A decoupled transcription platform enables tunable and predictable gene expression in yeast Authors: Yujie Chen, Hui Li, Lin Duan, Yang Liu, Jingrui Yan, Haoyang Chen, Jianguo Yang Published: 2026 ABSTRACT AbstractPredictable control of gene expression is essential for building genetic circuits and improving metabolic pathways, but conventional promoter libraries often behave unpredictably when genes are combined. Here we develop CRISPR-Activated Promoter-based Orthogonal expression (CAPO), a quantitative platform for controlling multiple genes in yeast. CAPO uses synthetic CRISPR-activated promoters that remain silent until matching guide RNAs recruit dCas9-VPR. We tune each gene by varying guide RNA abundance with defined T7 promoters, while keeping regulatory channels orthogonal. CAPO reaches expression levels comparable to strong native yeast promoters, maintains low background activity, and preserves promoter-strength order across different genes. We apply CAPO to program broad fluorescence color outputs and to rapidly optimize lycopene and 3-hydroxypropionic acid biosynthesis. These results establish CAPO as a scalable platform for predictable engineering of eukaryotic gene networks.Subject terms: Expression systems, Metabolic engineering, Synthetic biology, Genetic engineering Introduction Similar to how musical instruments require precisely positioned capos to produce harmonious melodies, cellular systems depend on the accurate and stable expression of massive gene sets to maintain physiological homeostasis1. Effective control of gene expression intensity ensures coordinated interactions within gene networks, thereby enhancing cellular function and mitigating metabolic burden or toxicity2. In bioengineering, this principle is manifested through meticulous genetic engineering strategies, including the fine-tuning of endogenous expression profiles or exogenous genetic elements. It serves not only to achieve targeted product expression but also to enable the reconstruction of complex biological networks within eukaryotic hosts3. Unlike prokaryotes, which primarily regulate gene expression via promoter sequences and ribosome-binding sites (RBS), eukaryotic cells rely on complex mRNA processing and modification4. Precise gene expression control in eukaryotes requires more advanced regulatory designs across both transcriptional and translational stages. Expression systems based on endogenous eukaryotic promoters remain the most prevalent, with promoter databases providing extensive libraries for expression tuning via screening5. In addition, extensive libraries of synthetic and hybrid promoters with graded activities have been generated in Saccharomyces cerevisiae and systematically characterized, providing standardized artificial promoter parts for tunable gene expression6,7. Combinatorial transcriptional engineering strategies, including promoter shuffling, rationally designed promoter libraries, and pathway-level promoter optimization, have also been established to enable multiplexed and coordinated tuning of gene expression in yeast6,8–10. The combinatorial pairing of exogenous genes with endogenous promoters of varying strengths to optimize product biosynthesis has been practiced in yeasts11–13. However, endogenous promoter-based regulation is limited by narrow expression gradients14, unpredictable output15, and susceptibility to interference from host physiological processes16, which restricts finer quantitative control. Moreover, in most artificial promoter and combinatorial frameworks, the promoter architecture and the encoded expression strength remain tightly coupled at the DNA level, so that each desired expression profile must be hard-wired into specific promoter–gene constructs7,8,10. The CRISPR activation (CRISPRa) technique, which involves the fusion of endonuclease-deficient Cas9 (dCas9) with transcriptional activation domains such as the tripartite VP64, p65, and Rta activation domains (VPR), emulates the modular structure of eukaryotic transcription factors. This advancement has facilitated the development of programmable regulation methods for the programmable regulation of endogenous gene expression by targeting dCas9-activator to specific genomic loci using guide RNA (gRNA)17,18. Nevertheless, this approach remains unable to overcome the influence of host physiology on endogenous promoters and is largely constrained by gRNA design limitations imposed by PAM sites. T7 RNA polymerase (T7 RNAP)-based orthogonal transcription systems have served as the primary platform for achieving precise and quantitative gene expression with minimal interference from native RNA polymerases in prokaryotes19,20. However, transcripts synthesized by T7 RNAP lack the 5’-cap modifications necessary for translation initiation in eukaryotic cells21, thereby restricting their application to prokaryotic systems. Recent advancements have involved the coupling of viral capping enzymes with T7 RNAP, facilitating the translation of T7 promoter-driven transcripts and providing a method for quantitative regulation of gene expression in eukaryotic contexts22,23. However, these quantitative expression systems exhibit reduced translation efficiency and may disrupt the dynamic regulation of mRNA capping and decapping22–26, diminishing their overall usability. T7 RNAP has also been employed for the expression of gRNA in CRISPRa to facilitate the regulation of endogenous promoters in eukaryotes27,28, although this approach cannot entirely circumvent the limitations of endogenous promoters29,30. Currently, Saccharomyces cerevisiae is extensively utilized as a chassis for the biosynthesis of numerous high-value compounds, such as lycopene, 3-hydroxypropionic acid (3-HP), artemisinin, and taxol, necessitating the expression of multiple heterologous genes31–33. In particular, lycopene and 3-HP biosynthetic pathways in yeast have been optimized through pathway regulation strategies, including promoter engineering, gene copy-number tuning, enhancement of precursor supply, and attenuation of competing pathways12,31,34–36. These studies underscore the importance of precise balancing of pathway gene expression for efficient metabolite production. However, the modulation of expression levels of these heterologous genes is primarily achieved using promoters exhibiting different strengths, derived from the host genome or synthetically designed. This strategy requires extensive reconstruction and assembly of the promoter, coding sequence, and terminator elements for each gene. As the number of target genes increases and the range of expression levels extends, the possible combinations of regulatory elements expand exponentially. This combinatorial explosion poses significant challenges, requiring substantial labor, resources, and time for the construction, screening, and validation of the optimized circuits. Here, we develop the CRISPR-Activated Promoter-based Orthogonal expression system (CAPO), which integrates the programmable activation mechanism of dCas9-VPR with the quantitative and orthogonal control enabled by T7 RNA polymerase. We engineer CRISPR-Activated Promoters (CAPs) by decoupling native promoter architecture from endogenous transcription-factor regulation, so that dCas9-VPR exclusively activates transcription. We regulate CAP output by modulating cognate gRNA concentrations, which orthogonal T7 RNA polymerase produces from defined T7 promoters. We further optimize CAP architecture and gRNA expression cassettes to achieve a broad dynamic range and minimal background activity. Finally, we apply CAPO to metabolic engineering pathways that require coordinated fine-tuning of multiple genes. Together, these results establish CAPO as a platform of tunable multigene regulation for eukaryotic synthetic biology and metabolic engineering. Results Creation and optimization of CRISPR-activated promoters Eukaryotic promoters are primarily comprised of core promoter regions and proximal control elements37,38 (Fig. 1a). The core promoter includes the TATA box and transcription start site (TSS), which are essential for the recruitment and assembly of the pre-initiation complex39. Proximal control elements contain binding sites for regulatory factors that play a crucial role in modulating gene expression levels40,41. In light of the activation mechanism of eukaryotic promoters, it is conceivable that replacing the proximal control elements with specific gRNA binding sites could result in a synthetic CRISPR-Activated Promoter (CAP), which can be activated solely via gRNA-targeted dCas9-activator (Fig. 1a, b). To evaluate this hypothesis, we initially selected the widely used PGK1 promoter from yeast as a model. By substituting the promoter sequences preceding 30 base pairs (bp) of the TATA box region, based on information from Eukaryotic Promoter Database5, with a 20 bp gRNA binding