheeses (12, 14, 17). Methanethiol (MTL) is the VSC most frequently detected in ripened cheeses. It is also the direct precursor of other sulfur aroma compounds, such as dimethyl sulfide, dimethyl disulfide (DMDS), dimethyl trisulfide (DMTS), and S-methylthioesters (4, 7, 18).

Although chemical degradation (Strecker degradation) can occur during the ripening process, the amino acid catabolism in cheese is due mainly to the action of microbial enzymes. In lactic acid bacteria, such as Lactococcus lactis, the first step of amino acid degradation is initiated by an aromatic or branched-chain aminotransferase (AraT or BcaT) (15, 16), which requires the presence of an α-keto acid as an amino group acceptor (28). Cheese-ripening yeasts, such as Geotrichum candidum, Kluyveromyces lactis, and Yarrowia lipolytica, are also known to contribute to the formation of various VSCs through the degradation of l-methionine (4, 25). Until recently, the involvement of an aminotransferase in the conversion of l-methionine to MTL in the yeasts was still speculative, since functional analysis of any gene encoding such an enzyme was not possible. Due to the recent availability of K. lactis and Y. lipolytica genomes (http://cbi.labri.fr/Genolevures/), several strategies using genome sequences can now be considered. For instance, all the putative aminotransferase-encoding genes from K. lactis were cloned in an overproducing vector, and their effects on the production of VSCs were analyzed (19). Two aromatic aminotransferase genes, KlARO8.1 and KlARO8.2, were found to be responsible for l-methionine aminotransferase activity in K. lactis. In Y. lipolytica, functional analysis of a branched-chain aminotransferase gene (YlBCA1) has recently shown that the corresponding enzyme is able to convert l-methionine to α-keto γ-methylthiobutyric acid (KMTBA) (8). KMTBA could be subsequently converted to MTL (4), a highly reactive sulfur compound that quickly reacts with itself, forming the oxidized and more stable compounds DMDS and DMTS (10).

Aminotransferases are widely distributed among microorganisms. Furthermore, two or more aminotransferases with overlapping specificities are generally found in the same microorganism (21, 27, 29). In the yeast Y. lipolytica, not all the putative genes encoding aminotransferases have been studied, and our understanding of their involvement in l-methionine catabolism remains incomplete. In the present study, all genes predicted to be involved in l-methionine catabolic pathways in the full genome of Y. lipolytica were searched. Then 18 oligonucleotides were designed and employed to perform transcriptional analysis of Y. lipolytica in the presence of l-methionine. Subsequently, in relation to the biochemical data (lactate and l-methionine consumption; KMTBA and VSC production), the transcriptional patterns of the genes were compared when Y. lipolytica cells were grown in a cheeselike medium under low- or high-l-methionine conditions.

MATERIALS AND METHODS

Media, growth conditions, and microbial analysis.

Y. lipolytica 370 was originally isolated from French cheeses. This strain belongs to our laboratory's collection (LGMPA, UMR INRA-INAPG, Thiverval-Grignon, France) and was stored in 5% glycerol-nonfat dry milk at −80°C. Y. lipolytica was precultured by inoculating a single colony into 20 ml of synthetic cheese medium in 100-ml Erlenmeyer flasks and incubating the flasks for 48 h at 25°C with agitation (150 rpm). The synthetic cheese medium contained (per liter of distilled water) 1 g of yeast extract (Gibco, Heidelberg, Germany), 15 g of Bacto Casamino Acids (Difco Laboratories), 38 ml of a 60% sodium lactate stock solution (Prolabo, Fontenay-sous-Bois, France), 0.1 g of CaCl₂ (Prolabo), 0.5 g of MgSO₄·7H₂O (Prolabo), 6.8 g of KH₂PO₄ (Sigma-Aldrich, St-Quentin Fallavier, France), and 10 g of NaCl (Prolabo). After the pH was adjusted to 5.5 ± 0.1 (using concentrated 2 N HCl), the medium was autoclaved (120°C, 20 min) and then supplemented with 20 g of lactose (Prolabo) per liter. One milliliter of the preculture was then used to inoculate (1%, vol/vol) 100 ml of synthetic cheese medium in 500-ml Erlenmeyer flasks. Prior to inoculation, l-methionine (Sigma-Aldrich) was added to the culture medium at a concentration of 1 g·liter⁻¹ (to obtain LM medium) or 6 g·liter⁻¹ (to obtain HM medium). Cultures were incubated for at least 96 h using the conditions described above. Viable cell counts, expressed in CFU·ml⁻¹, were determined by using a standard aerobic plate count procedure with yeast extract-glucose-chloramphenicol agar (Biokar Diagnostics, Paris, France). Surface inoculation was carried out by using a spiral plater (Interscience, Saint-Nom-La-Bretèche, France) and petri dishes to detect any microbial contamination. Plates were incubated at 25°C for at least 2 days before the numbers of CFU were determined.

In silico analysis.

Our first objective was to select Y. lipolytica genes encoding enzymes possibly involved in l-methionine catabolism by this cheese-ripening yeast. Possible related metabolic pathways were examined using the online service KEGG (http://www.genome.ad.jp/kegg/pathway.html). We used Saccharomyces cerevisiae sequences in the Saccharomyces Genome Database (http://www.yeastgenome.org/) as queries and employed BLAST (2) to screen sequence databases for homology. The Y. lipolytica sequence data were obtained from the Génolevures public database (http://cbi.labri.fr/Genolevures/). Even distantly related homologs of S. cerevisiae genes were identified in order to ensure that potential genes of interest were collected for Y. lipolytica. The results of our in silico analysis are shown in Table 1.

TABLE 1.

In silico analysis of target genes in the S. cerevisiae and Y. lipolytica genomes

