Introduction

The site-specific recombination system is a commonly used tool for genome manipulation, characterized by its specificity, high efficiency, and low off-target rate1,2,3. Recently, these systems have been applied in many areas, including crop improvement4, cell lineage tracing5,6 and metabolic engineering7. Recombinases catalyze site-specific recombination by attacking the DNA backbone to mediate double-strand exchanges and the process does not involve DNA synthesis, degradation, or cofactors such as ATP8. Site-specific recombination systems can generate deletion, inversion, or other structural variations between two specific sites by configuring the orientation and location of sites9. Multiplexed site-specific recombination systems can induce a large number of rearrangement events in a short time, thereby effectively expanding the diversity of structural variations10. However, existing eukaryotic multisite recombinases inevitably lead to undesired deletions of DNA fragments among sites when multiple sites coexist11 (Fig. 1a). For instance, in the Sc2.0 project, researchers have developed a Synthetic Chromosome Rearrangement and Modification by loxP-mediated Evolution (SCRaMbLE) system, which can rapidly drive genome structural rearrangements by interactions among thousands of loxPsym sites inserted into the synthetic chromosomes of Saccharomyces cerevisiae12,13,14. Meanwhile, the insertion of loxPsym sites into the genome causes extensive deletions of essential genes during recombination, leading to strain lethality15,16. Applying the Cre/loxP system for cell barcoding involves the insertion of multiple loxP sites among the barcodes. Although the diversity of barcode can be generated through random recombination among these sites, it will gradually decrease over time due to deletions between same-oriented sites5. DNA inversion system, a site-specific recombination system in bacteria, prefers inversions at reverse-oriented sites rather than deletions at same-oriented sites when multiple sites coexist17,18,19. However, our previous research indicates that the prokaryotic inversion system does not work in eukaryotes, and the multiplexed site-specific inversion system has not been found in eukaryotes20.

Fig. 1: Concept and screening of a eukaryotic multiplexed site-specific inversion system.
figure 1

a Comparison of conventional tyrosine recombinase (Cre enzyme) and DNA inversion enzyme (Rci enzyme). Cre enzyme typically induces deletion of DNA fragments between directly repeated sites when multiple sites are present. The Rci enzyme facilitates gene inversion between reverse-oriented sites. The black and yellow triangles, respectively, represent the loxP site and the sfx site and the red triangles indicate the sites that mediate recombination. b High-throughput screening system for multiplexed site-specific inversion enzyme. The promoters and open reading frames (ORFs) are arranged alternately but in opposing orientations, with the combination of promoter and corresponding terminator situated between a pair of inverted sfxa101 sites (yellow). Open arrows indicate silencing of selection markers, and solid arrows indicate expression of selection markers. The red triangles represent the sites that mediate recombination. c Flow diagram of two rounds of progressively stringent directed evolution in Saccharomyces cerevisiae. Fragment of the high-throughput screening system (7170 bp in length) was inserted into the chromosome X at the position of 639, 678 bp. The dashed line shows an example of the inversion situation of strains that can grow on the selective medium.

Full size image

In this study, we establish a multiplexed site-specific inversion system Rci51-5/multi-sfxa101 in eukaryotes that preferentially facilitates inversion at reverse-oriented sites rather than deletion at same-oriented sites. The mutant Rci51-5, with a ~ 1000-fold increase in total inversion efficiency against the wild-type, is obtained through a high-throughput screening strategy following two rounds of progressively stringent directed evolution. Five nonsynonymous mutations of Rci51-5 have been found to synergistically contribute to the enhanced efficiency of inversion. Our system of Rci51-5/multi-sfxa101 shows a significantly lower deletion rate than the Cre/multi-loxP system. We demonstrate the application of the Rci51-5/multi-sfxa101 system for optimizing the metabolic pathways of multiple genes by promoter substitution between an array of varied strength promoters and multiple genes of synthetic pathways. Using the biosynthetic pathway of β-carotene as an example, upon the Rci51-5 expression, the original promoters of exogenous genes are replaced by promoters in the array, leading to changes in transcriptional levels and improvement of β-carotene production. In conclusion, this eukaryotic multiplexed site-specific inversion system enriches the types of eukaryotic site-specific recombination systems and offers a special tool for genome engineering.

