In situ continuous evolution of native transporter HGT1 unlocks xylose utilization in Kluyveromyces lactis

Abstract

Efficient xylose valorization is a prerequisite for the industrial application of Kluyveromyces lactis in lignocellulosic biorefining, yet the inefficiency of transmembrane transport remains a significant bottleneck. To address this, a comprehensive "mining-evolving-engineering" strategy was established. Phylogenetic analysis and functional verification identified RAG1 and HGT1 as the predominant xylose transporters. To enhance their performance, a CRISPR-Cas9-mediated in vivo continuous evolution (CTRLE) system was developed in K. lactis, characterized by a continuous mutation window exceeding 300 bp and a broad mutational spectrum encompassing all possible transitions and transversions. Application of this system yielded a superior evolved mutant with significantly enhanced growth kinetics for xylose utilization, where the biomass accumulation within 72 h increased by 20% relative to the parental strain, representing a 100% improvement over the wild-type. The isolation of this variant validates the efficacy of the targeted mutagenesis strategy and marks the first successful application of an in vivo continuous evolution system in K. lactis. Molecular dynamics (MD) simulations further elucidated the structural mechanisms by which key mutations alter substrate affinity. This work provides robust chassis strains for xylose fermentation and establishes a versatile paradigm for membrane protein engineering in non-conventional yeasts.

1. Introduction

As fossil resources continue to be depleted, biorefining utilizing renewable lignocellulosic biomass has emerged as a pivotal strategy for establishing a sustainable bioeconomy [1,2]. Xylose is the second most abundant sugar component in lignocellulosic hydrolysates after glucose; consequently, its efficient conversion is a core factor determining the economic viability of biorefining processes [2,3]. Although the model organism Saccharomyces cerevisiae dominates industrial fermentation [4], its inherent Crabtree effect triggers a preference for ethanol fermentation under high sugar concentrations [5,6]. This phenomenon limits its application in the production of target metabolites derived from the tricarboxylic acid (TCA) cycle, which require high respiratory flux [7,8].

In contrast, Kluyveromyces lactis, a non-conventional Crabtree-negative yeast [9], exhibits superior aerobic respiratory capacity and a broad substrate spectrum [10]. It is increasingly regarded as an ideal cell factory for the production of organic acids, recombinant proteins, and secondary metabolites [11]. However, the efficiency of xylose utilization in wild-type K. lactis is extremely limited, as it faces systemic bottlenecks such as cofactor specificity mismatch, accumulation of intermediate metabolites, and low transport efficiency [12,13].

In recent years, evolutionary engineering—particularly in vivo continuous evolution systems based on in vivo mutagenesis mechanisms—has provided a new paradigm for addressing this challenge [14]. Unlike traditional Adaptive Laboratory Evolution (ALE), the construction of in vivo evolution platforms equipped with specific mutagenesis modules (OrthoRep [15], CRISPR-AID [16]) enables high-throughput mutation accumulation and screening targeting specific metabolic modules without compromising genomic stability. This approach significantly shortens the cycle of strain engineering. Notably, the recently developed CTRLE [17] in vivo continuous evolution system, mediated by CRISPR guidance and error-prone DNA polymerases derived from bacteriophages, possesses long-fragment windows and full-base mutation capabilities, offering robust technical support for improving xylose transport efficiency. In this study, we utilized this in vivo continuous evolution platform to achieve in situ target-directed evolution of native transporters. Furthermore, molecular docking and molecular dynamics (MD) simulations enable the elucidation of the molecular mechanisms underlying protein performance enhancements, establishing them as an essential paradigm in modern protein engineering research [18,19].

Based on this background, this study established a comprehensive engineering strategy spanning from gene mining to mechanistic elucidation. First, phylogenetic tree analysis was employed to mine members of the Major Facilitator Superfamily (MFS) within the K. lactis genome and predict potential xylose transporters. Second, gene knockout and overexpression experiments were conducted to verify the xylose transport functions of key candidate proteins and evaluate their transport capacities, thereby identifying the optimal evolutionary chassis. Subsequently, the CTRLE system was utilized to perform in situ directed evolution on the target transporter, resulting in the isolation of a mutant exhibiting a high xylose uptake rate. Finally, molecular docking and molecular dynamics (MD) simulations were applied to elucidate the molecular mechanisms by which key mutation sites alter substrate affinity and the dynamic properties of the transport channel.

2. Materials and methods

2.1. Strains, plasmids, and culture conditions

The microbial primers, plasmids and strains used in this study are listed in Supplementary Tables S1, S2 and S3, respectively. Escherichia coli JM109 was used for plasmid cloning and propagation. It was cultivated in Luria-Bertani (LB) medium (10 g/L tryptone, 5 g/L yeast extract, and 10 g/L NaCl) at 37 °C with orbital shaking at 220 rpm. Ampicillin (50 μg/mL) was added to the medium for plasmid maintenance when necessary.

K. lactis were routinely maintained and propagated for seed preparation in YPD medium consisting of 10 g/L yeast extract, 20 g/L peptone, and 20 g/L glucose. For the functional characterization of xylose transporters, strains were cultured in synthetic medium (YNBX) containing 6.7 g/L yeast nitrogen base (YNB) without amino acids and 20 g/L xylose. For adaptive evolution experiments, YPX medium (10 g/L yeast extract, 20 g/L peptone, and 20 g/L xylose) was employed.

For mixed-sugar co-fermentation assays, a formulated lignocellulosic hydrolysate medium was employed. A stock lignocellulosic hydrolysate stored in our laboratory, containing approximately 110 g/L glucose and 55 g/L xylose, was utilized. This stock hydrolysate was added to the YP medium base (10 g/L yeast extract and 20 g/L peptone) at a final volume concentration of 20% to replace the standard carbon source, yielding a final hydrolysate medium containing approximately 22 g/L glucose and 11 g/L xylose. Prior to inoculation, the formulated hydrolysate medium was sterilized by vacuum filtration through a 0.22 μm membrane filter to prevent the thermal degradation of sugars.

