Production of l-lactic acid by the yeast Candida sonorensis expressing heterologous bacterial and fungal lactate dehydrogenases

Background Polylactic acid is a renewable raw material that is increasingly used in the manufacture of bioplastics, which offers a more sustainable alternative to materials derived from fossil resources. Both lactic acid bacteria and genetically engineered yeast have been implemented in commercial scale in biotechnological production of lactic acid. In the present work, genes encoding l-lactate dehydrogenase (LDH) of Lactobacillus helveticus, Bacillus megaterium and Rhizopus oryzae were expressed in a new host organism, the non-conventional yeast Candida sonorensis, with or without the competing ethanol fermentation pathway. Results Each LDH strain produced substantial amounts of lactate, but the properties of the heterologous LDH affected the distribution of carbon between lactate and by-products significantly, which was reflected in extra-and intracellular metabolite concentrations. Under neutralizing conditions C. sonorensis expressing L. helveticus LDH accumulated lactate up to 92 g/l at a yield of 0.94 g/g glucose, free of ethanol, in minimal medium containing 5 g/l dry cell weight. In rich medium with a final pH of 3.8, 49 g/l lactate was produced. The fermentation pathway was modified in some of the strains studied by deleting either one or both of the pyruvate decarboxylase encoding genes, PDC1 and PDC2. The deletion of both PDC genes together abolished ethanol production and did not result in significantly reduced growth characteristic to Saccharomyces cerevisiae deleted of PDC1 and PDC5. Conclusions We developed an organism without previous record of genetic engineering to produce L-lactic acid to a high concentration, introducing a novel host for the production of an industrially important metabolite, and opening the way for exploiting C. sonorensis in additional biotechnological applications. Comparison of metabolite production, growth, and enzyme activities in a representative set of transformed strains expressing different LDH genes in the presence and absence of a functional ethanol pathway, at neutral and low pH, generated a comprehensive picture of lactic acid production in this yeast. The findings are applicable in generation other lactic acid producing yeast, thus providing a significant contribution to the field of biotechnical production of lactic acid.


Background
A variety of new products based on polymerized lactic acid are constantly being developed, increasing the demand for lactic acid. L-Lactic acid is typically produced in large quantities by carbohydrate fermentation by lactic acid bacteria. The fermentation is efficient at near neutral pH, controlled with neutralizing chemicals and generating lactate salts [1]. The undissociated (free) lactic acid rather than the salt of the acid is the required product for the polymerization reaction and additional processing is necessary to recover free lactic acid. Yeast are considered as attractive alternative hosts for lactic acid production at low pH because they are more acid tolerant than lactic acid bacteria. Low pH production would decrease the need for neutralizing chemicals. Several groups have demonstrated efficient production of L-lactic acid by S. cerevisiae expressing a heterologous gene encoding lactate dehydrogenase (LDH) [2][3][4]. The LDH gene has also been introduced into some nonconventional yeast species that have advantageous characteristics such as good acid tolerance or ability to metabolize carbohydrates that S. cerevisiae does not naturally consume. For example, Kluyveromyces lactis [5,6], Pichia stipitis [7], Candida boidinii [8] and Candida utilis [9] have been shown to produce high concentrations of lactic acid. In addition, e.g. Zygosaccharomyces bailii [10], and Kluyveromyces marxianus [11] expressing LDH have been shown to produce lactic acid.
One of the main issues related to lactic acid production using yeast, especially S. cerevisiae, is the ability of the yeast to produce ethanol in the presence of excess glucose. Even though the expression of the lactate dehydrogenase gene can itself decrease the conversion of glucose to ethanol to some extent [12], modification of the ethanol pathway, to remove competition with lactate dehydrogenase for pyruvate, has proved an effective way to increase the yield of lactic acid on glucose [2]. A single deletion of the pyruvate decarboxylase gene PDC1, encoding the main PDC isoenzyme in S. cerevisiae, decreased PDC activity moderately but the expression of PDC5 was enhanced in the absence of PDC1 [13,14]. A double deletion of PDC1 and PDC5 in lactic acid producing S. cerevisiae strains decreased ethanol production and increased lactic acid yield significantly, but still some ethanol was produced because the PDC6 gene was intact [2]. In addition, the growth of the PDC1 and PDC5 deleted strain was severely reduced on glucose medium [2], which may be undesirable in a production process. In contrast, the deletion of the only pyruvate decarboxylase encoding gene, PDC1, from K. lactis had only a mild effect on growth, was sufficient to eliminate ethanol production and improve lactate production [6].
Efficiency of lactic acid production will be affected not only by the choice of the host strain but also to some extent by the enzymatic properties of different LDH enzymes. LDH genes from different organisms result in different LDH activity levels and concentrations of produced lactic acid when expressed in the same S. cerevisiae host strain [4,10,15]. LDH activity level was also affected by the copy number of the LDH gene in the host [16].
We developed vectors and techniques for introducing genetic modifications into the non-conventional yeast C. sonorensis which enabled its genetic engineering for the first time. C. sonorensis is a methylotrophic yeast that readily ferments glucose to ethanol, utilizes several carbon sources including the pentose sugars xylose and arabinose, is relatively tolerant to acidic conditions, and has simple nutritional requirements [17,18]. The objective of the present work was to construct C. sonorensis strains expressing a heterologous LDH gene and containing an intact or modified ethanol fermentation pathway, and to characterize the effects of these modifications on lactic acid production. Strains expressing the L-lactate dehydrogenase encoding genes from Lactobacillus helveticus, Bacillus megaterium, and from the fungus Rhizopus oryzae were compared and evaluated for their relative efficiency in lactate production by C. sonorensis. The effect of increased LDH activity level as a result of expressing multiple LDH gene copies per genome was determined in strains containing a functional ethanol pathway and in strains deleted of the PDC genes. These studies revealed that production of lactate, ethanol and pyruvate was determined by the PDC modifications, the choice of LDH enzyme, and the LDH enzyme activity level, which varied with the LDH gene copy number.