S. cerevisiae open reading frame			Y. lipolytica ortholog			BLAST result (% similarity)	Expected value
Designation	Saccharomyces Genome Database annotationa	Size of predicted translation product (amino acids)	Accession no.b	Putative designation and function	Size of predicted translation product (amino acids)
Methionine or aromatic or branched-chain amino acid metabolism
YHR208w/BAT1	Mitochondrial branched-chain amino acid aminotransferase	393	YALI0D01265g	Similar to sp P38891 S. cerevisiae YHR208w BAT1 branched-chain amino acid aminotransferase (mitochondrial)	388	300/390 (76)	E-146
YJR148w/BAT2	Cytosolic branched-chain amino acid aminotransferase	376	YALI0F19910g	Similar to sp P47176 S. cerevisiae YJR148w BAT2 branched-chain amino acid aminotransferase (cytosolic)	373	209/374 (55)	1E-62
YGL202w/ARO8	Aromatic aminotransferase	500	YALI0E20977g	Similar to sp P53090 S. cerevisiae YGL202w ARO8 aromatic amino acid aminotransferase I	495	332/496 (66)	E-141
YHR137w/ARO9	Aromatic aminotransferase	513	YALI0C05258g	Weakly similar to sp P38840 S. cerevisiae YHR137w ARO9 aromatic amino acid aminotransferase II	532	259/521 (49)	6E-59
YLR244c/MAP1	Methionine aminopeptidase	387	YALI0D17138g	Similar to sp Q01662 S. cerevisiae YLR244c MAP1 methionine aminopeptidase (isoform 1)	371	267/363 (73)	E-127
YBL091c/MAP2	Methionine aminopeptidase	421	YALI0D05159g	Similar to sp P38174 S. cerevisiae YBL091c MAP2 methionine aminopeptidase (isoform 2)	471	295/468 (63)	E-122
YLR180w/SAM1	S-Adenosylmethionine synthase	382	No ortholog
YDR502c/SAM2	S-Adenosylmethionine synthase	384	YALI0B14509g*c	Highly similar to tr Q9P842 Candida albicans S-adenosylmethionine synthase 2 (CaSAM2)	386	340/380 (89)	E-177
YPL273w/SAM4	S-Adenosylmethionine-homocysteine methyltransferase	325	YALI0F25641g*	Weakly similar to tr Q8DZ17 Streptococcus agalactiae putative homocysteine S-methyltransferase MmuM	348	161/353 (45)	3E-29
YAL012w/CYS3	Cystathionine γ-lyase	394	YALI0F05874g	Similar to sp P31373 S. cerevisiae YAL012w CYS3 cystathionine γ-lyase	394	321/392 (81)	E-147
YGR155w/CYS4	Cystathionine β-synthase	507	YALI0E09108g	Similar to sp P32582 S. cerevisiae YGR155w CYS4 cystathionine β-synthase	538	362/514 (70)	E-146
YJR130c/STR2	Cystathionine γ-synthase	639	No ortholog
YLL058w	Molecular function unknown	575	YALI0D17402g*	Similar to tr Q86ZL4 Podospora anserina putative cystathionine γ-synthase	549	315/582 (54)	1E-94
YGL184c/STR3	Cystathionine β-lyase	465	YALI0D00605g*	Similar to tr Q96VU8 Neurospora crassa cystathionine β-lyase	434	269/419 (64)	6E-99
Amino acid and carbohydrate metabolism
YMR108w/ILV2	Acetolactate synthase (catalytic subunit)	687	YALI0C00253g	Highly similar to sp P07342 S. cerevisiae acetolactate synthase (mitochondrial precursor) (catalytic subunit)	681	514/658 (78)	0.0
YCL009c/ILV6	Acetolactate synthase (regulatory subunit)	309	YALI0C09636g	Similar to sp P25605 S. cerevisiae acetolactate synthase (mitochondrial precursor) (regulatory subunit)	288	226/283 (79)	E-102
YER178w/PDA1	E1α subunit of the pyruvate dehydrogenase complex	443	YALI0F20702g	Highly similar to sp P16387 S. cerevisiae YER178w PDA1 pyruvate dehydrogenase (lipoamide) (subunit E1α)	394	318/394 (80)	E-155
YBR221c/PDB1	E1β subunit of the pyruvate dehydrogenase complex	366	YALI0E27005g	Highly similar to sp P32473 S. cerevisiae YBR221c PDB1 pyruvate dehydrogenase (lipoamide) (subunit E1β)	330	282/329 (85)	E-132
			YALI0F05038g*	Similar to tr Q9P5K8 Neurospora crassa probable 3-methyl-2-oxobutanoate dehydrogenase (lipoamide) (subunit E1β)	398	195/355 (54)	1E-49
YLR044c/PDC1	Major isozymes of three pyruvate decarboxylase isozymes	563	YALI0D10131g	Similar to sp P06169 S. cerevisiae YLR044c PDC1 pyruvate	571	407/563 (72)	0.0
YLR134w/PDC5	Minor isoform of pyruvate	563	No ortholog	decarboxylases
	decarboxylase			(major isozyme)
YGR087c/PDC6	Minor isoform of pyruvate decarboxylase	563	YALI0D06930g	Similar to sp P26263 S. cerevisiae YGR087c	600	338/577 (58)	E-114
YDR380w/ARO10	Phenylpyruvate	635	No ortholog	PDC6 pyruvate
	decarboxylase			decarboxylases
YDL080c/THI3	Probable decarboxylase	609	No ortholog	(minor isozyme)

^aAnnotation in the Saccharomyces Genome Database (http://www.yeastgenome.org/).

^bAccession number of the Y. lipolytica sequence obtained from Génolevures (http://cbi.labri.fr/Genolevures/).

^cAn asterisk indicates a distantly related homolog of the S. cerevisiae open reading frame (see the putative designation).

Oligonucleotide probe design and DNA array construction.

DNA arrays that included 19 annotated protein-encoding sequences of Y. lipolytica were developed. One long oligonucleotide probe (70-mer) was designed for each gene of interest (Table 2) using the ROSO software (23), which was optimized to select the most specific sequence of each gene. Oligonucleotide sequences in the last 300 bases of each gene were searched, with no stable secondary structure and several optimal thermodynamic properties, as defined by ROSO (http://pbil.univ-lyon1.fr/roso/Home.php). As a final check, a low-homology search using BLASTN was performed for all Y. lipolytica genes to ensure that each probe was specific for a single gene and did not exhibit any cross-hybridization with any other gene. All of the oligonucleotides that were designed were commercially synthesized without modification by QIAGEN Operon (Alameda, CA). In addition, six Arabidopsis thaliana probes were used as external positive (with mRNA spiked) or negative (without mRNA spiked) controls, and they were provided by Stratagene (SpotReport oligonucleotide array validation system; Stratagene, La Jolla, CA). Spotting onto Corning UltraGAPS coated slides (γ-aminopropyl silane surface; Corning Life Sciences, New York) was performed at the Transcriptome-Biochips Platform in Toulouse, France (http://genopole.toulouse.inra.fr), using a VersArray ChipWriter Pro microarray (Bio-Rad). Probes were spotted in duplicate. After printing, DNA elements were cross-linked to the slides by UV irradiation (Stratalinker UV cross-linker; Stratagene) and stored in a vacuum chamber until they were used.

TABLE 2.

