Establishment of SRLC: a multiplex genome editing technology for Saccharomyces cerevisiae and its application in metabolic engineering of malonyl-CoA pathway

Abstract

The development of advanced genome engineering tools is crucial for optimizing metabolic pathways in Saccharomyces cerevisiae and achieving efficient biomanufacturing. This study proposes an enhancing multiplex genome editing strategy in S. cerevisiae by employing Escherichia coli-derived single-stranded annealing proteins (SSAPs) combined with S. cerevisiae-derived homologous recombinases (Rad51 and Rad52). The strategy utilizes an SSAP-Rad-Linearized CRISPR (SRLC) platform, which supports efficient simultaneous editing of multiple genomic loci without constructing complex multi-gRNA expression vectors. Co-overexpressing Rad51/Rad52 and E. coli SSAP proteins significantly enhances homologous recombination (HR), allowing precise multi-locus genome editing mediated by short homologous arms. Furthermore, SRLC employs a linearized CRISPR-Cas system to stimulate homologous recombination and enable counter-selection in S. cerevisiae, thereby improving precise multiplex genome editing efficiency. We applied SRLC to engineer the malonyl-CoA metabolic pathway in S. cerevisiae. Through a single round of editing and screening, we constructed a chassis strain with 9 targets simultaneously modification and achieved a 9.6-fold increase in intracellular malonyl-CoA. Using this chassis, 3-hydroxypropionic acid production increased 4.5-fold relative to wild-type S. cerevisiae. This platform offers a robust and scalable tool for S. cerevisiae manipulation and a practical pathway-engineering strategy for building for malonyl-CoA-derived factories.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12934-026-03068-w.

Introduction

Yeasts, and in particular Saccharomyces cerevisiae, underpin both ancient fermentations and modern bio-manufacturing, offering genetic tractability, industrial robustness, and a well-annotated eukaryotic genome that collectively make them attractive chassis for building cell factories [1]. Despite significant advances in the development and application of gene editing technologies for S. cerevisiae, the efficiency of genome-wide multi-locus editing remains a major bottleneck in building S. cerevisiae cell factories [2]. To address this bottleneck, it is crucial to establish accurate, rapid, and high-throughput genome editing and screening techniques suitable for multi-site engineering in S. cerevisiae.

CRISPR-Cas systems have transformed genome engineering by enabling programmable, RNA-guided double-strand breaks (DSBs) that can be repaired to generate targeted mutations or precise insertions [3–5]. In yeast, CRISPR-Cas9 rapidly became a general-purpose tool for markerless editing, leveraging the organism strong homologous recombination capacity to install deletions, substitutions, and cassette integrations [6–9]. However, the practical throughput of CRISPR engineering in yeast is frequently limited by the logistics of guide RNA (gRNA) construction and delivery, particularly when multiplex editing is needed for pathway refactoring, genotype-phenotype mapping, or combinatorial optimization. Multiplex CRISPR strategies in yeast typically express multiple gRNAs from plasmids or integrated arrays, using approaches such as RNA processing by heterologous endoribonucleases or tRNA-based architectures to liberate individual guides from polycistronic transcripts [10, 11]. Even when cloning is streamlined, plasmid-based multiplexing introduces additional constraints, including copy-number variation, context-dependent expression, and the need to recycle selectable markers and plasmid backbones across rounds of engineering [7, 8].

Cloning-free inspirations have emerged in other fungi and engineered systems, including tRNA promoter-driven guide expression and transient gRNA delivery strategies that reduce the overhead of constructing stable guide vectors [12, 13]. An appealing strategy is to deliver editing components as linear DNA, which can be generated rapidly by PCR and assembled on demand without intermediate cloning [14, 15]. Yeast genetics has long exploited recombination with linear fragments and synthetic oligonucleotides for site-directed modifications and chromosome engineering, demonstrating that short donors can, in principle, support precise genome rewriting when repair intermediates are appropriately managed [16, 17]. In bacteria, linear recombineering and related approaches have enabled rapid genome modifications using short homology arms and oligonucleotide donors, and have recently been combined with CRISPR selection to accelerate multiplex genome rewriting [18–21]. For yeast, linear delivery could similarly decouple guide design from plasmid construction, but this strategy faces two long-standing challenges: linear DNA is vulnerable to nucleolytic degradation, and the efficiency of HDR from short donors can be highly sensitive to the balance between competing DNA repair pathways and the kinetics of DSB repair.

Precise repair of DSB in S. cerevisiae is dominated by homologous recombination, which proceeds through end resection, Rad51-family recombinase filament assembly, homology search, and strand invasion, and is coordinated by a network of accessory factors that tune pathway choice and fidelity [17, 22–24]. Rad52 plays a central role in yeast HDR by mediating Rad51 loading and facilitating annealing reactions, while Rad51 paralogues such as Rad55/Rad57 modulate filament dynamics in opposition to anti-recombinase activities [25, 26]. These mechanistic layers help explain why HDR outcomes can differ markedly across loci, donor formats, and cellular states, and why engineering interventions that shift repair factor availability can yield large gains in precise editing. Single-strand annealing proteins (SSAPs) from bacteria and bacteriophages, exemplified by RecT and λRedβ, catalyze annealing of complementary single-stranded DNA (ssDNA) and are the enzymatic core of many recombineering systems [15, 20, 21]. Structural and mechanistic studies have revealed how these SSAPs form nucleoprotein filaments or clamps that stabilize ssDNA, accelerate homology recognition, and promote productive pairing even with relatively short regions of identity [27, 28]. Inspired by these properties, SSAPs have been repurposed beyond bacteria, including in mammalian contexts where SSAP-enhanced strategies can boost precise integration under Cas9 cleavage [29, 30]. Together, these advances suggest a general principle: transiently favoring annealing-competent intermediates and HDR-competent repair states can convert linear donor DNA from a weak substrate into an efficient template for genome rewriting.

