Two Hidden Enzymes May Be Draining the Fuel Behind Erythritol Production in Industrial Yeast

Erythritol has quietly become one of the most sought-after molecules in the food industry, a naturally brewed sugar alcohol that delivers sweetness without calories and without the digestive baggage of some artificial sweeteners. The workhorse behind much of its industrial production is not a plant or a factory vat of chemicals but an oleaginous yeast called Yarrowia lipolytica, a microbe famous for eating oils and excreting valuable products. Yet even the best industrial strains hit a ceiling on how much erythritol they can make, and a new open-access study in Biotechnology for Biofuels and Bioproducts argues that the bottleneck may be hiding in plain sight, in the form of two genes whose enzymes quietly siphon off the very cofactor that erythritol synthesis depends on.

The research, led by Muhammad Sarfaraz Iqbal and Hairong Cheng of Shanghai Jiao Tong University together with collaborators in Pakistan, Afghanistan and Saudi Arabia, took a transcriptome-guided approach to a problem that has long frustrated metabolic engineers. The central issue is NADPH, the reduced form of nicotinamide adenine dinucleotide phosphate. Erythrose reductase, the enzyme that performs the final step of converting erythrose into erythritol, consumes NADPH to do its job. But Yarrowia lipolytica carries a large repertoire of other NADPH-dependent oxidoreductases, enzymes that also bind and consume this cofactor for their own unrelated reactions. When erythritol production is pushed to industrial levels, these competing enzymes become silent drains on the NADPH pool, and the yields suffer. The specific contributors to this cofactor competition, however, had remained poorly defined.

To find the culprits, the team ran time-resolved RNA sequencing across 96 hours of fermentation, watching how every gene in the yeast genome changed its expression as erythritol accumulated. Two genes stood out with strikingly different temporal profiles. YALI0_F02013g, annotated as a putative pyridoxine 4-dehydrogenase, was progressively induced 3.4-fold over the course of the run, peaking at 5040 FPKM at 72 hours, precisely when erythritol production is typically at its most intense. YALI0_A19910g, annotated as a putative glycerol 2-dehydrogenase, moved in the opposite direction, being repressed 920-fold by the 96-hour mark. The fact that one candidate ramps up exactly when the product is being made, while the other is switched off as the process matures, suggested that both are functionally connected to erythritol metabolism, though in very different ways.

The expression trends were not left to stand on sequencing data alone. The researchers validated them by quantitative real-time PCR, achieving Pearson correlation coefficients greater than 0.94 between the RNA-seq and qRT-PCR measurements, a level of agreement that gives confidence the temporal profiles are real rather than artifacts of library preparation. The correlation analysis also revealed something more provocative. YALI0_A19910g expression correlated strongly negatively with erythritol titre, with a Pearson r of −0.858, meaning that as this gene’s expression fell, erythritol levels rose. YALI0_F02013g, by contrast, showed only a weak positive correlation with the product. The negative correlation for YALI0_A19910g is the kind of signal that immediately suggests a repressive relationship, as if the cell itself is trying to shut down this enzyme when erythritol production matters most.

But correlation alone cannot prove that an enzyme competes for NADPH. The team therefore turned to molecular docking, computationally fitting NADPH into the predicted three-dimensional structures of both enzymes to measure how tightly each would bind the cofactor. The results were unambiguous. YALI0_F02013g showed a predicted binding affinity of −9.8 kcal per mole for NADPH, while YALI0_A19910g came in at −8.6 kcal per mole, both figures indicating high affinity in the range typical of genuine NADPH-dependent enzymes. The docking analysis also identified conserved catalytic triads in both proteins, the characteristic arrangements of amino acids that oxidoreductases use to shuttle hydride ions from NADPH onto their substrates. In other words, both enzymes look structurally competent to grab NADPH and use it, exactly the property that would make them competitors in the cofactor economy of an erythritol-producing cell.