Results

Developing a high-throughput screening system for multiplexed site-specific inversion enzyme

The site-specific DNA inversion system is biased for DNA inversion and has previously been exclusively identified in prokaryotes17,21,22,23. In earlier studies, utilizing the wild-type Rci enzyme from Salmonella typhimurium and directed evolution, we have developed a DNA inversion system (Rci8/sfxa101) in eukaryotes, working at a pair of recombination sites20. To test whether this mutant (Rci8) can accomplish multisite inversion, we developed a screening system of multisite inversion in Saccharomyces cerevisiae (Fig. 1b). The system comprises three promoters — pGPM1, pTDH3, pTEF1 and corresponding terminators, along with three open reading frames (ORFs) that can be used for auxotroph or resistance screening — URA3, hphNT1 and kanMX. The promoters and ORFs are arranged alternately in opposing orientations, with inverted sfxa101 inversion sites inserted at both ends of each promoter-terminator pair. Upon inducing enzyme expression, inversions among the sfxa101 sites regulate on/off transcription of the selection markers, enabling yeast growth on corresponding selective media. To ensure stability, we integrated the sequence of the screening system into the X chromosome of Saccharomyces cerevisiae. Using the screening system of DNA inversion enzyme, we tested the multisite inversion capabilities of the wild-type Rci and the mutant Rci8. After 12 h of induced expression, yeast colonies containing the wild-type Rci or Rci8 can only grow on media of a single selection but not on media of two or three selections, indicating that neither enzyme could efficiently mediate multisite inversion (Supplemental Fig. 1).

Establishing a eukaryotic multiplexed site-specific inversion system through two rounds of progressively stringent directed evolution

Directed evolution is a regular approach for generating mutants and selecting desirable functions, which has become a powerful technology platform in protein engineering24,25. We expect to enhance the multisite inversion efficiency of the Rci enzyme through directed evolution and the high-throughput screening system for multiplexed site-specific inversion enzymes (Fig. 1c). Error-prone PCR was used to introduce random mutations into the coding sequence of wild-type Rci26. The plasmid libraries of mutants were transformed into the yeast strain yLJY001, which contained the screening system. Firstly, we plated the induced yeast on media of two selections (SC-URA+HYG-HIS-LEU, SC+HYG+G418-HIS-LEU, SC-URA+G418-HIS-LEU) to screen mutants capable of multisite inversion. In this round of directed evolution, 109 unique mutants were identified by Sanger sequencing and 12 mutants with a total inversion efficiency of over 0.1% were selected. The mutant Rci51 shows the highest total inversion efficiency up to 10.8%, which is a ~350-fold improvement compared with the wild-type, although the co-inversion efficiency of three markers remains low (Fig. 2a, b). To further improve multisite inversion efficiency, Rci51 was chosen as a template for a second round of directed evolution with increased selection pressure through a shorter induction time and three selection markers. Cells with the library of Rci51 mutants were induced for 8 h and plated on a medium of three selections (SC-URA+G418+HYG-HIS-LEU). After this second round, we identified 23 unique mutants, among which three were superior to Rci51 in both total inversion efficiency and three-markers co-inversion efficiency. Specifically, the mutant Rci51-5 shows the highest co-inversion efficiency of three markers, and the total inversion efficiency reaches 35.9% (Fig. 2a, b). To visually characterize the efficiency of multisite inversion for the wild-type Rci, Rci51, and Rci51-5, the induced yeast cells were plated on various selective media at the same concentration. The results show that Rci51-5 could generate a greater number of inverted colonies than wild-type Rci and Rci51 under various selective conditions (Fig. 2c). Especially, hundreds of clones containing Rci51-5 could grow on the media with three co-inversion selections, whereas almost none clones grew for Rci and Rci51. We also tried a third round of directed evolution, unfortunately, we did not obtain mutants with efficiency and specificity of multisite inversion higher than Rci51-5 (Supplemental Fig. 2).

Fig. 2: The outcome of two rounds of progressively stringent directed evolution and mutations analysis.
figure 2

a The total inversion efficiency of the mutants generated in the two rounds of directed evolution. The orange and green represent the mutants produced in the first and second rounds of directed evolution, respectively. The error bars represent standard deviation of three biological replicates. b The co-inversion efficiency of three markers for the mutants generated during the two rounds of directed evolution. The efficiency is calculated following 12 h of induction. The orange and green represent the mutants produced in the first and second rounds of directed evolution, respectively. The error bars represent standard deviation of three biological replicates. c Yeast colonies containing the wild-type Rci, Rci51, or Rci51-5 were grown on various selective media after inducing 12 h, respectively. Compared to the Rci and Rci51, more colonies of Rci51-5 were grown on selective media as the number of selection markers increased. The pictures were taken after 3 days incubation at 30 °C. d Sequence alignment of Rci51-5 and the wild-type Rci. The mutant Rci51-5 identified five nonsynonymous mutations (red rectangle). e The relative multisite inversion efficiency of mutants containing single or combinations of mutations. ‘+’ indicates the introduction of specific mutations in Rci sequence; ‘-’ indicates their absence. The error bars represent standard deviation of three biological replicates. f Structural details of the inversion enzyme/DNA system by AlphaFold3. Two sfxa101 DNA(s) are shown in the protein-nucleic acid complex with four Rci51-5 monomers (the four are in blue, green, yellow and magenta cartoon, respectively). Mutation residues are depicted in spheres. The dashed box marks the region zoomed in and showed in detail (Rci51-5 and the wild-type Rci are in red and black, respectively). The hydrogen bonds and salt bridges are depicted with yellow and magenta dashed lines, respectively. Source data for this figure are provided as a Source Data file.