site and a 150 bp spacer sequence, the promoter was constructed as CAPPGK1 (Supplementary Data 1). All spacer sequences consisted of random sequences initiated with a protospacer-adjacent motif site (PAM) to satisfy the binding requirements of dCas942. To assess the feasibility of CAPPGK1, EGFP was employed as the reporter, dCas9-VPR was chosen as the transcriptional activator, and T7 RNAP was used to express the corresponding gRNA (Fig. 1c and Supplementary Data 2). Significant EGFP expression was observed exclusively when both gRNA and transcriptional activation components were present (Fig. 1d), demonstrating the functional efficacy of our synthetic CAP design. Construction and verification of the CAPO system. a Design of CRISPR-Activated Promoter (CAP). Upper panel, classical eukaryotic promoter architecture. Lower panel, CRISPR-Activated Promoter architecture. TSS: Transcription Start Site. UTR: Untranslated Region. b Mechanism of CAP Activation. c Constructs for CAP feasibility verification. d Verification of CAP activation. EGFP was highly expressed when all three modules were present. Data were presented as mean ± SD from n = 6 biologically independent yeast colonies. Each dot represents one independent colony. Statistical significance was assessed by one-way ANOVA followed by two-sided Šídák’s multiple-comparisons test. ***P < 0.001. From bottom to top, P = 0.0003, P < 0.0001, P < 0.0001 and P = 0.0001. e CAP sequence optimization by spacer truncation analysis. The expression intensities of CAPs with the same spacer length are stacked for display. Data were presented as mean ± SD. CAPPGK1 were analyzed with n = 2 biologically independent yeast cultures, whereas CAPTDH3 and CAPGAL2 were analyzed with n = 6 biologically independent yeast cultures. Statistical significance was assessed by one-way ANOVA followed by two-sided Dunnett’s multiple-comparisons test and indicated by multiple-comparisons letters. Groups labeled with different letters are significantly different (P < 0.05). Exact P values and test statistics are provided in the Source Data file. f Relative expression intensities of CAPs were constructed based on different eukaryotic core promoters. PC positive control, TDH3 promoter. NC negative control, empty vector. Data were presented as mean ± SD from n = 6 biologically independent yeast cultures. Each dot represents one independent culture. Statistical significance was assessed by one-way ANOVA followed by two-sided Dunnett’s multiple-comparisons testing; all P values for the indicated comparisons are <0.0001. Error bars indicate SD. Source data are provided as a Source Data file. Given the critical dependence of transcriptional activation efficiency on the spatial proximity between transcription-factor binding sites and transcriptional start sites43, we implemented systematic truncation testing with 10-bp resolution to optimize the spacer length in our engineered CAPs. Functional assessment of CAPPGK1 constructs containing spacer sequences ranging from 150 to 0 bp revealed optimal transcriptional activation efficiency when spacer lengths were maintained between 10 and 60 bp (Fig. 1e). Subsequent transcriptional activation profiling of CAPGAL2 and CAPTDH3 (constructed analogously to CAPPGK1) variants with randomized spacer lengths (60–10 bp) revealed a conserved length-dependent activation profile, indicating distinct optimal spacer ranges for robust promoter activation (Fig. 1e). Through a systematic evaluation of both architectural simplicity and transcriptional output, we ultimately established 30-bp spacers as the optimal configuration, balancing promoter minimization with robust activation efficiency. This standardized design framework was successfully extended to five additional CAPs (Fig. 1f), with an average length of 230 bp (Supplementary Data 1), demonstrating robust transcriptional activation and negligible basal activity in yeast. In summary, we developed and validated a compact CAPO system, achieving a reduced CAP length compared to traditional Class II promoters in eukaryotes. Importantly, our design features an expression system that employs discrete CAP and T7 promoter-driven gRNA modules, effectively decoupling the transcriptional activation from the CAP sequence. This modular architecture offers a strategy for the precise regulation of gene expression intensity through gRNA dosage modulation, thereby eliminating the need for repeated reconstruction of expression cassettes when adjusting multiple gene expression levels. Enabling the CAPO fine-tuning ability by gRNA expression module design During the implementation of the CAPO system for multiple gene expressions, we identified a discrepancy between gRNA expression and the corresponding targeted gene activation intensities (Supplementary Fig. 1a, b). This inconsistency may arise from the varying binding affinities of gRNAs to their cognate CAPs, leading to an uneven distribution of dCas9-VPR and consequently resulting in mismatched transcriptional outputs. To normalize gRNA-CAP interaction kinetics, we required a rapid method for quantifying the relative binding efficiencies of gRNAs. Noting that the gRNA expression cassette and prototype CAP spacer possess nearly identical sequence lengths, we hypothesized that replacing the CAP spacer with the T7-based gRNA expression cassette could create integrated transcription units, enabling simultaneous gRNA production and target gene activation. This conceptual framework led to the development of the T7-driven CRISPR-activated promoter (T7DCAP; Fig. 2a and Supplementary Data 1), which enables synchronous modulation of the gRNA and its target, as both share a common sequence. The engineered T7DCAP activates targeted gene expression through a biphasic mechanism: first, T7 RNAP directs the initial transcription of the gRNA; second, the resulting gRNA complexes with dCas9-VPR to target the original gRNA coding locus, ultimately activating a secondary transcription to express the downstream gene (Fig. 2a, b). The mechanism of T7DCAP enables the rapid screening of gRNAs with high activation potency. To accommodate the PAM sequence requirement of dCas9-VPR, the first three nucleotides, GTT, of the gRNA scaffold were strategically replaced with either AGG or GGG (Supplementary Fig. 2). To preserve the specific structure of the gRNA, the corresponding nucleotides that paired with GTT were also substituted (Supplementary Fig. 2). The results revealed that substituting the 5′-terminal trinucleotide of the scaffold with AGG restored full functionality of T7DCAP, achieving transcriptional outputs comparable to those of native CAP (Fig. 2c). In contrast, the original GTT and its GGG substitute exhibited minimal activation of T7DCAP (Fig. 2c). Subsequently, a dual-reporter system coupled with AGG-T7DCAP was used for subsequent gRNA screening (Fig. 2d). In this dual-reporter system, the TagBFP module expressed from the gRNA T (gTDH3) incorporated T7DCAP was used as an internal reference, whereas the mCherry-associated T7DCAP serves as a screening platform for candidate gRNAs (Fig. 2d). The fluorescence ratio of mCherry to TagBFP, each activated by gRNA T, was normalized to 1.0 (Fig. 2e). When the ratios of random gRNAs activating mCherry to gRNA T-activated TagBFP were near 1.0, it suggested a binding affinity similar to that of gRNA T. Using this selection strategy, we obtained gRNAs with normalized binding affinities compared to prototype gRNA T from a curated gRNA library containing hundreds of candidates (Fig. 2e and Supplementary Fig. 3). The selected gRNAs with binding affinities similar to that of gRNA T were subsequently integrated into the refined CAP architectures to generate an optimized set of CAPs showing improved activation strength (Supplementary Fig. 4). When used for multiple gene expression, the new CAPs circumvented the incongruent outputs observed with original unoptimized CAPs (Supplementary Fig. 1c). Optimization of CAPO for multiplex gene expression. a Mechanism of T7-Driven CRISPR-Activated Promoter and verification vector. The T7DCAP system functions via a biphasic transcriptional activation process. In the first round of T7DCAP transcription, the T7 promoter is recognized by T7 RNA polymerase, which transcribes gRNA from the TSS1. b Second round of T7DCAP transcription. The resulting gRNA forms a complex with dCas9-VPR and is subsequently retargeted to the specific guide sequence region within the CAP promoter, thereby recruiting the endogenous eukaryotic transcriptional machinery and activating downstream transcription from the TSS2. c Verification of T7-Driven CRISPR-Activated Promoter. When the first three base pairs of the gRNA scaffold were changed from GTT to AGG, T7DCAP was successfully activated by dCas9-VPR. Significance was assessed by one-way ANOVA followed by two-sided Šídák’s multiple-comparisons