ORFs selected and corresponding oligonucleotide probes used for Y. lipolytica microarray construction

Y. lipolytica open reading framea,b	Putative gene designationb	Oligonucleotide probe (70-mer)c
YALI0F19910g	BAT2	TACAAGAACAAGGCTGGCAACAAGCTGATTCTGTTCCGACCCGAGCTCAACATGGCTCGTCTCAACGACT
YALI0E20977g	ARO8	ATCGACTTCATCAAGGAGCACACCACCATGATCCACAACATTCCTTACAAGAACTGGGACATTTGTGCCA
YALI0C05258g	ARO9	GCTGATATCGACCATTACACCTCCTTGATCGCTCCCTCTTACCTCACCATTGATGATGTAGGACTGGTTG
YALI0D17138g	MAP1	ACAGTTTGAGCACACTCTGTTAGTCACTGAGACTGGTGTGGAGATTCTCACCGCAAGATACAAGAACTCT
YALI0D05159g	MAP2	TGTTCCTATTGTCAAGTCTGCTGACATGACCAAGATGGAAGAGGGTGAGACCTTTGCCATCGAGACTTTT
YALI0B14509g*d	SAM2	ACTCCGAGAAGGGTTTCGACTACAAGACCTGTAACGTGCTTGTTGCCATTGAGCAGCAGTCTCCCGATAT
YALI0F25641g*	SAM4	GGTCAGCCAAGTTCTTGCTTCCTATCAACTCCGATTTCCTCAAATGTATCGAAAGCGCCGATCCATGGAC
YALI0F05874g	CYS3	TCTCCTTCCGAATCAAGGGAGGAGCCGAGTCTGCCAACAAGTTCTGCCAGAACACCAAGATCTTCACTCT
YALI0E09108g	CYS4	CTTTCCATTGGCACTCCTTCTGGCGATAAGAAGGCCTCTAAGCAGCGACAGTTTGTGGAGGTTACTGTTG
YALI0D17402g*		CTCGTGTCGATTGTTTTGAACGACGAGGACTCAGCCAAGCGGTTTTTTGACACTCTTCAGACTGCCAAGG
YALI0D00605g*	STR3	AACGTGACTTCCCTGAGGATCTCATTCGATTGTGTGTTGGTATCGAAGACCCCGATGATCTGATTGATGA
YALI0C00253g	ILV2	CTATTGTCCTCAACAACGAGGAACAGGGTATGGTCACCCAGTGGCAGTCTCTCTTCTACGAGAACCGATA
YALI0C09636g	ILV6	GCTTCAACATTGATTCGCTGGTAGTGTGCAACACGGAGGTTGCCGATCTGTCGCGAATGACCATTGTTCT
YALI0F20702g	PDA1	CCAGTACATGCGATCCCACAACGATCCTATTTCTGGTCTCAAGGCCCACATCCTGGAGCTTAATTTCGCC
YALI0E27005g	PDB1	AACTGTCAAGCCTCTCGACACTGAGACCATTTTCAACTCCATCAAGAAGACTAACCGACTTGTCTCTGTC
YALI0F05038g*		ATTTGAAGATCTCCAGGTCCCCAACGTGACCCGAATCTTCCACTCCATCAAAAAGGCCATTGAGTATTAG
YALI0D10131g	PDC1	GGAATTGCATTCGGACCACGTCAAAATCAAGTCCGCCACCTACAACAACGTCGGCATGAAAATGCTGTTC
YALI0D06930g	PDC6	TGTTCGGATGTCAAGTTCAACACCAACAACTCACTTCTCACCCAGATCTTCTACTCGTGCATTGGTTTCA

^aY. lipolytica sequences were obtained from the Génolevures public database.

^bFor functions, see Table 1.

^cOligonucleotide probes (70-mers) were designed using the ROSO software (see Materials and Methods).

^dAn asterisk indicates a distantly related homolog of the S. cerevisiae gene.

Cell lysis, total RNA extraction, and labeling.

All solutions were prepared with 0.1% diethyl pyrocarbonate (Sigma-Aldrich)-treated water and autoclaved at 121°C for 20 min. Plastic ware, surfaces, and pipettors were decontaminated by using RNase AWAY (Molecular BioProducts, San Diego, CA).

Sample preparation.

Cells were centrifuged for 5 min at 8,200 × g and 4°C. The pellets were washed with 1× Tris-EDTA buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0; Sigma-Aldrich) and then resuspended in 150 μl of 10% N-lauroylsarcosine (Sigma-Aldrich) and 1 ml of RNeasy lysis buffer (RLT buffer) (QIAGEN, Hilden, Germany)-β-mercaptoethanol (Sigma-Aldrich) (1:0.01). The suspension was vortex mixed for approximately 3 min, poured into sterile 2-ml Eppendorf tubes (each aliquot contained 800 μl of suspension), and then stored at −80°C or used for extraction.

Total RNA isolation.

For RNA extraction, 200 mg of zirconium beads (diameter, 0.1 mm; BioSpec Products, Bartlesville, OK) and 800 μl of RLT buffer-β-mercaptoethanol were added to an aliquot containing 800 μl of a cell suspension (see above). The mixture was shaken with a FastPrep FP120 bead beating system (Bio 101, Vista, CA) for 30 s at a machine speed setting of 6.0 m·s⁻¹. Samples were cooled on ice for 1 min, and the shaking procedure was repeated a second time. Phase separation was carried out after centrifugation for 5 min at 1,700 × g and 4°C. The aqueous phase was transferred to a fresh tube, and an equal volume of 70% ethanol was added, after which extraction was performed with an RNeasy Midi kit (Qiagen) used according to the manufacturer's instructions. Total RNA was eluted directly from the RNeasy silica gel membrane in 500 μl of diethyl pyrocarbonate-treated water and immediately precipitated by addition of 50 μl of 3 M sodium acetate and 400 μl of absolute isopropanol (4°C). The contents of the tubes were mixed by inversion and incubated at −20°C for at least 2 h. The RNA was collected by centrifugation (30 min, 20,800 × g, 4°C), and the pellets were washed twice with 250 μl of cold 70% ethanol, dried for 30 min at room temperature, and resuspended in 50 μl of 1× Tris-EDTA buffer. Samples were then hydrated overnight at 4°C after addition of 0.5 μl (20 U) of RNase inhibitor (RNasin; Promega, Madison, WI). The RNA integrity was visualized by electrophoresis at 6 V·cm⁻¹ in a 1% agarose gel, which was stained with 0.3 μg·ml⁻¹ ethidium bromide (Sigma-Aldrich) and photographed under UV light. The quantity and purity were assessed by measurement of absorbance and calculation of A₂₆₀/A₂₃₀ and A₂₆₀/A₂₈₀ ratios using a spectrophotometer (Beckman DU640B; Beckman Instruments, Fullerton, CA).

Labeling of cDNA targets.

Total RNA was fluorescently labeled with a CyScribe first-strand cDNA labeling kit (Amersham Biosciences, Piscataway, NJ) without any amplification, using both anchored oligo(dT) and random nonamer priming methods together. Reverse transcription labeling reactions were performed at 42°C for 1.5 h with a thermocycler (GeneAmp PCR System 9700; Perkin-Elmer Applied Biosystems, Foster City, CA) using direct incorporation of dCTP-Cy3 (Amersham Biosciences) according to the manufacturer's instructions. The RNA template and unincorporated fluorescent nucleotides were then eliminated by chemical treatment (15 min at 37°C with 2 M NaOH). After neutralization with 2 M HEPES (pH 6.8) (Sigma-Aldrich), labeled cDNA was purified on GFX columns (CyScribe GFX purification kit; Amersham Biosciences) and then concentrated using a Microcon YM-30 filter (Millipore, Bedford, MA).

DNA array hybridization, washing, and scanning.