Rapid and multiplexable genome editing is crucial for engineering S. cerevisiae, as constructing efficient cell factories often requires coordinated optimization of distributed traits such as precursor supply, cofactor balance, metabolite transport, and stress tolerance [1]. Taking the synthesis of malonyl-CoA-derived chemicals as an example, malonyl-CoA serves as a key precursor for fatty acids and polyketides. Its intracellular level is tightly regulated and frequently becomes a limiting factor for heterologous pathway productivity [31]. To address this, we developed a linearized CRISPR editing workflow for S. cerevisiae, which enables efficient and simultaneous cleavage at multiple genomic loci by circumventing the need for constructing complex multi-guide RNA co-expression vectors. Furthermore, we integrated the endogenous homologous recombination (HR) repair system of yeast with the molecular logic of bacterial recombineering. By combining Rad51/Rad52-dependent HR with a single-strand annealing protein (SSAP)-assisted annealing mechanism derived from bacteria, we significantly enhanced the precision and efficiency of repairing multiple double-strand breaks (DSBs). Based on this strategy, we established a robust platform for precise multiplex genome editing in S. cerevisiae, termed SRLC (SSAP-Rad51/52-Linearized CRISPR).

Using the SRLC platform, we rapidly engineered the malonyl-CoA metabolic network in S. cerevisiae. By employing a biosensor that converts intracellular malonyl-CoA levels into a fluorescent signal [32, 33], we enabled high-throughput screening of engineered strains. Results demonstrate that SRLC supports simultaneous, multiplex genomic rewriting in yeast, with the editing scale being highly scalable and readily adaptable to downstream phenotypic analysis needs. This platform makes systematic, multi-locus optimization of complex metabolic pathways in S. cerevisiae a routine operation, providing a powerful tool for accelerating metabolic engineering efforts in yeast.

Results

Establishment of the linearized CRISPR system in S. cerevisiae

Drawing on strategies established in bacteria that use linear PCR-derived gRNA expression fragments and enhance their stability through terminal modifications [18, 19], we developed in S. cerevisiae S288C a linearized CRISPR workflow based on plasmid-borne, constitutively expressed Cas9 combined with transient delivery of linear gRNA cassettes. Specifically, the gRNA expression unit was generated by PCR as a short linear fragment comprising promoter, gRNA, tracrRNA and terminator sequences, and protective phosphorothioate modifications were introduced at both ends of the fragment to reduce the risk of exonucleolytic degradation of nucleic acids in the eukaryotic nucleus. In parallel, a donor DNA template was provided as double-stranded DNA carrying homology arms flanking the target locus (Fig. 1a-c). In comparison with conventional CRISPR-Cas9 approaches, the editing efficiency of this linearized CRISPR framework was essentially maintained, indicating that transient delivery of stability-protected linear gRNA expression cassettes can achieve performance comparable to standard implementations (Fig. 1d). This preservation of efficiency provides a practical rationale for deploying the linearized CRISPR strategy in multiplex genome editing, where repeated or simultaneous delivery of multiple guide constructs is required and the simplicity of linear fragment preparation can be leveraged without incurring a substantial penalty in editing outcome. To optimize delivery conditions, we performed a dose-gradient comparison of the amount of linear gRNA fragment used for transformation and observed a pronounced optimal window. At low input levels, homologous recombination efficiency was reduced, whereas excessive input was associated with impaired cell growth and a concomitant decline in homologous recombination efficiency (Fig. S1a). Notably, alterations to the homology arms of the donor DNA markedly affected editing efficiency, indicating that system performance is highly sensitive to the availability and activity of homologous recombination mechanisms (Fig. S1b). Overall, this linearized CRISPR route integrates stability-enhanced transient gRNA delivery with stable Cas9 expression to support efficient genome editing in yeast, and highlights the importance of balancing nucleic-acid dosage with recombination capacity to achieve precise editing outcomes.

Fig. 1.

Establishment of the linearized CRISPR system in S. cerevisiae S288C. a Schematic diagram of the conventional CRISPR-Cas9 system setup in S. cerevisiae. b Workflow of linear gRNA expression and donor DNA delivery for genome editing in S. cerevisiae. The system involves the transient delivery of linear gRNA expression fragments and a donor DNA template with homology arms flanking the target locus. c Cas9 introduced a double-strand break at the genomic target site, and precise genome editing was achieved through donor DNA-directed homology dependent repair. d Comparison of genome insertion efficiencies between conventional CRISPR-Cas9 and linearized CRISPR systems. Values are means of the biological replicates, and the error bars indicate the standard deviations of all (n = 4) biological replicates. P values were obtained using the two-tailed Student’s t-test: ****P < 0.0001

Previous studies have demonstrated that co-delivering multiple short gRNA expression PCR fragments in bacteria and fungi can alleviate the burden of constructing multi-gRNA expression systems [12, 13, 19]. Building on this, we tested the co-transformation delivery of multiple linear gRNA fragments in S. cerevisiae (Fig. 2a). The results showed that co-delivery of four linear gRNA fragments successfully enabled simultaneous editing of four target loci (Fig. 2b). Compared to traditional multi-gRNA plasmid constructs, the co-delivery of linear gRNA fragments significantly reduced the construction time and enabled parallel activation of multiple targets. This approach provides a technical foundation for the delivery step in the development of multi-locus engineering workflows.

Fig. 2.

Co-delivery of multiple linear gRNA fragments for multiplex genome editing in S. cerevisiae S288C. a Schematic representation of the co-transformation strategy used to deliver multiple linear gRNA fragments in S. cerevisiae. Multiple linear gRNA fragments targeting different loci are simultaneously introduced into the S. cerevisiae cells for parallel editing. b Efficiency of quadruple integration by the linearized CRISPR system. Values are means of the biological replicates, and the error bars indicate the standard deviations of all (n = 4) biological replicates. P values were obtained using the two-tailed Student’s t-test: ****P < 0.0001

Evaluating the overexpression of Rad family genes for enhancing homologous recombination on the linearized CRISPR platform