Full size image

To compare the efficiency of Rci51-5 at different inversion sites, we tested the inversion efficiency of the Rci51-5 at sfxa101, sfxa102, sfxa107, sfxa108, sfxa109, sfxa11227. The result indicated that the sfxa101 site shows the highest efficiency among the sfx recombination sites (Supplemental Fig. 3), which is consistent with results in bacteria27,28,29. Therefore, a multiplexed inversion system was established in yeast based on the inversion enzyme Rci51-5 and sfxa101 sites. Furthermore, by constructing a double fluorescence detection system, we preliminarily verified that the system of Rci51-5/multi-sfxa101 could work in HEK-293T cells (Supplemental Fig. 4). In addition, the wild-type Rci enzyme belongs to the tyrosine recombinase family and we tested the orthogonality of the Rci51-5/sfxa101 system against the commonly utilized site-specific recombination systems (Cre/loxP30, Flp/frt31, and Vika/vox32). The results showed that Rci51-5 was exclusively active at the sfxa101 site, while other recombinases exhibited no activity at the sfxa101 site (Supplemental Fig. 5), indicating that the Rci51-5/sfxa101 system is orthogonal to those above site-specific recombination systems. To improve the robustness of the system, we used a genetic AND gate15 to achieve a tighter regulation of Rci51-5 expression, and integrated the transcriptional unit of Rci51-5 into the yeast genome (Supplemental Figs. 6, 7).

Synergistic effect of five mutations for the improved inversion efficiency of the evolved enzyme

To clarify the mutations of Rci51-5, we sequenced the full-length of Rci51-5 plasmid and the result of Sanger sequencing indicates the presence of five nonsynonymous mutations: Q30K, L109P, F246V, A295S and A330T (Fig. 2d). To determine which of these mutations or their combinations contributed to the enhanced inversion efficiency of Rci51-5, we constructed 23 mutants, each containing single mutation or various combinations of mutations. Then, we tested the multisite inversion efficiency of these mutants. The results from a single mutations demonstrated that only the F246V mutation could facilitate multisite inversion, with an efficiency of 0.021 relative to the Rci51-5, while others could not (Fig. 2e). In the combination of two mutations, the efficiency of the L109P & F246V increases by 15-fold compared to the F246V mutation. Some combinations of three or four mutations (Q30K & L109P & F246V, Q30K & L109P & F246V & A295S) can further improve the efficiency of inversion, while others show a decrease in inversion efficiency. A similar phenomenon has also been observed in other protein engineering studies of multi-mutational variants33,34. The multisite inversion efficiency of the Rci51-5 mutant, which contains five mutations, is higher than all other mutants. We then employed the AlphaFold3 to predict structure of the protein-nucleic acid complex of Rci51-5/sfxa101 for visualization35 (Fig. 2f). The precited structure of Rci51-5/sfxa101 exhibits a tetrameric complex structure, similar to the structure of Cre/loxP36. By closely looking at the predicted structure, we observed that the F246V may reduce the steric hindrance, leading to a closer proximity of the two helices in the middle of the protein chain, and mutants containing the L109P show a tendency to expand central channel of the tetrameric complex. These are likely important for the efficient formation of tetrameric complex of the Rci enzyme, which may account for the significant improvement in efficiency for the combination of L109P & F246V. In addition, the other three mutants are found to potentially exhibit additional interactions in the complex of tetrameric-DNA (a salt bridge of protein-DNA for K30, new hydrogen bonds for S295 and T330). Previous studies have shown that the change of amino acid residues for nucleases is related to the protein activity conformation and protein anchoring ability to DNA37. We speculate that these five nonsynonymous mutations may synergistically affect the interaction of tetrameric Rci51-5 and DNA molecules, thus improving recombination efficiency, although it is still difficult to reveal the explicit mechanism of the synergistic effects for five mutations especially in the in vivo environment of yeast cells34.

