Glucose assimilation rate determines the partition of flux at pyruvate between lactic acid and ethanol in Saccharomyces cerevisiae

Engineered Saccharomyces cerevisiae expressing a lactic acid dehydrogenase can metabolize pyruvate into lactic acid. However, three pyruvate decarboxylase (PDC) isozymes drive most carbon flux toward ethanol rather than lactic acid. Deletion of endogenous PDCs will eliminate ethanol production, but the resulting strain suffers from C2 auxotrophy and struggles to complete a fermentation. Engineered yeast assimilating xylose or cellobiose produce lactic acid rather than ethanol as a major product without the deletion of any PDC genes. We report here that sugar flux, but not sensing, contributes to the partition of flux at the pyruvate branch point in S. cerevisiae expressing the Rhizopus oryzae lactic acid dehydrogenase (LdhA). While the membrane glucose sensors Snf3 and Rgt2 did not play any direct role in the option of predominant product, the sugar assimilation rate was strongly correlated to the partition of flux at pyruvate: fast sugar assimilation favors ethanol production while slow sugar assimilation favors lactic acid. Applying this knowledge, we created an engineered yeast capable of simultaneously converting glucose and xylose into lactic acid, increasing lactic acid production to approximately 17 g L−1 from the 12 g L−1 observed during sequential consumption of sugars. This work elucidates the carbon source‐dependent effects on product selection in engineered yeast.

S. cerevisiae exhibits the Crabtree effect such that respiration is shut down in favor of fermentative metabolism when glucose is present in high concentrations even under aerobic conditions. [11] Because of this, production of large amounts of lactic acid from glucose has generally required deletion of pyruvate decarboxylase (PDC) enzymes encoded by the PDC1, PDC5, and PDC6 genes. [12][13][14] However, this deletion of PDCs comes with critical drawbacks such as C 2 auxotrophy which requires supplementation with either acetate or ethanol, and reduced growth rate and fermentation speed. As such, it is desirable to engineer a yeast strain capable of efficient production of non-ethanol compounds derived from pyruvate, such as lactic acid, without deleting PDC enzymes.
Prior reports have established methods to completely rewire yeast metabolism, bypass the fermentative metabolic structure regardless of aeration, and engineer Crabtree-negative yeast. [15] However, the prior reports showing superb yields of lactic acid from xylose without deletion of any PDC genes suggest alternative approaches are possible. As the preferred production of ethanol over lactic acid is closely tied to glucose uptake, glucose signaling and metabolism may play a role in the partition of flux at the pyruvate branch point.
Perception of glucose in yeast is performed through the sensors Snf3 and Rgt2, which respond to low and high concentrations of glucose, respectively. [16] The glucose signal leads to degradation of Mth1 and Std1 [17] and phosphorylation of the repressor Rgt1, [18] releasing it from promotors and derepressing its target genes. There is some evidence that high concentrations of xylose can be sensed by Snf3 and simulating a glucose signal by deletion of RGT1 has been shown to improve xylose fermentation. [19] Although the Snf3/Rgt2 signaling pathway has been extensively mapped as to its role in regulating expression of hexose transporters, the perception of glucose is also known to affect the growth rate and thus metabolism of yeast. [20] However, the exact effects of the Snf3/Rgt2 signaling pathway on yeast metabolism remain largely unexplored.
S. cerevisiae also exhibits accelerated glycolysis in the presence of high levels of glucose. [21] While accelerated glycolysis leads to rapid production of ethanol in wild-type strains, controlling central carbon metabolism by modulating expression of HXK1 can paradoxically increase yields of a pathway competing for pyruvate flux without deletion of endogenous PDC genes. [22] By exploring the properties of glucose perception and glycolytic flux, we report here a deeper understanding of the mechanisms underlying enhanced yields of lactic acid from cellobiose and xylose fermentations. Deletion of the glucose sensors Snf3 and Rgt2 enhances yields of lactic acid from glucose and this improvement is replicated in a strain possessing all native sensors by controlling glycolytic flux. We also fully replicate the high yields of lactic acid observed during cellobiose fermentations solely by manipulating glycolytic flux. However, this methodology remains unable to match the high yields of lactic acid observed during xylose fermentations, suggesting still-unknown physiological or regulatory mechanisms present during xylose fermentation which promote the production of lactic acid over ethanol. Ultimately, we apply the knowledge gained here to create a strain capable of producing high yields of lactic acid during glucose/xylose co-consumption.