test; ****P < 0.0001 for all indicated comparisons. d, e High-throughput gRNA screening. Screened gRNAs were compared with the best-performing gRNA, gRNA T. The fluorescence intensity ratio of mCherry/TagBFP resulting from gRNA T-activated expression of mCherry was normalized to 1.0 (red square). f T7 promoter library screening for multiplex quantification capability of the CAPO system. Upper panel, regulation module; lower panel, expression module. Mutated bases in the T7 promoter sequence are marked in red, and the transcription start site is underlined. g, h Screened T7 promoter library with different expression intensities and sequences. Fluorescence intensities were normalized to the T7 wild-type (WT) group and are presented as relative values. The BFP/GFP ratio, indicated by black lines, served as a metric of expression stability. NC: Negative control, plasmid only express gRNAs for BFP and GFP. i Comparison of expression robustness between CAPs and eukaryotic endogenous promoters (EEPs). CAPs were sorted using a T7 promoter library from R10 to R1; EEPs were sorted according to EGFP expression levels. Data were presented as mean ± SD; error bars indicate SD. For (c, i), n = 4 biologically independent colonies; for (d, e, g, h), n = 3 biologically independent colonies. Source data are provided as a Source Data file. To achieve precise control over transgene expression within the CAPO system, a T7 promoter mutant library was engineered to titrate the gRNA transcriptional flux. Degenerate mutations were introduced in the −17 to −13 region of the T7 promoter, which is responsible for expressing gRNA targeting the CAP-mCherry cassette for strength screening. CAP-sfGFP and CAP-TagBFP cassettes activated by gRNAs expressed from the wild-type T7 promoter were used as internal controls (Fig. 2f). Twenty-six T7 promoter mutants were selected for sequencing and re-evaluation in triplicate (Fig. 2g and Supplementary Fig. 5). Ultimately, ten mutants were ranked and renamed according to their discernible expression gradients (Fig. 2h). To further determine whether the engineered T7 promoter gradient was reflected at the transcriptional level, RT-qPCR analysis was performed on the ranked mutant series. A progressive decrease in cognate gRNA abundance was observed along the gradient, accompanied by a corresponding reduction in the mRNA level of the target gene, indicating that the transcriptional output of CAPs was positively associated with the abundance of the corresponding gRNA (Supplementary Fig. 6). Notably, one hyperactive mutant, renamed T7 promoter Rank 1 (PT7R1), exhibited a 30% increase in expression compared to that of the wild-type T7 promoter (Fig. 2g). Importantly, the CAPs activated by gRNAs expressed from these T7 mutants maintained linear expression profiles for LacZ and GFP (Fig. 2i), whereas to endogenous eukaryotic promoters, which showed promoter strength inversion when expressing LacZ compared to GFP (Fig. 2i and Supplementary Data 1). This finding demonstrates that CAPs effectively address the promoter strength inversion issue commonly encountered in conventional eukaryotic expression systems7, thereby enabling precise and predictable transcriptional regulation. The fidelity of the expression gradient for LacZ and EGFP was consistent with mCherry screening data (Fig. 2g, i), highlighting the quantitative regulatory capacity of CAPO across different exogenous genes. Evaluation of coordinated multiple gene expression control of the CAPO system A foundational principle of color vision, initially described by Thomas Young in 1802, posits that all discernible colors can be generated by combining three primary colors (red, green, and blue, RGB) at varying intensities, a concept later formalized as the trichromatic theory44. To investigate the capability of the CAPO system in precisely orchestrating multiple gene expression, an integrated RGB reporter system was developed, wherein mCherry, sfGFP, and TagBFP served as proxies for the red, green, and blue channels, respectively (Fig. 3a). By applying the principles of trichromatic theory to replicate conventional RGB display mechanisms, the CAPO platform was employed to independently and continuously modulate the expression levels of each fluorescent protein through a customized T7 promoter library (Fig. 3a). This approach enables the production of diverse and predictable color outputs, thereby illustrating the capacity of this system for fine-tuned multiple gene expression. To further verify that the observed activation phenotype was attributable to the introduced CAPO regulatory elements rather than to the original strain background, the RGB reporter system was reconstituted in a clean background strain, where fluorescence activation was detected only when both the gRNA module and the activator module, consisting of dCas9-VPR and T7 RNAP, were simultaneously present, whereas either module alone produced only minimal background fluorescence (Supplementary Fig. 7). The CAPO platform supports two orthogonal regulatory paradigms: rational color programming via deliberate adjustment of CAP-specific expression intensities of each fluorescent protein to achieve the designed final color (Fig. 3b) and stochastic color screening by mixing 11 distinct expression states of each CAP-specific expressed fluorescent protein, together generating >10³ possible combinations of fluorescent protein expression levels (Fig. 3c). Because all color variations arise solely from differential expression ratios governed by the selection of T7 promoters in the gRNA expression plasmids, it is possible to rapidly determine the expression profile corresponding to any chosen color by PCR amplification and sequencing of an approximately 1000 bp region from the relevant gRNA expression plasmid (Fig. 3d). This streamlined approach enabled the rapid decoding of gRNA-driven color programming (Fig. 3e). These color combinations enabled the generation of diverse colony patterns using only three chromatic elements (Fig. 3f). CAPO-mediated RGB palette. a Schematic illustration of the quantitative regulation of fluorescent protein expression ratios via gRNA expression plasmids, implemented under a single RGB expression plasmid framework using the CAPO system. b Fluorescence images of yeast colonies expressing multiple fluorescent proteins. The annular sectors indicate the quantified expression ratio of each fluorescent protein as determined by gRNA, with the central region representing the actual observed color of the colony. c Fluorescence imaging of a mixed-colony plate displaying a broad spectrum of colony colors. Distinct gRNA expression plasmids combined with the same RGB expression plasmid yielded diverse color outputs across colonies. d Overview of the workflow for color-based colony identification and sequencing: colonies with varying colors were isolated, and barcode DNA fragments were amplified by PCR from the gRNA expression plasmids, followed by sequencing to determine the expression ratio. e Relationship between colony color and relative expression combinations of the three fluorescent proteins. The abundance of each gRNA was controlled by the strength of the corresponding T7 promoter mutants, as illustrated in Fig. 2h. f Artistic rendering constructed by arraying yeast colonies exhibiting different colors. g Three-dimensional scatter plot illustrating the distribution of colony colors in the RGB space. Each point represents one biologically independent colony, plotted according to the normalized fluorescence intensity, in arbitrary units, of sfGFP, mCherry and TagBFP measured using a microplate reader; n = 1004 colonies. Point colors were computed from Commission Internationale de l’Éclairage (CIE) color coordinates using MATLAB. h Two-dimensional projections of the colony color coverage in the RGB space, visualized by collapsing the three-dimensional data in (g) onto the principal planes. i Functional validation of the CAPO system in regulating endogenous transcriptional activity. Upper panel, the positions targeted by different gRNAs; lower panel: the resulting expression levels of the fluorescent protein following regulation by these distinct gRNAs. Data were presented as mean ± SD from n = 6 biologically independent colonies per group. Significance were analyzed by one-way ANOVA with two-sided Dunnett’s multiple-comparisons test. **** denotes P < 0.0001 for all comparisons. Source data are provided as a Source Data file. To explore the color boundary achievable with the CAPO system-based RGB platform, quantitative mapping of the generated color space was performed using 1000 independent colonies. The fluorescence intensities of sfGFP (λex 485/λem 510 nm), mCherry (λex 587/λem 610 nm), and TagBFP (λex 402/λem 457 nm) were measured using a microplate reader, with each channel serving as an orthogonal measurement