Rci51-5/multi-sfxa101 system shows a lower rate of deletion than Cre/multi-loxP system

The Cre/loxP system is one of the most widely used site-specific recombination systems, which can efficiently mediate the deletion between same-oriented sites or inversion between reverse-oriented sites without accessory proteins30. However, when applying the Cre/loxP system at multiple sites, such as in cell barcoding construction5 and metabolic pathway optimization7, deletions may result in decreased library diversity and loss of the targeted genes. The multiplexed site-specific inversion system we developed in eukaryotes may provide a tool to address this issue. To compare the Rci51-5/multi-sfxa101 with the Cre/multi-loxP systems, we constructed two circuits containing multiple sfxa101 and loxP sites, respectively. The circuits also contain a metabolic pathway capable of generating color compounds, enabling the rapid verification of recombination outcomes. We assembled an array containing ten promoters and corresponding terminators and selected three transcription units involved in the β-carotene pathway. The array and the transcription units are oriented in opposing directions, with the sfxa101 or the loxP sites inserted into 3’ of promoters, respectively (Fig. 3a). The deletion events mediated by recombinase might disrupt the metabolic pathway, leading to the appearance of white yeast colonies38. The induced yeast containing the Rci51-5/multi-sfxa101 or Cre/multi-loxP were plated and the proportion of white colonies was counted at different times. At 24 h, the number of white colonies in the Cre/multi-loxP system is significantly higher compared to the Rci51-5/multi-sfxa101 system (Fig. 3b). Within 48 h, the proportion of white colonies for the Rci51-5/multi-sfxa101 system is notably lower than the Cre/multi-loxP system, although it increases slowly over time (Fig. 3c). The results showed that the deletion rate of the Rci51-5/multi-sfxa101 system was notably lower than Cre/multi-loxP. Furthermore, we counted the proportions of dark orange colonies for two systems at various time points. Within 48 h, the numbers of dark orange colonies for Rci51-5/multi-sfxa101 system increased over time, whereas the Cre/multi-loxP system demonstrated a decline after 24 h, which was most likely attributed to the accumulation of deletions (Supplemental Fig. 8).

Fig. 3: Comparison of the Rci51-5/multi-sfxa101 system and Cre/multi-loxP system.
figure 3

a Two circuits design to validate the recombination outcome of different systems. The 3’ and 5’ of the circuit consist of the promoter array and exogenous genes, respectively. The recombination site is inserted into 3’ of each promoter. The promoter array encompasses a range of promoters with different strengths (color scale): strong promoters pGMP1, pTDH3, pTEF1, medium-strength promoters pPGK1, pPDC1, pTEF2, pTPI1, and weak promoters pCHO1, pADH1, pCYC1. The yellow triangles represent the sfxa101 site, while the green triangles represent the loxP site. b Yeast colonies induced by different recombination systems. The magnified region shows different colony colors and the white colonies of the Rci51-5/multi-sfxa101 system are significantly lower than the Cre/multi-loxP system. The cultures induced by 24 h were spread on synthetic complete medium lacking leucine and uracil (SC-Leu-Ura). The pictures were taken after 7 days incubation at 30 °C. c The proportion of white colonies to total colonies at different induction times. Bars and error bars represent average and standard deviation of three biological replicates, respectively. Source data for this figure are provided as a Source Data file.

Full size image

Applying multiplexed site-specific inversion system to promoter substitution for metabolic engineering

In the field of metabolic engineering, the production of target compounds involves the co-expression of multiple genes along a metabolic pathway39,40. Identifying the optimal combination of gene expression levels is crucial for increasing the yield of target products, and promoter engineering is an important tool in achieving this goal41,42,43. By inserting recombination sites among promoters of varying strengths and exogenous genes, respectively, site-specific recombination systems can be utilized for metabolic engineering by generating new transcriptional combinations. The multiplexed site-specific inversion system we developed exhibits a bias for inversions rather than deletions, which decreases the risk of loss of exogenous genes, giving it a unique advantage in constructing multi-gene promoter regulation systems for metabolic pathways.