Development of tools for C. sonorensis transformation
Growth inhibition tests in YPD medium supplemented with antibiotics in a range of concentrations suggested that ≥200 μg/ml of G418 was inhibitory and thus could probably be used for the selection of transformants. Furthermore, C. sonorensis was melibiase (α-galactosidase) negative suggesting that transformants could be selected based on growth on the disaccharide melibiose, or screened on the chromogenic substrate X-α-gal.
Initial attempts to transform C. sonorensis with pTEF/ Zeo, pMI203 and pMI205, containing the zeocin resistance gene expressed under heterologous promoters, did not yield selectable transformants. For this reason, a genomic library was constructed to isolate C. sonorensis promoters to direct the expression of LDH and marker genes. Genes encoding highly expressed glycolytic phosphoglycerate kinase (PGK1) and glyceraldehyde-3-phosphate dehydrogenase (TDH1) were isolated by hybridization with the C. albicans PGK1 and the S. cerevisiae TDH1 probes, respectively. Sequences upstream of the predicted open reading frames (i.e. promoters) of the PGK1 and TDH1 genes were subsequently cloned upstream of the ORFs of the marker genes MEL5 and G418 R . C. sonorensis was successfully transformed with each of the four linearized constructs using the lithium acetate method. Both the direct selection for MEL5-containing transformants on minimal medium containing melibiose as the sole carbon source and the detection of blue colour on nonselective X-α-gal plates were suitable methods for the isolation of transformants. Southern analyses indicated that the integration sites varied between the transformants (data not shown).
Isolation of PDC1 and PDC2 and demonstration of their functional roles in ethanol production A 0.6 kb fragment of a PDC sequence homologue was amplified by PCR from C. sonorensis DNA using degenerate primers for PDC. The fragment was used as a probe to isolate the corresponding full length PDC1 gene (acc. AM420319) from the genomic library. Additional PCR reactions with the same degenerate PDC primers revealed another putative PDC sequence present in C. sonorensis, and a full length PDC2 gene (acc. AM420320) was isolated. The predicted open reading frames of the PDC1 and PDC2 genes code for 575 and 568 amino acids, respectively, and have 62% amino acid sequence identity to each other, 68% and 59% identities with Ogataea parapolymorpha PDC (acc. EFW96140.1), and 61% and 63% identities with Candida boidinii PDC1 (acc. BAI43440), respectively, as the best hits identified in database searches (BLASTP 2.2.26). This supports the hypothesis that the two genes code for pyruvate decarboxylases.
To assess the functional role of the two PDC genes, strains deleted of either one or both PDC genes were constructed. C. sonorensis PDC1 was replaced by the MEL5 marker using pMI267, while PDC2 was replaced by the G418 R marker using pMI287. Screening of transformants for decreased ethanol production enabled detection of candidate PDC2-deleted strains, but candidate PDC1-deleted strains could not be distinguished. Southern analyses were used to screen for PDC1-deletion and to confirm PDC2 deletion by the absence of PDC1or PDC2-specific hybridization signals ( Figure 1A) and the appearance of transformation marker-specific signals of appropriate size. Transformants deleted of PDC1 or PDC2 were found at 15% and 5% frequency, respectively.
The parent and the various PDC deletion strains were grown in YP-5% glucose and pyruvate decarboxylase enzyme activity was measured. PDC activity was similar in the wild type and pdc1Δ strains, but was reduced to 20% of the wild type activity or lower in the pdc2Δ strain. The pdc1Δ pdc2Δ strain had essentially no detectable PDC activity ( Figure 1B).
Deletion of PDC1 or PDC2 affected ethanol production differently. Strains with an intact PDC2 produced similar amounts of ethanol irrespective of the presence or absence of PDC1 ( Figure 1D). Deletion of PDC2 alone caused a large decrease in ethanol production, but deletion of both PDC1 and PDC2 was necessary to eliminate ethanol production ( Figure 1D). These data, together with the enzyme activity measurements ( Figure 1B) and Northern analyses on the expression of C. sonorensis PDC1 and PDC2 genes (data not shown) demonstrate that PDC2 codes for a PDC isoenzyme that is abundant and is the main enzyme responsible for directing pyruvate to acetaldehyde and further to ethanol production.
The pdc1Δ pdc2Δ and pdc2Δ strains excreted significantly more pyruvate than strains with an intact PDC2 ( Figure 1E). However, the pdc2Δ strain consumed pyruvate when ethanol was being produced, while no net reduction in pyruvate concentration was observed with the ethanol non-producing pdc1Δ pdc2Δ strain.
The pdc1Δ pdc2Δ strains utilized glucose the slowest ( Figure 1F), the two strains with an intact PDC2 the fastest, and pdc2Δ showed an intermediate glucose consumption rate. PDC2 deletion also resulted in an approximately 50% decrease in the final biomass, while PDC1 deletion did not affect biomass accumulation ( Figure 1C).
Lactate and ethanol production with strains expressing different LDH genes LDH genes from three different sources, L. helveticus, B. megaterium and R. oryzae, were separately expressed in C. sonorensis under control of the C. sonorensis PGK1 promoter. Integration of the LDH gene was targeted into the PDC1 locus to provide a uniform set of strains, which produce both ethanol and lactic acid, for comparison.
Strains expressing LhLDH (pMI257 transformants), BmLDH (pMI265 transformants), RoLDH (pMI266 transformants) or no LDH (pMI267 transformants), produced different amounts of lactate, ethanol and biomass and consumed glucose at different rates ( Figure 2). The pdc1Δ::LhLDH and pdc1Δ::BmLDH strains produced similar lactate concentrations, whereas the pdc1Δ::RoLDH strain produced significantly less lactate than the two other LDH strains ( Figure 2A). Ethanol was produced at the highest rate by wild type C. sonorensis and the pdc1Δ strain without LDH, and ethanol production by the pdc1Δ::RoLDH strain was only slightly slower ( Figure 2B). These strains consumed glucose at a higher rate than the pdc1Δ::BmLDH and pdc1Δ::LhLDH strains ( Figure 2C). The pdc1Δ::BmLDH strain produced ethanol and consumed glucose at higher rates than the pdc1Δ::LhLDH strain ( Figure 2B and C). However, they produced comparable maximum lactate and ethanol concentrations (Figure 2A and B) and final yields on glucose ( Figure 2D) even though lactate yield on glucose for the pdc1Δ:: LhLDH strain was higher than that for the pdc1Δ::BmLDH strain during the first 40 h of the cultivation. When lactate production per gram biomass was assessed, the pdc1Δ:: RoLDH strain was the least and the pdc1Δ::LhLDH strain the most efficient in converting glucose to lactate (data not shown). The final biomass produced by the strains lacking LDH was higher (OD 600 =22) than that of the LDH strains, in particular when compared to the pdc1Δ::LhLDH and pdc1Δ::BmLDH strains (OD 600 =10) ( Figure 2D).
The effect of multiple LDH gene copies on lactate and ethanol production Strains containing 1 to 3 copies of the LhLDH or BmLDH gene integrated at non-homologous sites in the genome were identified by Southern analysis (data not shown). LDH enzyme activity increased with increasing LDH copy number, but the volumetric lactate production did not increase (data for LhLDH shown in Figure 3A and 3B). The yield of lactate on glucose did increase with increasing LDH activity and copy number (e.g. at 48 h 0.28, 0.34, and 0.40 g lactate / g glucose with 1, 2, and 3 LDH copies) owing to significant reduction in ethanol production, glucose consumption ( Figure 3C and D) and biomass production (data not shown).