axis. All fluorescence signals were normalized to their respective maximum values (Fig. 3g). Given the non-uniformity of Euclidean distances in the RGB space, the effective number of distinguishable colors was assessed by mapping the RGB values of each colony to the CIELAB color space and calculating their Euclidean distances. Colors were considered distinct if their ΔE exceeded 2.3, which corresponds to the threshold for human visual discrimination. Among the 1004 unique colonies, 877 distinguishable color variants were identified, demonstrating that a high level of color diversity was captured within the colony collection. (Fig. 3g). This demonstrates that the CAPO system can generate an RGB color space with primary color discrimination comparable to entry-level display technologies, highlighting its precise regulatory capacity. To further investigate the distribution of the RGB color space, a three-dimensional plot was projected onto a two-dimensional plane (Fig. 3h). Using the normalized coordinate (0.5, 0.5) as the center, the plot was divided into four quadrants. The majority of colony data points were clustered in the region where both x < 0.5 and y < 0.5, whereas fewer colonies populated the other areas. This suggests that when all three fluorescent proteins are expressed simultaneously, competition for the cell resources leads to reduced absolute expression levels of each of the proteins. However, relative protein expression remained robust due to precise gRNA-mediated regulation (Fig. 3e). Following the quantitative analysis of exogenous gene expression capacity via the RGB color space of CAPO system, the regulatory ability of CAPO over endogenous gene circuits in eukaryotes was further investigated. A chromosomally integrated, strong constitutive promoter TDH3 driving mCherry expression was used for proof-of-concept validation. gRNA target sites were designed both upstream and downstream of the transcription start site (TSS) to emulate the functions of CRISPRa and CRISPR interference (CRISPRi) (Fig. 3i). Among the designed gRNAs, four for CRISPRa and three for CRISPRi, three successfully elicited the expected regulatory effects, with relative positions to the TSS consistent with previous studies. These results provide preliminary evidence that the CAPO system can modulate endogenous gene expression, thereby expanding its range of potential applications in eukaryotic cell engineering45. Application of the CAPO system in metabolic pathway engineering To illustrate the capacity of CAPO in systematic metabolic pathway optimization, the lycopene biosynthesis pathway was selected as a validation framework. S. cerevisiae inherently possesses the mevalonate pathway, which produces central precursors for lycopene synthesis (Fig. 4a). To eliminate the rate-limiting effect of HMG1 in the mevalonate pathway31,34,46 and facilitate a focused analysis of the heterologous enzymes, a truncated variant of HMG1 was overexpressed. This approach elevates intracellular IPP/DMAPP levels and minimizes the confounding influence of endogenous regulatory feedback, thereby allowing for an accurate functional assessment of downstream components. Capitalizing on this endogenous metabolic foundation, the CAPO system was employed to drive the expression of three heterologous genes, crtE, crtB, and crtI, for lycopene synthesis, each expressed from CAPT, CAPA, and CAPB, respectively (Fig. 4b and Supplementary Data 2). Subsequently, all expression cassettes were assembled to generate the plasmid pYJ1168 and genomically integrated into the S. cerevisiae SA1236 strain to develop the YJ1168 strain, which served as a platform for metabolic applications targeting lycopene biosynthesis (Supplementary Data 3). Precise tuning of crtE, crtB, and crtI expression levels was achieved by modulating gRNA expression levels using the CAPO framework. To intuitively illustrate the expression levels of each gene, the rank number of T7 promoter mutants used for the corresponding gRNA expression was arranged in sequence following R (refer to rank combination), with a smaller value indicating a higher intensity for each T7 promoter mutant (Fig. 4b). Rapid optimization of lycopene and 3-hydroxypropionic acid biosynthesis in S. cerevisiae enabled by the CAPO system. a Schematic representation of the lycopene and 3-hydroxypropionic acid (3-HP) biosynthetic pathway in yeast. Exogenous enzymes required for lycopene and 3-HP production are labeled in red and gray, endogenous enzymes responsible for precursor synthesis are labeled in green, and endogenous enzymes that mediate branch point metabolism are labeled in blue. b Strategy for rational screening of key exogenous genes using the CAPO system. Individual exogenous pathway genes were selectively upregulated using programmable CAP promoters to identify putative rate-limiting steps in a single strain background. c YJ1168 Strain construction and Lycopene yield after elevated expression of individual exogenous pathway genes by gRNA plasmids. d Schematic of random screening for optimal lycopene expression using the CAPO system. After identifying CrtI as the critical enzyme, crtI expression was maximized, and crtE and crtB expression levels were determined using a T7 promoter library for random screening. e Combinatorial screening of CrtE and CrtB expression levels. f YJ1464 Strain construction and 3-HP yield after elevated expression of individual exogenous pathway genes by gRNA plasmids. g Heatmap of 3-HP titers from a focused gRNA-expression plasmid library constructed based on the diagnostic screen in (f). Forty-eight randomly selected colonies were cultivated in 48-well plates and quantitatively analysed. h Validation of top candidates from the focused screen. i Strategies for further enhancement of CrtI expression to promote lycopene biosynthesis. j Enhancement of lycopene production by tuning endogenous gene expression using the CAPO system. For quantitative bar plots in (c, e, f, h–j), data were presented as mean ± SD; error bars indicate SD. Each dot represents one colony-derived culture from a biologically independent colony. For (c, e, i, j), n = 3; for (f, h), n = 4. In (g), each heatmap cell represents one randomly selected colony, with n = 48 colonies. Statistical significance was assessed by one-way ANOVA followed by two-sided Dunnett’s multiple-comparisons test; different letters indicate statistically significant differences at P < 0.05. Source data and exact P value are provided as a Source Data file. Given that the resulting number of potential combinations for three genes, each with ten distinct expression intensities, exceeded the capacity of our transformation efficiency, this may have resulted in the inability to obtain the optimal combination. Consequently, we used a hierarchical optimization strategy rather than a blind full-factorial screen to identify the rate-limiting enzyme among these three proteins. The expression of crtE, crtB, and crtI was manipulated such that one gene was controlled by a gRNA driven by the wild-type T7 promoter (T7 promoter Rank 2, PT7 R2) at high intensity, whereas the other two were regulated by gRNAs expressed from the moderate-strength PT7 R6. A control was established with all three genes controlled by gRNAs expressed from PT7 R6 (Fig. 4b). The results revealed that only the strain with elevated CrtI expression (R662) exhibited a significant increase in lycopene production compared to the control strain (R666), indicating that CrtI is the rate-limiting enzyme (Fig. 4c and Supplementary Fig. 8). Subsequently, crtI expression was maximized by gRNA expressed from the hyperactive PT7R1, while a comprehensive exploration of crtE and crtB expression ratios was performed utilizing the complete T7 promoter library (PT7R1 to PT7R10) to express gRNAs for CAPT and CAPA. The gRNA screening plasmid library was then transformed into the YJ1168 strain to screen for maximized lycopene production (Fig. 4d). From the resultant screening plates, 48 single colonies exhibiting visible red pigmentation were selected for cultivation in 24-well plates, and the top 12 gRNA combinations with the highest lycopene production were chosen for sequencing. Following sequencing, representative high-producing gRNA combinations were selected for quantitative analysis, which identified R851 and R351 as the optimal enzyme-expression stoichiometry for lycopene production (Fig. 4e and Supplementary Fig. 9). Because R851 achieved a lycopene production level comparable to that of R351 with a lower overall gRNA expression input, R851 was selected for subsequent experiments. To further evaluate how the designed expression patterns were translated into pathway output, RT-qPCR analysis was performed on representative lycopene strains R222, R662, and R851 (Supplementary Fig. 10). The relative mRNA abundances of crtE, crtB, and crtI were found to be consistent with the preset gRNA ratios in these strains, indicating that the multiplex CAPO system enabled programmable regulation of pathway gene expression at the transcriptional level. Although the lycopene screen demonstrated that CAPO can identify productive enzyme-expression stoichiometries, lycopene provides an intrinsic color phenotype that facilitates colony preselection. To test whether CAPO could also be applied to products without a visual readout, we selected 3-hydroxypropionic acid (3-HP) as a second validation product. 