2.2. Transformation and screening

Competent cell preparation and DNA transformation of K. lactis were performed using electroporation as previously described [20]. Briefly, harvested yeast cells were treated with 7 mL of conditioning buffer (10 mM LiAc, 10 mM Tris-HCl, and 0.6 M sorbitol) supplemented with 40 μL of 2 M dithiothreitol (DTT). After centrifugation at 4,800g for 5 min the cell pellet was washed and resuspended in ice-cold 1 M sorbitol.

For transformation, 500 ng of the CRISPR/Cas9 plasmid harboring the specific sgRNA and 1000 ng of linear donor DNA were mixed with the competent cells and transferred to a pre-chilled 0.2-cm electroporation cuvette. Electroporation was performed using a Gene Pulser (Bio-Rad Laboratories, Hercules, CA, USA) with the following parameters: voltage, 2.0 kV; time constant, 5 ms. Immediately after the pulse, 1 mL of ice-cold 1 M sorbitol was added to the cuvette. The cell suspension was incubated at 30 °C with shaking at 220 rpm for 1 h to allow recovery. The cells were then spread onto YPD agar plates supplemented with 250 μg/mL Hygromycin B. After 3–4 days of incubation, gene knockout strains were screened and verified by colony PCR. To cure the CRISPR/Cas9 plasmid, the verified colonies were subcultured on non-selective YPD plates for 2–3 consecutive rounds.

2.3. Bioinformatics mining and phylogenetic analysis of sugar transporters

The proteome data of K. lactis GG799 was retrieved from the NCBI Assembly platform. To identify putative sugar transporters, the obtained sequences were aligned against the Transport DB 2.0 database [21]. The transmembrane domains (TMDs) of the filtered proteins were predicted using the TMHMM server (v.2.0) [22]. Proteins containing 11 or more TMDs were retained as potential transporter candidates. From this subset, sequences belonging to the Major Facilitator Superfamily (MFS) were extracted, resulting in 18 putative transporters for functional annotation.

To elucidate evolutionary relationships, homologous sequences from S. cerevisiae and K. marxianus, along with previously reported yeast xylose transporters, were retrieved. Multiple sequence alignment was performed using MUSCLE [23]. Based on the alignment, a phylogenetic tree was constructed using MEGA 11 [24] via the Maximum Likelihood (ML) method [25]. The reliability of the tree topology was assessed with 500 bootstrap replications. Finally, the phylogenetic tree was visualized and annotated using the Interactive Tree Of Life (iTOL) [26] tool to compare the evolutionary distances between the putative transporters and known xylose transporters.

2.4. Construction of the sugar transport protein deletion strain

Gene deletions were performed using a CRISPR-Cas9-mediated genome editing system. Specific sgRNAs targeting the sugar transporter gene were designed and cloned into the CRISPR vector. The donor DNA for homologous recombination was constructed by fusing the 1000 bp upstream and downstream flanking sequences of the target ORF via overlap extension PCR (OE-PCR). The competent cells were co-transformed with 200 ng of the circular CRISPR plasmid and 300 ng of the fused donor DNA fragment via electroporation. Transformants were selected on YPD plates supplemented with Hygromycin B. Genotypes were verified by diagnostic PCR using primers flanking the deletion site. Isolates exhibiting the expected band shift were streaked onto non-selective media for 2–3 rounds to cure the plasmid, yielding the markerless deletion strain.

2.5. Construction and transformation of the modified pKLAC1-Hyg vector

The commercial integrative vector pKLAC1 was modified by replacing the acetamidase (amdS) marker with a Hygromycin B phosphotransferase cassette, yielding the vector pKLAC1-Hyg. The coding sequence of gene was cloned into pKLAC1-Hyg under the control of the P_LAC4 promoter. Instead of restriction enzyme linearization, the linear donor DNA for integration was generated via high-fidelity PCR amplification. Specific primers were designed to amplify the entire integration cassette from the recombinant plasmid, excluding the bacterial backbone sequences. The purified PCR product was transformed into K. lactis competent cells via electroporation. Transformants were selected on YPD agar plates supplemented with 200 μg/mL Hygromycin B. Successful integration at the LAC4 locus was confirmed by diagnostic PCR.

2.6. Evaluation of xylose utilization capacity

To evaluate the xylose utilization capability of the engineered strains, both spot assays and shake-flask fermentations were performed.

Spot Assay: Strains were initially cultivated in 2 mL of YPD medium until the logarithmic growth phase. The cells were harvested by centrifugation and washed twice with sterile water to remove residual carbon sources. The cell suspensions were normalized to an OD₆₀₀ of 1.0 and serially diluted (usually 10-fold gradients). Aliquots (2 μL) of each dilution were spotted onto YNB agar plates or YNBX plates. The plates were incubated at 30 °C for 2–4 days and photographed to record growth phenotypes.

Shake-Flask Fermentation: For growth curve and metabolite analysis, the washed pre-cultures were inoculated into YNB liquid medium supplemented with 20 g/L xylose (YNBX) to an initial OD₆₀₀ of 0.1. Cultivation was performed at 30 °C with shaking at 220 rpm. Samples were withdrawn every 24 h to monitor cell growth OD₆₀₀ and to determine the residual xylose concentration in the supernatant via HPLC. The relative growth capacity was defined as the ratio of the final OD₆₀₀ reached by the engineered strain after 192 h of cultivation (stationary phase) to the corresponding final OD₆₀₀ of the wild-type strain, expressed as a percentage.