Building on the previously established and validated linearized CRISPR platform, we further explored the overexpression of key homologous recombination factors in S. cerevisiae S288C (Rad51, Rad52, Rad54, Rad55, and Rad57), to systematically compare their effects on the efficiency of homologous recombination repair following Cas9-induced double-strand breaks (DSB) (Fig. 3a). This investigation provides a quantifiable basis for enhancing repair pathways in the development of multi-locus editing workflows in yeast. Compared to the linearized CRISPR control group, overexpression of Rad51 significantly increased the rate of correct editing. Under standard donor conditions (100 bp homology arms), the correct editing rate improved from 18% to 28%, a 1.5-fold increase. Under donor-limiting conditions (homology arms shortened by 30 bp), the control group showed a dramatic decrease in correct editing rate to 0.67%, whereas the Rad51 overexpression group maintained a rate of 1.67%, indicating that Rad51 overexpression more effectively directs DSB repair toward homologous recombination-mediated precise outcomes (Fig. 3b). Rad52 overexpression also provided a moderate improvement under standard conditions, with the correct editing rate rising from 18% to 22% (approximately 1.2-fold). However, under donor-limiting conditions (homology arms shortened by 30 bp), the correct editing rate in the Rad52 group remained nearly the same as the control, suggesting that its overall robustness was weaker than that of Rad51 (Fig. 3c). Notably, when donor homology arms were longer than 500 bp, the enhancement in editing efficiency by Rad51 and Rad52 overexpression disappeared, highlighting the critical importance of overexpressing these proteins when using short-homology-arm donors.

Fig. 3.

The effect of overexpression of Rad family genes combined with linearized CRISPR system on homologous recombination Repair in S. cerevisiae S288C. a Experimental setup for evaluating the impact of Rad family gene overexpression on homologous recombination efficiency in the linearized CRISPR system. Key genes in the homologous recombination pathway, such as Rad51, Rad52, Rad54, Rad55, and Rad57, are overexpressed in S. cerevisiae to determine their effect on precise editing outcomes. Overexpression of Rad51 (b), Rad52 (c), and Rad54 (d) were compared with linearized CRISPR in terms of the editing efficiency using donors of different homologous arm lengths. e Comparison of gene editing efficiencies promoted by overexpression of different RAD family gene combinations. Values are means of the biological replicates, and the error bars indicate the standard deviations of all (n = 4) biological replicates. P values were obtained using the two-tailed Student’s t-test: ****P < 0.0001

Upon overexpression of Rad54, Rad55 and Rad57, we observed modest increases in the frequency of correctly edited clones under both standard donor conditions with 100-bp homology arms and donor designs with shortened homology arms (Fig. 3d and Fig. S2). By contrast, under donor conditions with long homology arms, Rad54 overexpression markedly suppressed homologous recombination (Fig. 3d). Previous studies have demonstrated that bridge-induced chromosome translocations in yeast rely heavily upon a Rad54/Rdh54-dependent, Pol32-independent pathway [34]. In the context of multiplex genome editing, where multiple simultaneous double-strand breaks (DSBs) are generated, the non-physiological overexpression of Rad54 may drive hyperactive chromatin remodeling and non-specific strand invasion. This could significantly increase the propensity for ectopic recombination and fatal chromosomal translocations. Furthermore, homologous recombination is a highly coordinated process requiring a strict stoichiometric balance among repair proteins. An excess of Rad54 might disrupt its delicate equilibrium with other essential factors, such as Rad51 and Rad52, thereby leading to the formation of non-productive repair complexes. Consequently, the diversion of cellular repair resources towards non-specific chromosomal interactions, coupled with potential toxicity, likely accounts for the diminished efficiency observed with Rad54 overexpression. Mechanistically, Rad51 functions as the central recombinase of the homologous recombination pathway, acting as a RecA homolog to catalase homology search, DNA pairing and strand invasion. Rad52 promotes the loading of Rad51 onto single-stranded DNA, whereas Rad55 and Rad57 form a Rad51 paralogue complex that stabilizes Rad51-ssDNA filaments and counteracts anti-recombination factors such as Srs2, thereby increasing the likelihood of productive strand invasion and subsequent gene conversion [25]. Guided by these roles, we next evaluated combinatorial overexpression of Rad-family genes with standard donor conditions (100 bp homology arms). Co-overexpression of Rad51 and Rad52 produced the largest gain in editing efficiency, increasing the correct editing rate from 28% to 52%, corresponding to an approximate twofold improvement. The next most effective combination was co-overexpression of Rad51, Rad55 and Rad57, which increased editing efficiency from 28% to 41%, corresponding to an approximately 1.5-fold improvement (Fig. 3e). Notably, simultaneous overexpression of all four genes resulted in lower editing efficiency than the Rad51-Rad52 combination, and qRT-PCR analysis showed that, under the simultaneous overexpression of four RAD genes, the transcript levels of Rad51, Rad52, Rad55 and Rad57 were largely comparable to those observed when only two Rad genes were co-overexpressed. Therefore, the reduced genome-editing efficiency in the four-gene condition is unlikely to be attributable to decreased transcription of these Rad genes (Fig. S3). Rad51-Rad52 co-expression may already drive homologous recombination towards a near-saturated regime, or that excessive Rad55/57 levels promote non-productive complex occupancy that imposes additional cellular burden and perturbs assembly dynamics, thereby diminishing repair outcomes.

Establishment of multiplex genome editing strategy in S. cerevisiae

Given that SSAP-mediated homologous recombination recombineering coupled with CRISPR-Cas9 has been shown in bacteria and eukaryotic cells to promote homologous recombination [19, 29], we reasoned that expressing bacterial single-strand annealing protein (SSAP) in yeast could substantially improve multiplex editing efficiency. Using the previously established linearized CRISPR-Rad platform in S. cerevisiae S288C as a starting point, we co-expressed the E. coli and phage-derived SSAPs Redβ or RecT with the endogenous HR factors Rad51 and Rad52, and then quantified their impact on homology-directed editing efficiency and extension to multiple loci (Fig. 4a). It is important to note that the donor DNA used for integration in these assays was the exact same KanMX cassette that was utilized and described in the preceding experiments (Figs. 1, 2 and 3). In the single-locus editing assay, the incorporation of bacterial SSAPs in the Rad51-Rad52 background led to a clear increase in precise editing efficiency. The Redβ condition produced the most pronounced improvement (with the correct editing rate rising from 52% to 90%), whereas the RecT combination also outperformed Rad51-Rad52 (Fig. 4b). Under donor-limiting conditions using short homology arms of 30 bp, the Redβ combination maintained a comparatively higher rate of correctly edited clones, indicating superior robustness when donor homology is constrained (Fig. 4c). Collectively, these results support a model in which Rad51 and Rad52 raise the baseline capacity of endogenous homologous recombination, while Redβ further enhances effective donor utilization by promoting single-strand annealing and donor capture, thereby amplifying the advantage when donor availability is reduced or homology arms are shortened. Under the conditions tested here, the Rad51-Rad52-Redβ module constituted the most effective combination for improving both efficiency and stability of linearized CRISPR-mediated precise editing (Fig. 4d).