Taking the biosynthetic pathway of β-carotene as an example, we applied the multiplexed site-specific inversion system to promoter engineering and evaluated the substitution of promoters via a PCR-based rapid verification strategy (Fig. 4a). This strategy confirms whether the original promoters of exogenous genes have been replaced and identifies which promoters have replaced them. The diversity of promoter substitution can be assessed through mixed colonies PCR. The yeast culture induced for 24 h in a liquid medium served as the mixed template for PCR, while the uninduced yLJY023 was the control. For each exogenous gene, forward primers (F-crtYB, F-crtI, F-crtE) are combined with the reverse primers of their original promoters (R-1, R-2, R-3), as well as with primers F1-F10 for the ten promoters in the array, and a total of 33 PCR assays are conducted. The results of PCR showed that for each exogenous gene, all 11 primer pairs in the mixed template yielded positive bands, while in the control, only the primers of exogenous genes and the primers of original promoters could amplify positive bands. These findings demonstrate that in the experimental group, original promoters of exogenous genes could be replaced by any promoter from the array (Fig. 4b). To characterize the efficiency of promoter substitution, yeast cultures were plated at various induction times, and 256 single clones were randomly selected and validated using PCR to confirm the substitution of the original promoters. The results showed that the efficiency of promoter substitution improved with increasing induction time (Fig. 4c). At 48 h, the proportions of yeasts with one, two, and three promoters replaced are 44%, 8.1%, and 4.0%, respectively, and the total efficiency of promoter substitution is up to 56.1%. In order to assess the stability of the system, we performed a 14 day consecutive subculturing of the strain yLJY023, which harbors the plasmid of multiplexed site-specific inversion system (Supplemental Fig. 9a). The results of PCR and whole genome sequencing confirmed the long-term stability and capability of the Rci51-5/multi-sfxa101 system (Supplemental Fig. 9b, c).

Fig. 4: The multiplexed site-specific inversion system applied to promoter substitution for metabolic engineering.
figure 4

a Schematic diagram of PCR strategy for validating whether the original promoter has been replaced. The horizontal arrows indicate the orientation of primers. The promoters P1-P10 consist of pGPM1, pTDH3, pTEF1, pPGK1, pPDC1, pTEF2, pTPI1, pCHO1, pADH1 and pCYC1. Taking the original promoter pHXT7 of crtI is replaced by promoter P8 as an example to illustrate the verification strategy: when P8 replaces pHXT7, the forward primer F-crtI becomes a reverse primer, which can amplify a short positive band with F-8 (green ‘√’) through PCR reaction, whereas F-crtI and R-2 can not (red ‘×’). b Three sets of PCR results for mixed colonies. The top, middle, and bottom panels show the PCR results of primer of crtYB, crtI, crtE with the primers of promoters (one primer of original promoter, ten primers of promoters in the array), respectively. The top of each set of gel represents the experimental group (induced liquid cultures) and the bottom represents the control (uninduced liquid cultures). The detailed sequence of primers is shown in Supplemental Table 3. c The efficiency of promoter substitution. The error bars represent standard deviation of three biological replicates. d β-carotene biosynthesis pathway in Saccharomyces cerevisiae. Bold characters on the arrows represent exogenous genes targeted for expression optimization. Full/dashed arrows indicate one/multiple enzymatic conversion(s). e The multi-gene promoter regulation system for β-carotene pathway. The correlation among genotype, production of carotenoid and the expression levels of exogenous genes in strains with replaced original promoters. The control yLJY023 is not induced, and other strains are arranged in descending order based on color intensity. The results for exogenous gene promoter substitution are verified by PCR analysis. The transcriptional levels of exogenous genes are evaluated by qPCR (the transcriptional levels of three exogenous genes in the control yLJY023 are set as the base value). The error bars represent standard deviation of three biological replicates. The yields of carotenoid are determined by high-performance liquid chromatography (HPLC). Quantification is performed in biological triplicate for each strain as shown. The error bars represent standard deviation of three biological replicates. Source data for this figure are provided as a Source Data file.

Full size image

Next, we evaluated the ability of the multiplexed site-specific inversion system to diversify gene transcription and boost production of β-carotene in metabolic pathways. The synthesis of β-carotene involves multiple genes (Fig. 4d) and our previous research has demonstrated a correlation between phenotypic color and carotenoid yield in yeast38. In this study, we selected colonies with various colors and identified 11 strains with unique combinations of exogenous genes and promoters by PCR. Through a systematic analysis of genotypes and production of carotenoids for all tested strains, we found that strains with enhanced transcription of the crtI (yLJY045 and yLJY028) or crtYB (yLJY044 and yLJY055) showed a 4.9–6.3 fold-change and 1.9–2.6 fold-change for the production of β-carotene, respectively. The strain yLJY045 increased the production of β-carotene with a 6.3-fold (1.16 µg/mg DCW) compared to the control yLJY023, with its original promoter pHXT7 of crtI replaced by a strong promoter pTDH3. The qPCR results indicate a significant enhancement in the expression of the crtI gene for yLJY045 (Fig. 4e). This result confirms that crtI is a rate-limiting enzyme for the biosynthetic pathway of β-carotene in Saccharomyces cerevisiae, as compared to crtYB and crtE, which is consistent with previous studies44,45. Moreover, strain yLJY081 exhibits three inversions, in which original promoters of all exogenous genes have been replaced, resulting in a deep orange colony color. The results of high-performance liquid chromatography (HPLC) showed that the production of β-carotene and lycopene for yLJY081 is 1.4 and 0.32 µg/mg DCW, corresponding to a 7.6-fold and 12.4-fold increase in yield compared to the original construct. Our result demonstrates that the multiplexed site-specific inversion system can be used to optimize gene expression for metabolic engineering. We also applied the multiplexed site-specific inversion system to the biosynthetic pathway of violacein (Supplemental Fig. 10).