The fermentation samples were centrifuged at 10,000 rpm for 5 min, and the supernatant was diluted with ultrapure water and filtered at 0.22 μm for detection. Concentrations of xylose were analyzed by a Sugar-Pak I column (Waters, USA) equipped with RID systems, which used 50 mg/L EDTA-Ca as a mobile phase and was eluted at 80 °C with a flow rate of 0.4 mL/min.

2.7. Fluctuation analysis and mutation rate calculation

Fluctuation analysis was performed according to previously reported protocols. Briefly, 500 ng of the mutator plasmid was electroporated into competent cells. After recovery at 30 °C for 1 h, 1 mL of the transformation culture was transferred into 25 mL of fresh YPD liquid medium supplemented with Hygromycin B and grown to saturation at 30 °C. The cultures were then plated on synthetic defined (SD) agar plates either lacking or supplemented with uracil to determine the number of revertants and total viable cells, respectively. Mutation rates were estimated using the Ma-Sandri-Sarkar Maximum Likelihood method [27] via the online tool Falcor. For the URA3 gain-of-function assay targeting the stop codon (TAA) [28], the calculation was adjusted to account for the fact that 7 out of the 9 possible single-base substitutions result in a functional reversion.

2.8. Sample preparation and high-throughput sequencing

The mutant plasmid library was transformed into K. lactis GG799-URA3∗ competent cells. The transformants were diluted 1:1000 into fresh YPD medium supplemented with 250 μg/mL Hygromycin B and cultivated at 30 °C for 7 days.

For sequencing analysis, 3 mL of the culture was harvested. Genomic DNA was extracted using the SanPrep Column Yeast Genomic DNA Extraction Kit (Sangon Biotech). The extracted DNA served as the template to generate amplicons of the targeted gene via PCR, with the pre-evolution genomic DNA serving as the control. The PCR products were purified using the SanPrep Column DNA Gel Extraction Kit (Sangon Biotech). Library construction, quality control, and high-throughput sequencing were performed on the Illumina NovaSeq platform by GENEWIZ (Suzhou, China).

2.9. Construction of the evolution library and mutagenesis

To introduce diversity into the HGT1 gene, a tiling sgRNA strategy was employed. Multiple sgRNAs were designed to target distinct loci along the HGT1 coding sequence, with a spacing interval of approximately 300 bp. These sgRNAs were individually cloned into the T7-nCas9 evolution system plasmid, which enables localized mutagenesis. The constructed plasmids were transformed into the starting strain WT-XR.

For mutagenesis, the transformants were initially cultured in liquid YPD medium. The cultures were incubated at 30 °C with shaking until the stationary phase was reached, allowing for the accumulation of genetic variations within the target gene.

To enrich for mutants with enhanced xylose utilization, the mutagenized population was subjected to adaptive evolution. The stationary-phase cultures were harvested, normalized to an initial OD₆₀₀ of 0.1, and inoculated into liquid YPX medium containing 20 g/L xylose and Hygromycin B. The evolutionary culture was carried out by serial passaging. Every 3 days, cell growth was monitored by measuring OD₆₀₀. Subsequently, the culture was transferred into fresh YPX medium at a standardized initial cell density.

Following the enrichment cycles, the culture broth was serially diluted and spread onto solid YPX agar plates. After incubation for 3 days, single colonies exhibiting larger sizes compared to the population average were picked as potential positive mutants. The genotype of these variants was verified by colony PCR targeting the HGT1 locus, followed by Sanger sequencing to identify specific mutations.

2.10. Fermentation evaluation in different carbon sources

To comprehensively evaluate the fermentation performance of the engineered strains, shake-flask cultivations were performed using three distinct carbon source conditions: glucose, xylose, and mixed-sugar lignocellulosic hydrolysate. Seed cultures were initially grown in YPD medium, harvested at the exponential growth phase, washed twice with sterile water, and inoculated into the respective fermentation media to an initial OD₆₀₀ of 0.1.

For single-sugar evaluations, the strains were cultivated in YP medium (10 g/L yeast extract and 20 g/L peptone) supplemented with either 20 g/L glucose (YPD) or 20 g/L xylose (YPX). For the mixed-sugar co-fermentation, cultivations were conducted using the formulated lignocellulosic hydrolysate medium described in Section 2.1.

All fermentations were conducted in 250 mL Erlenmeyer flasks at 30 °C with orbital shaking at 220 rpm. Samples were periodically collected to monitor cell growth OD₆₀₀ and to quantify the residual sugar concentrations in the supernatant via HPLC, following the identical analytical method described in Section 2.6.

2.11. Molecular dynamics simulation protocol

The three-dimensional structures of the transporters were predicted using AlphaFold 3 [29]. The chemical structure of xylose was obtained from the PubChem database [30]. To explore the protein-ligand interactions, molecular docking was performed using AutoDock [31]. The resulting docking conformations were visualized and analyzed using PyMOL [32] to identify key interacting residues.

Molecular dynamics (MD) simulations were conducted using the GROMACS [33] 2024.3 software package with the Amberff14sb force field, and the solvent environment was described by the TIP3P water model. The initial binding pose of xylose was determined based on previous molecular docking results. To mimic physiological conditions and ensure overall charge neutrality, an appropriate number of sodium (Na⁺) and chloride (Cl⁻) ions were added to the simulation box. Prior to production simulations, the system was subjected to energy minimization using the steepest descent algorithm to remove unfavorable steric clashes and relax the initial geometry. Subsequently, a 200 ps restrained molecular dynamics (MD) simulation was performed. During this equilibration stage, the integration time step was set to 1 fs to enhance stability, and positional restraints were applied to the protein backbone atoms as well as the ligand molecule, with ligand restraints generated using the Genrestr module. This procedure allowed sufficient equilibration of the solvent and ions while maintaining the structural integrity of the protein–ligand complex. Following equilibration, all restraints were removed, and the time step was increased to 2 fs. A 200 ns production MD simulation was then carried out under isothermal–isobaric (NPT) conditions at 303.15 K (30 °C) and 1 atm pressure.