Strains, media recipes, and culture conditions
All strains and plasmids used in this study are listed in Tables 1 and 2, respectively. Preculture was performed in 5 mL of YP medium (10 g L −1 yeast extract, 20 g L −1 Bacto peptone) aerobically at 30 • C for 36 h with 40 g L −1 of glucose as a carbon source. Fermentation experiments were performed in 125 mL flasks containing 25 mL YP medium at 30 • C and 100 rpm, with initial carbon sources indicated in fermentation profiles. All fermentations were performed using biologically independent duplicate cultures. All fermentations with lactic acid as a product contained 20 g L −1 CaCO 3 to maintain pH near neutral levels. Doxycycline-controlled expression of hexokinases was performed as described previously. [23] Briefly, when using the D452-iH1L and D452-iH2L strains, preculture was performed in 5 mL of YP medium aerobically at 30 • C for 36 h with 40 g L −1 of galactose along with 1, 3, 6, or 12 µg mL −1 of doxycycline. Doxycycline was added to fermentations with the D452-iH1L and D452-iH2L strains to final concentrations of 1, 3, 6, and 12 µg mL −1 of fermentation medium to control expression of hexokinase and the overall consumption rate of glucose.

Genetic techniques
Standard restriction digestion and molecular cloning techniques were employed for plasmid creation. [24,25] The structure and target sequences of all guide RNAs used in this study are listed in Table S1 and all primers used in this study are listed in Table S2. For doxycycline induction experiments, the rtTA(S2) variant [26] was synthesized from IDT as a gBlock. Two integrative plasmids were created to house the synthetic transactivator rtTA(S2) [26] and the TetO7-driven target gene as described previously. [23] The pRS405-LdhA plasmid was created by PCR amplifying the pPGK1-LdhA-tPGK1 expression cassette from pITY-LdhA [4] using primers SL13 and SL14. To create strains D452-2L, D452-2AL, D452 iH1L, and D452 iH2L, the pRS405-LdhA plasmid was linearized with AgeI then integrated into the yeast genome and selected using the leucine auxotrophic marker. To create the EJ4L and SR8L strains, a single copy of an LdhA expression cassette was inserted into the PBN1-SBP1 intergenic region using CRISPR/Cas9 genome editing (

Expression analysis
RNA was extracted from three biologically independent replicates of strains SR8 and SR8#22 grown to mid-exponential phase with glucose as a carbon source. RNA-seq experiments and analysis were performed as previously described. [23,27]

RESULTS
To examine the effects of a carbon source on lactic acid and ethanol production, we selected three yeast strains within the same lineage with different sugar consumption abilities: D452-2, the parental yeast strain, SR8, an engineered xylose-consuming strain derived from F I G U R E 1 Carbon source affects lactic acid yield. Fermentation of the D452-2L strain on glucose (top left), the SR8L strain on xylose (top right), and the EJ4L strain on cellobiose (bottom left) along with yields from each fermentation (bottom right). Fermentations were inoculated to an initial cell density equal to an optical density (OD) of 1 and performed in 25 mL YP media in 125 mL flasks with 20 g L −1 CaCO 3 at 100 RPM and 30 • C. Data points indicate the optical density at 600 nm (yellow circles) or the concentration of glucose (blue squares), xylose (purple squares), cellobiose (green squares), lactic acid (red upward triangles), or ethanol (black downward triangles). Data points are the average of biologically independent duplicate cultures with standard deviations indicated by error bars, which are not visible when the standard deviation is smaller than the size of the data point D452-2; and EJ4, a cellobiose-and xylose-consuming strain derived from SR8. [28][29][30] The R. oryzae lactic acid dehydrogenase encoded by LdhA was then introduced into each strain to enable production of lactic acid, yielding D452-2L, SR8L, and EJ4L. [4] The three resulting strains were then cultured: D452-2L on glucose, SR8L on xylose, and EJ4L on cellobiose.