Fig. 4.

The effect of overexpression of SSAPs combined with Rad proteins and linearized CRISPR system on multiplex genome editing in S. cerevisiae S288C. a Schematic representation of the SSAP-Rad-Linearized CRISPR system (SRLC), where bacterial SSAPs (Redβ or RecT) are co-expressed with Rad51 and Rad52 to promote homologous recombination from short linear donors. The donor DNA integrated in these experiments is identical to the KanMX cassette detailed in Figs. 1, 2 and 3. b Comparison of gene editing efficiencies enhanced by different combinations of SSAPs with Rad51-Rad52. c Comparison of the enhancement of Rad51-Rad52-mediated genome editing by Redβ under donor templates with different homology arm lengths. d Redβ preferentially binds single-stranded DNA and facilitates strand annealing, with enhanced affinity for donor templates carrying short homology arms. This activity promotes the recruitment and engagement of RAD family proteins, thereby improving genome editing efficiency. e Efficiency of four-locus genome editing using the SRLC system. Values are means of the biological replicates, and the error bars indicate the standard deviations of all (n = 4) biological replicates. P values were obtained using the two-tailed Student’s t-test: ****P < 0.0001

On the basis of these comparisons, we selected Rad51-Rad52-Redβ as an HR-enhancement module and establish the SSAP-Rad-Linearized CRISPR-mediated multiplex editing technology (SRLC). In this framework, multiple linear gRNA expression fragments were pooled with their corresponding donor DNAs and co-transformed in a single step, enabling concurrent induction of double-strand breaks across multiple target sites and parallel repair through homology-directed recombination. In the four-locus editing experiment, SRLC yielded substantially more mutants than the linearized CRISPR system (with the correct editing rate rising from 3% to 20%) (Fig. 4e). The SRLC strategy enables more efficient generation of pathway-scale combinatorial mutants in S. cerevisiae and supports iterative, multi-round genome engineering. Overall, incorporation of the Rad51-Rad52-Redβ module improves linearized CRISPR-mediated genome editing and increases the efficiency of multiplex editing, with the most pronounced gains observed when short homology-arm donors are used to streamline donor preparation. To further dissect the editing accuracy and multiplexing distribution of the SRLC platform, we analyzed the genotype of randomly selected clones from the 4-locus co-transformation experiment. Among the analyzed population, while 28% of the clones contained all four intended edits (complete editing), a significant proportion exhibited partial editing: approximately 35% contained three edits, 22% contained two edits, and 15% contained a single edit (Fig. S4). This distribution demonstrates that SRLC not only facilitates high-efficiency simultaneous multi-locus editing but also generates a diverse combinatorial mutant library in a single transformation step. This feature is particularly advantageous for pathway optimization, where fully edited genotypes are not always the optimal phenotype due to metabolic imbalances. Furthermore, within the partially edited populations, we observed locus-to-locus variability in editing frequencies. This variability is consistent with established CRISPR mechanics and is likely driven by inherent differences in specific gRNA cleavage efficiencies, local chromatin accessibility at the targeted genomic loci, and variations in GC content affecting homology arm annealing. Despite this locus-dependent bias, the SRLC platform consistently yielded a highly diverse combinatorial pool.

Because multiplex editing constitutes a primary advantage of the SRLC platform, we sought to rigorously evaluate the editing outcomes across the cellular population using a comprehensive sequencing-based approach. Following the multiplex transformation targeting four distinct genomic loci simultaneously, we subjected the resulting population to targeted deep sequencing (NGS). This analysis provided a quantitative profile of the population’s genetic landscape. Our data revealed that the fully edited genotype, in which all targeted loci were successfully modified comprising 26.5% of the total population. Partially edited populations containing only one, two and three targeted modifications were observed at frequencies of 14.3%, 23.7% and 34.2%, respectively, while the unedited wild-type fraction remained minimal (1.3%) (Table S1). Furthermore, sequencing confirmed high junction fidelity at the integration sites with negligible insertion/deletion (indel) artifacts. These population-level sequencing results comprehensively validate the robustness and high efficiency of the SRLC system, demonstrating its capacity to drive complex, multi-locus genomic modifications simultaneously while significantly reducing the downstream screening burden.

To further explore the utility of the SRLC platform for advanced strain engineering, such as the integration of multi-gene metabolic pathways, we investigated the impact of donor DNA length on editing efficiency. We designed and tested linear donor DNA fragments of varying lengths (1 kb, 2 kb, 5 kb, 8 kb, and 10 kb) for targeted genomic integration. As shown in Figure S5, the SRLC system successfully facilitated the precise integration of all tested fragments. As anticipated for homologous recombination-based approaches, an inverse correlation was observed between donor length and editing efficiency. While robust editing efficiencies were maintained for the 2 kb and 5 kb insertions, a significant decline was observed when the donor length exceeded 5 kb (i.e., for the 8 kb and 10 kb fragments). We attribute this sharp decrease primarily to the inherent physical constraints of yeast transformation. It is well documented that the cellular uptake efficiency of large intact linear DNA fragments (> 5 kb) is exceedingly low. Consequently, the drastic reduction in the intracellular molarity of the donor DNA limits the available substrates for the recombination machinery. Therefore, the efficiency bottleneck for large fragment integration is largely dictated by the low entry efficiency of large linear DNA into the yeast cell, rather than an intrinsic limitation of the SRLC system’s recombination capacity.