2.12. Statistical analysis

All quantitative experiments in this study, including shake-flask fermentations, metabolite concentration measurements, and mutation rate calculations, were performed in independent biological triplicates (n = 3). Data are presented as the mean ± standard deviation (SD).

3. Results

3.1. Characterization of carbon source utilization and identification of xylose metabolic bottlenecks

To characterize the substrate utilization capabilities of K. lactis GG799, we evaluated its growth performance in YNB medium containing 20 g/L of different sole carbon sources (Fig. 1a). As expected, the strain efficiently assimilated glucose, lactose, and galactose, showing rapid exponential growth and reaching maximum OD₆₀₀ values exceeding 9.0 by 48 h. Interestingly, a distinct phenotype was observed with maltose, where the strain exhibited a prolonged lag phase of approximately 24 h before entering a rapid growth phase, ultimately yielding an OD₆₀₀ of 7.1. Conversely, the strain demonstrated poor utilization of xylose. The growth trajectory on xylose was comparable to the no-carbon control (OD₆₀₀ < 0.5 at 72 h), confirming the inability of K. lactis GG799 to support growth using xylose as the sole carbon source.

Fig. 1.

Identification of physiological bottlenecks in xylose utilization by K. lactis GG799. (a) Growth profiles of the wild-type K. lactis strain on different carbon sources (20 g/L). (b) Schematic representation of the xylose metabolic pathway and the pentose phosphate pathway (PPP) in yeast. Key enzymes include xylose reductase (XR), xylitol dehydrogenase (XDH), and xylulose kinase (XK). (c) Relative growth capacity of engineered strains with different gene reinforcements. The relative growth capacity is defined as the ratio of the final OD₆₀₀ reached by the engineered strain after 192 h of cultivation (stationary phase) compared to that of the wild-type (WT) strain (set as 100%). All experiments were performed in independent biological triplicates, and error bars represent standard deviations.

The catabolism of xylose in K. lactis proceeds via an oxidoreductive pathway (Fig. 1b): xylose is first reduced to xylitol by xylose reductase (XR), then oxidized to xylulose by xylitol dehydrogenase (XDH), and finally phosphorylated by xylulose kinase (XK) to yield xylulose-5-phosphate, which enters the pentose phosphate pathway (PPP). Although the genome encodes this complete enzymatic machinery, the natural assimilation efficiency remains suboptimal. To identify the rate-limiting steps in this process, we enhanced the expression of key genes involved in xylose transport and metabolism in the wild-type (WT) background. Specifically, endogenous genes (XR, XDH, XK, TAL) and a heterologous xylose transporter gene from K. marxianus (KMAR_10531, designated as XT in this study) were cloned into the pKLAC1 expression vector and integrated into the LAC4 locus via homologous recombination. Growth assays in YNB medium with xylose as the sole carbon source revealed that overexpression of these genes universally improved growth performance compared to the WT (Fig. 1c). Notably, the strains WT-XR and WT-XT exhibited the most significant enhancement, with their relative growth capacities reaching 128% and 122% of the WT, respectively. These findings suggest that both initial xylose reduction and transmembrane transport are critical bottlenecks restricting xylose utilization in K. lactis GG799.

3.2. Identification and functional validation of xylose transporters in K. lactis

Research on xylose transporters remains limited. To identify potential candidates in K. lactis, we performed a genome-wide phylogenetic analysis using sugar transporter sequences from K. lactis, K. marxianus, and S. cerevisiae. The resulting phylogenetic tree clustered these proteins into distinct functional clades (Fig. 2). In K. marxianus, xylose transport is primarily mediated by HGT1 (a high-affinity glucose transporter present in six copies) and the HXT homologs KHT1 (a low-affinity glucose transporter, two copies) and KHT2 (a high-affinity glucose transporter, four copies) [34]. Additionally, the K. marxianus glycerol transporter encoded by KMAR_60179 has been reported to exhibit low-affinity xylose transport activity [35]. Similarly, in S. cerevisiae, most proteins capable of xylose transport cluster within the Hxt-like family. Based on sequence homology to these characterized transporters in K. marxianus and S. cerevisiae, we identified four orthologous candidates in the K. lactis genome: KLLA0_A11110g (HGT1), KLLA0_D13310g (RAG1, a homolog of KHT1 and KHT2), and KLLA0_F26246g (STL1-1) and KLLA0_A03223g (STL1-2), which correspond to the glycerol transporter lineage. Consequently, we hypothesized that these four proteins possess xylose transport capabilities.

Fig. 2.

Genome-wide phylogenetic analysis of sugar transporters in K. lactis and related yeasts. The phylogenetic tree was constructed using the Maximum Likelihood method based on protein sequences of the Major Facilitator Superfamily (MFS). The analysis includes transporters from three yeast species: S. cerevisiae (indicated in blue), K. marxianus (indicated in purple), and K. lactis (indicated in red). The outer colored rings delimit functional clades predicted based on the substrate specificity of characterized homologs, classifying the transporters into clusters such as Hgt-like, glycerol, xylose, Hxt-like, sensors, maltose, inositol, fructose, and lactose.