Carbon sources strongly influenced lactic acid and ethanol production. D452-2L cultured on glucose produced almost entirely ethanol, EJ4L cultured on cellobiose produced slightly more lactic acid than ethanol, and SR8L cultured on xylose produced almost entirely lactic acid ( Figure 1). Xylose cultures led to the highest production of lactic acid (10.6 g L −1 ), followed by cellobiose (7.3 g L −1 ) and lastly glucose with the lowest production (1.1 g L −1 ).
The transmembrane sensors Snf3 and Rgt2 alter the growth rate and metabolism of yeast in response to extracellular glucose. [20] While there is some evidence that an active Snf3/Rgt2 sensing path-way enhances xylose consumption, [19] the role these two sensors play in cellobiose and xylose fermentations remains unclear. [31] We hypothesized that the Snf3/Rgt2 pathway might be inactive during consumption of non-native sugars xylose and cellobiose. Therefore, inactivating this pathway during glucose fermentation through deletion of SNF3 and RGT2 may enhance lactic acid production from glucose. We therefore derived a series of strains with Δsnf3Δrgt2 and GLK1) in D452-2, leading to the strain D452∆hxk 0 which is unable to consume glucose. Then, the synthetic transactivator rtTA(S2) [26] was introduced followed by reintroduction of hexokinase expression under control of the rtTA(S2)-controlled TetO7 promoter. As a result, expression of hexokinase was controlled by the addition of doxycycline in a titratable manner. This technique was previously found to enable complete control over the glucose consumption rate. [23] We created two sets of strains from this, one with controllable expres-sion of HXK1 (D452 iH1) and the other with controllable expression of HXK2 (D452 iH2). We next introduced R. oryzae LdhA and named the resulting strains D452 iH1L and D452 iH2L.
These strains were then cultured in flasks with various amounts of doxycycline to observe lactic acid production from the strains exhibiting a wide range of glucose uptake rates ( Figure S5). We found that the yields of lactic acid and ethanol were clearly linked to the sugar uptake rates (Figure 3). Slower glucose-consuming strains tended to produce more lactic acid while faster glucose-consuming strains pro- Consumption of glucose/xylose mixtures by engineered yeast usually results in two stages: rapid consumption of glucose followed by slow xylose assimilation. However, reduced glycolytic flux allows simultaneous uptake of both glucose and xylose. [23] This coupled with the information shown here-that xylose is a superior carbon source for lactic acid production and that slow glucose consumption is better than fast glucose consumption for yielding lactic acid-led us to hypothesize Therefore, the R. oryzae LdhA gene was introduced into the glucose/xylose co-fermenting engineered yeast strain SR8#22. [23] Then, the SR8L strain and the SR8#22L strain were compared in fermen-

DISCUSSION
The Snf3/Rgt2 sensors control glucose perception in yeast [20] and initiate a signaling cascade which enables precise expression of hexose transporters in response to extracellular glucose concentrations. [32] As expected, deletion of these sensors negatively impacts the fermentation of glucose ( Figure 2, Figure S1). Yeast cells without functional glucose sensors tend to produce less ethanol, accumulate more biomass, and require longer periods of time to complete a fermentation. Although we found that deletion of the glucose sensors benefits the production of lactic acid in both galactose and glucose cultures, we were able to match the increased yields in strains with active sensors by reducing the glucose assimilation rate. Nonetheless, the inherently increased biomass yields indicate that the glucose sensor deletions may be a viable strategy to increase yields of growth-associated products in engineered yeasts.