Comparison of the different LDH genes in a PDC negative strain background
Representative PDC1-deleted strains, each expressing a different LDH gene, or no LDH, were transformed with the PDC2 replacement cassette from pMI287 to enable comparison of the different LDH strains in the absence of PDC enzyme activity and ethanol production.
The origin of the LDHs had a greater effect on the efficiency of lactate production in the ethanol nonproducing pdc1Δ pdc2Δ transformants ( Figure 4) than in the pdc1Δ PDC2 transformants ( Figure 3). The pdc1Δ:: LhLDH pdc2Δ strain produced 2-fold and 3-fold higher lactate concentrations than the pdc1Δ::BmLDH pdc2Δ and the pdc1Δ::RoLDH pdc2Δ strains, respectively ( Figure 4A). Lactate contributed 92% (LhLDH), 72% (BmLDH) or 59% (RoLDH) of the detected extracellular metabolites. All strains excreted pyruvate but the concentrations differed depending on the presence and type of LDH ( Figure 4B). The highest pyruvate concentration was produced by the pdc1Δ pdc2Δ strain without LDH. Of the LDH strains, the LhLDH strains produced the lowest and the RoLDH strains the highest pyruvate concentration ( Figure 4B and D), analogous to the ethanol concentrations produced by the pdc1Δ PDC2 strains containing the corresponding LDH gene ( Figure 2B). Introduction of any of the three LDH genes enhanced glucose consumption, compared with the pdc1Δ pdc2Δ strain lacking LDH, the LhLDH strain being the most efficient in this respect ( Figure 4C).  Production of lactate and pyruvate was accompanied with a decrease in the pH of the culture media to pH 3.3 -3.5 (data not shown).
LDH enzyme activities were measured with or without fructose-1,6-diphosphate at two different pH values due to the differences in the optimal conditions for the individual enzymes [19,20]. The LDH enzyme activity measured in vitro in the PDC negative strains containing one copy of LhLDH, BmLDH or RoLDH directly correlated with the lactate amount measured; the LhLDH strain that produced the highest final lactate concentration also had the highest enzyme activity ( Figure 4E).
Addition of a second BmLDH copy, integrated in the PDC2 locus, increased the final lactate concentration by 30% ( Figure 5), reduced pyruvate accumulation by 15%, and enhanced glucose consumption compared to the single copy BmLDH strain. Based on these data it appears that the level of BmLDH enzyme activity in the single copy BmLDH strain restricted lactate production. In comparison, a second copy of LhLDH resulted in small (< 9%) but significant increase in lactate ( Figure 5) and decrease in pyruvate production (not shown). Even so, cessation of lactate production still occurred and was not overcome by increasing LDH copy number, which suggests that factors other than LDH dosage prevented lactate accumulation in the C. sonorensis cultures.