Single and combinatorial gene deletion mutants were constructed utilizing the CRISPR/Cas9 system to elucidate the physiological roles of the identified candidates in xylose transport. Spot assays revealed that on glucose medium (Fig. 3a), only the ΔRAG1 strain exhibited a noticeable growth defect with smaller colony morphology, suggesting RAG1 may also participate in glucose transport, whereas other single mutants grew comparably to the wild-type (WT). In contrast, on xylose medium, the strains displayed distinct phenotypes: ΔHGT1 showed the most severe growth impairment, being barely visible at high dilutions, followed by ΔRAG1, while ΔSTL1-1 and ΔSTL1-2 were only slightly affected. Notably, the growth capability on xylose declined progressively with the accumulation of gene deletions (2D, 3D, 4D), with the quadruple mutant (4D) completely losing the ability to grow on xylose. Notably, although the maximum specific growth rates (μ_max) of the single-deletion strains did not differ significantly from the wild-type (Fig. S1a), their maximum xylose consumption rates revealed distinct functional impairments (Fig. S1b). Specifically, the ΔHGT1 strain exhibited the lowest rate (0.04 g/L/h), less than half that of the WT (0.10 g/L/h), while the ΔRAG1 strain showed an intermediate reduction. In contrast, deletions of STL1 genes had negligible impact on uptake capacity. These physiological disparities translated into marked deviations in their overall growth curves. Liquid fermentation kinetics further corroborated these findings (Fig. 3b). After 168 h of cultivation, the WT strain reached an OD₆₀₀ of around 6.6. Conversely, the ΔHGT1 strain exhibited severely retarded growth (OD₆₀₀ of approximately 3.8), representing a roughly 42% reduction in biomass compared to the WT. The ΔRAG1 strain showed sluggish growth in the late phase (OD₆₀₀ close to 4.5), while ΔSTL1-1 and ΔSTL1-2 showed only minor reductions, indicating auxiliary roles in xylose transport. Regarding multiple deletion strains, the 2D strain showed significantly suppressed growth, whereas the 3D and 4D strains exhibited similarly abolished growth with overlapping profiles, indicating that the presence of a single remaining candidate transporter is insufficient to support effective growth on xylose. Consistent with the growth profiles, xylose consumption analysis (Fig. 3c) showed that while the WT utilized xylose most efficiently, single-deletion strains displayed varying degrees of delay. Crucially, the residual xylose concentration in both the 3D and 4D strain cultures remained virtually unchanged over 168 h. Collectively, these results confirm that HGT1 and RAG1 serve as the primary xylose transporters in K. lactis, and all four proteins cooperatively mediate xylose uptake.

Fig. 3.

Functional characterization of K. lactis xylose transporters via CRISPR-Cas9 mediated combinatorial deletion and single-gene complementation. (a) Spot plate assays of wild-type (WT) and deletion mutants on YNB agar with 20 g/L glucose or xylose. "2D", "3D", and "4D" denote double, triple, and quadruple deletion strains, respectively. (b) Growth profiles OD₆₀₀ in liquid xylose medium. (c) Xylose consumption kinetics during fermentation. (d) Spot assays of the transporter-null chassis (4D) rescued by individual candidate genes expressed. (e) Growth kinetics of the complemented strains. All experiments were performed in independent biological triplicates, and error bars represent standard deviations.

Aiming to delineate the individual contributions of the four candidate transporters to xylose uptake, single-gene complementation assays were performed in the quadruple deletion (4D) background. The coding sequences of the target genes were cloned into the integrative vector pKLAC1-Hyg and linearized to facilitate integration into the LAC4 locus. This strategy placed each transporter under the control of the strong, constitutive P_LAC4 promoter, generating four rescued strains designated as 4D-RAG1, 4D-HGT1, 4D-STL1-1, and 4D-STL1-2. Spot assays on xylose medium revealed distinct growth phenotypes among the complemented strains (Fig. 3d). The 4D-RAG1 strain exhibited the most robust growth, comparable to the wild-type, confirming the dominant role of RAG1. Notably, despite its annotation as a glycerol transporter, 4D-STL1-1 showed substantial growth recovery, with colony size and density clearly surpassing those of 4D-HGT1 and 4D-STL1-2. In contrast, 4D-STL1-2 displayed only marginal growth recovery. Liquid growth kinetics further quantified this hierarchy (Fig. 3e). By 216 h, 4D-RAG1 achieved the highest biomass (OD₆₀₀ approximately 7.0). The 4D-STL1-1 strain followed with a final OD. reaching 5.2, effectively outperforming both 4D-HGT1 and 4D-STL1-2, which plateaued at a lower biomass (OD₆₀₀ around 3.0) with overlapping profiles. An intriguing discrepancy emerged regarding HGT1. While the single deletion of HGT1 resulted in the most severe growth defect among the single mutants (Fig. 3a), its constitutive overexpression in the 4D background failed to confer a commensurate growth advantage, performing significantly worse than STL1-1. This paradox suggests that the critical physiological role of HGT1 in the wild-type strain is likely attributable to high native transcriptional abundance. While RAG1 remains the most efficient xylose transporter, STL1-1 shows significant potential under constitutive expression. Furthermore, the data underscores a marked functional divergence between the STL1 family members, STL1-1 and STL1-2, regarding xylose metabolism.