Hybridization was performed using a slide for each biological replicate. To reduce the nonspecific adsorption of fluorescent probes to the surface, DNA array slides were prehybridized by adding 5 μl of a 10-mg·ml⁻¹ herring sperm DNA (Promega) suspension previously heated at 95°C for 2 min and 30 μl of DIGeasy hybridization buffer (Roche Diagnostics GmbH, Mannheim, Germany) to each slide covered with a coverslip (22 by 40 mm; LifterSlip premium printed cover glass; Erie Scientific Company, Portsmouth, NH). Each slide was then put in an individual hybridization chamber (Corning, Avon, France) and immersed in a water bath for 1 h at 60°C. It was then washed in 0.1× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and dried by centrifugation at 150 × g for 3 min at room temperature before hybridization. Labeled targets and 5 μl of herring sperm DNA were heated at 95°C for 5 min for denaturation and then snap-cooled on ice. Twenty microliters of the hybridization buffer was added to the mixture and then injected under a new coverslip. The hybridization chamber was incubated in a 60°C water bath for 18 h. Slides were then immersed in 2× SSC-0.1% SDS to remove the coverslip and washed for 5 min in 2× SSC-0.1% SDS, 1× SSC, and 0.2× SSC successively, before they were dried by centrifugation. Slides were scanned at 532 nm (wavelength for Cy3 fluorescence) using a ScanArray 4000 robot (Packard Biosciences, Boston, MA) with 5-μm pixel resolution. Pictures were generated by using appropriate gains on the photomultiplier tube to obtain the highest signal intensity without saturation.

Image quantification and DNA array data analysis.

Hybridization signals were analyzed using the software QuantArray (Packard BioChip Technologies, Billerica, MA), and the mean fluorescence intensity of each spot was quantified. Prior to data analysis, we did not subtract background values or use ratio measurements (24). In order not to introduce normalization conditions based on underlying biological hypotheses (for example, using the average of some housekeeping genes with the assumption that the expression level of such genes would not vary), the signal intensity for each spot for a given condition was normalized by the median (22). The need for alternative normalization techniques arose with the realization that genes assumed to be housekeeping and “designated” by the manufacturers as such on arrays are not reliable for accurate data normalization (26). The expression level of each gene was then calculated by using the average of all individual hybridizations.

Real-time PCR conditions.

The RNA extraction and purification procedures used are described above. The cDNAs were subsequently synthesized using the SuperScript III first-strand synthesis system (Invitrogen). A mixture containing up to 5 μg of total RNA, oligo(dT)₂₀ (50 μM), and deoxynucleoside triphosphate (10 mM) was prepared, incubated at 65°C for 5 min, and then placed on ice for at least 1 min. A cDNA synthesis mixture containing 10× reverse transcriptase buffer, MgCl₂ (25 mM), dithiothreitol (0.1 M), RNaseOUT (40 U μl⁻¹), and SuperScript III reverse transcriptase (200 U μl⁻¹) was added to each RNA-primer mixture and then incubated for 50 min at 50°C. The reaction was stopped by incubation for 5 min at 85°C.

The primers used for real-time PCR were designed so that they were about 20 to 25 bases long, had a G+C content of more than 50%, and had a melting temperature of about 60°C. The length of the PCR products ranged from 90 to 150 bp. LightCycler (Roche, Mannheim, Germany) software was used to select primer sequences. All the primers were synthesized by Eurogentec (Seraing, Belgium) (Table 3).

TABLE 3.

Primers used in this study

Primer	Accession no.	Sequence (5′-3′)	Putative functiona
ARO8-R	YALI0E20977g	GGCTCCGACCCAGTTGT	Aromatic amino acid aminotransferase
ARO8-F		TTCTCCTCCGCCATCGAGTG
ARO9-R	YALI0C05258g	ACTCGGCAAACTTGAGGATCT	Aromatic amino acid aminotransferase
ARO9-F		GACCTCGATTTGTGCCTGCTG
BAT1-R	YALI0D01265g	TCCAGCTCAAGAGGACCGTAGG	Branched-chain amino acid aminotransferase
BAT1-F		CCAAGCTCCAGATCACCAAGAC
BAT2-R	YALI0F19910g	TCCAACGGCCTTGGAGTTCT	Branched-chain amino acid aminotransferase
BAT2-F		CCTCAAGCTCTACTGCTCCGA
ACT-R	YALI0D08272g	GGCCAGCCATATCGAGTCGCA	Actin
ACT-F		TCCAGGCCGTCCTCTCCC

^aAnnotations were obtained from Génolevures (http://cbi.labri.fr/Genolevures/).

SYBR green I PCR amplification was performed using a LightCycler (Roche). Amplification was carried out in a 10-μl (final volume) mixture containing 250 ng of RNA sample, 4 mM MgCl₂, 0.5 μM primer, and 1 μl of LightCycler-FastStart DNA Master SYBR green I (Roche). Five dilutions of cDNA were used to determine the efficiencies of real-time PCR. A negative control without cDNA was systematically included. The amplification procedure involved incubation at 95°C for 8 min for the initial denaturation, followed by 45 cycles consisting of (i) denaturation at 95°C for 10 s, (ii) annealing at a temperature that was 5°C below the melting temperature of the primers for 7 s, (iii) extension at 72°C for 6 s, and (iv) fluorescence acquisition (530 nm) at the end of extension. The temperature transition rate was 20°C/s for each step. After real-time PCR, a melting curve analysis was performed by continuously measuring fluorescence during heating from 65 to 95°C at a transition rate at 0.1°C/s. The threshold cycle values were determined with the LightCycler software (version 3.3), using the second derivative method. Standard curves were generated by plotting the threshold cycle values as a function of the initial log RNA concentration. PCR efficiency (E) was then calculated using the following formula: E = 10^−1/slope. The actin gene was chosen as a suitable internal control gene to normalize the results. The Pffafl method was used to calculate the change in transcript abundance normalized to the actin gene.

HPLC analyses.

Culture samples stored at −20°C were thawed at 4°C, centrifuged (2,060 × g, 15 min), and filtered using a polyethersulfone membrane filter (pore size, 0.22 μm; diameter, 33 mm; Dutscher, Brumath, France) before analysis. The lactose, glucose, galactose, lactate, KMTBA, and α-hydroxy-γ-methylthiobutyric acid contents of the filtrates were determined by high-performance liquid chromatography (HPLC) (Waters TCM HPLC; Waters, Saint Quentin en Yvelines, France) with a cation-exchange column (Aminex HPX-87H; diameter, 7.8 mm; length, 300 mm; Bio-Rad, Ivry-Sur-Seine, France) maintained at 65°C with a thermostat. The mobile phase was sulfuric acid (0.01 N) at a flow rate of 0.6 ml·min⁻¹. Most compounds of interest were detected with a Waters 486 tunable UV/visible detector regulated at 210 nm; the only exception was lactose, for which a Waters 410 differential refractometer was used. Methionine was analyzed with a reversed-phase column (Symmetry C₁₈; pore size, 100 Å; diameter, 4.6 mm; length, 100 mm; Waters). A gradient consisting of H₂O plus acetonitrile at a flow rate of 0.6 ml·min⁻¹ was applied as follows: 100% H₂O for 2.5 min, 100 to 90% H₂O for 0.5 min, 90 to 60% H₂O for 7 min, and 60 to 100% H₂O for 4 min. UV detection at 210 nm was used. All compounds were quantified using calibration curves established with pure chemicals.