Discussion

The multiplexed site-specific inversion system produces complex DNA rearrangements through random inversions among multiple sites, which plays a crucial role in generating genetic diversity and phenotypic adaptation17,46,47. Here, we developed a multiplexed site-specific inversion system in yeast based on Rci51-5 and sfxa101 sites, which can manipulate gene inversions among multiple reverse-oriented sites and shows a significantly lower deletion rate than Cre/multi-loxP. Traditional approaches for altering gene expression profiles of metabolic pathways generally involve multiple rounds of combinatorial assembly48,49,50, whereas our system only requires a simple initial one-step assembly process to generate a diverse library of cells upon the induction of inversion enzyme. Kevin et al. established GEMbLeR (multiplexed Gene Expression Modification by LoxPsym Cre Recombination), facilitating promoter/terminator substitution in Saccharomyces cerevisiae for the rapid optimizing of heterologous biosynthetic pathways7. Comparing with Cre/multi-loxP system, our Rci51-5/multi-sfxa101 system could reduce the rate of deletions when applied to multi-gene optimization of metabolic pathways, and this advantage can be further expanded when applied to multiple rounds of metabolic engineering. In addition, the generation of diverse library of transcriptional combinations poses additional challenges for downstream high-throughput screening. In this study, we rely on phenotypic color-based screening, while it is not applicable to the screening of colorless products. Other advanced high-throughput screening methods51,52, such as the combination of transcription factor-based biosensors and fluorescence-activated cell sorting (FACS), have the potential to expand the application of our system to broader metabolic productions. In the multi-gene promoter regulation system, the original promoter of crtYB is more easily substituted, potentially indicating a positional preference among multiple inversion sites (Supplemental Fig. 11). A feature of the multiplexed site-specific inversion system is to change the relative position of DNA through inversion, and it may have the capacity to address this issue by iterative rounds of inversion.

Although the rate of deletion for the Rci51-5/multi-sfxa101 system is notably lower than the Cre/multi-loxP system, instances of deletion leakage have been detected in our study. In the future, the system might be further optimized in terms of recombinases, recombination sites, and co-factors: (1) The AI-assisted rational design of enzymes might accelerate the acquisition of recombinases with lower deletion rate53; (2) The sequence of the sfx can affect the inversion frequency and specificity27 and the design and modification of the sequence of the sfx site might facilitate the inversion events; (3) The specificity of the system may be further optimized by introducing factors, such as nucleoid-associated protein HU and nucleoid protein Fis (Factor for inversion stimulation)17. In addition, when applied the multiplexed site-specific inversion system to metabolic engineering, incorporating a backup pathway may eliminate the risks caused by the deletion events. While we have shown that the inversion system is stable within a period of time, for a longer-term stability in the application of industrial-scale production, we suggest that removing the inversion system after determining the desired mutant.

The promoters of varying types and transcriptional strength derived from promoter engineering54,55 can be incorporated into the multiplexed site-specific inversion system to promoter substitution, which holds the potential to generate more diverse combinations of gene expression. Moreover, many other metabolic engineering strategies might be combined with our system to further enhance the productivity of useful chemicals56,57,58,59. Based on the capability of the Rci51-5/multi-sfxa101 system to invert at multiple sites in HEK-293T cells, it offers potential applications in constructing cellular barcodes and genetic circuits, providing an alternative to the Cre/multi-loxP system5,60,61. As the system is orthogonal to other commonly used site-specific recombination systems, it can be combined with other systems to enhance the scale and flexibility when manipulating larger-scale complex gene rearrangements.

In conclusion, we have developed a multiplexed site-specific inversion system capable of inducing DNA inversions among multiple sites in eukaryotes. This system provides an approach to altering multiplexed gene expression for metabolic engineering and a unique tool for genome engineering in eukaryotes.