3.3. Systematic characterization of the T7DNAP3M in vivo mutagenesis system in K. lactis

To achieve targeted evolution of native transporters, we employed the T7DNAP3M mutagenesis system of CTRLE previously established by our group [17]. Mechanistically, this system utilizes the nCas9 nickase to recruit an error-prone T7 DNA polymerase, thereby facilitating targeted hypermutation in the region proximal to the nick site (Fig. 4a). While the fundamental architecture was constructed in our prior work, its performance in K. lactis remains to be systematically and quantitatively evaluated, which is crucial to verify its applicability for transporter engineering. Here, the off-target rate, mutation window, and mutation spectrum were comprehensively characterized using a URA3 reporter allele harboring a premature TAA stop codon at position 131. Specifically, by designing sgRNAs to target either the CAN1 locus (located on a separate chromosome from the reporter) or sites positioned at varying distances from the reporter codon, we rigorously evaluated the off-target effect and the effective length of the mutation window, respectively. In this assay, any mutation converting the TAA stop codon to a sense codon restores the strain's ability to grow on uracil-deficient medium. Using the spontaneous mutation rate of the wild-type strain (1.0 × 10⁻¹⁰ mutations/nucleotide/generation) as a baseline, the T7DNAP3M mutagenesis system demonstrated exceptional efficiency. The on-target mutation rate exhibited a dramatic 6 × 10⁷-fold increase compared to the wild-type level (Fig. 4b). In contrast, the off-target mutation rate showed only a moderate 200-fold increase. The effective range of mutagenesis was further evaluated (Fig. 4c). Notably, although the mutation frequency gradually declined with increasing distance from the nicking site, the rates remained consistently above 10⁻⁶ mutations/nucleotide/generation within a 325 bp region. These data indicate that the T7DNAP3M system possesses an effective mutation window extending beyond 300 bp.

Fig. 4.

Systematic characterization of the T7DNAP3M in vivo mutagenesis system in K. lactis. (a) Mechanism of continuous evolution system in vivo. (b) Quantitative assessment of on-target and off-target mutation rates compared to the wild-type spontaneous mutation baseline (control). (c) Analysis of the effective mutation window. (d) Spatial distribution of substitution frequencies across a genomic region spanning −200 bp to +500 bp relative to the nick site, derived from high-throughput sequencing data. (e) The mutational spectrum of T7DNAP3M. The stacked bar chart illustrates the percentage of specific base substitutions for each original base (A, T, C, G). All experiments were performed in independent biological triplicates, and error bars represent standard deviations.

To comprehensively characterize the mutagenic profile of the T7DNAP3M system, high-throughput sequencing was performed on the target locus. The sequencing data revealed a broad and robust mutational footprint. The T7DNAP3M system induced a substantial increase in substitution frequencies compared to the wild-type control (Fig. 4d). Notably, this hyper-mutation was distributed across a wide genomic window, with elevated mutation frequencies detectable from −100 bp to +400 bp relative to the nick, confirming the system's capability to introduce mutations over long DNA fragments. Base substitution patterns were further analyzed to determine the mutational spectrum (Fig. 4e). Crucially, the data demonstrated that the T7DNAP3M system is capable of generating all possible types of base substitutions across all four nucleotides. This confirms that the system can access a diverse sequence space for protein engineering.

3.4. Directed evolution of HGT1 for enhanced xylose transport

A growth-coupled, in vivo mutation-driven continuous evolution strategy was employed to augment xylose transport capacity. The previously constructed WT-XR strain served as the starting chassis. In this engineered strain, the downstream xylose metabolic pathway had been reinforced to prevent intracellular xylose accumulation. This metabolic design ensures that transmembrane transport remains the rate-limiting step for cellular growth, thereby establishing a direct linkage between transport efficiency and growth rate to drive effective selective pressure.

HGT1 was selected as the target for evolution based on three primary criteria. First, our deletion assays identified it as the dominant endogenous xylose transporter, and its orthology to the high-affinity transporter from K. marxianus suggests favorable basal activity. Second, combined evidence from knockout and overexpression experiments indicates that the HGT1 locus likely supports high expression levels, making it an ideal template for evolution to achieve significant gains in total transport flux. Third, and most crucially, while RAG1 exhibited robust specific activity for xylose, its deletion severely impaired native glucose transport (Fig. 3a). Because our ultimate goal is the efficient co-utilization of mixed sugars in lignocellulosic hydrolysates, evolving RAG1 might undesirably compromise the yeast's inherent efficiency in glucose assimilation. Therefore, HGT1 was selected as the optimal and safer target to maximize xylose transport without sacrificing robust native glucose metabolism.

A mutagenesis plasmid specifically targeting the HGT1 locus (Fig. 5a) was constructed and transformed into the WT-XR strain. The resulting population was then subjected to serial subculturing in medium containing xylose as the sole carbon source (YPX). Population-level growth monitoring (Fig. 5b) revealed that the evolved population targeting HGT1 exhibited a progressive growth advantage. While the difference was marginal during the first three transfers, a substantial surge in biomass was observed starting from the fourth transfer. By the end of the fifth transfer, the OD₆₀₀ of the evolved population reached approximately four-fold that of the control (the WT-XR strain transformed with an off-target mutagenesis plasmid, ensuring an identical baseline of metabolic burden and systemic evolutionary pressure). These results indicate that, in the transport-limited WT-XR background, continuous evolution successfully enriched phenotypes with enhanced xylose utilization capabilities. Subsequently, the culture was diluted and plated onto YNBX medium; sequencing of large colonies led to the isolation of a mutant strain harboring an HGT1-I463F mutation.

Fig. 5.

Directed evolution of HGT1 for improved xylose utilization and phenotypic characterization. (a) Schematic overview of the in vivo continuous mutagenesis and selection strategy for evolving the xylose transporter HGT1. Cells harboring the mutagenesis system were transformed with the target gene and subjected to iterative rounds of cultivation using xylose as the sole carbon source. Improved variants were enriched through serial passaging, followed by screening on xylose plates and confirmation by phenotypic assays. (b) Growth performance (OD₆₀₀) of the control strain(WT-XR transformed with an off-target evolution plasmid) and the strain expressing evolved HGT1 across successive generations during adaptive enrichment. (c) Time-course profiles of cell growth (OD₆₀₀) for wild-type (WT), WT expressing xylose reductase (WT-XR). (d) Corresponding xylose consumption profiles of WT, WT-XR, and XR-HGT1M strains over time. All experiments were performed in independent biological triplicates, and error bars represent standard deviations.