Solid-phase microextraction-gas chromatography-mass spectrometry analyses.

The VSC production was analyzed by an automatic solid-phase microextraction method using a gas chromatograph (Varian CP-3800; Varian Inc., Walnut Creek, CA) and a single quadrupole mass spectrophotometer equipped with an impact electronic source (model 1200; Varian Inc.). Automation of extraction and injection was achieved with a CombiPAL autosampler (CTC Analytics, Zwingen, Switzerland). Defrosted samples (5 ml) kept at 4°C were preincubated for 2 min at 40°C with agitation at 250 rpm. The extraction was carried out using a 100-μm polydimethylsiloxane fiber (Supelco, Bellefonte, PA) for 40 min at 40°C and equivalent agitation conditions. The sample was injected by desorption at 250°C for 60 s in splitless mode using a standard Varian split/splitless injector (model 1177; Varian Inc.). The volatile compounds were carried to a nonpolar capillary column (HP-5 MS; 30 m by 0.25 mm; film thickness, 0.25 μm) by helium at a constant flow rate (1.2 ml/min). The compounds were then separated using the following temperature program. First, the temperature was maintained at 15°C for 8 min. Subsequently, the temperature was increased to 220°C at a rate of 5°C/min. Separated compounds were detected with a mass spectrometry detector. Data were collected in the range from 30 to 400 atomic mass units at a rate of 2 scans/s. Volatile compounds were identified by comparison of their ion chromatograms with the NIST/02 Mass Spectral Library (National Institute of Standards and Technology, Gaithersburg, MD). Data were analyzed using the Statgraphics Plus software (Statistical Graphics Corp., Englewood Cliffs, NJ). Values are expressed as means ± standard deviations of three replicates.

RESULTS

Substrates assimilated by Y. lipolytica for growth under conditions that mimic a cheese-ripening environment.

The abilities of Y. lipolytica to degrade lactose, lactate, and l-methionine were evaluated using a cheeselike medium. The effects of low and high concentrations of l-methionine (1 g·liter⁻¹ [LM medium] and 6 g·liter⁻¹ [HM medium], respectively) in the culture medium were also compared. Data for Y. lipolytica growth and pH evolution under both culture conditions are presented in Fig. 1. Regardless of the initial concentration of l-methionine, Y. lipolytica (inoculated at a concentration of 8 × 10³ CFU·ml⁻¹) grew exponentially during the first 36 h of cultivation and formed a population that contained about 10⁸ CFU·ml⁻¹ at 48 h. Although the l-methionine in LM medium was depleted after 48 h (Fig. 2), the final cell yield was even higher than that in HM medium (Fig. 1). This result is consistent with the greater utilization of lactate in LM medium at 72 and 96 h. This suggests that there was a shift from l-methionine catabolism to lactate catabolism after 48 h in LM medium. An increase in the pH (initially adjusted to pH 5.5 ± 0.1) was observed after 24 h of cultivation in both culture media, but the increase was slightly less in HM medium (0.3 pH unit at the end of the experiment [96 h]). The kinetics of lactose, lactate, and l-methionine consumption were also monitored during exponential and stationary phases for both culture conditions (Fig. 2). Y. lipolytica did not consume lactose in either culture medium. Since no other sugar (glucose or galactose) was initially provided in the culture medium, the major carbon sources available for Y. lipolytica growth were lactate and amino acids. As determined by HPLC analyses of culture supernatants, l-methionine was the most efficiently degraded carbon substrate in both culture conditions, independent of its initial concentration. Indeed, during growth in HM medium, almost 88% of the amino acid was consumed by Y. lipolytica, while the initial amount of lactate (∼23 g·liter⁻¹) remained unchanged until the end of the experiment (96 h). When Y. lipolytica cells were grown in LM medium, the lactate concentration started to decrease significantly at the end of the exponential phase, once l-methionine had been completely depleted. Consequently, Y. lipolytica did not assimilate lactate in the presence of l-methionine, and this amino acid was efficiently used by Y. lipolytica as a substrate for growth.

FIG. 1.

Y. lipolytica growth and pH evolution in a cheeselike medium initially containing 6 g·liter⁻¹ (filled symbols) or 1 g·liter⁻¹ (open symbols) of l-methionine. The error bars indicate standard deviations calculated using the average values for three independent determinations. Squares, growth; circles, pH.

FIG. 2.

Effect of l-methionine concentration on the degradation of lactose, lactate, and l-methionine. The error bars indicate standard deviations calculated using the average values for three independent determinations.

In silico prediction of the Y. lipolytica genes that could participate in l-methionine catabolism.

Based on the results described above, our main objective was to identify several target genes involved in the l-methionine catabolic pathway in Y. lipolytica. Since the l-methionine transamination pathway was found to be active in cheese-ripening yeasts, such as G. candidum (9), K. lactis (19), and Y. lipolytica (8), all genes that putatively encode aromatic or branched-chain aminotransferases were searched against the Y. lipolytica genome using the Génolevures public database (Table 1). Two open reading frames encoding aromatic aminotransferases were identified on the basis of sequence homology with the known ScARO8 and ScARO9 genes of S. cerevisiae. The YlARO8 (YALI0E20977g) and YlARO9 (YALI0C05258g) genes are located on the E and C chromosomes of Y. lipolytica and exhibit 50 and 28% identity at the protein sequence level with ScARO8 and ScARO9, respectively. Y. lipolytica also has two branched-chain aminotransferase genes, one with a mitochondrial targeting signal (YlBAT1) and one which is cytosolic (YlBAT2) (8). Lactic acid bacteria, mainly L. lactis, possess two enzymes, cystathionine β-lyase and cystathionine γ-lyase, which are able to convert l-methionine to MTL in a single step (1, 13). To our knowledge, in the cheese-ripening yeasts the ability of these cystathionine lyases to carry out both the α,β-elimination and α,γ-elimination reactions with the l-methionine substrate has not been studied yet. A demethiolating activity is also suspected in certain yeasts, including Y. lipolytica, under conditions simulating cheese ripening (4), but the genes involved have never been studied. By using BLASTP and S. cerevisiae cystathionine β/γ-lyase genes as queries, two homologs of ScSTR3 and ScCYS3 were found in Y. lipolytica. One gene (YALI0D00605g*) exhibits 48% identity at the protein sequence level with the cystathionine β-lyase ScSTR3 gene, and another gene (YALI0F05874g) exhibits 66% identity with the cystathionine γ-lyase ScCYS3 gene. Nevertheless, the Y. lipolytica DNA sequences had no significant homology with the l-methionine γ-lyase gene (mgl), which catalyzes the one-step degradation of l-methionine to MTL in Brevibacterium linens (3). Since Y. lipolytica has strictly oxidative metabolism, catabolism of all carbon sources occurs through the tricarboxylic acid (TCA) cycle. Consequently, pyruvate metabolism is an important branch point in carbohydrate utilization. It may control different pathways crucial for determining the fate of the imported carbohydrates. For a more detailed analysis of l-methionine catabolism, a second group of genes involved in pyruvate metabolism was searched in the Y. lipolytica genome. Table 1 shows the results of our in silico comparative analysis of the target genes of S. cerevisiae and Y. lipolytica.