3-HP is a chemically distinct, non-chromogenic organic acid, and its production must be evaluated quantitatively rather than by colony color36. A CAPO-regulated 3-HP chassis strain, YJ1464, was constructed using CAPT, CAPA, and CAPB to express YDFG, BAPAT, and PAND, respectively. In this pathway, PAND supplies β-alanine from aspartate, BAPAT converts β-alanine to malonic semialdehyde, and YDFG reduces malonic semialdehyde to 3-HP (Fig. 4a). Because the 3-HP pathway presented a similar challenge of multigene expression balancing as the lycopene pathway but lacked a visual readout, we applied the previously established reduced diagnostic workflow instead of performing an exhaustive expression search. The four diagnostic gRNA expression plasmids were introduced into the YJ1464 strain to selectively enhance one CAPO-controlled module at a time, while maintaining moderate activation levels in the other two modules (Fig. 4b). Among these, R662, with enhanced PAND expression produced the highest 3-HP titer, indicating that PAND-dependent β-alanine formation is a major limiting step in the engineered pathway (Fig. 4f). Based on this result, we constructed a focused gRNA-expression plasmid library, maintaining PAND activation within a high-expression range, while further exploring the expression balance of the YDFG and BAPAT modules. Specifically, gRNA B was expressed from PT7R1to PT7R4, whereas gRNA T and gRNA A were expressed from PT7R1 to PT7R7. The designed plasmid library was transformed into YJ1464, and 48 colonies were randomly selected from the transformation plates and cultivated in 48-well plates for quantitative analysis (Fig. 4g and Supplementary Fig. 11). Five candidates produced higher 3-HP titers than the best-performing plasmid in the diagnostic screen. A subsequent validation round confirmed R622 and R712 as the top-performing gRNA-expression configurations, achieving ~30% higher 3-HP production than the R222 group (Fig. 4h). These results demonstrate that CAPO is not restricted to visually screenable products but can also facilitate the optimization of non-chromogenic metabolites through quantitative, hierarchically reduced screening. In light of the visual readout property of lycopene and the presence of substantial upstream pathway components, we chose the lycopene system to further investigate whether the insights derived from CAPO screening can guide subsequent pathway engineering beyond promoter-level tuning. Considering that CrtI serves as the rate-limiting enzyme in lycopene synthesis in yeast, we next tested whether further increasing the CrtI expression level could enhance lycopene production. However, the expression of the corresponding gRNA was already driven by the strongest PT7R1, precluding further increases using a more potent one. Consequently, we explored alternative strategies, including introducing an additional expression cassette of crtI driven by CAPB and incorporating an extra copy of gRNA B to activate CAPB. The experimental results revealed that the addition of an extra crtI expression cassette (strain YJ1218) resulted in a significant increase in lycopene yield compared to the original YJ1168 strain when transformed with the R851 combination for gRNA expression (Fig. 4i). Conversely, the inclusion of an additional gRNA B copy did not enhance lycopene production (Fig. 4i), leading us to hypothesize that the gRNA B binding site on CAPB may have been saturated because of its expression from the hyperactive PT7R1. In contrast, the additional crtI expression cassette provided additional binding sites for gRNA B, facilitating further activation of crtI expression. Consequently, the strain with an additional crtI copy carrying the R851 combination was selected for subsequent investigations. After improving the downstream heterologous lycopene module through CAPO-guided expression tuning and crtI copy-number expansion, we next investigated whether lycopene production could be further enhanced by modulating endogenous precursor supply. In S. cerevisiae, over ten genes are involved in the conversion of acetyl-CoA to farnesyl pyrophosphate (FPP) and dimethylallyl pyrophosphate (DMAPP) within the mevalonate pathway. The upregulation of these related genes can theoretically enhance the precursors necessary for lycopene biosynthesis. Conversely, squalene synthase (ERG9) diverts FPP into squalene biosynthesis, potentially affecting lycopene yield. Given that the capacity of the CAPO system to directly activate or repress endogenous promoters has been validated (Fig. 3i), we sought to further increase lycopene yield by regulating endogenous genes using CAPO. Due to PAM-site restrictions and diversity in promoter regulatory structures, conventional CRISPRa strategies typically require extensive gRNA screening to ensure stable activation of target genes17,45. Consequently, in accordance with the principle of upstream activation and downstream inhibition of target genes, we designed 16 activation gRNAs (gRNAa) targeting BTS1, ERG8, ERG10, ERG12, ERG13, ERG19, ERG20, and IDI1, and four inhibitory gRNAs (gRNAi) targeting ERG9 (Supplementary Data 1). These 16 gRNAa and four gRNAi expression cassettes were randomly assembled on a plasmid with a CEN6/ARS4 replication origin, resulting in 64 two-gRNA (g2) combinations. The plasmid mixture was co-transformed with the R851 gRNA combination into the YJ1218 strain for screening. Sixteen colonies with pronounced red pigmentation were selected for HPLC quantification (Supplementary Fig. 12), and plasmids from the top eight lycopene-producing strains were sequenced to identify the gRNA combinations. Notably, two of the five selected combinations exhibited ~20% increase in lycopene yield during the second round of verification conducted in triplicate (Fig. 4j). To evaluate the potential for enhanced production through broader endogenous regulation, gRNA3 and gRNA5 combinations were constructed based on the two most effective g2 plasmids identified in the previous selection round. This approach allowed us to assess whether stacking the most favorable perturbations would yield an additive improvement in lycopene production. HPLC quantification showed that gRNA3 led to a modest 7.1% increase in lycopene accumulation relative to R851, whereas gRNA5 did not significantly improve production (Fig. 4j). This lack of improvement may be attributed to gRNA5 causing an imbalanced expression of endogenous genes, thereby impacting the substrate supply for lycopene biosynthesis. In conclusion, these findings demonstrate that the CAPO system retains the capability to regulate endogenous genes by CRISPRi- and CRISPRa-directed patterns and enables the programmable regulation of heterologous pathway modules when combined with CAPs, thereby expanding its utility for metabolic flux optimization. Evaluation of CAPO capacity for pathway-level ON/OFF control Finally, the scope of CAPO was extended from pathway optimization to encompass higher-order coordination of multiple engineered metabolic modules. The discrete ON/OFF control across a complete heterologous pathway was initially evaluated using the violacein biosynthetic pathway, a chromogenic model in which the production of the final purple pigment requires the coordinated expression of VioABCDE, whereas the interruption of individual enzymatic steps gives rise to distinct, visually distinguishable states (Fig. 5a)13,47. Within the CAPO system, combinatorial gRNA plasmids driven by the T7 promoter were constructed to target orthogonal CRISPR-activated promoters (Fig. 5b). In strain YJ1283, vioA, vioB, vioC, vioD, and vioE were individually placed under the control of these CAP modules (Fig. 5a). In the absence of cognate gRNAs, the engineered pathway remained inactive, resulting in non-pigmented colonies, consistent with the minimal basal activity of the CAP system. When the defined gRNA combinations were introduced, selective activation of individual pathway nodes was achieved, producing distinct colony colors (Fig. 5b). Upon simultaneous expression of all five cognate gRNAs, the complete violacein pathway was activated and the characteristic purple colony phenotype was observed, thereby demonstrating that CAPO can be used as a modular controller for multigene metabolic pathways (Fig. 5c and Supplementary Fig. 13a). Building on this