The correlation to lactate concentration of the concentration of CaCO 3 added as a neutralizing agent
In the previous experiments less than 14 g/l lactate was produced in medium with no pH buffering (final pH 3.3). The pdc1Δ::LhLDH pdc2Δ::LhLDH strain, was also grown in YP-10% glucose medium supplemented with calcium carbonate (CaCO 3 ) concentrations from 5 to 30 g/l as a neutralizing agent to control the pH and to determine the relationship between free lactic acid and total lactate production. The final pH in the cultures was between pH 3.5 and 4 ( Figure 6A), around the pKa of lactic acid (pH 3.8). The total lactate concentration (24 to 66 g/l), lactate yield on glucose, and final pH increased with increasing CaCO 3 concentration, but the concentration of free lactic acid varied relatively little between the conditions and was maximal, 19 g/l, at low CaCO 3 concentrations ( Figure 6A). The proportion of free lactic acid in the total lactate decreased with increasing CaCO 3 concentration and final pH from~80% at final pH 3.5 to~20% at final pH 4. A similar correlation between CaCO 3 and lactate concentrations was also observed on YNB-10% glucose medium (Figure 6 B) but the lactate concentration was 10 g/l higher in rich YP medium than in YNB-medium at each CaCO 3 concentration between 5 and 25 g/l ( Figure 6). The pH in YP and YNB media were similar at each CaCO 3 concentration although the lactate concentrations differed.
Lactate production in cultivations containing a nonlimiting concentration of CaCO 3 LhLDH, BmLDH and RoLDH strains, with and without PDC modifications, were studied for lactate production on YNB-10% glucose minimal medium using a twostage cultivation protocol with separate biomass generation and lactate production phases. The production phase was inoculated with a biomass concentration of 5 g/l cell dry weight and the pH was maintained above 5.7 by addition of CaCO 3 . The LhLDH, BmLDH and RoLDH strains produced up to 93, 88 and 78 g/l lactate, respectively (Table 1). Glucose consumption and lactate production for representative pdc1Δ::LDH pdc2Δ are shown in Figure 7. The lactate production rate during the first 24 hours was the highest with the pdc1Δ::LhLDH pdc2Δ strain (3.3 g/l/h) followed by the pdc1Δ::BmLDH pdc2Δ (2.0 g/l/h) and pdc1Δ::RoLDH pdc2Δ (1.5 g/l/h). A visible calcium lactate precipitate was formed in some cultivations of the LhLDH or BmLDH strain, but was never formed in the RoLDH cultivations. The LhLDH strains produced on average 5 g/l more lactate than the corresponding BmLDH strains when the PDC1 gene or both PDC genes were deleted, but the difference was not statistically significant. The optical purity of the lactate was high, since the concentration of D-lactate was below 0.6 g/l (determined enzymatically from samples containing the maximum lactate concentration). Thus more than 99% of the total lactate was L-lactate.
The ethanol concentrations were small even for the strains with intact PDC1 and/or PDC2 (Table 1) in YNB-10% glucose medium with CaCO 3 . The RoLDH strains with an intact PDC2 produced significantly (p < 0.05) more ethanol than the LhLDH or BmLDH strains with an intact PDC2 (Table 1).  Pyruvate concentrations produced by the pdc1Δ::LDH pdc2Δ strains were significantly different (p < 0.05) for LhLDH, BmLDH, and RoLDH strains. The pyruvate concentration was also lower for the strains containing two copies of BmLDH or LhLDH than for the corresponding strains with a single LDH gene copy (Table 1), as in the non-buffered cultivations (see above).
In the absence of both PDC genes no ethanol was produced. As the by-product concentrations for all strains were extremely small relative to lactate concentrations, and decrease in ethanol concentration was accompanied with an increase in pyruvate concentration, double PDC deletions did not result in an increase in lactate concentration or yield on glucose.