Gene expression profiles in the presence of l-methionine.

We carried out a DNA array analysis of some Y. lipolytica genes that could participate in l-methionine catabolism under conditions simulating cheese ripening. A list of the selected genes and the corresponding designed oligonucleotide probes is shown in Table 2. For gene expression measurement, samples were taken at three different times, the early (48 h), mid- (72 h), and late (96 h) stationary phases, when Y. lipolytica cells were grown in medium initially containing 6 g·liter⁻¹ of l-methionine (HM medium). Hybridization signals were quantified using the software QuantArray (Packard BioChip Technologies) and were normalized by using the median (for a detailed statistical analysis, see Materials and Methods). To evaluate the quality of the DNA array data, the reproducibility of the duplicate spots on the arrays and the variations observed in replicate experiments were assessed (see Fig. S1 in the supplemental material). The data obtained from duplicate spots on the same array were highly reproducible (R² > 0.97) (see Fig. S1a in the supplemental material). To obtain an indication of the reproducibility, spot intensities were also compared for pairs of arrays hybridized with two sets of Cy3-cDNA prepared from the same RNA extract (i.e., technical repeats). The results showed that there was good reproducibility between arrays (see Fig. S1b in the supplemental material); in all cases, a correlation coefficient of >0.96 was generated. Spot intensities were also compared for pairs of cDNA arrays hybridized with two sets of Cy3-cDNAs prepared from two different RNA extracts from Y. lipolytica cells grown in identical conditions (i.e., biological repeats). As a representative example (see Fig. S1c in the supplemental material), the results also showed that there was good reproducibility between arrays (R² > 0.94). Low variability was obtained (from spot to spot, slide to slide, and cDNA synthesis to cDNA synthesis), and this served as a validation test for the use of our DNA array. Table 4 shows the changes in gene expression during the stationary phase when the transcript levels for each gene at 72 and 96 h were compared to that at 48 h. For the samples from cultures in the mid-stationary phase (72 h), the expression of fewer genes was altered compared to the expression at 48 h. In contrast, the expression levels appeared to rebound in the late exponential phase since 32% of the candidate genes were up-expressed more than twofold at 96 h. Among these genes were two genes predicted to be involved in the amino acid transamination pathway in Y. lipolytica, YlBAT2 and YlARO8, and four genes related to pyruvate metabolism, YlILV2, YlPDB1, YALI0F05038g*, and YlPDC6. Other presumptive genes associated with l-methionine catabolic pathways (YlCYS3, YlCYS4, YALI0D17402g*, and YlSTR3) were down-expressed. These first results indicate that transcript levels in Y. lipolytica cells varied during the stationary phase, and quantitative differences were apparent. Concomitant with l-methionine catabolism, identification of the YlARO8 gene, which was strongly up-expressed (2.1- to 4.3-fold), along with possible underestimation of the calculated expression ratios due to highly saturating signals at 96 h, indicates that the l-methionine transamination pathway is highly induced in Y. lipolytica.

TABLE 4.

ORFs predicted to be up-expressed or down-expressed during the stationary phase when Y. lipolytica was grown in a cheeselike medium initially containing 6 g·liter⁻¹ of l-methionine (HM medium)^a

Y. lipolytica open reading frame	Putative designation	Functionb	Change in expression during stationary phase atc:
			72 h (72 h/48 h ratio)	96 h (96 h/48 h ratio)
Methionine or aromatic or branched-chain amino acid metabolism
YALI0F19910g	BAT2	Branched-chain aminotransferase	−2.4	2.4
YALI0E20977g	ARO8	Aromatic aminotransferase	2.1	4.3d
YALI0C05258g	ARO9	Aromatic aminotransferase	−1.2	1.8
YALI0D17138g	MAP1	Methionine aminopeptidase	−1.7	−3.0
YALI0D05159g	MAP2	Methionine aminopeptidase	−1.9	−2.2
YALI0B14509g*	SAM2	S-Adenosylmethionine synthetase	−1.4	−8.5
YALI0F25641g*	SAM4	S-Adenosylmethionine-homocysteine methyltransferase	−1.3	−1.0
YALI0F05874g	CYS3	Cystathionine γ-lyase	−1.1	−1.5
YALI0E09108g	CYS4	Cystathionine β-synthase	1.1	−1.1
YALI0D17402g*		Putative cystathionine γ-synthase	−2.2	−1.9
YALI0D00605g*	STR3	Cystathionine β-lyase	−2.9	−3.2
Amino acid and carbohydrate metabolism
YALI0C00253g	ILV2	Acetolactate synthase (catalytic subunit)	−1.3	2.7
YALI0C09636g	ILV6	Acetolactate synthase (regulatory subunit)	−1.6	1.9
YALI0F20702g	PDA1	Pyruvate dehydrogenase (E1α subunit)	−1.4	1.8
YALI0E27005g	PDB1	Pyruvate dehydrogenase (E1β subunit)	1.2	5.7d
YALI0F05038g*		Probable 3-methyl-2-oxobutanoate dehydrogenase	1.1	2.2
YALI0D10131g	PDC1	Pyruvate decarboxylase	−1.6	1.4
YALI0D06930g	PDC6	Pyruvate decarboxylase	−1.4	2.1

^aThe transcript levels at 72 and 96 h (mid- and late stationary phase) were compared to those at 48 h (early stationary phase).

^bFor detailed annotation, see Table 1.

^cA negative value indicates that the transcript was repressed, and a positive value indicates that the transcript was induced. Values for transcripts that were repressed or induced more than twofold (ratio, >2.0) are indicated by bold type. At 72 h there was one gene that was induced more than twofold, and there were three genes that were repressed more than twofold. At 96 h there were six genes that were induced more than twofold and four genes that were repressed more than twofold.

^dSpot with highly saturating signals at 96 h, leading to possible underestimation of the calculated expression ratio.

Effects of l-methionine on gene expression and VSC production.