Interestingly, we also observed that deletion of Snf3 and Rgt2 negatively impacts fermentation of xylose ( Figure S2). Recent observations indicate that xylose interacts with at least one of these sensors and can induce signal transmission, as deletion of either glucose sensor leads to decreased HXT1 expression during xylose fermentations. [19] As such, yeast strains lacking the glucose sensors may have altered transporter expression profiles which are sub-optimal for xylose uptake.
This possibility is compatible with our observation that deletion of glucose sensors had little effect on cellobiose fermentation ( Figure S3).
As cellobiose fermentation depends on transgenic expression of the cellobiose transporter CDT-2, altering native transporter expression leaves cellobiose uptake unaffected.
Using the D452 iH1L and D452 iH2L strains with direct control over hexokinase expression, we identified a connection between glycolytic flux and the yields of lactic acid and ethanol from glucose. There must be some other physiological mechanism or factors that lead to enhanced lactic acid yields from engineered yeast strains cultured on xylose. For instance, glucose induces degradation of the Jen1 transporter, [9] which along with Ady2 is known to modulate lactic acid production in yeast. [33] Cells cultured on xylose display increased expression of JEN1 and ADY2 compared to cells cultured on glucose and deletion of these two transporters reduces lactic acid yields from strains cultured on xylose. [8,27] A low rate of sugar metabolism coupled with strong potential for exporting lactic acid may explain the nearly theoretically maximum yields of lactic acid observed during mid-exponential phase of xylose fermentations. As such, coupling overexpression of JEN1 and ADY2 with a reduced glycolytic flux may enable high lactic acid yields from glucose without deleting endogenous PDCs. Notably, the K m of R. oryzae LdhA on pyruvate is approximately 0.5 mM while the major S. cerevisiae PDC isoforms Pdc1, Pdc5, and Pdc6 exhibit K m values of 4.7, 9.9, and 8.2 mM, respectively. [4,34] Similarly, S. cerevisiae pyruvate dehydrogenase exhibits a much higher affinity for pyruvate than PDC. These varied kinetic parameters have been proposed as a determining factor of the flux distribution at the pyruvate branch point in yeast. [35] A slow metabolic rate may contribute the kinetics of the enzymes using pyruvate as a substrate. [35] It is thus possible that some currently unknown intracellular glucose signaling mechanisms modify the kinetics of the PDC enzymes and impact the flux distribution at the pyruvate branch point.
13C-MFA of S. cerevisiae has shown that fermentative conditions result in about 75% of pyruvate flux moving toward acetaldehyde. [36] In contrast, during a reduced Crabtree effect condition, only about 1% of pyruvate flux is directed toward acetaldehyde. [37] While in fermentative conditions only about 6% of pyruvate enters the mitochondria, a reduced Crabtree effect condition leads to approximately 75% of pyruvate entering the mitochondria. [36,37] The high yields of lactic acid from xylose thus strongly agree with prior reports that xylose metabolism elicits a respiratory response in engineered yeast. [38] Expression analysis of the fast glucose-consuming strain SR8 and slow glucose-consuming strain SR8#22 reveals no significant differences in expression of PDC1, PDC5, PDC6, nor the transcription factor encoded by PDC2 ( Figure S7). Additionally, no difference in expression of PGK1, whose promoter drives expression of LdhA in our study, was observed. Taken together, these observations indicate that transcriptional regulations alone cannot explain our observations and some post-translational regulations are key to explain the control glucose assimilation rate exerts over flux distribution at the pyruvate branch point.