Intracellular lactate and pyruvate concentrations
Intracellular and extracellular lactate concentrations were correlated in PDC positive BmLDH strains cultivated in CaCO 3 -buffered minimal media. Cells in CaCO 3 -buffered medium had more intracellular lactate compared to extracellular lactate at the beginning of the cultivations (0 and 8 hours). At the end of cultivation (48 hours) the intra-and extracellular lactate concentrations were similar (up to 80 g/l) ( Figure 8A).
The intracellular pyruvate concentration was higher in pdc1Δ pdc2Δ strains than in strains containing an intact PDC2 (pdc1Δ PDC2 or PDC1 PDC2), as was the extracellular concentration ( Figure 8B).

Discussion
New molecular tools have enabled genetic engineering of C. sonorensis for the first time. The antibiotic marker gene G418 R and the non-antibiotic marker MEL5 were expressed under the control of endogenous PGK1 or TDH1 promoters, and the PGK1 promoter was additionally used to express three different LDH genes. Targeted integration into PDC1 and PDC2 through homologous recombination was common but not as frequent as integration into non-homologous sites in the genome. In addition to PDC loci, homologous integration into the PGK1 locus occurred in some transformants when the marker gene was located between two identical PGK1 promoter copies in the construct, e.g. in pMI257 or pMI265.
C. sonorensis has two non-allelic PDC genes, PDC1 and PDC2, both of which contribute to ethanol production. The PDC2 gene encodes the major isoenzyme. PDC2 deletion resulted in a decrease in growth, glucose utilization and ethanol production rates, and in an increase in pyruvate levels. In contrast, the PDC1 deleted strain did not noticeably differ from the parent strain in respect to these parameters, and 85% of PDC activity was retained. Both intra-and extracellular pyruvate concentrations were significantly increased in the C. sonorensis pdc1Δ pdc2Δ strain, compared with strains containing an intact PDC2 gene. The expression of any of the three LDH genes in the C. sonorensis pdc1Δ pdc2Δ strain background provided an alternative route for pyruvate metabolism and NAD + regeneration, and was accompanied with a significant decrease in pyruvate accumulation, particularly in the LhLDH strain. LDH expression in a pdc1Δ pdc2Δ strain also enhanced glucose consumption in non-buffered medium, with LhLDH having a greater positive effect than the other two LDHs had. Even so, glucose consumption by a pdc1Δ::LhLDH pdc2Δ strain was slow relative to the ethanol producing LDH strains in non-buffered medium, as also observed in S. cerevisiae [2]. C. sonorensis strains expressing LDH from L. helveticus, B. megaterium or R. oryzae, showed characteristic differences in the conversion of glucose to lactate and byproducts, demonstrating that the properties of the LDH enzyme have a fundamental impact on carbon distribution at the pyruvate branch point. In general, the concentration and yield on glucose of the by-products ethanol (PDC+) or pyruvate (pdc1Δ pdc2Δ) were the highest with the RoLDH strains and the lowest with the LhLDH strains. The efficiency of lactate production corresponded to the LDH enzyme activity, with the LhLDH strain having the highest, the BmLDH strain intermediate, and the RoLDH strain the lowest in vitro activity in single-copy LDH strains. S. cerevisiae strains expressing a LDH from Lactobacilli had higher LDH activity than a BmLDH expressing strain [15]. Data on the properties of the LDH enzymes used are limited, but significant differences have been reported [15,19,20]. The kinetic properties such as the pH optimum, affinity for pyruvate and the cofactor NADH, inhibition by a high substrate or product concentration, and the equilibrium of the reaction would determine the effectiveness of each enzyme in vivo. Indeed, the C. sonorensis strains with different LDH enzymes produced different final extracellular lactate and pyruvate concentrations, which were shown to correlate with intracellular concentrations.
Concentrations of intracellular lactate have not been reported for LDH expressing yeasts to date. The present initial work found that lactic acid producing C. sonorensis cells harvested from un-buffered cultures contained a significant intracellular lactic acid concentration. This may interfere with multiple cellular functions, but in spite of this, the cells were able to excrete lactate. It has been proposed that lactate export is energy dependent and uses ATP in S. cerevisiae [21]. The lactate and acetate transporters JEN1 and ADY2 that are known to import lactate are also involved in lactate excretion, but another presently unknown lactate export mechanism also exists in S. cerevisiae [22,23].
Lactic acid accumulation decreases the pH of the culture medium leading to an increase in the proportion of undissociated lactic acid in the medium. At pH 4, a fraction of the lactic acid will be undissociated. Undissociated acid is believed to re-enter the cell also via passive diffusion. In the cytosol, at neutral pH, it will dissociate to form the lactate anion and proton, thus increasing the ATP demand for lactate export [21]. Cytosolic acidification caused by lactic acid may eventually result in cell death [24].
Lactic acid accumulation causes also weak acid stress to the cells. Cells exposed to weak acids adapt to some extent, for instance by up-regulating excretion of the acid, by blocking re-entry, or by metabolizing the acid [25]. Different yeast species may use different strategies to maintain cellular pH and ion homeostasis. The present data showed that lactate concentration in the culture medium decreased in prolonged cultivations indicating that the cells consumed lactate (see e.g. Figure 2).
An optimal lactic acid production host should tolerate acidic conditions and produce a high concentration of undissociated lactic acid in order to reduce the need for neutralizing chemicals, and ethanol production should be eliminated. As shown in Figure 6, the media composition (YP or YNB) and the amount of CaCO 3 determine how much lactic acid a strain can produce. The medium composition is an important consideration in production process because of cost, downstream processing, and product quality, which was why minimal medium (YNB) was mainly used in the present work. It is evident that differences in the culture conditions used by different groups complicate fair comparison between the species. Not surprisingly, the highest total lactate production levels by yeasts have been obtained at higher pH using non-limiting concentration of neutralizing chemicals.
When benchmarked against published results where media and operating conditions are disclosed, this C. sonorensis strain compares favorably. Representative results for lactate producing yeast strains, which produce little or no ethanol, at neutral pH have been reported for C. boidinii that produced 86 g/l lactic acid in the presence of non-limiting CaCO 3 and final pH 6.15 [8], and C. sonorensis that produced 92 g/l lactate at 0.94 g/g yield on glucose in less than 48 h. S. cerevisiae pdc1Δ pdc5Δ strain produced 82 g/l lactate at 0.82 g/g yield in YP-10% glucose in the presence of 30 g/l CaCO 3 , but the pH was not reported [2]. C. utilis produced 103 g/l lactic acid in YP medium containing 109 g/l glucose (0.95 g/g yield on glucose) and 45 g/l CaCO 3 in in 33 h, with final pH of 4 [9]. S. cerevisiae wine yeast produced 40 g/l lactate [3], and diploid S. cerevisiae produced 50 g/l lactate below pH 4 [14]. C. boidinii produced 50 g/l lactic acid in YP-10 g/l glucose medium containing 30 g/l CaCO 3 . In comparison, C. sonorensis LhLDH strains produced in YP medium 66 g/l lactate (0.73 g/g yield) with 30 g/L of CaCO 3 and final pH 4.0. The LDH and PDC modifications are the necessary basis for further yeast development towards an industrial lactic acid process.