For a more detailed analysis of the l-methionine catabolic pathway in Y. lipolytica, we decided to identify genes which are transcriptionally up-regulated in the presence of l-methionine and which may affect conversion of l-methionine to VSCs. The expression profiles of 19 genes were analyzed and compared at three different times during the stationary phase (48, 72, and 96 h) when Y. lipolytica was grown in HM or LM medium. It must be pointed out that complete depletion of l-methionine was observed in LM medium at 48 h (Fig. 2), whereas 3.7 g·liter⁻¹ (61%) of l-methionine was detected in HM medium at 48 h, 1.9 g·liter⁻¹ (31%) was detected at 72 h, and 0.7 g·liter⁻¹ (12%) was detected at 96 h. Table 5 shows the results of the comparative analysis of gene expression in Y. lipolytica cells at different times for the two culture media (high- and low-l-methionine conditions). It seems clear that for the majority of the genes, there was greater expression in the presence of l-methionine (74 to 84% of the genes). For instance, the aromatic aminotransferase YlARO8 gene was found to be strongly up-regulated (11.0- to 15.1-fold) in HM medium compared to the expression in LM medium (Fig. 2). The YlBAT2 gene, which encodes a branched-chain aminotransferase, was also positively regulated in the presence of l-methionine (5.6- to 15.0-fold). In the same way, the YlARO9 transcript levels, although higher in HM medium than in LM medium, remained fairly stable over time (Table 5). The YlARO8, YlBAT2, and YlARO9 expression levels were confirmed by an independent method, real-time PCR (Table 6). Furthermore, real-time PCR showed that YlBAT1 was also up-regulated by l-methionine. No quantitative increases (ratios of expression in HM medium to expression in LM medium of <2) were observed for the YlCYS3, YlCYS4, YALI0D17402g*, and YlSTR3 transcripts in the presence of l-methionine. However, the situation seems to be more complicated than simple induction. It appears that the YlARO8 and YlBAT2 transcript levels are higher in the presence of l-methionine than in the absence of l-methionine. However, the transcript levels in the presence of l-methionine actually seem to be inversely proportional to the remaining l-methionine concentration. It is clear from Fig. 2 that in HM medium, l-methionine is being degraded at 48 h, yet the levels of the YlARO8 and YlBAT2 transcripts are higher at 72 and 96 h, when the l- methionine levels are lower. The results described above show that the transamination pathway is very important in the breakdown of l-methionine in Y. lipolytica and that induction of this pathway is modulated by the l-methionine concentration. Concomitant with these transcriptional analyses, biochemical measurements were obtained for the cultures (Table 7). In HM medium, l-methionine degradation was accompanied by important KMTBA accumulation after 24 h. The maximal concentration of the latter compound (3.07 ± 0.07 mmol·liter⁻¹) was detected at 60 h, and then the compound was gradually degraded. Since the enzymatic reduction of KMTBA to α-hydroxy-γ-methylthiobutyric acid, previously observed in S. cerevisiae and Kluyveromyces sp. (20), was not detected in Y. lipolytica cultures, the production of VSCs at 72 and 96 h was measured and compared in both culture conditions (Table 7). MTL, DMDS, and DMTS were detected only in HM medium at 72 and 96 h, which indicates that there is a strong relationship between VSC production and KMTBA disappearance. Traces of methylthiopropanal (methional) were also detected in HM medium at 72 h.

TABLE 5.

Genes up- and down-regulated by l-methionine^a

Y. lipolytica open reading frame	Putative gene designation	Functionb	Change (fold)(HM medium/LM medium ratio) atc:
			48 h	72 h	96 h
Methionine or aromatic or branched-chain amino acid metabolism
YALI0F19910g	BAT2	Branched-chain aminotransferase	1.5	5.6	15.0
YALI0E20977g	ARO8	Aromatic aminotransferase	1.1	11.0	15.1d
YALI0C05258g	ARO9	Aromatic aminotransferase	1.7	1.7	2.2
YALI0D17138g	MAP1	Methionine aminopeptidase	1.2	1.0	−2.6
YALI0D05159g	MAP2	Methionine aminopeptidase	1.1	1.9	1.3
YALI0B14509g*	SAM2	S-Adenosylmethionine synthase	−1.0	−1.6	−6.7
YALI0F25641g*	SAM4	S-Adenosylmethionine-homocysteine methyltransferase	1.7	1.8	1.5
YALI0F05874g	CYS3	Cystathionine γ-lyase	−1.3	−1.2	−1.5
YALI0E09108g	CYS4	Cystathionine β-synthase	1.4	1.2	1.5
YALI0D17402g*		Putative cystathionine γ-synthase	1.5	1.7	−1.0
YALI0D00605g*	STR3	Cystathionine β-lyase	1.6	−1.5	−2.1
Amino acid and carbohydrate metabolism
YALI0C00253g	ILV2	Acetolactate synthase (catalytic subunit)	6.0	1.1	1.7
YALI0C09636g	ILV6	Acetolactate synthase (regulatory subunit)	3.2	1.1	2.4
YALI0F20702g	PDA1	Pyruvate dehydrogenase (E1α subunit)	2.0	−1.0	1.3
YALI0E27005g	PDB1	Pyruvate dehydrogenase (E1β subunit)	1.5	4.6	11.9d
YALI0F05038g*		Probable 3-methyl-2-oxobutanoate dehydrogenase	−3.2	1.4	1.4
YALI0D10131g	PDC1	Pyruvate decarboxylase	2.4	1.2	1.3
YALI0D06930g	PDC6	Pyruvate decarboxylase	2.1	2.7	2.9

^aSamples were taken at early (48 h), mid- (72 h), and late (96 h) stationary phase when Y. lipolytica was grown in a cheeselike medium under high- or low-l-methionine conditions. Complete depletion of l-methionine was observed in LM medium at 48 h, whereas 3.7 g·liter⁻¹ (61%) of l-methionine was detected in HM medium at 48 h, 1.9 g·liter⁻¹ (31%) of l-methionine was detected at 72 h, and 0.7 g·liter⁻¹ (12%) of l-methionine was detected at 96 h (Fig. 2).

^bFor detailed annotation, see Table 1.

^cA negative value indicates that the transcript was repressed, and a positive value indicates that the transcript was induced. Values for transcripts that were repressed or induced more than twofold (ratio, >2.0) are indicated by bold type. At 48 h there were five genes that were induced more than twofold and one gene that was repressed more than twofold. At 72 h there were four genes that were induced more than twofold, and no gene was repressed more than twofold. At 96 h there were six genes that were induced more than twofold and three genes that were repressed more than twofold.

^dSpot with highly saturating signals at 96 h, leading to possible underestimation of the calculated expression ratio.

TABLE 6.

Comparison of transcript level ratios for aminotransferase-encoding genes obtained using DNA array or real-time PCR^a

Y. lipolytica open reading frame	Putative designation	Functionb	Change (fold) (HM medium/LM medium ratio)
			DNA array	Real-time PCR
YALI0D01265g	BAT1	Branched-chain aminotransferase	NDc	15
YALI0F19910g	BAT2	Branched-chain aminotransferase	15	13
YALI0E20977g	ARO8	Aromatic aminotransferase	15d	26
YALI0C05258g	ARO9	Aromatic aminotransferase	2.2e	0.5

^aSamples were taken at late stationary phase (96 h) when Y. lipolytica was grown in a cheeselike medium initially containing 6 g·liter⁻¹ of l-methionine (HM medium) or 1 g·liter⁻¹ of l-methionine (LM medium).