pathway-level programmability, lycopene and 3-HP biosynthetic modules were subsequently integrated into a single yeast strain, YJ1471, to examine whether distinct metabolic outputs could be independently controlled within the same cellular chassis (Fig. 5c). In this design, the lycopene pathway genes were driven by CAPA, CAPB, and CAPT, whereas the 3-HP pathway genes were controlled by CAPC, CAPD, and CAPE. When gRNAs A, B, and T were introduced (plasmid gABT), lycopene production was specifically activated, while the 3-HP module remained at background levels. Conversely, when gRNAs C, D, and E were introduced (plasmid gCDE), 3-HP production was activated, whereas lycopene production was not detected. In the absence of gRNAs, neither lycopene nor 3-HP was produced at an appreciable level (Fig. 5c and Supplementary Fig. 13b). Collectively, these results indicate that CAPO is not restricted to promoter-level tuning or optimization of individual pathways, but can also provide a scalable framework for orthogonal and programmable control of multiple heterologous metabolic modules within a single strain, thereby supporting the development of multifunctional production chassis for multiplex metabolic engineering. Programmable biosynthetic product diversification in a single CAP-integrated chassis strain achieved via gRNA combination switching. a Schematic illustration of the violacein biosynthetic pathway from L-tryptophan to violacein, including the major intermediates formed through sequential catalysis by VioABCDE. In strain YJ1283, vioA, vioB, vioC, vioD, and vioE genes were placed under the control of orthogonal CAP promoters. b Construction of gRNA expression plasmids and programmable activation of the violacein pathway. Individual or combinatorial gRNA arrays were assembled on 2 μm-origin plasmids, and each gRNA was transcribed from a wild-type T7 promoter. Colored circles indicate gRNAs targeting the corresponding CAP promoters. Gray circles indicate spacer sequences, and the cross indicates no pathway activation in the empty-vector (EV) control. Different gRNA combinations activated distinct subsets of vioA–vioE, producing characteristic colony pigmentation patterns and demonstrating programmable control of discrete violacein pathway states. c Construction and expression performance of the integrated lycopene/3-HP production strain. In strain YJ1471, the lycopene pathway genes were driven by CAPA, CAPB, and CAPT, whereas the 3-HP pathway genes were driven by CAPC, CAPD, and CAPE. The introduction of gABT specifically activated lycopene production, while only a background-level 3-HP signal was observed. In contrast, the introduction of gCDE activated 3-HP production without detectable lycopene. In the empty vector control, neither product was produced at an appreciable level. The right panel shows product-associated chromatographic peak areas, and the lower-right images show the corresponding liquid-culture phenotypes under different plasmid conditions. Data are presented as mean ± SD from n = 3 biologically independent colonies per group. Significance was assessed by one-way ANOVA followed by two-sided Dunnett’s multiple-comparisons test, compared to the EV group; ****P < 0.0001. **P = 0.0034. Source data are provided as a Source Data file. Discussion Eukaryotic promoters possess intricate cis-elements influenced by transcription factor availability that govern the strength and timing of gene expression, complicating the avoidance of interference from host physiology and resulting in unpredictable outputs. To address these limitations, researchers have designed eukaryotic promoters by integrating artificial or native UASs with core sequences, thereby generating synthetic promoters with shortened sequences, inducible properties, or varying strengths7,48–50. However, in these eukaryotic promoter systems, the promoter sequence and its encoded expression strength remain intrinsically coupled, so that strength control is embedded directly in the promoter architecture and repeated reconstruction of expression cassettes is required when gene expression levels need to be adjusted (Fig. 6a). In this study, a CAPO expression system was developed that enables high-level, tunable, and orthogonal multigene expression in yeast. In this architecture, transcriptional activation is structurally decoupled from promoter regulation by employing CRISPR-activated promoters, which are responsive solely to a dCas9-VPR activator, whose availability is controlled by gRNAs transcribed by orthogonal T7 RNAP, so the promoter–gene cassette functions as an expression module that is separated from the strength-control module encoded by a regulatory axis involving T7 RNAP, gRNA, and dCas9-VPR. Comparison between endogenous eukaryotic transcription systems and CAPO transcription systems. a Comparison of endogenous eukaryotic promoters and CRISPR-activated promoters in terms of transcriptional regulation. Endogenous promoters rely on distinct native transcriptional activators, which limits the precision of independent modulation of gene expression strengths; regulation typically requires promoter exchange to adjust target gene expression levels. In contrast, CRISPR-activated promoters universally recruit dCas9-VPR as a shared artificial transcriptional activator, allowing the independent regulation of promoter strength by varying the concentration of gRNAs specific to each promoter. b A schematic comparison of strain optimization workflows for target product synthesis is presented via the design-build-test-learn (DBTL) cycle. Traditional approaches require promoter swapping for each gene to maximize the production of target metabolites. This methodology necessitates the construction and validation of numerous strains to identify optimal expression combinations, rendering it labor-intensive and time-consuming. Conversely, the CAPO approach, which uses CRISPR-activated promoters, enables one-shot strain construction. Subsequent tunable gene expression is achieved by applying a preassembled library of gRNA combinations targeting different CAPs. This method significantly reduces construction costs and accelerates the strain optimization. It was demonstrated that CAPO achieved expression strengths comparable to those of strong native eukaryotic promoters, with the strongest CAPO configuration reaching 1.75-fold of the expression level driven by the native TDH3 promoter (Supplementary Fig. 4). In addition, CAPO provided a 20-fold tunable expression range across the tested gRNA input levels, and expression output showed a direct quantitative association with gRNA abundance (Fig. 2g). By recruiting the host’s endogenous RNA polymerase II machinery for transcription, CAPO bypasses the bottleneck of inefficient translation associated with non-native T7 transcripts, a limitation of systems relying on fusions between T7 RNAP and viral capping enzymes22–26,51. In contrast to conventional strategies that require extensive screening of constitutive promoter libraries to achieve the desired expression levels, the CAPO system enables predictable and continuous modulation of gene expression through combinatorial gRNA plasmid libraries. These libraries are specifically designed to operate with CAPs and, once constructed, can be reused to fine-tune the expression strength of diverse target genes without requiring the construction of additional promoter libraries. Using the same gRNA library framework, lycopene biosynthesis and 3-HP biosynthesis were optimized, resulting in a 3.8-fold yield improvement in lycopene production and a 1.3-fold yield improvement in 3-HP production during rapid pathway screening (Fig. 4e, h). In this framework, the combinatorial design space associated with promoter engineering is reduced, gene expression is insulated from host physiological fluctuations by an orthogonal T7-driven gRNA layer, and regulatory precision, scalability and modularity in eukaryotic systems are enhanced, thereby addressing the limitations of traditional promoter-based platforms, complementing the artificial promoter and combinatorial regulation strategies summarized in the introduction, and facilitating scalable metabolic engineering and genetic circuit design (Fig. 6b). Standard CRISPRa and CRISPRi techniques are effective in modulating endogenous genes; however, their application to multigene pathways often requires complex gRNA design and poses the risk of off-target genomic interactions. Additionally, due to the constraints of PAM sites and target location relative to the transcription start site, identifying suitable gRNA targets within a specific genomic region can be challenging45. CAPO circumvents these challenges by employing episomal expression cassettes and synthetic promoters, providing a self-contained, plug-and-play solution for exogenous gene circuits that minimizes interference