Regarding throughput and scalability, SRLC offers distinct advantages over established multiplexing platforms. Plasmid-based systems, such as tRNA-gRNA arrays and Csy4 platforms, are highly efficient but theoretically and practically constrained by the maximum size of the gRNA array that can be stably cloned and maintained, typically limiting routine multiplexing to 4–6 targets per cycle [8, 10, 11]. Furthermore, scaling these methods to test combinatorial subsets requires the laborious construction of individual plasmids for every permutation. High-throughput platforms like yeast CREATE can achieve genome-wide single-nucleotide mapping but rely on large-scale oligonucleotide synthesis and specialized plasmid library construction [2]. In contrast, SRLC decouples the gRNA and donor from the plasmid backbone. By utilizing a mix-and-match pool of linear PCR fragments, SRLC can effortlessly scale to target dozens of genes simultaneously (as demonstrated by the pooling of 30 targets in our malonyl-CoA pathway engineering), yielding complex combinatorial libraries in a single day of preparation. The successful isolation of strain C73, which contained 9 simultaneous edits, underscores the platform’s exceptional scalability for high-throughput phenotypic screening.

The modification and application of the malonyl-CoA metabolic pathway in S. cerevisiae

As one of the most widely used chassis microorganisms in industrial biotechnology, the supply capacity of key metabolic precursors in S. cerevisiae S288C is a critical factor limiting the synthesis efficiency of target products. Malonyl-CoA, an essential central precursor for the biosynthesis of fatty acids, polyketides, and various biofuels, holds a pivotal position in metabolic engineering. Studies have shown that enhancing the intracellular availability of malonyl-CoA in S. cerevisiae can significantly promote the production of these important metabolites, making this molecule a key target for metabolic pathway optimization [35, 36]. Based on a systematic analysis of the malonyl-CoA metabolic network in S. cerevisiae, this study selected 30 key genes from eight major metabolic pathways as engineering targets and designed up or down-regulation strategies according to their functions (Table S2). Specifically, for genes requiring up-regulation, their native promoters were replaced with the strong TRP1 promoter to enhance transcription; for genes requiring down-regulation, an LVA protein degradation tag was added to the C-terminus of the coding sequence to accelerate target protein degradation. Preliminary single gene modification experiments showed that perturbation of each target gene could increase intracellular malonyl-CoA accumulation to varying degrees (Figure S6), confirming the validity of the selected targets. To obtain strains with stronger malonyl-CoA accumulation capacity, we employed simultaneous and combinatorial editing technology (SRLC) to modify these 30 target genes in parallel, thereby constructing a mutant library of S. cerevisiae with engineered malonyl-CoA metabolic networks. For efficient screening of high malonyl-CoA producing phenotypes, we introduced a biosensor based on a malonyl-CoA responsive promoter. This sensor utilizes the FapR-fapO regulatory system [32]: when intracellular malonyl-CoA levels are low, the transcriptional repressor FapR binds to the fapO operator sequence, inhibiting reporter gene expression; when malonyl-CoA accumulates to a certain concentration, it binds to FapR and induces a conformational change, causing FapR to dissociate from the DNA, thereby derepressing and activating reporter gene expression and ultimately leading to enhanced fluorescence signal. This sensor enables high-throughput monitoring of intracellular malonyl-CoA levels in the mutant library, providing a reliable tool for rapid identification of high producing mutants.

Through one round of SRLC editing combined with flow cytometry sorting, we obtained a mutant Saccharomyces cerevisiae strain C73 (Fig. 5a, b). The fluorescence intensity of C73 was significantly increased by 18-fold compared to the wild-type strain of S. cerevisiae, indicating a systematic reprogramming of its malonyl-CoA-related metabolic network (Fig. 5c). Genotypic analysis further revealed that nine target genes were successfully edited in C73. Among these edited genes, those showing upregulated expression included: PGI (catalyzing the phosphorylation of glucose to glucose-6-phosphate, the first rate-limiting step in glucose utilization), PYK (a key enzyme in the final step of glycolysis, responsible for transferring high-energy phosphate to ATP), ACS (converting acetate to acetyl-CoA, an intermediate involved in multiple biochemical pathways), ACC (the rate-limiting enzyme in fatty acid synthesis, catalyzing the conversion of acetyl-CoA to malonyl-CoA), TALA (participating in the generation of ribose-5-phosphate for nucleotide synthesis and NADPH for reductive biosynthesis and oxidative stress defense), FBP (catalyzing the hydrolysis of fructose-1,6-bisphosphate to fructose-6-phosphate), and ZWF (the first rate-limiting enzyme in the oxidative phase of the pentose phosphate pathway, responsible for oxidizing G6P while generating NADPH). The downregulated genes were MS (catalyzing the condensation of acetyl-CoA and glyoxylate to malate) and ICL (a key enzyme in the glyoxylate cycle, responsible for the cleavage of isocitrate) (Figs. S7, S8). The results above demonstrate that SRLC can achieve multi-target synergistic genome modifications in S. cerevisiae and stably transmit these genetic changes into selectable phenotypes. Further metabolite quantification supports these conclusions: the intracellular malonyl-CoA concentration in strain C73 reached 0.1339 nmol mg^− 1 DCW, representing a 9.6-fold increase over wild-type yeast (0.0140 nmol mg^− 1 DCW) (Fig. 5d). These findings indicate that combining SRLC with metabolic sensor-based high-throughput screening enables efficient and parallel optimization of complex metabolic pathways in S. cerevisiae, providing a powerful tool for precise and targeted combinatorial metabolic engineering.

Fig. 5.

Application of SRLC in malonyl-CoA metabolic pathway engineering. a Schematic representation of SRLC-driven metabolic engineering and malonyl-CoA biosensor for high-yield strain selection. b Flow cytometry analysis of fluorescence intensity (FI) in S. cerevisiae C73, the mutant strain with enhanced malonyl-CoA biosynthesis. The FI is used as a proxy for intracellular malonyl-CoA levels. c Comparison of FI between wild-type S. cerevisiae S288C and mutant strains (S. cerevisiae C73). d Quantification of malonyl-CoA concentration in wild-type and C73 strains. e 3-hydroxypropionate (3-HP) production in wild-type and C73 strains. Values are means of the biological replicates, and the error bars indicate the standard deviations of all (n = 4) biological replicates. P values were obtained using the two-tailed Student’s t-test: ****P < 0.0001

A critical concern associated with multiplex genome engineering, particularly when utilizing systems that generate simultaneous multiple double-strand breaks (DSBs) alongside the overexpression of recombination factors, is the potential induction of genome instability and off-target structural variations. To comprehensively evaluate the genomic integrity of our system, we performed Whole-Genome Sequencing (WGS) on the final engineered strain, C73, which harbors precise simultaneous edits across 9 distinct genomic loci (Fig. S9). The WGS analysis corroborated the exceptional specificity and fidelity of our editing platform. Sequence alignment against the reference genome confirmed that all 9 targeted insertions/modifications were flawlessly executed. More importantly, a genome-wide survey for unintended alterations revealed a highly stable genomic landscape. We found absolutely no evidence of large-scale chromosomal rearrangements, translocations, or severe structural variations typically associated with multiple simultaneous DSBs. The only detectable unintended collateral mutations across the entire genome were restricted to four minor, localized deletions (spanning only tens of base pairs). Crucially, all four of these small deletions occurred entirely within non-coding regions, ensuring no disruption to essential ORFs or cellular functions. Taken together, these WGS results provide compelling evidence that our multiplex editing strategy can achieve extensive, high-order genomic rewiring while robustly maintaining overall genome stability.