^bFor detailed annotation, see Table 1.

^cND, not detected.

^dSpot with highly saturating signals, leading to possible underestimation of the calculated expression ratios.

^eSpot with very low signal intensity (close to the detection level), leading to possible inaccuracy in the calculated expression ratio.

TABLE 7.

Production of KMTBA and VSCs by Y. lipolytica cells growing in a cheeselike medium under high- or low-l-methionine conditions

Compound	Time (h)	Production ina:
		HM medium	LM medium
KMTBA	0	ND	ND
	24	0.60 ± 0.03	tr
	48	2.90 ± 0.12	tr
	60	3.07 ± 0.07	ND
	72	2.86 ± 0.10	ND
	96	2.10 ± 0.11	ND
MTL	72	19.0 ± 1.4	ND
	96	22.9 ± 2.6	ND
DMDS	72	125.7 ± 47.2	0.2 ± 0.1
	96	52.8 ± 3.5	0.1 ± 0.0
DMTS	72	0.1 ± 0.0	tr
	96	0.5 ± 0.1	tr
Dimethyl sulfide	72	0.1 ± 0.0	ND
	96	0.3 ± 0.0	ND
Methylthiopropanal	72	tr	ND
	96	ND	ND

^aThe production of KMTBA is expressed in mmol·liter⁻¹, and the production of the other compounds is expressed in 10² peak area·CFU⁻¹. ND, not detected; tr, trace (<0.1 mmol·liter⁻¹ or <10 peak area·CFU⁻¹).

DISCUSSION

Cheese-ripening yeasts develop during the early stage of ripening, and it is generally thought that they neutralize the curd by consuming lactate (11) and may also contribute to the formation of VSCs from l-methionine (4, 8, 19, 20). In the present study, we assessed the ability of Y. lipolytica to grow, deacidify, and generate VSCs under culture conditions that mimic the cheese-ripening environment. The typical pH range of the curd is 4.4 to 5.5 depending on the type of cheese. It is possible to determine a range of maximum methionine concentrations available in the curd (4.8 to 6 g/kg), which correspond to the maximum methionine content of the curd. Depending on the proteolysis of the curd, the amount of methionine available may vary considerably.

As expected (6), Y. lipolytica is unable to assimilate lactose as a carbohydrate substrate. Genes that encode β-galactosidase (LAC4) and lactose permease (LAC12) in Kluyveromyces sp. were not detected in the Y. lipolytica genome. Consequently, lactate and l-methionine were evaluated as readily available carbon sources for Y. lipolytica growth in a cheeselike medium. The effects of low and high concentrations of l-methionine in the culture medium were evaluated. For both culture media, HPLC analyses revealed that Y. lipolytica prefers l-methionine to lactate as a carbon source. Interestingly, in silico analysis revealed a striking redundancy of putative high-affinity l-methionine permease MUP1 genes in the Y. lipolytica genome (YALI0D16137g, YALI0F03498g, YALI0F25795g, YALI0D19646g, and YALI0F07018g), and the number is higher than the number in other cheese-ripening yeast genomes, like the genomes of S. cerevisiae, K. lactis, and Debaryomyces hansenii. This suggests that the methionine transporters may provide a competitive advantage to Y. lipolytica, allowing it to grow rapidly on l-methionine. Once the l-methionine in the medium has been completely depleted, Y. lipolytica assimilates lactate as a carbon source. Consequently, two independent phenomena could explain the great increase in pH in both culture conditions: (i) an increase in the ammonia release concomitant with the l-methionine degradation and (ii) deacidification of the medium due to lactate consumption. The observation that Y. lipolytica could rapidly grow and neutralize the two culture media, regardless of the carbon source used for growth, is also interesting. It suggests that Y. lipolytica has a positive effect on the development of acid-sensitive microorganisms, such as the coryneform bacteria (e.g., B. linens) during ripening.

Furthermore, several genes associated with l-methionine catabolism and pyruvate metabolism in Y. lipolytica were simultaneously investigated at the transcriptional level. Gene expression profiles were analyzed and compared when Y. lipolytica cells were grown in HM and LM media. Changes in the expression of some genes was also observed when the l-methionine concentration in the growth medium was changed. The YlARO8, YlBAT1, and YlBAT2 genes (predicted to be involved in the amino acid transamination pathway) were found to be strongly modulated by l-methionine, clearly indicating their involvement in the l-methionine transamination step in Y. lipolytica. One of these genes, YlARO8, is most strongly modulated in HM medium. The induction of the YlBAT1 gene by its substrate, l-methionine, is in good agreement with results showing that the overexpression of the YlBAT1 gene significantly increased l-methionine transamination, as well as VSC production (8). In contrast, the expression levels of the YlARO9 gene were hardly modulated by the l-methionine concentration. Therefore, the observation that the l-methionine transamination step is highly active in Y. lipolytica was confirmed; there was transient accumulation of the transamination product, KMTBA, in HM medium after 48 h of cultivation, which coincides with the maximum rate of l-methionine consumption. Furthermore, the YlARO8 gene product has been overproduced in Escherichia coli and purified. We found that YlAro8p had transaminase activity and was highly active with l-methionine (data not shown). Other studies to evaluate the specificity of other Y. lipolytica transaminases with amino acids, including l-methionine, will be done. In K. lactis, KlARO8.1 and KlARO8.2, two homologs of YlARO8, were found to be responsible for l-methionine aminotransferase activity (19). We therefore suspect that YlAro8p is involved in l-methionine transamination in Y. lipolytica.

Alpha-ketoglutarate is generally the preferred amino group acceptor for transamination reactions and is usually generated through sugar catabolism. Since Y. lipolytica has strictly oxidative metabolism, alpha-ketoglutarate is most probably provided through the TCA cycle. Transcriptional analysis of genes related to pyruvate metabolism showed that YlILV6, YlPDB1, and YlPDC6 were strongly up-regulated in the presence of l-methionine. The fact that the YlPDB1 and YlPDC6 genes were also highly expressed at late stationary phase in HM medium may reflect the continued need for TCA cycle intermediates as precursors for l-methionine catabolism. It may also be attributable to the utilization of pyruvate as an amino group acceptor for YlAro8p, YlBat1p, and YlBat2p, as previously described for ScAro9p (21).

Our results suggest that the conversion of l-methionine to MTL is initiated by a transamination step. This suggestion is supported by biochemical data which show that in HM medium, the transamination product KMTBA yields were significantly increased together with VSC production. In the yeast G. candidum, KMTBA was found to transiently accumulate, and its degradation corresponded to an overall increase in the production of VSCs (5, 9). Our data suggest that in Y. lipolytica, l-methionine transamination involves three major genes, YlARO8 (YALI0E20977g), YlBAT1 (YALI0D01265g), and YlBAT2 (YALI0F19910g), whose levels of expression were increased in HM medium. Although a KMTBA-demethiolating activity has been measured in Y. lipolytica (4), the mechanism which leads to the demethiolation of KMTBA to MTL could be studied at the transcriptional level in Y. lipolytica.