Deletion of the PDC genes to enhance production of non-ethanol products can be undesirable as the resulting PDC-negative strain will be C 2 auxotrophic and require supplementation of either acetate or ethanol, in addition to a reduced growth rate. An alternative method has been proposed to prevent ethanol production without the C 2 auxotrophy through deletion of the acetaldehyde dehydrogenase isoforms. [39,40] Expression of a lactate dehydrogenase in a strain with deletions in the acetaldehyde dehydrogenases ADH1, ADH2, ADH3, ADH4, ADH5, and SFA1, the glycerol-3-phosphate dehydrogenases GPD1 and GPD2 and the PDC PDC1 enabled production of lactic acid from glucose without ethanol accumulation or any requirement for supplementation of C 2 compounds. However, the resulting strain completed fermentations extremely slowly: over 40 h were required for complete consumption of 2 g L −1 glucose. In contrast, our highest-yielding strain SR8L was able to convert 20.9 g L −1 xylose into 10.6 g L −1 lactic acid in under 40 h.
In our experiments with co-fermentation of glucose and xylose, we employed medium containing approximately equivalent amounts of the sugars at 20 g L −1 each. Optimizing these concentrations and the ratio of glucose to xylose can aid in maximizing yield and productivity of lactic acid in a laboratory setting. For industrial applications, lignocellulosic hydrolysates typically contain approximately 60%-70% glucose and 30%-40% xylose. [41] Nonetheless, the results presented here serve as a general proof-of-concept and it is reasonable to expect that the trends we observed here will be maintained regardless of the medium sugar concentrations and ratio.
In addition to the increased lactic acid yields shown here, reducing the glucose phosphorylation rate has been shown to enable simultaneous consumption of glucose/xylose and glucose/galactose mixtures. [23] After expressing the LdhA gene in the glucose/xylose co-consuming strain SR8#22, we observed simultaneous bioconversion of glucose and xylose into lactic acid and ethanol. Enabling equal co-consumption of both glucose and xylose avoids a two-phase fermentation where glucose is rapidly converted to ethanol, followed by a slower conversion of xylose into predominantly lactic acid. Such a two-phase fermentation is incompatible with techniques such as continuous fermentation, as glucose is continuously fed into the reaction chamber and will constantly be preferred over the accumulating xylose.
Although the results here represent a step in the right direction, additional challenges remain before S. cerevisiae is a viable production host for commercial biorenewable lactic acid. The experiments performed here employed calcium carbonate as a neutralizing agent, which is too costly for an economically sustainable biorefinery. Further engineering to enhance S. cerevisiae tolerance to low pH conditions or inhibitors found in common feedstock hydrolysates will be essential to any viable commercial efforts. It is also possible that S. cerevisiae may never be an optimum production host for biorenewable lactic acid. Identification and engineering of non-conventional yeasts such as Issatchenkia orientalis have revealed the possibility for incredible natural resistance to inhibitors and low pH inherent to certain species. [42] Despite the vast differences between baker's yeast and many nonconventional yeasts, some exist which are metabolically very similar to the highly studied S. cerevisiae. Future researchers may aim to strike a balance between identifying new yeasts with optimum natural characteristics and the ability to transfer over the vast amounts of knowledge gained from years of research into S. cerevisiae. Alternatively, the mechanisms underlying increased resistance of these non-conventional yeasts may be identified and engineered into S. cerevisiae. For example, the I. orientalis GPI-anchored protein encoded by IoGAS1 confers low pH tolerance when expressed in engineered S. cerevisiae strains. [13,43] Future researchers aiming to create lactic acid and other pyruvate-derived molecules in S. cerevisiae strains expressing native PDC enzymes may find some inspiration from the physio-logical changes present during xylose metabolism. Using xylose as a carbon source has been shown to increase yields of a variety of products, such as isobutanol, [44,45] 2,3-butanediol, [46,47] poly-3-Dhydroxybutyrate, [48,49] squalene, and amorphadiene. [50] A deeper understanding of the exact phenomena underlying enhanced product yields from xylose may allow adapting this knowledge to other carbon sources, such as sucrose, glucose, and fructose, which are cheaper and more abundant than xylose. While our findings demonstrate that flux, but not perception, contribute to enhanced yields of lactic acid, additional research will need to be performed to investigate whether these conclusions stay valid for other products derived from pyruvate.