Malonyl-CoA is a key precursor for synthesizing various high-value metabolites, including 3-hydroxypropionic acid (3-HP), an important bio-based platform chemical for producing acrylic acid and its derivatives [37]. Previous studies attempted to increase intracellular malonyl-CoA levels by exogenous addition of cerulenin, a potent fatty acid synthesis inhibitor, but this method often caused significant growth inhibition and was thus unsuitable for cell growth-dependent production systems [38]. To evaluate the ability of the engineered strain to channel malonyl-CoA toward target product synthesis, we introduced the 3-HP biosynthetic plasmid pJFE3-mcr into both the wild-type and mutant C73 strains, and assessed 3-HP yield as a functional output indicator of malonyl-CoA flux. The results showed that C73 achieved a 3-HP titer of 183 mg/L, which was 4.5 times higher than that of the wild-type strain (Fig. 5e). This confirms that SRLC-mediated multiplex genome editing effectively promotes malonyl-CoA accumulation, thereby enhancing the biosynthesis capacity of related high-value metabolites.

Discussion

This study establishes the SSAP-Rad-linearized CRISPR (SRLC) technology, a rapid, multi-locus simultaneous genome editing platform for S. cerevisiae S288C that eliminates the need for complex multi-gRNA expression vector construction. The technique achieves simultaneous multi-site precise editing in the yeast genome by transient delivery of PCR derived, stability protected gRNA expression fragments to enable multiplexed Cas9 cleavage, coupled with programmed enhancement of homologous directed repair (HDR) and SSAP assisted donor utilization. The platform is defined by three core features: (i) linear gRNA expression cassettes that maintain Cas9 guided cutting performance while bypassing plasmid assembly; (ii) targeted up regulation of key homologous recombination factors to boost the cell’s ability to perform precise repair of Cas9 induced breaks using short donors; and (iii) heterologous SSAPs that further improve donor capture efficiency and robustness when sequence homology is limited.

While established multiplex CRISPR strategies in S. cerevisiae, such as Csy4-mediated RNA processing [10] and tRNA-gRNA arrays [11], have successfully achieved high-order multiplexing, they remain fundamentally dependent on plasmid construction. The assembly of complex multi-gRNA vectors requires labor-intensive cloning steps, including DNA assembly, bacterial propagation, and sequence validation, which limits the speed of iterative Design-Build-Test cycles. Furthermore, plasmid-based systems are often constrained by copy-number variation and the necessity to recycle selectable markers across continuous engineering rounds [8]. Compared to these established approaches, the SRLC platform provides a highly practical, cloning-free alternative, with less time and effort required (Table 1). By directly employing PCR-amplified, end-protected linear gRNAs, SRLC uncouples guide delivery from plasmid backbones, enabling the on-demand generation and combinatorial mixing of targets. It is important to note that the delivery of exogenous linear DNA poses potential risks of nucleolytic degradation and cellular toxicity if the concentration is too high; therefore, balancing the nucleic-acid dosage is critical for optimal viability. Crucially, while traditional CRISPR systems often exhibit low efficiency when using short-homology linear donors due to repair pathway competition, SRLC overcomes this bottleneck. The synergistic co-expression of endogenous Rad51/Rad52 and the bacterial SSAP (Redβ) stabilizes linear repair intermediates, significantly boosting the multiplex precise editing efficiency even with severely shortened homology arms.

Table 1.

Comparison of SRLC with established multiplex CRISPR platforms in S. cerevisiae

Feature	SRLC (this study)	tRNA-gRNA arrays	Csy4-mediated systems	Iterative Conventional CRISPR
gRNA delivery format	Linearized PCR fragments	Plasmid	Plasmid	Plasmid (Single/Dual gRNA)
Cloning requirement	None (PCR-based)	High (Golden Gate/Gibson)	High (Golden Gate/Gibson)	Moderate to High
Preparation time (Pre-transformation)	< 1 day	3–5 days	3–5 days	2–3 days per cycle
Multiplexing capacity	High (Tested up to 9 loci)	High (Often 4–6 loci)	High (Often 4–6 loci)	Low (1–2 loci per cycle)
Donor DNA format	Linear (Short homology arms enabled by SSAP/Rad)	Linear/Plasmid (Usually requires longer homology)	Linear/Plasmid (Usually requires longer homology)	Linear/Plasmid
Flexibility for combinatorial editing	High (Mix-and-match PCR products)	Low (Requires re-cloning entire array)	Low (Requires re-cloning entire array)	Low
Key advantage	Eliminates cloning bottleneck; scalable	High efficiency per transformation	High efficiency per transformation; precise processing	Established, highly reliable for single targets
Reference	This work	[11]	[10]	[8]