Methods

Strains, plasmids, and growth medium

The yeast strains and plasmids used in this study are listed in Supplemental Tables 1 and 2, respectively. Yeast strains were grown at 30 °C in YPD medium (20 g/L glucose, 20 g/L peptone, 10 g/L yeast extract). For selective growth, the plasmid-harboring yeast strains were cultured in Synthetic Complete (SC) medium supplemented with a 2 g/L dropout mixture lacking leucine, histidine, tryptophan, and uracil, with additional components added according to selective requirements. Liquid culture was grown at 30 °C and 220 rpm in incubator. Expression of Rci mutants, regulated by the inducible promoter GAL1, was induced in SC medium containing 2% galactose. For Cre recombinase, fusion with an Estrogen Binding Domain (EBD) enabled estradiol-dependent regulation, with induction achieved using SC medium supplemented with 1 μM β-estradiol. Modified SC medium included SC+GAL (20 g/L galactose replacing glucose), SC+EST (1 μM β-estradiol), SC+HYG (0.3 g/L hygromycin), and SC+G418 (1 g/L geneticin). Escherichia coli Trans-T1 (Beijing Biomed Co., Ltd.) was used for plasmid transformation. Escherichia coli was grown at 37 °C in LB medium (15 g/L agar, 10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl). 100 mg/L ampicillin or kanamycin was added for selection. The sequence of the screen system was integrated into the right arm of X chromosome in yeast by homologous recombination62.

Construction of mutant library

Random mutations were introduced into the Rci encoding sequence via error-prone PCR, generating a library of 1.0 × 107 unique colonies with 1–4 mutations per kb in the Rci gene26. The library capacity and mutation frequency in the second and third rounds of directed evolution were the same as in the first round, using Rci51 and Rci51-5 as templates for mutagenesis, respectively. The Rci mutant library was constructed and tested following the standard protocol of GENEWIZ, and the sample was sent on dry ice to GENEWIZ, Suzhou.

Selection of multiplexed site-specific inversion enzyme

The library of mutant plasmid was introduced into yLJY001, which contains the screening system for multiplexed site-specific inversion enzyme63. The gene of Rci variants was cloned on a centromeric plasmid and transcriptionally controlled by an inducible promoter GAL1. The transformed strains were cultured in a liquid medium (SC-HIS-LEU+GAL) for 12 h. Induced yeast cells were plated in triplicate on media (SC+HYG+G418-HIS-LEU, SC-URA+HYG-HIS-LEU, SC-URA+G418-HIS-LEU) to screen for mutants that underwent multisite inversion. Strain colonies capable of growing on media of two selections were selected for yeast plasmid extraction, followed by enrichment of plasmids containing Rci mutants through E. coli transformation64. The mutation sites of the enzyme were determined by Sanger sequencing. During the second round of directed evolution, yeast transformation and induction culture are the same as the first round, but the induction time of Rci51 mutants is reduced from 12 h to 8 h. The induced cells are directly plated on SC-HIS-LEU-URA+G418+HYG to increase the selection pressure (Fig. 1c).

In the third round of directed evolution, the mutant plasmid library of Rci51-5 was transformed into yeast strains containing a specific screening system for the inversion enzyme by Saccharomyces cerevisiae lithium acetate transformation63. The transformed cells were cultured in a liquid medium (SC-LEU+HYG+GAL) for 12 h. The induced yeast cells were spread on SC-URA-LEU+HYG to screen for mutants undergoing inversion reaction. A total of 5000 mutants were randomly selected and replicated on SC-HIS-LEU+HYG+GAL medium to assess deletion efficiency. Among them, mutants that can barely grow on SC-HIS-LEU+HYG+GAL medium were selected, indicating inversion capability with decreased deletion efficiency. The plasmids containing Rci51-5 mutations were extracted and enriched64. These plasmids were then transformed into yeast strain yLJY001 to validate the efficiency of multisite inversion. Mutant plasmids capable of growing on SC-URA+G418+HYG-HIS-LEU after induction expression with galactose for 12 h were selected for sequencing verification (Supplemental Table 2).

Inversion efficiency assay for Rci mutants

To test the inversion efficiency of mutants, yeasts carrying the Rci mutants were individually plated on media of one, two, or three selections and medium without selections. We defined the total inversion efficiency as the ratio of colonies on media of a single selection (SC-URA-HIS-LEU, SC+G418-HIS-LEU, SC+HYG-HIS-LEU) to those on medium without selection. The efficiency of co-inversion three markers was defined as the ratio of colonies on the medium of three selections (SC-URA+G418+HYG-HIS-LEU) to those on the medium without selection.

Inversion efficiency assay for different sfx sites

The different sfx sites (sfxa101, sfxa102, sfxa107, sfxa108, sfxa109, sfxa112) were cloned into the recombination reporter vectors. Plasmids containing different sfx sites were introduced separately into yeast that contained the plasmid of the Rci51-563. The transformed strains were cultured in a liquid medium (SC-HIS-LEU+GAL) for 12 h, and the induced yeast cells were plated in triplicate on SC-URA-HIS-LEU and SC-HIS-LEU. The inversion efficiency is the ratio of colonies on SC-URA-HIS-LEU to those on SC-HIS-LEU.