Conclusions
We developed an organism without previous record of genetic engineering to efficiently produce L-lactic. Genetic modification of C. sonorensis opens the possibility to exploit this novel host organism in the production of useful biochemicals. The frequent occurrence of both targeted and non-homologous integration into the genome gives flexibility to strain design and construction. Both PDC1 and PDC2 enzymes contributed to ethanol production, but PDC2 encodes the main isoenzyme. The possibility to generate knock out strains allowed us to demonstrate the significance of each of the PDC genes in the context of lactic acid production in C. sonorensis. Unexpectedly, LDH strains with intact PDC genes produced very little ethanol and as much lactate as the PDC deleted strains in the presence of CaCO 3 , although the same strains produced more ethanol than lactate in nonbuffered conditions. This indicated that not only the genotype but also the culture conditions had a large influence on carbon distribution between ethanol and lactate.
The Cargill commercial implementation of a yeast for lactic acid production has demonstrated the high potential of yeasts as hosts for organic acid production [26]. This present work showed that glucose could be converted to highly pure L-lactate at an excellent yield by C. sonorensis expressing a LDH in minimal medium in the presence of CaCO 3 . The purity of the product i.e. taking into account the formation of by-products ethanol and pyruvate, and to some extent the concentration of lactate differed between strains expressing different LDH genes. The lactate production parameters (concentration, yield, production rate) observed with C. sonorensis strains expressing LhLDH compare favorably with other lactic acid producing yeasts, illustrating that Candida yeasts have high potential as lactic acid production hosts. Among the LDH genes studied, LhLDH was the most suitable one to produce lactic acid with C. sonorensis in the conditions studied. Thus, the choice of the LDH is an important consideration in the development of improved production hosts.

Microbial strains
E. coli strains DH5α (Gibco BRL, Gaithersburg, MD) and XL-1 Blue (Stratagene, La Jolla, CA) were used as hosts for cloning and plasmid propagation. C. sonorensis ATCC32109 (American Type Culture Collection), was used throughout the study and was the parental strain of the transformants generated in this work.
Non-buffered cultivations in YP -5% (w/v) glucose were inoculated to an optical density (OD 600 ) of 0.2 with cells grown on YPD agar. In some experiments the YP medium contained 10% (w/v) glucose and 5 to 30 g/l of calcium carbonate (CaCO 3 ) for pH control.
For two-stage cultivations, the biomass was grown in yeast nitrogen base medium (YNB w/o amino acids; Difco, Sparks, MD) supplemented with 5% glucose and buffered to pH 5.5 with 0.5 M 2-[N-Morpholino] ethanesulfonic acid (MES). After overnight cultivation at 30°C and 250 rpm, cells were harvested by centrifugation and transferred into YNB medium supplemented with 10% (w/v) glucose, to give an initial cell density of OD 600~1 5 corresponding to approximately 5 g/l cell dry weight. 80 g/l of calcium carbonate was added for pH control in some of the cultures.
Cultures were incubated at 30°C with 100 rpm shaking in 250 ml Erlenmeyer flasks containing 50 ml medium.