To rank the two candidates against each other, the researchers built an integrated prioritisation model that weighed expression dynamics, correlation with product titre, docking scores and functional annotations. The model scored YALI0_F02013g at 5.6 out of 10 and YALI0_A19910g at 5.0 out of 10, naming the putative pyridoxine 4-dehydrogenase as the higher-priority engineering target. The ranking may seem counterintuitive given the strong negative correlation of YALI0_A19910g with erythritol titre, but the logic becomes clear when flux is considered. YALI0_F02013g is induced precisely during the production window, meaning its enzyme is present and active exactly when NADPH is most precious, so its cofactor demand is the one that actually drains the pool in real time.

The team then used in silico metabolic modelling to estimate what would happen if YALI0_F02013g were repressed. The flux estimates suggest that knocking down this single gene could conserve approximately 67 percent of the competing NADPH demand, potentially increasing erythritol yield by around 14 percent. A 14 percent yield improvement may sound modest in the abstract, but in industrial biotechnology, where margins on commodity products are razor thin, a double-digit gain achieved by repressing one gene rather than rebuilding an entire pathway is the kind of result that gets attention. The authors are careful to note that these projections require experimental validation, and the modelling is presented as a hypothesis-generating exercise rather than a demonstrated outcome.

The broader significance of the study lies in its methodology. Rather than blindly overexpressing erythrose reductase or deleting genes at random, the researchers combined transcriptomics, correlation analysis, molecular docking and metabolic modelling into a single pipeline that converges on specific, testable targets. This is the kind of rational cofactor engineering that is increasingly seen as the next frontier in industrial strain development. NADPH is not just a currency for erythritol; it is required for fatty acid synthesis, for amino acid biosynthesis, for oxidative stress defence and for a host of other anabolic reactions. Every metabolic engineer who has tried to boost a NADPH-dependent product knows the frustration of raising cofactor supply only to watch competing enzymes absorb the gain. Identifying which enzymes are the biggest absorbers is the prerequisite for fixing the problem, and this study offers a template for doing exactly that.

There are also caveats worth keeping in mind. The study was conducted in a specific strain background of Yarrowia lipolytica, and the authors explicitly frame YALI0_F02013g as a candidate metabolic bottleneck in the strain examined, not a universal law of erythritol fermentation. Different industrial strains may carry different complements of oxidoreductases, and the relative importance of these two genes could shift with the carbon source, the fermentation conditions or the genetic modifications already present in a production strain. The annotations themselves are putative, based on sequence similarity rather than purified enzyme assays, so the actual substrates of YALI0_F02013g and YALI0_A19910g in vivo remain to be confirmed experimentally.

Even so, the study delivers something the field has lacked: named, ranked, structurally characterised candidates for the NADPH competition that limits erythritol yields. If experimental follow-up confirms the in silico projections, repressing YALI0_F02013g could become a standard engineering move in industrial Yarrowia strains, a single-gene edit that redirects roughly two-thirds of a competing cofactor demand toward the product. And if the pipeline itself proves robust, the same transcriptome-guided strategy could be applied to other NADPH-dependent products, from xylitol to lipids, in microbes across the industrial biotechnology landscape. The sweetest result of all may be the demonstration that the path to higher yields sometimes runs not through adding more enzyme, but through finding and silencing the quiet competitors that have been there all along.

Subject of Research: Metabolic Engineering

Article Title: Transcriptome-guided identification of YALI0_F02013g and YALI0_A19910g as candidate NADPH-dependent oxidoreductases associated with enhanced erythritol production in Yarrowia lipolytica

Article References: Iqbal, M. S., Iqbal, M. H., Babakarkhil, T., Alkhanani, A. F., Alsubhi, W. A., Anwar, Y., & Cheng, H. (2026). Transcriptome-guided identification of YALI0_F02013g and YALI0_A19910g as candidate NADPH-dependent oxidoreductases associated with enhanced erythritol production in Yarrowia lipolytica. Biotechnology for Biofuels and Bioproducts. https://doi.org/10.1186/s13068-026-02827-6

Image Credits: AI Generated

DOI: 10.1186/s13068-026-02827-6

Keywords: Yarrowia lipolytica, erythritol, NADPH, oxidoreductase, transcriptomics, molecular docking, metabolic engineering, biotechnology, YALI0_F02013g, cofactor efficiency, Transcriptome-guided, identification