Orthogonality of the Rci51-5/sfxa101 system assay

The sfx sites in Supplemental Fig. 3a were replaced with vox, frt, or loxP sites, respectively. Plasmids containing different recombination sites were introduced separately into yeast that contained the plasmid of the Rci51-5. The transformed strains were cultured in a liquid medium (SC-HIS-LEU+GAL) for 12 h. Induced yeast cells were plated in triplicate on SC-URA-HIS-LEU. The plasmids containing Cre, Flp and Vike recombinase expression genes were introduced separately into yeast containing the recombinant reporter vector containing sfxa101 sites. Induced different recombinant enzymes for 12 h and induced yeast cells were plated medium with URA selection, respectively.

Cell culture, plasmid transfection, and fluorescence microscopy

HEK-293T cells (ATCC) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Procell) supplemented with 10% FBS (Procell) at 37 °C with 5% CO2. For plasmid transfection, 293T cells were seeded in 6-well plates (Corning) at 70% confluence. Transfection was performed by mixing 5 μg of plasmid with 250 μL of Opti-MEM (Gibco) and 7 μL of Liposomal Transfection Reagent (Yeasen) with 250 μL of Opti-MEM (Gibco) separately for 5 min. The two solutions were thoroughly mixed and incubated at room temperature for 20 min, then the final mixture was added dropwise into each well of the 6-well plate. Cells were imaged at 48 h post-transfection using the EVOS M7000 live-cell fluorescence imaging system equipped with a fluorescence microscope.

Molecular modeling and structural analysis for protein-nucleic acid complexes

The three-dimensional structure model of the Rci, Rci51 and Rci51-5 were constructed based on their amino acid sequence using the AlphaFold3 program with default settings, and the structural information of the protein-nucleic acid complexes was obtained35. PyMOL software (Delano, 2010) was used to analyze structures of the complexes65.

Deletion rate assay

The yLJY022 and yLJY023, which contain a multi-gene promoter regulation system for the β-carotene metabolic pathway using Cre/multi-loxP and Rci51-5/multi-sfxa101 respectively, were cultured in SC-URA-LEU+GAL and SC-URA-LEU+EST liquid medium for 12 h. A new medium was replaced every 12 h, and samples were taken at 8 h, 12 h, 24 h, 36 h, and 48 h, respectively. The samples were streaked on SC-LEU-URA at different times to calculate the proportion of white colonies.

Mixed colonies PCR

The yeast yLJY023 was cultured in SC-URA-LEU+GAL and SC-URA-LEU liquid medium for 24 h, respectively. The strains cultured in the two media were used as templates for PCR. Each promoter substitution of the exogenous gene was verified by 11 pairs of PCR primers. The first set of primers included the forward primer for the exogenous gene and the reverse primer for its original promoter, while the subsequent 10 sets consisted of the forward primer for the exogenous gene and the forward primer for promoters in the array. A total of 33 PCR reactions were conducted.

Promoter substitution efficiency assay

The yLJY023 was cultured in SC-URA-LEU+GAL for 12 h. A new liquid medium SC-HIS-LEU+GAL was replaced every 12 h, and samples were taken at 8 h, 12 h, 24 h, 36 h, and 48 h and plated on SC-LEU-URA. Colonies on the medium were randomly picked to verify the original promoter substitutions of crtE, crtI, and crtYB by the strategy described in Fig. 4a.

HPLC measurement of carotenoid production

The induced and control yeast strains were cultured in 5 mL of SC-Ura liquid medium at 220 rpm, 30 °C in a shaking incubator. The saturated cultures were diluted to an initial OD600 of 0.1 in 30 mL of SC-Ura liquid medium and grown for 72 h with the same condition. Cells from 10 mL culture were collected and then washed with double-distilled water and harvested by centrifugation, repeated twice. The cells were dried in a freeze dryer for measurement of the dry cell weight. The dry cells were resuspended in 700 μL of acetone and then shaken with quartz sand for 30 min. The acetone extracts were centrifuged and filtered with a 0.22 μm filter for subsequent analysis. The analysis of carotenoids was performed by HPLC (Waters 2695) equipped with BDS HYPERSIL C18 column (15 cm × 4.6 mm) and UV detection at 460-470 nm. The mobile phase consisted of acetonitrile-water (9:1 v/v) and methanol-2-propanol (3:2 v/v) with a flow rate of 1 mL per min at 25 °C. The content of the carotenoids was expressed as µg per mg dry cell weight. Each sample was performed on technical triplicates.

Statistics and reproducibility

All experiments included in this study were performed three times at least. All data are presented as mean ± standard deviation. No statistical method was used to predetermine sample size. No data were excluded from the analyses. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.