DNA manipulations
Plasmid DNA was isolated using Qiagen kits (Qiagen Corp, Chatsworth, CA, USA). Recombinant DNA work was carried out using conventional techniques [27]. Oligonucleotides were purchased from Sigma-Genosys (Little Chalfont, UK). PCR was performed using Dynazyme EXT polymerase (Finnzymes, Espoo, Finland) with an initial incubation for 3 min at 94°C, followed by 29 cycles of 45 sec at 94°C, 45 sec at 55°C, 2 min at 72°C, with a final incubation for 10 min at 72°C.
Isolation of PGK1, TDH1, PDC1 and PDC2 genes from C. sonorensis Yeast DNA was isolated by phenol extraction from cells broken with glass beads [28]. The genomic library of C. sonorensis ATCC32109 was prepared using partially Sau3A digested size fractionated genomic DNA that was cloned into the BamHI digested lambda DASH™ vector (Stratagene, La Jolla, CA, USA) as described previously [29]. The library was screened by colony/plaque hybridization. C. albicans PGK1, amplified by PCR from genomic DNA with primers 5092 and 5091 was used as a probe to isolate the C. sonorensis gene for 3-phosphoglycerate kinase (PGK), and S. cerevisiae TDH1, amplified with primers 4125 and 4126 (Table 2) was used as a probe to isolate the gene for glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Fragments of PDC1 and PDC2, were amplified from genomic DNA of C. sonorensis with primers 5116 and 5118 ( Table 2) which were designed from conserved regions in the known pyruvate decarboxylase amino acid sequences, WAGNANELNA and DFNTGSFSY, of P. stipitis PDC1 (U75310) and PDC2 (U75311), S. cerevisiae PDC1 (X04675), and C. albicans PDC11 and PDC12 (sequence data for C. albicans was obtained from the Candida Genome Database website at http://www.candidagenome.org/). The C. sonorensis PDC1 and PDC2 fragments obtained with primers 5116 and 5118 (Table 2) were used as probes for the isolation of the corresponding genes from a genomic library. The identity of the purified genomic clones was verified by DNA sequencing.

Southern and colony/plaque hybridizations
Southern blots were prepared using conventional techniques and hybridized with probes labeled with [α-32 P] dATP or [α-32 P]dCTP (Amersham Pharmacia, Little Chalfont, UK) or with digoxigenin-11-dUTP (Roche, Mannheim, Germany). The presence and copy number of the LDH gene was verified by Southern analysis of HindIII digested yeast DNA using the corresponding LDH gene as the probe (see Table 2 for probe PCR primers). The PDC1-or PDC2 deletions were verified by Southern analyses by the absence of PDC1-or PDC2specific hybridization signals and the appearance of transformation marker-specific signals of appropriate size. PDC1 or PDC2 probes correspond to nucleotides in the deleted area and were amplified by PCR. Radioactive hybridization signals were detected by scanning exposed storage phosphor screens using the Typhoon 8600 variable mode imager (Molecular Dynamics, Sunnyvale, CA). Non-radioactive signals were detected colorimetrically with NBT and BCIP (Promega, Madison, WI).
Plasmids for replacement of the PDC2 locus containing C. sonorensis PDC2 5′ homology region (0.8 kb), C. sonorensis GPD1 promoter, E. coli G418 R , S. cerevisiae MEL5 terminator, C. sonorensis PGK1 promoter, one of the LDH genes, S. cerevisiae GAL10 terminator and C. sonorensis PDC2 3′ homology region (0.9 kb), were prepared as follows. The BmLDH from pMI265 and the G418 R expression cassettes were joined to form pMI278 ( Table 3). The region upstream of PDC2 ORF was amplified by PCR using the primers Cs26 and Cs27 (Table 2), and the genomic copy of the C. sonorensis PDC2 as the template (GenBank accession number AM420320), and the PCR product was inserted upstream of the LDH expression cassette resulting in plasmid pMI279. Then the 0.9 kb PDC2 3′ homology region amplified by PCR as above using primers Cs29 and Cs30 (Table 2) was added to form pMI286. BmLDH in pMI286 was replaced by LhLDH resulting in pMI288. pMI287 was constructed by removing BmLDH from pMI286.