with the host genome. This architecture also supports the construction of different metabolic pathways within the same host strain because pathway-specific CAP modules can be installed and selectively activated by cognate gRNAs without repeated chromosomal engineering. Compared with conventional multi-pathway-switching strategies, such as promoter replacement7, inducible-promoter rewiring52, or recombinase-based genetic switches53, CAPO provides a simpler gRNA-defined ON/OFF layer with near-zero basal activity. In the pathway-switching experiment, cognate activation increased the lycopene-associated signal from 58.37 to 4822.26 mAU·s, whereas the non-cognate lycopene signal remained at 95.67 mAU·s, corresponding to only 2.0% of the activated output and an ON/OFF ratio of 50.4-fold. For 3-HP, cognate activation increased the signal from 82.71 to 418.21 mAU·s, while the non-cognate signal remained at 58.63 mAU·s, below the no-gRNA basal background. These results indicate that pathway-specific activation was achieved with limited non-target output, thereby reducing unintended expression from inactive pathway modules and lowering the need for pathway-specific reoptimization (Fig. 5c). The intracellular resource pool available for the expression of a heterologous pathway is inherently limited. Consequently, enhanced production cannot be achieved simply by maximizing the expression of all heterologous genes. Instead, optimal metabolite synthesis requires a balanced allocation of cellular resources between host growth and gene expression, as demonstrated for the lycopene and 3-HP pathways (Fig. 4). The modular and scalable architecture of CAPO has significant implications for the stoichiometric control of multigene expressions. The separation between the transcriptional activator dCas9-VPR and the gRNA-based targeting signal allows for the optimization of these components in a partially independent manner. This design enables straightforward multiplexing, as distinct CAP-gRNA pairs can be deployed to independently regulate multiple genes within a single cell. This level of granular control is important for metabolic engineering, where balancing enzyme expression is essential to maximize product flux and minimize the accumulation of toxic intermediates. Given that the gRNA binding site within the CAP can be designed with any sequence, it is possible to evaluate it using software to minimize crosstalk with the host. However, it was observed that differences in gRNA binding affinity to CAP could cause expression outputs to deviate from those predicted solely by gRNA abundance, particularly under multigene expression conditions (Supplementary Fig. 1). This issue was subsequently addressed through the implementation of a T7DCAP-based screening strategy, which facilitated the identification of several gRNAs with similar affinities to the CAPs (Fig. 2e and Supplementary Fig. 3). Furthermore, gRNA binding affinity may offer an additional regulatory dimension to the CAPO system, where high affinity correlates with elevated expression levels. The concurrent utilization of both T7 RNAP and dCas9 within the CAPO system inherently establishes an AND gate functionality. Extensive research has been conducted on T7 RNAP and dCas9, including split T7 RNAP systems that respond to light or chemical stimuli and the fusion of G protein-coupled receptors (GPCRs) to enable dCas9 to respond to hormonal signals54,55. Integrating these findings with the CAPO system could broaden its activation patterns, thereby allowing the construction of more complex logic gates. Furthermore, given the conserved core principles of transcription across eukaryotes, adaptation of the CAPO system to additional eukaryotic hosts is expected to be feasible, although host-specific optimization of promoter, terminator, and chromatin context will be required. Despite its advantages, the CAPO system presents opportunities for future investigation and optimization. The expression of dCas9-VPR and T7 RNAP imposes a potential metabolic burden on host cells, which may be mitigated by engineering or de novo designing more compact and efficient Cas9-VPR and T7 RNAP orthologs. Integrating core CAPO components into the host genome could also enhance long-term stability and reduce cell-to-cell variability in expression. Looking forward, the CAPO system could be integrated with biosensors to create dynamic feedback control loops that enable the autonomous regulation of metabolic pathways in response to cellular states or environmental cues. In conclusion, the CAPO system constitutes an approach in the field of quantitative gene expression technology for eukaryotes. By effectively integrating the programmable activation of CRISPRa with the orthogonal control of the T7 system, we have developed a platform that addresses the need for both expression strength and regulatory precision. This study paves the way for the rational design and engineering of increasingly complex biological systems in eukaryotic hosts, may support future applications in metabolic engineering and genetic circuit design. Methods Strains and cultivation conditions Escherichia coli JM109 was used for routine cloning and cultivated at 37 °C in Luria Broth medium, which comprised 5 g/L yeast extract, 10 g/L tryptone, 10 g/L NaCl, and 15 g/L agar for solid medium. The medium was supplemented with 50 mg/L 5-Bromo-4-chloro-3-indolyl β-D-galactoside (X-Gal) and 20 μM Isopropyl β-D-thiogalactoside (IPTG) for blue-white screening. Antibiotics were administered at the following concentrations, 50 mg/L carbenicillin, 25 mg/L kanamycin, and 25 mg/L chloramphenicol. Saccharomyces cerevisiae W303a and its recombinant derivatives served as host strains and were cultivated at 30 °C in yeast extract peptone dextrose (YPD) medium, containing 10 g/L yeast extract, 20 g/L peptone, 20 g/L dextrose, 0.04 g/L adenine, and 20 g/L agar for solid medium. Alternatively, synthetic dropout (SD) medium was used, consisting of 6.7 g/L yeast nitrogen base, 20 g/L glucose, appropriate dropout supplements, 0.04 g/L adenine, and 20 g/L agar for solid medium. For induction, glucose in SD medium was replaced with galactose or a suitable combination of glucose and galactose. Plasmid construction The plasmids used in this study are listed in Supplementary Data 4 and were verified by Sanger sequencing before use in further experiments. The 2×Phanta Flash Master Mix (Vazyme, China) and I-5™ 2×High-Fidelity Master Mix (Tsingke, China) were used to amplify DNA fragments. The 2×MultiF Seamless Assembly Mix (ABclonal, China), M5 HiPer pTOPO-Blunt Simple Cloning Kit (Mei5bio, China), Type II restriction endonucleases, and T4 DNA ligase (Thermo Scientific, USA) were used for plasmid construction using Gibson Assembly56, TOPO cloning57, or Golden Gate assembly58,59. Primers and codon-optimized genes were chemically synthesized by Tsingke Biotech Co., Ltd. (Tianjin, China). Golden Gate assembly utilizes both type IIS restriction endonucleases and T4 DNA ligase within a single reaction system, enabling iterative cycles of enzymatic digestion and ligation. To standardize plasmid construction, modular overhangs were designed for promoters, coding sequences (CDSs), terminators, and vector backbones. In addition, specific recognition sites associated with different hierarchical vector levels were incorporated. Such standardization enables modularity, interoperability, and improved usability of genetic elements for downstream applications (Supplementary Fig. 14). All constructions were verified by Sanger sequencing. DNA or amino acid sequences can be found in the Supplementary Data files. Yeast strain construction The yeast strain was constructed via homologous recombination, utilizing S. cerevisiae W303a as the parental strain. The non-replicative plasmid pSA1236, which carries dCas9-VPR, URA3, and T7 RNAP expression cassettes flanked by homologous arms corresponding to the YCRWΔ12 locus on chromosome III of S. cerevisiae, was introduced into the W303a strain to establish the stable strain SA1236 (Supplementary Data 3). Subsequently, plasmids pYJ733, pYJ1283, pYJ1168, pYJ1218, and pYJ1464 were integrated into chromosome XIV at the YNRCΔ9 locus of the yeast strain SA1236 to independently generate strains YJ733, YJ1283, YJ1168, YJ1218, and YJ1464. Plasmid pYJ1471 was integrated into chromosome VIII at the YHRCΔ14 locus of the yeast strain SA1236 to generate strain YJ1471. (Supplementary Data 3 and 4). Yeast transformation The yeast transformation protocol was adapted from the lithium acetate method60. A single yeast colony was inoculated into 5 mL of YPD liquid medium and cultured overnight at 30 °C with shaking at 220 rpm. Subsequently, an appropriate volume of culture (1:50 dilution) was transferred to a fresh YPD medium. After incubation at 30 °C until the OD600 reached ~0.5, the cells were harvested