The performance of linearized CRISPR in yeast is often assumed to be limited chiefly by linear DNA instability and by competition between repair pathways [14, 17]. Studies indicate that these limitations can be partially overcome through systematic coordination of donor design and repair state modulation, such as the SRLC strategy. In our experiments, shortening donor homology arms markedly reduced correct editing rates in the baseline linearized CRISPR context, indicating that donor capture and/or efficient strand invasion became rate-limiting. Under conditions of restricted donor design, elevating Rad51 protein expression produced the most significant and consistent positive effect, which aligns with Rad51 core functions in forming nucleoprotein filaments, conducting homology search, and mediating strand invasion [24, 25]. In contrast, overexpression of Rad52 also improved editing efficiency, but the enhancement was relatively limited. Especially with shortened homology arms, Rad52 only partially restored editing efficiency, further suggesting that the main bottleneck under these conditions lies in the effectiveness or filament stability of Rad51. Notably, when longer homology arms (> 500 bp) were used, overexpression of either Rad51 or Rad52 showed no clear advantage, indicating that sufficient donor homology length can saturate the cellular homologous recombination capacity. Therefore, Rad51 hold the greatest potential for improving editing efficiency losses caused by excessively short homology arms. The non‑additivity observed when multiple Rad genes were simultaneously overexpressed further suggests that HR is governed by stoichiometric and kinetic balances, too much of certain accessory factors may create non‑productive complexes or cellular burden, an effect that will be important to consider when transferring SRLC to different strain backgrounds or edit classes. A second mechanistic contribution is the transplantation of bacterial SSAP activity into S. cerevisiae S288C to improve HDR from short linear donors. SSAPs such as Redβ and RecT are core catalysts of bacterial recombineering, enabling efficient incorporation of oligonucleotides and linear DNA fragments, and they have been combined with CRISPR selection to accelerate multiplex genome rewriting [18–20]. In SRLC, Redβ provided the largest enhancement beyond Rad51-Rad52, approximately doubling correct editing in single‑locus assays and improving robustness when homology arms were reduced to 30 bp. This behavior is consistent with structural and single‑molecule studies showing that SSAPs stabilize ssDNA and secure homology during annealing [27, 28]. More broadly, the convergence between SRLC and recent mammalian SSAP‑enabled knock‑in strategies, such as recombinase‑enhanced DNA integration (RED‑CRISPR) [29] and dCas9-SSAP cleavage‑free knock‑in [30] supports an emerging principle: transiently increasing the abundance and lifetime of annealing‑competent intermediates can make linear donors efficient across domains, even when donor design would otherwise be suboptimal.

While the generation of a library targeting 30 nodes successfully demonstrated the high-throughput capability of our platform, the identification of a top-performing strain harboring a specific combination of only nine edits warrants deeper metabolic consideration. Although a systematic experimental reconstruction and quantitative flux analysis of all combinations are beyond the scope of this technology-focused study, a mechanistic analysis of these nine specific targets (glucokinase, pyruvate kinase, transketolase, glucose-6-phosphate dehydrogenase, fructose-1,6-bisphosphatase, acetyl-CoA synthase, acetyl-CoA carboxylase, malate synthase, and isocitrate lyase) reveals a highly synergistic metabolic rewiring. Specifically, the modifications to acetyl-CoA synthase and acetyl-CoA carboxylase directly augment the intracellular pools of essential precursors (acetyl-CoA and malonyl-CoA). Concurrently, the involvement of glucose-6-phosphate dehydrogenase and transketolase suggests a redirection of carbon flux towards the pentose phosphate pathway, crucial for maintaining the NADPH cofactor balance required for intensive biosynthesis. Furthermore, the selection of malate synthase and isocitrate lyase indicates an activation of the glyoxylate shunt, a classic metabolic bypass that conserves carbon by minimizing CO2 evolution during the TCA cycle [31]. The emergence of this specific 9-edit combination from a 30-target library elegantly highlights the inherent trade-offs between cellular growth and product synthesis. Metabolic networks are highly constrained; forcefully perturbing all 30 nodes simultaneously would likely impose a severe metabolic burden and disrupt vital homeostatic balances, leading to growth defects. The combinatorial exploration enabled by our platform effectively navigated this complex landscape, converging on an optimized sweet spot, a subset of modifications that synergistically maximizes precursor and cofactor fluxes toward the target product without fatally compromising the core metabolic fluxes required for biomass accumulation. This multi‑node solution complements previous sensor‑coupled malonyl‑CoA engineering efforts which identified targeted interventions through fewer edits per cycle or alternative discovery modalities [32, 33], and it provides a mechanistic bridge to chassis‑level strategies that enhance production phenotypes by rewiring global metabolism and stress responses [35, 36]. Importantly, SRLC provides a route to scale such hypothesis testing by rapidly constructing variant combinations, quantitatively mapping genotype and product relationships.

While we introduced targets for 30 distinct loci in a single transformation, the highest number of simultaneous edits recovered was 9 (Strain C73). This ceiling is likely dictated by the cellular toxicity associated with simultaneous Cas9-induced double-strand breaks (DSBs) and the saturation of the host’s DNA repair machinery. However, in the context of metabolic engineering, achieving up to 9 simultaneous edits in a single round is highly advantageous. The objective of highly multiplexed targeting (e.g., 30 targets) is rarely to achieve complete editing at all sites, which could compromise cell viability, but rather to generate a rich combinatorial landscape from which optimal, multi-locus synergistic variants can be rapidly screened.

SRLC is particularly suited to applications in which the dominant constraint is the rapid construction of high order genotypes, rather than maximizing the efficiency of a single edit. This includes iterative pathway and chassis optimization, combinatorial rewiring of precursor supply and competing sinks, and multiplex perturbation for genotype to phenotype mapping in Saccharomyces cerevisiae. In these settings, SRLC enables scalable multiplex editing while minimizing cloning workload, and it integrates naturally with high throughput selection or screening strategies, including genetically encoded biosensors, to accelerate the identification of improved variants. More broadly, the platform has strong universality because it builds on established CRISPR enabled homologous recombination frameworks in yeast and is conceptually compatible with widely used multiplex guide expression strategies such as guide RNA arrays and tRNA based processing systems. As a result, SRLC should generalize across loci and common edit types in yeast, including knock ins, knock outs, and regulatory element replacement, with practical performance primarily determined by standard design parameters such as guide choice, donor design, and cellular repair capacity [6, 11, 39, 40].

In summary, SRLC offers a modular framework that matches mutant construction speed to the throughput of downstream screening and selection. By replacing plasmid-based guide assembly with PCR-derived linear guides and engineering the repair landscape through Rad factors and SSAPs, the method makes short homology linear donor editing substantially more tractable in S. cerevisiae S288C. The combination with metabolite biosensors illustrates how SRLC can accelerate discovery of polygenic, high‑performance genotypes for industrial biotechnology. Future work that further stabilizes linear components, improves multiplex viability, and quantitatively benchmarks off‑target and genome stability outcomes should expand SRLC applicability to broader classes of edits and to diverse yeast chassis strains.