Transformation of C. sonorensis
All plasmids were digested with restriction enzymes prior to transformation to facilitate integration into the genome, unless otherwise stated. C. sonorensis was transformed using the lithium acetate method [32,33]. After 3 hours incubation in liquid YPD medium, cells were spread onto agar-solidified YPD medium containing 200 μg/ml G-418 sulfate, or 40 μg/ml X-α-Gal (ICN Biochemicals, Aurora, OH, USA), a chromogenic substrate of α-galactosidase. Cells transformed with NotI digested pMI257 (LhLDH), pMI265 (BmLDH), pMI266 (RoLDH), or pMI267 (no Table 3 Plasmids for C. sonorensis transformations constructed in this work pMI288 CsPDC2 5′-CsP GPD1 -G418 R -ScT MEL5 -CsP PGK1 -LhLDH-ScT CYC1 -CsPDC2 3′ LDH) were selected for melibiase activity, and with pMI286 (BmLDH), pMI287 (no LDH) or pMI288 (LhLDH) were selected for G418 resistance. The PDC2 gene was replaced by the G418 R gene in the pdc1 deleted, melibiase positive C. sonorensis transformants containing the LhLDH, BmLDH, RoLDH, or no LDH by transformation with pMI287 (no LDH). In addition, PDC2 was replaced by the G418 R and LDH genes by introducing a second copy of LhLDH or BmLDH into strains containing the LhLDH or BmLDH gene, respectively, integrated in the PDC1 locus. Putative PDC2 deletants were screened for decreased ethanol production. Replacements of PDC1 or PDC2 genes and the presence of LDH were verified by Southern analyses. Strains constructed in this work are listed in Table 4.

PDC and LDH enzyme activity measurements
Enzyme activities were measured from freshly prepared cell extracts. Cells from 5 ml samples were harvested by centrifugation, washed with 1 ml of ice-cold 10  2) and in 50 mM imidazole-HCl (pH 6.5) buffer, each containing 0.4 mM NADH, 5 mM fructose-1, 6-diphosphate (FBP) and 2 mM pyruvate. R. oryzae LDH activity was measured in the presence and in the absence of FBP at pH 6.5. The activities are expressed in units per milligram protein (U/mg). 1 U was defined as the amount of enzyme required to reduce 1 μmol of substrates per min. Protein concentrations were measured using a protein assay reagent (Bio-Rad 500-0006) and bovine serum albumin (Sigma) as the protein standard.

Analytical methods
The culture supernatants were analyzed by HPLC for lactic acid, glucose, pyruvic acid, acetic acid, glycerol and ethanol using a Waters 2690 Separation Module and Waters System Interphase Module liquid chromatography coupled with a Waters 2414 differential refractometer and a Waters 2487 dual λ absorbance detector (Waters, Milford, MA). A Fast Juice Column (50 mm × 7.8 mm, Phenomenex, Torrance, CA) and a Fast Acid Analysis Column (100 mm × 7.8 mm, Bio-Rad, Hercules, CA) or, alternatively, a Fast Acid Analysis Column (100 mm × 7,8 mm, Bio-Rad) and an Aminex HPX-87H Organic Acid Analysis Column (300 mm × 7.8 mm, Bio-Rad) were equilibrated with 2.5 mM H 2 SO 4 in water at 60°C and samples were eluted with 2.5 mM H 2 SO 4 in water at a 0.5 ml/min flow rate. Data were acquired with Waters Millennium software. Undissociated lactic acid was determined from supernatant samples diluted in ethyl acetate. The standard was prepared by dissolving lithium lactate in 0.5 M HCl and further diluting it in ethyl acetate. Samples and standards were eluted with the Fast Juice Column (50 mm × 7.8 mm, Phenomenex) and Fast Acid Analysis Column (100 mm × 7.8 mm, Bio-Rad) as above at 1.0 ml/min flow.
Lactate and ethanol yields were calculated as the amounts of accumulated products per amount of consumed sugar. Yields are reported for the sample time when sugar concentration was first observed to be below 1.5 g/l, unless otherwise stated.
Intracellular concentrations of lactic acid and pyruvate were measured from cells harvested from 1 ml of culture by centrifugation, washed with 1 ml 1 M Tris-HCl pH 9.0, resuspended in 1 ml of ice cold 5% (w/v) trichloroacetic acid by vortexing for 1 min and incubated on ice for 30 min. Samples were vortexed for 1 min, centrifuged at 13 000 rpm for 30 min at +4°C, and L-lactic acid in the supernatant was measured with the L-lactic acid UV method (#10139084035, Roche, Mannheim, Germany) method or by HPLC. Pyruvate was measured enzymatically using a pyruvate kit (Sigma Diagnostics, St. Louis, MO). Intracellular concentrations of lactic acid and pyruvate were calculated assuming that one gram of cell dry weight corresponds to 2 ml cell volume [34].
D-lactate was determined enzymatically with the Llactate UV-method (#10139084035, Roche, Mannheim, Germany) using d-LDH instead of L-LDH in the assay.

Statistical analyses
Data are given as means. Where appropriate, values were compared by analysis of variance (ANOVA) and significant differences determined using Fisher's multiple range test. P values < 0.05 were considered statistically significant.