The specific contribution of the HGT1-I463F mutation to the evolved phenotype was delineated by validating its effect via reverse engineering. The mutant allele was introduced into the parental WT-XR background to generate the reconstructed strain, XR-HGT1M (Fig. 5c). Within 72 h, it achieved an OD₆₀₀ of 25, surpassing the WT-XR strain by approximately 20% and reaching a biomass nearly double that of the WT. Although this biomass enhancement was more moderate than the ∼4-fold difference observed in the evolved population under screening pressure (Fig. 5b)—suggesting the potential contribution of synergistic genomic adaptations in the population—the fact that the single I463F mutation alone conferred a distinct growth advantage confirms it as a significant contributing factor of the improved phenotype. Furthermore, xylose consumption analysis (Fig. 5d) corroborated these findings, showing that the XR-HGT1M strain utilized xylose at a notably faster rate than the parental strain, thereby directly linking the growth advantage to enhanced transport efficiency.

To further evaluate the industrial robustness of the evolved transporter, we assessed the performance of the engineered strains using glucose as the sole carbon source and a formulated lignocellulosic hydrolysate medium (containing approx. 22 g/L glucose and 11 g/L xylose) as a mixed-sugar source. When cultured with glucose alone (Fig. S2), the growth and glucose consumption kinetics of the XR-HGT1M mutant were identical to the parental strains, confirming that the HGT1-I463F mutation preserves the native glucose utilization capability. Interestingly, during co-fermentation in the formulated hydrolysate medium (Fig. S3), the profiles revealed that both the parental and engineered strains initiated continuous xylose consumption (Fig. S3c) while glucose was still abundant and being rapidly assimilated (Fig. S3b). This indicates that K. lactis inherently possesses a relatively weak carbon catabolite repression (CCR) mechanism, enabling a natural degree of true simultaneous co-utilization. Within this co-utilization context, because the yeast's intrinsic metabolic capacity for glucose is vastly higher than for xylose, glucose is consumed at a much faster rate, which heavily dominates the early macroscopic fermentation profile. However, despite this intense kinetic competition from glucose, the XR-HGT1M mutant exhibited a measurable and statistical advantage in xylose utilization. Specifically, the xylose consumption of the XR-HGT1M mutant at 24 h was higher than that of the WT (p < 0.01) and WT-XR (p < 0.05) strains (Fig. S3d). This robust evidence demonstrates that the HGT1-I463F mutation successfully enhances xylose uptake even in complex mixed-sugar environments.

3.5. Molecular docking and molecular dynamics simulation of transporters

To unravel the molecular mechanism underlying the enhanced transport efficacy of the mutation I463F, molecular docking and molecular dynamics (MD) simulations were performed.

Structural docking analysis (Fig. 6a and b) revealed that although I463F is located distally from the substrate-binding pocket, it induces a significant conformational rearrangement of the active site via allosteric effects. Comparing the binding modes of the wild-type (WT) and the mutant, the I463F variant exhibits a more compact binding pocket with strengthened interactions between key residues and xylose. Specifically, the hydrogen-bonding distance between Asp45 and the xylose molecule was shortened from 3.3 Å/2.7 Å in the WT (Fig. 6a) to 2.9 Å/2.5 Å in the mutant (Fig. 6b), facilitating a more robust hydrogen bond network. Furthermore, the side chain of Asn442 underwent a conformational shift, transitioning from a weak interaction with the ether oxygen of the hemiacetal group in the WT to a stronger hydrogen bond with the hydroxyl group in the mutant. Collectively, this remodeling of the hydrogen bond network contributes to enhanced substrate recognition and stability within the binding pocket.

Fig. 6.

Structural and molecular dynamics analysis of HGT1 and the I463F mutant. (a) Overall structural model of the wild-type HGT1 transporter and a zoomed-in view of the substrate-binding pocket. (b) Structural model of the HGT1 I463F mutant showing altered interactions within the substrate-binding pocket. (c) Time evolution of backbone root-mean-square deviation (RMSD) for wild-type HGT1 and the HGT1 I463F mutant during 200 ns molecular dynamics simulations. (d) Root-mean-square fluctuation (RMSF) profiles of individual residues for HGT1 and the HGT1 I463F mutant. (e) RMSD of bound xylose relative to the transporter during molecular dynamics simulations for HGT1–xylose and HGT1 I463F–xylose complexes.

MD simulations provided further insights into the dynamic properties of the transporter. Analysis of the backbone Root Mean Square Deviation (RMSD) (Fig. 6c) showed that the I463F mutant reached equilibrium more rapidly and stabilized at a lower value (∼0.7 nm) compared to the WT, which drifted to ∼1.1 nm, indicating that the mutation significantly enhances global protein stability. This is corroborated by the Root Mean Square Fluctuation (RMSF) analysis (Fig. 6d), where the mutant displayed markedly reduced residue flexibility, particularly in the C-terminal region, where the fluctuation peak dropped from 2.5 nm in the WT to below 0.6 nm. These findings align with recent studies on ancestral xylose transporters, where reduced structural fluctuations were similarly associated with enhanced transport efficiency [36].

Crucially, the ligand RMSD profile (Fig. 6e) highlighted a distinct kinetic behavior. While the xylose in the WT complex remained tightly bound until ∼120 ns, the xylose in the I463F complex exhibited significant positional fluctuations (RMSD >10 nm) starting as early as ∼50 ns. In the context of transport mechanics, this earlier substrate destabilization suggests a facilitated release mechanism.