by centrifugation at 1500×g for 5 min at room temperature and washed twice with double-distilled water (ddH₂O). For every 50 mL of initial culture, the cell pellet was resuspended in 1 mL of freshly prepared lithium acetate solution (containing 100 mM LiAc, 10 mM Tris-HCl, and 1 mM EDTA, pH 7.5). A 200 μL aliquot of the lithium acetate-cell suspension was combined with 10 μL each of ssDNA and plasmid mixture (pre-mixed; for genomic integration, plasmids were linearized with restriction enzymes before mixing). Subsequently, 1.2 mL of PEG solution (freshly prepared with 40% PEG4000, 100 mM LiAc, 10 mM Tris-HCl, and 1 mM EDTA, pH 7.5) was added. The transformation mixture was incubated at 30 °C with shaking at 220 rpm for 30 min, followed by a 15-min heat shock at 42 °C. Cells were collected by centrifugation at 4000×g for 5 min and resuspended in 50 μL of TE buffer, containing 10 mM Tris-HCl, 1 mM EDTA, pH 7.5. The suspension was plated onto selective SD agar plates and incubated at 30 °C for approximately 36 h until colonies emerged. Promoter gradient stability assay The promoter gradient stability assay was performed to evaluate the robustness and consistency of gene expression for both native eukaryotic promoters and CAPs (Supplementary Data 1, 4). A series of promoters with predefined activity gradients were cloned upstream of reporter genes (GFP and LacZ) using the standardized Golden Gate cloning assembly method. CAPs regulated by distinct gRNA expression gradients, as well as native eukaryotic promoters, were separately introduced into S. cerevisiae strains SA1236 and W303a. Reporter gene expression was measured according to the protocols described above. Stability of the promoter gradient was assessed by comparing the relative reporter expression among different promoters across biological replicates. Experiments were performed in triplicate. β-Galactosidase activity assay The β-galactosidase activity assay for yeast was modified from the traditional method used to measure LacZ activity in bacteria61. Cell pellets from 2 mL cultures were collected by centrifugation at 12,000×g for 1 min, washed, and resuspended in 2 mL saline buffer (containing 0.9% NaCl). OD600 was measured using saline buffer as a blank, with dilution adjusted to maintain values between 0.3 and 0.8. To initiate the reaction, 200 μL of the cell suspension was mixed with 800 μL of saline in 10 mL tubes, alongside a reference tube containing 1 mL of saline. Then, 1 mL of freshly prepared Z buffer (composed of 60 mM Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, and 1 mM MgSO4), supplemented with 50 mM β-mercaptoethanol, 0.1% SDS, and 40 μL CHCl₃ were added to all tubes. This was followed by vortexing for 10 s and a 5-min incubation at 28 °C. Next, 400 μL of substrate O-nitrophenyl β-D-galactopyranoside (ONPG, 4 mg/mL in Z buffer) was added to each tube at 15-s intervals, and the reactions were allowed to proceed at 28 °C until a visible yellow coloration appeared. The reaction was terminated by sequentially adding 1 mL of 1 M Na₂CO₃ at 15-s intervals. Finally, OD420 and OD550 were measured using the reference tube as a blank, and reactions with OD420 values outside the 0.2–0.8 range were re-analyzed using optimized incubation times. The activity was quantified in Miller units and calculated using the formula, Miller units = (OD420 − 1.75 × OD550)/(t × v × OD600) × 1000, where t represents the reaction time (minutes), v represents the volume of yeast solution in the reaction system (mL), OD600 represents the absorbance of the original yeast solution at 600 nm, and 1.75 × OD550 is corrected light scattering. Quantitative and visual assay of fluorescence The multiplexed regulatory capability of our system was validated using a tri-fluorescence assay employing sfGFP, mCherry, and TagBFP as reporters, which enabled the visual and quantitative assessment of exogenous gene expression ratios. For fluorescence visualization, yeast cells transformed with a mixture of gRNA expression cassettes for sfGFP, mCherry, and TagBFP activation were plated on solid SD medium with 1% glucose and 1% galactose. After incubation at 30 °C for approximately 40 h, plates with visible colonies were captured using an SH-Compact 523 (HISUMA) with the following optical configurations: mCherry (excitation 530 nm/emission filter 590 nm), sfGFP (excitation 470 nm/emission filter 530 nm), and TagBFP (excitation 400 nm/emission filter 470 nm). The final image of the colorful colonies was generated by merging pictures captured from different fluorescence parameters using ImageJ. To quantitatively assess fluorescence intensity, yeast strains harboring the respective plasmids were inoculated into 400 μL SD medium in a 48-well plate and incubated with shaking at 30 °C. Subsequently, the output fluorescence signals were monitored every 10 min for 48 h using a BioTek Synergy H1 microplate reader with the following parameters: green fluorescence (sfGFP, excitation 485 nm/emission 528 nm), red fluorescence (mCherry, excitation 570 nm/emission 610 nm), and blue fluorescence (TagBFP, excitation 402 nm/emission 457 nm). The maximum ratio of the fluorescence signal to OD600 was selected to represent the output for further analysis. Extraction and analysis of lycopene and 3-hydroxypropionic acid To extract lycopene, 500 μL of the yeast cell culture was collected by centrifugation at 12,000×g for 5 min. The supernatant was discarded, and the resulting pellet was resuspended in 500 μL of acetone. Subsequently, glass beads with a diameter of 500 μm, occupying one-third of the tube volume, were added. The tube was then placed into the adapter of a TissueLyser (Shanghai Jingxin), which had been precooled with liquid nitrogen. The cells were disrupted at a frequency of 70 Hz for 300 s to facilitate lycopene extraction. The acetone phase containing the extracted lycopene was filtered and analyzed using high-performance liquid chromatography (HPLC). Lycopene signals were detected at a wavelength of 471 nm by an HPLC system (Agilent 1260) equipped with a C18 column (4.6 × 150 mm, 5 μm) and a UV/VIS detector, utilizing methanol/acetonitrile/dichloromethane (21:21:8, v/v) as the mobile phase at a flow rate of 1 mL/min at 30 °C62. For 3-hydroxypropionic acid quantification, 1 mL of yeast culture broth was directly filtered through a 0.22-μm membrane filter and transferred to an HPLC sample vial. A 25 μL aliquot of each prepared sample was injected and analyzed over a 30-min run on an Aminex HPX-87H ion exclusion column (Bio-Rad) at 60 °C, using 1 mM H₂SO₄ as the mobile phase at a flow rate of 0.6 mL/min. The concentration of 3-HP was determined by monitoring the UV absorbance at 220 nm. Yeast RNA extraction and quantitative real-time PCR Yeast cells cultured overnight in YPD or SD medium were subcultured into the same medium at an OD600 of 0.2–0.3. Cells were collected by centrifugation, and total RNA was extracted using a Yeast RNA Kit (Omega, R6870). The RNA was then used to synthesize cDNA using the PrimeScript™ RT Reagent Kit (Takara, RR037A). Quantitative real-time PCR (RT-qPCR) was performed using Taq Pro Universal SYBR qPCR Master Mix (Vazyme, Q712). The primers used for quantifying the genes of interest, along with the internal controls, are listed in Supplementary Data 1. The results were analyzed using the 2−ΔΔCT method and are presented as relative changes in the expression of target genes63. Statistics and reproducibility Unless otherwise stated, biological replicates were independent cultures initiated from separately picked colonies derived from the same transformation. For each experiment, measurements were confirmed in three or more independent repeat experiments performed on different days from fresh starts, as specified in the corresponding figure legends. Data were shown as the mean ± SD. Statistical significance was determined using one-way ANOVA followed by two-sided Šídák’s multiple-comparisons test or one-way ANOVA followed by two-sided Dunnett’s multiple-comparisons test. The statistical tests and data compositions used for each figure are noted in the corresponding figure legend. Significance is indicated as follows: *P < 0.05, **P < 0.01, ***P < 0.0001, and ****P < 0.0001. All exact P values are provided in the figures, figure legends or Source Data files. Genetic design and sequence reading were performed by SnapGene (v6.0.2). Three-dimensional RGB color space figure (Fig. 3g, h) was analyzed and graphed using MATLAB (R2022a). All data were analyzed and graphed by GraphPad Prism v8, and Microsoft 2019 (Word, Excel, PowerPoint). 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. Supplementary information Supplementary Information Description of Additional Supplementary Information Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Reporting Summary Transparent Peer Review file Source data Source Data