Methods

Strains and culture conditions

All strains used in this study are listed in Table S3, and S. cerevisiae S288C are derived from CEN.PK2-1C [32]. E. coli was cultured in Luria-Bertani (LB) broth or on LB agar plates (1.2% agar) with ampicillin (100 µg mL^− 1), kanamycin (15 µg mL^− 1), chloramphenicol (15 µg mL^− 1) or gentamicin (5 µg mL^− 1) as required. S. cerevisiae and mutants were cultured in YPD broth or on YPD agar plates (1.2% agar). YPD medium containing 10 g/L yeast extract and 20 g/L peptone, supplemented with 20 g/L glucose, was used to culture yeast cells without any plasmid. The growth medium contained 1.7 g L^− 1 yeast nitrogen base, 5 g L^− 1 ammonium sulfate, synthetic complete (SC) dropout components lacking uracil, leucine and/or histidine, and 20 g L^− 1 glucose. Where indicated, selective agents were supplemented at final concentrations of 800 mg L^− 1 G418, 60 mg L^− 1_L-canavanine, or 5-fluoroorotic acid (5-FOA). All Saccharomyces cerevisiae strains were cultured at 30 °C with agitation at 200 rpm. Starter cultures were grown in 5 mL medium in 40-mL tubes and subsequently scaled up to 40 mL in 100-mL shake flasks for experimental cultivation.

Plasmids, primers and genetic manipulations

The plasmids and primers used in this study are listed in Table S4 and Table S5, respectively. E. coli GBdir was used for plasmid cloning [41]. LLHR was used for constructing recombineering vectors; template plasmids were adopted from the reported Red/ET system [19]; the gRNAs were predicted by the software: https://chopchop.cbu.uib.no/. LLHR assembly was adopted for constructing other plasmids. Gene sequences ScRad51, ScRad52, ScRad55 and ScRad57 were cloned from S. cerevisiae CEN.PK2-1 C, redβ and recT were adopted from E. coli GBdir and E. coli GBred [41], and constructed by overlapping PCR.

Genomic integration of S. cerevisiae S288C was performed by electrotransformation [11]. For each transformation, either 0.5 µg of the recombinase plasmid or donor DNA fragments at a ratio of 1 µg per kilobase were added. During genome editing, varying amounts of gRNA fragments were co-transformed together with the donor DNA. Plates were incubated on selective medium for 2–3 days, after which individual colonies were picked for PCR verification. All oligonucleotide primers used for plasmid construction, donor preparation and clone verification are listed in Table S5.

Calculation of gene insertion efficiency

To assess insertion efficiency, 20 colonies were randomly selected from each biological replicate (with at least three independent biological replicates) and streaked onto antibiotic-containing plates. SC agar plates supplemented with 800 mg l^− 1 G418 were used to select for kanMX insertion. All strains were verified by PCR amplification of the insertion locus using one primer pair flanking the upstream and downstream regions of the target site to confirm kanMX integration. For all experiments (n ≥ 3 biological replicates), mean values and s.d. were calculated. The statistical analyses were performed using unpaired t Test: p values were calculated with two sides/tails and hypothesis of two sample assuming equal variances.

Malonyl-CoA quantification

The extraction and quantification of malonyl-CoA were implemented as previous [42]. Cultures were grown to OD₆₀₀=0.4, and 1.8 mL samples were chilled, centrifuged (4,000 g, 15 min, 4 °C), washed with PBS, and extracted on ice with 120 µL acetonitrile: methanol: water (45:45:10, v/v/v) containing 0.1 M formic acid. After 15 min extraction, samples were neutralized with ammonium hydroxide, centrifuged (15,000 g, 3 min, 4 °C), and supernatants were analyzed by LC-MS. Analyses were performed on an LC-DAD and ESI-MS system (Bruker Impact HD microTOF Q III, positive mode) using a Thermo™ Acclaim™ RSLC 120 C₁₈ column. A tributylamine-based aqueous phase and isopropanol were applied at 0.3 mL min^{− 1} with diode array and MS detection.

Flow cytometry

Strains were recovered from glycerol stocks by streaking onto solid growth medium and incubated at the strain-specific optimal temperature until single colonies appeared. Colonies were collected and grown in liquid medium to an OD₆₀₀ of approximately 0.8. Cells were fixed by resuspension in 1.5 mL PBS, followed by addition of 1.5 mL fixative, vortex mixing, and centrifugation at 12,500 g for 5 min at 4 °C. The supernatant was removed and the fixation step was repeated prior to analysis. Flow cytometry was performed following instrument startup and quality control according to the manufacturer instructions. Fixed cell suspensions were loaded into tubes compatible with the cytometer sample port. Forward scatter (FSC) and side scatter (SSC) parameters were optimized to ensure full visualization of the bacterial population. Cell populations were gated, and doublets were excluded based on FSC-area versus FSC-height. Fluorescence acquisition was performed by generating histograms of GFP signals from mutant strains and corresponding wild-type controls. At least 10,000 events were recorded per sample. For cell sorting, appropriate gates and collection modes were defined, after which sorted populations were collected and the system was cleaned and flushed with PBS according to standard procedures.

Metabolite analysis

Strains were initially grown in 5 mL YPD medium and subsequently transferred to shake flasks containing 40 mL medium at a starting OD₆₀₀ of 0.2 for cultivation over 48 h. Production of 3-hydroxypropionic acid (3-HP) was quantified using high-performance liquid chromatography (HPLC; Shimadzu, Japan) fitted with an Aminex HPX-87 H column (Bio-Rad). Separation was performed with 2.5 mM H₂SO₄ as the mobile phase at a flow rate of 0.6 mL min^− 1, with the column temperature maintained at 65 °C.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Wentao Zheng: Conceptualization, Formal analysis, Investigation, Methodology, Writing - original draft. Miao Wang: Conceptualization, Formal analysis, Investigation. Qiang Tu: Investigation. Xiaoying Bian: Investigation. Youming Zhang: Formal analysis. Xue Wang: Conceptualization, Formal analysis, Project administration, Supervision, Writing - review & editing, Funding acquisition.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.