4. Discussion

The efficient conversion of lignocellulosic biomass relies on the coordinated uptake and catabolism of xylose [37,38]. While K. lactis possesses a native pathway for xylose assimilation, its wild-type efficacy is negligible [12,39]. To address the bottleneck of xylose uptake in K. lactis, we characterized the endogenous transporter landscape. While RAG1 exhibits the highest specific activity, HGT1 presents a kinetic paradox: its critical physiological role is likely sustained by high transcriptional abundance rather than intrinsic turnover, underscoring the necessity for evolutionary enhancement. Additionally, we observed marked functional divergence within the STL1 family, where constitutively expressed STL1-1 unlike the inactive STL1-2 demonstrated significant transport potential. The identification of these native transporters established the necessary baseline for further optimization, highlighting the potential of mining endogenous genetic resources before resorting to heterologous expression.

A conventional strategy for enhancing yeast xylose fermentation typically involves the heterologous expression of highly efficient transporters derived from naturally xylose-fermenting species. However, the introduction of foreign membrane proteins frequently encounters systemic bottlenecks, such as misfolding, mistargeting, and heterologous toxicity, which can induce severe membrane stress and metabolic burden. In contrast, the "mining-evolving" strategy employed in this study presents inherent physiological advantages. By engineering an endogenous membrane protein like HGT1, the evolved transporter ensures superior structural folding and seamless integration into the host cell membrane. Furthermore, native transporters are already fully integrated into the host's pre-existing regulatory networks, chaperone systems, and vesicular trafficking pathways. By applying continuous evolution to these native targets, we can significantly enhance their transport kinetics while effectively circumventing the physiological incompatibilities associated with heterologous expression, ultimately yielding a more robust, compatible, and genetically stable chassis for industrial biorefining.

Currently, CRISPR-guided base editors [16] and yEvolvR [28] are representative targeted in vivo mutagenesis systems established in yeast. By tethering deaminases to Cas9 variants (dCas9 or nCas9), CRISPR-based base editors enable targeted hypermutation at specific genomic loci. However, this method is constrained by a narrowed mutational window (∼100 bp) and bias base substitution types provided by deaminase in S. cerevisiae. The yEvolvR system based on error-prone CRISPR-guided DNA polymerase I is another powerful tool for in vivo continuous evolution in S. cerevisiae, reported to diversify all four nucleotides within a mutational window primarily concentrated within 50 bp surrounding the nicking site. In this study, the T7DNAP3M mutagenesis system of CTRLE [17] in K. lactis, a yEvolvR-derived technology, was demonstrated a sustained mutational window exceeding 300 bp, which is particularly advantageous for evolving large transporter proteins (>600 amino acids). Moreover, T7DNAP3M system exhibits superior mutagenic performance: it achieves a high on-target mutation rate while simultaneously maintaining a low off-target frequency. Collectively, the T7DNAP3M system offers a robust, efficient, and safe platform for the continuous evolution of industrial yeast strains. To validate the efficacy of this system, we applied it to the continuous evolution of HGT1, marking the first successful implementation of an in vivo continuous evolution platform in K. lactis.

Remarkably, the effective mutant HGT1-I463F was isolated after only five rounds of serial passage, significantly reducing the time consumption compared to traditional adaptive laboratory evolution (ALE) [40,41]. Although the isolated HGT1-I463F mutant significantly improved biomass yield, it did not fully recapitulate the performance of the evolved population. This observation aligns with evolutionary principles where adaptation is often driven by the synergistic action of major driver mutations and background polygenic variations [42]. Nevertheless, the rapid isolation of a dominant beneficial mutation within such a short timeframe serves as a robust proof-of-concept for the system's high efficiency. Crucially, this pioneering application establishes a foundational framework for the future mining and evolutionary engineering of endogenous genes in this industrially important yeast.

Furthermore, our co-fermentation assays using actual lignocellulosic hydrolysate revealed the significant industrial potential of the evolved transporter. While K. lactis naturally exhibits a weak carbon catabolite repression (CCR) phenotype [43] that allows for a certain degree of simultaneous sugar uptake, the massive disparity in the inherent metabolic kinetics between glucose and xylose typically dominates the fermentation profile. Remarkably, despite this competitive environment, the XR-HGT1M mutant demonstrated a statistically significant enhancement in xylose consumption at 24 h compared to the parental strains. This indicates that the HGT1-I463F mutation effectively raises the baseline for xylose transport in complex mixed-sugar streams without compromising native glucose utilization. In the future, continuing to enhance the capacity of xylose transporters and accelerating downstream xylose metabolic flux may fully bridge this kinetic gap, ultimately achieving efficient synchronous co-utilization.

Finally, MD simulations provided a mechanistic basis for the enhanced performance of the I463F mutant. A major challenge in transporter engineering is overcoming the trade-off between substrate affinity and turnover rate [44]. Our analysis revealed that while the I463F mutation tightens the binding pocket to enhance specificity, it paradoxically induces earlier substrate destabilization, as evidenced by high-amplitude ligand RMSD fluctuations. This finding highlights the power of combining continuous evolution with computational analysis to resolve complex kinetic bottlenecks.

In summary, this study presents a comprehensive framework for strain engineering: spanning from the identification of endogenous targets (RAG1/HGT1), to the application of a high-efficiency, wide-window evolutionary tool (T7DNAP3M), and culminating in the mechanistic resolution of transport kinetics. This integrated strategy provides a robust toolkit for mining and optimizing critical functional loci in K. lactis.

CRediT authorship contribution statement

Xiangdi Chen: Writing – original draft, Methodology, Investigation, Formal analysis, Data curation. Shuaili Chen: Writing – review & editing, Methodology, Investigation, Formal analysis, Data curation. Jingwen Zhou: Supervision, Project administration. Guocheng Du: Supervision, Project administration. Guoqiang Zhang: Writing – review & editing, Supervision, Resources, Funding acquisition.

Data availability statement

The data supporting the findings of this study are available within the paper and its supplementary information files. The datasets generated and analyzed during the current study are available from the corresponding author upon request.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: