ERAD-level gene switches for on-demand protein secretion and rapidly controlled gene therapies Authors: Ting Gao, Shichao Li, Yuxuan Fan, Wenyi Yan, Minghui He, Qihao Zhang, Yilin Li, Qi Liu, Pengli Wang, Jian Jiang, Lihang Zhang, Yuting Zhou, Liuqi Zhao, Qiqi Xiong, Yuhang Wu, Sichen Yuan, Shaocong Fang, Hongyun Tang, Martin Fussenegger, Hui Wang, Jiawei Shao, Mingqi Xie Published: 2026 ABSTRACT On-demand regulation of therapeutic activities is critical to broaden the clinical utility of gene therapies. Here, we employ an Endoplasmic Reticulum-Associated Degradation (ERAD)-centered strategy to develop trigger-inducible gene switches tailor-designed for rapid protein secretion. By temporarily attaching conditional degron motifs to various therapeutic proteins of interest (POIs), inducible, tunable and reversible control of native POI secretion into the bloodstream was achieved through oral administration of specific clinically licensed small-molecule drugs. This simple and generalizable design principle is highly compatible with adeno-associated virus (AAV)-mediated gene delivery, enabling long-term and remote-controlled transgene expression in male mice in vivo. To showcase potential therapeutic benefits of ERAD-level gene switches, we describe gene therapy approaches for cardiovascular diseases and chronic pain - two classes of disease treatments that may most urgently require on-demand drug actions. This study achieves dose- and time-dependent control of transgene activities without requiring the design of overly complex gene circuits, and may form important basis to move synthetic biology-based regulation systems towards therapeutic usage. Introduction A long-lasting goal of gene therapies and modern medicine is the capability to treat various intractable diseases through “smartened therapeutics” created on the basis of genetic engineering1. Therefore, regulation strategies allowing for precisely controlled therapeutic activities with pre-programmed time- and context-specificity are critical to expand the clinical utility of next-generation gene therapies2,3. However, most gene therapy products currently approved for clinical usage can only provide constitutive transgene activities when introduced in vivo4–11, rendering treatment compatibility generally confined to diseases that arise from particular gene defects12,13. To treat more complex diseases that require precisely dosed, accurately timed and feedback-controlled therapeutic activities according to instant metabolic and cellular states, a multitude of trigger-inducible gene switches responsive to light14–16, ultrasonic waves17, electrical signals18, temperature19,20, disease markers21–23, food additives2,24 or clinically licensed drugs25–27were engineered in recent years. Although all these gene switches enable robust and reliable remote control of transgene expression in diverse preclinical models, the underlying regulation mechanisms remain at the conventional gene expression stages of transcription16,28, translation15,25,29, and RNA splicing26,30. As a consequence, several hours are typically required to observe therapeutic protein secretion upon stimulation, which restricts treatment eligibility to chronic diseases such as type-2 diabetes14 or cancer22. To treat acute pathological conditions that strictly call for on-demand drug action, novel regulation systems operating at faster timescales are therefore required. To engineer rapid-acting gene switches capable of approaching the goal of instantaneous protein secretion, various major post-translational engineering strategies have emerged in recent years. For example, target proteins can be first retained in the ER by fusion to a conditional aggregation domain (CAD), allowing the presence of a small-molecule drug to activate protein secretion by selectively inhibiting CAD-mediated self-aggregation31. Similarly, the protein of interest (POI) can be sequestered inside the endoplasmic reticulum (ER) through a Tobacco etch virus (TEV) protease-cleavable ER anchor fused to the POI preprotein, allowing the presence of full-length32 or inducible protease variants33–36 to mediate removal of the ER anchor and concomitant release of the POI into the secretory pathway. Using these ER-centered strategies, protein secretion rates were increased to reach the timescale of several minutes, with significant therapeutic effects being recorded within an hour after exposure to the trigger signal(s) in vivo35. Even faster timescales were achieved through synthetic exocytosis-centered strategies. Because insulin exocytosis in pancreatic β-cells is naturally regulated by spontaneous bursts of intracellular calcium ions (Ca2+), various trigger-inducible insulin secretion strategies were assembled through ectopic overexpression of voltage- or ligand-gated calcium channels in β-cell lines18,19, ultimately achieving real-time control of blood glucose levels in diabetic mice18. When moving these strategies towards clinically relevant settings, however, the exocytosis strategy may be technically limited to cell-based delivery relying on specialized host cell lines that originate from the neuroendocrine lineage, which are either in short supply or may further complicate regulatory approval37. The ER retention strategy is compatible with both cell-based and in vivo gene therapy delivery, but the reliance on exogenous proteases may require additional and carefully arranged experimental investigation to exclude potential safety concerns when moving towards therapeutic usage in humans38. Therefore, engineering of simple, gene therapy-compatible, and ideally protease-free switches may be highly desired to enable on-demand transgene action within clinically eligible settings. To close this gap, we speculate that the Endoplasmic Reticulum-Associated Degradation (ERAD) pathway of mammalian cells may be an attractive accession point to realize rapid-acting and “switchable” behaviors of protein secretion. ERAD plays a central role in governing quality control of protein expression and is responsible for rapidly identifying and eliminating misfolded proteins from the secretory pathway with high fidelity39,40. Thus, user-defined control over protein secretion may be achieved by directly fusing conditional degron motifs to the POI, whose properly folded and misfolded configuration(s) can be precisely monitored by specific control signals. Compared to other ER-centered strategies such as drug-31 or protease-repressible retention35,36, this design ensures that no target proteins are stably expressed in the basal state, thus posing minimal burdens on ER physiology and cellular metabolism. Functionally, we were able to achieve trigger-inducible control over POI secretion, showing comparable kinetic profiles to ER-retention systems and exocytosis, while ensuring similar expression strengths typically seen with constitutive expression strategies. Furthermore, although we experimentally show that fusion of conditional degrons may not significantly distort the functionality of most POIs, we carefully designed different secretion strategies by introducing furin-cleavable sites into the linker domains to ensure that POIs are secreted in its native form upon stimulation. Using a Trimethoprim (TMP)-controlled system for proof-of-concept41, we further show how ERAD-level gene switches can be readily installed onto AAV vectors to enable long-lasting and reversible expression of transgenes with close to 4 kb in size in vivo, thus offering the perspective of using FDA-approved small molecules to regulate different types of therapeutic proteins in different organs. Following AAV-mediated delivery into the liver of male mice, ERAD-level gene switches enabled reversible control of TMP-stimulated transgene expression over a year without showing significant adverse effects on the overall liver physiology or ER stress. This TMP-regulated gene switch was also transferred into muscle- and central nervous system (CNS)-selective AAV vectors to enable trigger-inducible secretion of therapeutics treating cardiac abnormalities and chronic pain - two major classes of diseases that may most benefit from on-demand drug actions. Taken together, this work establishes the mammalian ERAD pathway as an attractive accession point to design trigger-inducible gene switches, describes a hitherto fastest gene therapy approach for therapeutic transgene delivery, and discusses the types of diseases that may be effectively treated with gene therapies that provide rapid therapeutic responses. Results Design and validation of ERAD-level gene switches in mammalian cells Intuitively, regulated exocytosis is regarded as the fastest means of protein secretion in human cells42. In pancreatic β-cells, for example, insulin is stored within secretory granules following protein synthesis, and is released into the extracellular space within a few minutes upon stimulation by Ca2+ entry from glucose-dependent membrane depolarization43. Indeed, when we stimulated a monoclonal β-cell line stably transgenic for co-secretion of insulin and Oplophorus gracilirostris luciferase (NanoLuc) in a 1:1 stoichometry18, a steep rise in protein production was observed within 2 min after exposure to 60 mM potassium chloride (KCl) (Fig. 1a). However, this accumulation in protein secretion was almost completely attenuated within 2 h, and showed relatively low absolute expression levels upon reaching saturating states (Fig. 1a). As an alternative, we hypothesize that constitutive expression systems may also achieve sufficiently fast secretion rates at steady-state to provide supply for in situ production of therapeutic transgenes. In fact, when we recorded the kinetic profiles of NanoLuc secretion using a constitutive expression vector in non-endocrine human HEK-293 cells, we noticed that protein secretion triggered by exocytosis was indeed faster within the first 2–5 min upon stimulation, but the secretion capacity of constitutive expression systems could be precisely adjusted by their transfected plasmid amounts to match exocytosis-like increases of secreted protein amounts recorded per time unit (Fig. 1a). Notably, we show that constitutive secretion regimen, as soon as fully activated, are able to yield higher absolute expression levels over an extended and therapeutically relevant timescale of 1–2 h (Fig. 1a). Therefore, the design of a gene switch that may rapidly toggle between constitutively active (ON) and inactive states of protein secretion (OFF) could be instrumental for on-demand transgene expression from a practical perspective.Fig. 1Design principles of an ERAD-level gene switch.a Comparative analysis of protein secretion by constitutive expression and regulated exocytosis. Time course of NanoLuc secretion from HEK-293 cells with a constitutive NanoLuc expression vector. NanoLuc secretion from stable ElectroBeta cells treated with KCl was measured as a positive control. NanoLuc levels were normalized against cell numbers. b Molecular mechanisms of ERAD-controlled protein secretion. In the OFF state, degron-tagged proteins are degraded. Upon ERAD inhibitor treatment or signal-induced stabilization (ON), protein of interest (POI) secretion is restored. c Comparative analysis of constitutive degradation efficiencies. HEK-293 cells were transfected with constitutive expression vectors for different degron-tagged NanoLuc constructs before NanoLuc levels were measured after 36 h. d Stimulation of target protein secretion by various chemical inhibitors of protein degradation. NanoLuc secretion from HEK-293 cells expressing ecDHFRN18T, A19V-tagged NanoLuc (pGT69) after treatment with DMSO, CB-5083, CP26, NMS-873, MG132 or CQ. e Kinetics of CB-5083-inducible NanoLuc secretion. Time course of NanoLuc secretion from HEK-293 cells transfected with pGT69 after exposure to CB-5083. f CB-5083 concentration-dependent impact on gene expression and cytotoxicity in mammalian cells. HEK-293 cells were transfected with pGT69 or native untagged secreted alkaline phosphatase SEAP (white, pSEAP2-Control), and NanoLuc and SEAP levels in culture supernatants were measured at 6 h after CB-5083 exposure. g Optimization of Trimethoprim (TMP)-inducible protein secretion. NanoLuc secretion from HEK-293 cells transfected with different amounts of pGT69 and cultured with 0 or 40 μM TMP. Numbers above bars indicate fold changes of NanoLuc expression between TMP-treated cells and cells treated with DMSO (vehicle control). Data are presented as mean ± SD; n = 3 (a, c, g) or n = 4 (d, e, f) independent experiments; circles show individual results. Statistical significance was determined using one-way ANOVA with Dunnett’s multiple comparison test (d, f) or two-way ANOVA with Bonferroni’s multiple comparison test (e). The p values indicate the significance of differences in the mean values versus the non-induced group (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); ns, not significant (p > 0.05). “Const.” in (a) denotes constitutive. Experimental details are shown in Supplementary Data 4. Source data are provided as a Source Data file. Design principles of an ERAD-level gene switch. a Comparative analysis of protein secretion by constitutive expression and regulated exocytosis. Time course of NanoLuc secretion from HEK-293 cells with a constitutive NanoLuc expression vector. NanoLuc secretion from stable ElectroBeta cells treated with KCl was measured as a positive control. NanoLuc levels were normalized against cell numbers. b Molecular mechanisms of ERAD-controlled protein secretion. In the OFF state, degron-tagged proteins are degraded. Upon ERAD inhibitor treatment or signal-induced stabilization (ON), protein of interest (POI) secretion is restored. c Comparative analysis of constitutive degradation efficiencies. HEK-293 cells were transfected with constitutive expression vectors for different degron-tagged NanoLuc constructs before NanoLuc levels were measured after 36 h. d Stimulation of target protein secretion by various chemical inhibitors of protein degradation. NanoLuc secretion from HEK-293 cells expressing ecDHFRN18T, A19V-tagged NanoLuc (pGT69) after treatment with DMSO, CB-5083, CP26, NMS-873, MG132 or CQ. e Kinetics of CB-5083-inducible NanoLuc secretion. Time course of NanoLuc secretion from HEK-293 cells transfected with pGT69 after exposure to CB-5083. f CB-5083 concentration-dependent impact on gene expression and cytotoxicity in mammalian cells. HEK-293 cells were transfected with pGT69 or native untagged secreted alkaline phosphatase SEAP (white, pSEAP2-Control), and NanoLuc and SEAP levels in culture supernatants were measured at 6 h after CB-5083 exposure. g Optimization of Trimethoprim (TMP)-inducible protein secretion. NanoLuc secretion from HEK-293 cells transfected with different amounts of pGT69 and cultured with 0 or 40 μM TMP. Numbers above bars indicate fold changes of NanoLuc expression between TMP-treated cells and cells treated with DMSO (vehicle control). Data are presented as mean ± SD; n = 3 (a, c, g) or n = 4 (d, e, f) independent experiments; circles show individual results. Statistical significance was determined using one-way ANOVA with Dunnett’s multiple comparison test (d, f) or two-way ANOVA with Bonferroni’s multiple comparison test (e). The p values indicate the significance of differences in the mean values versus the non-induced group (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); ns, not significant (p > 0.05). “Const.” in (a) denotes constitutive. Experimental details are shown in Supplementary Data 4. Source data are provided as a Source Data file. To gain control over constitutive protein secretion, the mammalian ER-associated degradation (ERAD) pathway may be an attractive accession point. When proteins enter the secretory pathway, ERAD identifies and targets misfolded proteins for immediate removal from the endoplasmic reticulum (ER) and initiation of proteasomal degradation40. Thus, manipulation of the ERAD machinery could be decisive to rapidly control whether constitutively expressed proteins shall eventually complete the secretory pathway (ON), or be effectively discarded from the cells at this stage (OFF) (Fig. 1b). To test this hypothesis, we fused a broad selection of well-established degron signals to the N- or C-terminus of the target protein NanoLuc, expecting that the degron motif may use its misfolded or intrinsically disordered signature to exploit ERAD-mediated removal of the entire fusion protein (Fig. 1c). Among all candidates tested, such as ubiquitin tags44, hPEST signals, a self-excising Hepatitis C virus derived NS3/NS4A complex45, or mutants of E. coli-derived or murine dihydrofolate reductase (DHFR), ecDHFRN18T, A19V showed strongest constitutive degradation efficiency (Fig. 1c). ERAD-involvement was confirmed by shRNA-mediated knockdown of characteristic genes such as human synoviolin1 (SYVN1, also known as HRD1) (Supplementary Fig. 1a, b) or when exposing the transfected cells to selective inhibitors of VCP/p97 such as NMS-873 or CB-5083, or the Hrd1-specific CP26 (Fig. 1d and Supplementary Fig. 1c). Similarly, when we tested different types of chemical inhibitors of eukaryotic protein degradation, only CB-5083, NMS-873 and CP26 accounted for significant upregulation of ecDHFRN18T,A19V-tagged NanoLuc expression (Fig. 1d and Supplementary Fig. 1c), while no difference in NanoLuc secretion was observed when cells were exposed to proteasome inhibitors such as MG132 (Fig. 1d and Supplementary Fig. 1d) or the ER-phagy inhibitor chloroquine (Fig. 1d). As anticipated, ERAD-dependent protein secretion was indeed operating at very fast timescales, with stimulation of NanoLuc secretion being recorded within 30 min of CB-5083 exposure (Fig. 1e). There was also a correlation between absolute NanoLuc secretion levels and the administered CB-5083 dose (Fig. 1f), though the concentration range of CB-5083 used in our experiments appeared to significantly compromise cell physiology and overall secretion capacity of the mammalian host cell (Fig. 1f and Supplementary Fig. 1e). Therefore, to engineer ERAD-centered gene switches of real clinical utility, the use of non-toxic and patient-compliant trigger compounds (other than CB-5083) must be preferred (Fig. 1b). Trimethoprim (TMP)-inducible protein secretion in mammalian cells ecDHFRN18T, A19V, like most E. coli-derived dihydrofolate reductase (ecDHFR) mutants, are known to respond to the FDA-approved drugs Trimethoprim (TMP) and Methotrexate (MTX) as chemical control compounds that protect ecDHFR from proteasomal degradation41,46. Therefore, based on our results showing that ecDHFRN18T, A19V could effectively engage a constantly active ERAD pathway and enable trigger-inducible protein secretion (Fig.1d-f), we sought to use ecDHFRN18T, A19V as the degron module within the frame of an ERAD-centered gene switch tailor-designed for therapeutic usage (Fig. 1b). Hence, in the basal state, ecDHFRN18T, A19V-tagged target proteins of interest (POIs) may be spontaneously degraded by the cell’s endogenous ERAD pathway, until the presence of TMP or MTX promotes POI secretion by rapidly protecting ecDHFRN18T, A19V from ERAD-mediated degradation (Supplementary Fig. 1f). In fact, when we constitutively expressed ecDHFRN18T, A19V-tagged NanoLuc in mammalian cells, regulation performances of over 1000-fold in TMP-stimulated NanoLuc expression were achieved, with basal secretion levels remaining constantly low even when increasing the plasmid levels of the expression vector (Fig. 1g). We also compared NanoLuc expression levels between intracellular and secreted fractions either under basal or TMP-stimulated conditions by using an arbitrary NanoLuc-Coh2 fusion protein as a positive control indicative for maximal secretion capacity, high molecular stability as well as minimal accessibility to ERAD-mediated degradation (Supplementary Fig. 2a). Results in HEK-293 cells suggest that most target proteins must have been degraded in the basal state, while TMP could “rescue” approximately 10% of proteins for secretion in their bioactive forms (Supplementary Fig. 2a). Though, as we were able to show that intracellular NanoLuc not rescued by TMP still had some enzymatic activity (Supplementary Fig. 2a), we hypothesize that fusion to the ecDHFRN18T, A19V will not cause misfolding of the entire target protein, where ERAD is primarily engaged by an intrinsic mechanism specific to ecDHFRN18T, A19V (Fig. 1b). In contrast to CB-5083 (Fig. 1f and Supplementary Fig. 1e), TMP also showed almost no signs of cytotoxicity either in the short or longer term (Supplementary Fig. 2b). Also, no significant impact on human folate metabolism was detected after RNA sequencing and gene set enrichment analysis (GSEA) of adipose-derived mesenchymal stromal cells cultivated in TMP-containing medium (Supplementary Fig. 2c). To directly demonstrate ERAD-dependent turnover, we performed cycloheximide (CHX)-based chase experiments using a His-tagged NanoLuc-ecDHFRN18T,A19V construct. Throughout all results, we could observe a distinct protein degradation profile after CHX treatment, which was partially prevented by CB-5083 but not by TMP (Supplementary Fig. 3a–h). The same effects were observed either when control compounds were added at 4 hours after CHX exposure (Supplementary Fig. 3a-d), or simultaneously applied with CHX (Supplementary Fig. 3e-h). Hence, these data would corroborate a mode of action where NanoLuc-ecDHFRN18T,A19V is primarily degraded through ERAD in its native state (with CB-5083 triggering NanoLuc secretion by means of ERAD inhibition), whereas TMP permits protein secretion by inducing a “properly folded” protein configuration that evades a constitutively active ERAD pathway rather than causing direct ERAD inhibition (Fig. 1b). Furthermore, we showed that overexpression of NanoLuc-ecDHFRN18T, A19V and/or the presence of TMP did not affect the levels of endogenous ERAD substrates such as PIGK (Supplementary Fig. 3i) or cellular markers of ER stress such as ERP72 or HSPA5 (Supplementary Fig. 3j). Impacts on ER stress and ER morphology were only observed when treating native or ecDHFRN18T, A19V-transgenic cells with the classical ER stress inducer Tunicamycin (TM) (Supplementary Fig. 3k and Supplementary Fig. 4), suggesting that TMP may qualify as a potent, safe and clinically compatible trigger signal for dose-dependent regulation of transgene expression (Fig. 2a).Fig. 2Regulation performance of ERAD-level gene switches.a TMP concentration-dependent NanoLuc expression. HEK-293 cells were transfected with pGT69 and cultured with different TMP concentrations. NanoLuc levels were measured at 48 h post-transfection. Numbers above bars are fold changes relative to DMSO-treated cells (0 µM TMP). b Activation kinetics of TMP-inducible protein secretion. HEK-293 cells were transfected with pGT69. NanoLuc levels were monitored over 4 h after TMP exposure. Numbers above bars indicate fold changes between TMP-treated and DMSO-treated cells at the same time point. c TMP-regulated protein secretion in different mammalian cell lines. U251, KPC, U-87 MG, N2a and Hepa1-6 were transfected with pGT69 and cultured with 0 or 40 μM TMP. NanoLuc levels were profiled at 48 h post-transfection. Numbers above bars indicate fold changes of NanoLuc expression between TMP-treated and DMSO-treated cells. d Reversibility of TMP-regulated protein secretion. HEK-293 cells were transfected with pGT69 and cultured with TMP for 2 h (red), before the medium was exchanged into fresh TMP-free medium every 2 h. After repeating the medium exchange step (black) for three times, TMP-containing medium was re-administered. NanoLuc levels were profiled every 30 min. e Schematic diagram of different ecDHFRN18T, A19V-dependent secretion strategies. For Type-C secretion, degron motifs are placed at the C-terminus of the POI through a furin-cleavable linker, whereas Type-N secretion inserts the degron motif between a signal peptide and the furin cleavage site at the N-terminus. Type-N secretion enables POI production in its native format. f, g Applicability of the ERAD-level gene switch strategy to regulate various therapeutic proteins. HEK-293 cells were transfected with constitutive expression vectors for FCS-ecDHFRN18T, A19V-linked Bevacizumab, Rituximab, B6H12 (f), or shGLP-1, mTSLP, ssANP, ssTNK-TPA, ssHwTx and ssADPN (g), and cultured with 0 or 40 μM TMP. Therapeutic protein levels were measured by ELISA at 48 h post-transfection. Numbers above bars indicate fold changes to DMSO-treated cells. Data are shown as mean ± SD; n = 3 independent experiments; circles show individual results. Experimental details are shown in Supplementary Data 4. Source data are provided as a Source Data file. Regulation performance of ERAD-level gene switches. a TMP concentration-dependent NanoLuc expression. HEK-293 cells were transfected with pGT69 and cultured with different TMP concentrations. NanoLuc levels were measured at 48 h post-transfection. Numbers above bars are fold changes relative to DMSO-treated cells (0 µM TMP). b Activation kinetics of TMP-inducible protein secretion. HEK-293 cells were transfected with pGT69. NanoLuc levels were monitored over 4 h after TMP exposure. Numbers above bars indicate fold changes between TMP-treated and DMSO-treated cells at the same time point. c TMP-regulated protein secretion in different mammalian cell lines. U251, KPC, U-87 MG, N2a and Hepa1-6 were transfected with pGT69 and cultured with 0 or 40 μM TMP. NanoLuc levels were profiled at 48 h post-transfection. Numbers above bars indicate fold changes of NanoLuc expression between TMP-treated and DMSO-treated cells. d Reversibility of TMP-regulated protein secretion. HEK-293 cells were transfected with pGT69 and cultured with TMP for 2 h (red), before the medium was exchanged into fresh TMP-free medium every 2 h. After repeating the medium exchange step (black) for three times, TMP-containing medium was re-administered. NanoLuc levels were profiled every 30 min. e Schematic diagram of different ecDHFRN18T, A19V-dependent secretion strategies. For Type-C secretion, degron motifs are placed at the C-terminus of the POI through a furin-cleavable linker, whereas Type-N secretion inserts the degron motif between a signal peptide and the furin cleavage site at the N-terminus. Type-N secretion enables POI production in its native format. f, g Applicability of the ERAD-level gene switch strategy to regulate various therapeutic proteins. HEK-293 cells were transfected with constitutive expression vectors for FCS-ecDHFRN18T, A19V-linked Bevacizumab, Rituximab, B6H12 (f), or shGLP-1, mTSLP, ssANP, ssTNK-TPA, ssHwTx and ssADPN (g), and cultured with 0 or 40 μM TMP. Therapeutic protein levels were measured by ELISA at 48 h post-transfection. Numbers above bars indicate fold changes to DMSO-treated cells. Data are shown as mean ± SD; n = 3 independent experiments; circles show individual results. Experimental details are shown in Supplementary Data 4. Source data are provided as a Source Data file. From a gene therapy perspective, an ideal gene switch should permit high fold-changes, rapid and reversible activation kinetics, multi-organ functionality, regulation of different types of therapeutic proteins, and compatibility with various viral and non-viral gene delivery strategies47. Indeed, with 15 min being sufficient to reveal a two-fold stimulation of target protein secretion (Fig. 2b), the Trimethoprim (TMP)-regulated ERAD-level gene switch performed at significantly faster timescales when directly compared to conventional gene switches acting at the levels of transcription (Supplementary Fig. 5a; inspired by ref. 16), RNA-splicing (Supplementary Fig. 5b; inspired by ref. 26) or translation (Supplementary Fig. 5c; according to ref. 29), while showing a comparable kinetic profile to regulation strategies based on TEV-repressible ER-retention (Supplementary Fig. 5d; inspired by ref. 35). The ERAD-level gene switch was also functional in a variety of mammalian cell lines, consistently showing negligible basal secretion levels when cultivating different cells in TMP-free medium (Fig. 2c). In addition, the gene switch not only showed rapid activation kinetics (Fig. 2b), but also fast inactivation profiles upon TMP withdrawal (Fig. 2d), indicating potential utility for on-demand transgene regulation. Low basal expression levels could also be maintained over the course of several days, showing ultimate advantages over state-of-the art transcription-level gene switches in terms of low background activities (Supplementary Fig. 6a). Lastly, we inserted a furin cleavage site (FCS) into the linker domain used to connect the target protein and the degron signal, allowing furin-mediated proteolysis to mediate spontaneous degron removal within the Golgi-apparatus (Fig. 2e). In theory, while proteins engineered with C-terminal furin-cleavable linkers (denoted ERAD-level gene switch with Type-C inducible secretion; Fig. 2e) may still contain amino acid residues originating from the furin cleavage site, target proteins where the furin-cleavable linker was placed after the N-terminal signal peptide may no longer contain any additional residues (denoted Type-N secretion; Fig. 2e). In practice, our experimental findings confirm that target proteins were indeed processed at the designated FCS (Supplementary Fig. 6b), suggesting that (i) both Type-C and Type-N secretion may compromise absolute expression levels to a some extent (Supplementary Fig. 6c), that (ii) Type-N secretion may produce native therapeutic proteins but Type-C secretion allows for higher ERAD-efficiencies (Supplementary Fig. 6d), and that (iii) not all therapeutic proteins are compatible with both secretion types (Fig. 2f, g and Supplementary Fig. 6e). Furthermore, we anticipate that degradation efficiencies and basal expression may be most significantly affected by the molecular weight, the presence of stabilizing domains such as hIgG-Fc fused to the POI, and the theoretical solvent-accessible surface area (SASA) of the degron(s) (Supplementary Data 2). Taken together, when developing ERAD-centered gene switches for biomedical applications, the furin-independent strategy may be sufficient when high fold-changes are the prime engineering goal and if the output protein may tolerate the presence of a C-terminal appendix (such as in the case of reporter proteins; Supplementary Fig. 6b, c). In contrast, when therapeutic applications in vivo are the ultimate aim, N-type secretion may be favored in order to reduce safety concerns arising from the potentially modified therapeutic protein (Supplementary Fig. 6d). In any case, optimization of fold-changes may be completed by empirical testing of different expression levels of a single protein expression unit by avoiding the need to titrate particular plasmid ratios (Supplementary Fig. 7a, c), before establishing an optimal TMP-dose to reach a targeted efficacy window for each therapeutic protein (Supplementary Fig. 7b, d). Compatibility of ERAD-level gene switches with AAV-mediated gene delivery Next, we sought to transfer the ERAD-level gene switch into various AAV vectors and test its regulation performance in various gene therapy settings. Because the engineering of ERAD-level gene switches only requires attachment of a < 200 amino acids peptide motif to the protein of interest (Fig. 1b), AAV-based delivery of therapeutic cargos of close to 4 kb in size may be achieved in this manner (Fig. 3a). In particular, after confirming that dose-dependent (Fig. 3b) and rapid stimulation of target gene expression was also achieved in mice (Fig. 3c), we created AAV2/8, AAV2/Myo and AAV2/9n particles that were previously shown to mediate efficient gene delivery into the liver, muscle and central nervous system (CNS), respectively. Indeed, 14 days after intravenous (for AAV2/8 and AAV2/Myo) or intrathecal injection (for AAV2/9n) of 3*1011 vg AAV particles expressing ecDHFRN18T, A19V-tagged SEAP into mice, a strict TMP-dependent SEAP production profile was observed in the animals’ bloodstream (Fig. 3d). Fluorescence imaging of muscle (Supplementary Fig. 8) and spinal cord tissues (Supplementary Fig. 9) of mice receiving isogenic EGFP-expressing AAV particles confirm that the gene switches were indeed delivered into the respective body parts (Fig. 3a). Notably, delivery of an ERAD-level gene switch into wild-type male mice using AAV2/8-particles resulted in a long-lasting profile of TMP-dependent transgene expression for over one year (Fig. 3e). In contrast, hepatic expression of a state-of-the art transcription-level gene switch using the same AAV2/8 serotype failed to reveal trigger-inducible transgene expression in the longer term (Supplementary Fig. 10), even though this gene switch was also regulated by a clinically licensed drug (i.e. grazoprevir) and showed excellent performances both in vitro (Supplementary Fig. 5a) and in mice following hydrodynamic injection (Supplementary Fig. 10a). Grazoprevir-inducible gene switch characteristics were restored when applying the same AAV2/8-particles to immunodeficient NSG mice (Supplementary Fig. 10b), suggesting the involvement of immune-related factors of transgene silencing48. Such issues were not observed with the simpler designed ERAD-level gene switch comprising markedly reduced xenogenic payloads (Fig. 3e). Also, no signs of hepatic side effects were observed when analyzing the overall liver physiology of mice treated with the AAV2/8 particles for over one year, compared to non-treated mice of equal ages indicative for “normal” aging (Supplementary Fig. 11). Although hepatic inflammatory activity as evaluated with the METAVIR scoring system49 was indeed observed when comparing the histopathology data of AAV-treated mice with those of untreated mice, there was no significant difference in the inflammation profiles between mice receiving or not receiving TMP (Supplementary Fig. 12), suggesting that most immune reactions could result from the transgene delivery process in general50 (Supplementary Fig. 11 and Supplementary Fig. 12). Furthermore, we show that ERAD-level gene switches appeared to not significantly induce ER stress when compared to conventional constitutive gene expression strategies both at a cellular level (Supplementary Fig. 3j) and in mice (Fig. 3f). Instead, elevation of ER stress markers was observed with the protein secretion system previously engineered on the basis of FK506-repressible intracellular protein aggregation31 (Supplementary Fig. 13b), whose design appeared highly similar to ERAD-level gene switches at a first glance (Supplementary Fig. 13a). Though, using fluorescent imaging, we showed that both systems were indeed mechanistically different, where the ERAD-level gene switch showed minimal protein expression if not required (Supplementary Fig. 13c), while transgene activity was permanently active when using the drug-repressible aggregation strategy (Supplementary Fig. 13c, d). Furthermore, we show that both regulation strategies were highly effective in permitting high fold-changes of trigger-inducible protein secretion (Supplementary Fig. 13e, f), where ERAD-level gene switches generally resulted in lower basal expression (Supplementary Fig. 13e), with drug-repressible aggregation being superior in regulating insulin secretion (Supplementary Fig. 13f). Nevertheless, ERAD-level gene switches may fulfill many promises to develop gene therapy products with reliable safety and efficacy profiles of on-demand transgene expression.Fig. 3ERAD-level gene switches for AAV-mediated gene therapy.a Application potential. ERAD-level gene switches can be installed onto single vector AAV systems, allowing delivery of POI cargos of close to 4 kb into target tissues through flexible choice of AAV serotypes. b Validation of TMP-regulated protein secretion in vivo. Mice were hydrodynamically injected with ecDHFRN18T, A19V-tagged SEAP (pGT68) and fed with TMP 3 times per day (total daily dose indicated). SEAP levels in bloodstream were measured at 24 h after the first TMP administration. Numbers above bars are fold changes relative to mice receiving saline. c Activation kinetics. 24 h after hydrodynamic injection of pGT68, mice received a single oral dose of TMP (4 mg/kg) before SEAP levels were tracked for 2 h. Numbers above bars are fold changes relative to mice receiving saline. d Compatibility of the TMP-regulated ERAD gene switch with various therapeutically relevant AAV serotypes. A constitutive expression unit for ecDHFRN18T, A19V-tagged SEAP (from pGT68) was transferred into AAV2 particles with tropisms for the liver (AAV2/8), muscle (AAV2/Myo) or central nervous system CNS (AAV2/9n). 14 days after intravenous (AAV2/8, AAV2/Myo) or intrathecal injection (AAV2/9n) of AAV particles, SEAP levels in the bloodstream or cerebral fluid of mice receiving or not receiving TMP were measured. e Long-term control of TMP-stimulated protein secretion in mice. TMP-stimulated SEAP production from mice treated with AAV2/8 particles was monitored for 1 year by recurrent TMP feeding at different timepoints. Numbers above bars are fold changes relative to mice receiving saline. f Potential impact of ERAD-level gene switches on ER stress in target tissues. One year after AAV2/8-particles injection, relative mRNA expression levels in liver were profiled by qRT-PCR using primers listed in Supplementary Table 1. Untreated mice of equal age were used as controls. Data are shown as mean ± SEM; n = 7 (b, c), n = 6 or 8 (d), or n = 5 or 6 (e, f) mice per group; circles show individual results. Statistical analysis was performed using one-way ANOVA with Tukey’s multiple comparison test (f). ns, not significant. Source data (Source Data file) and experimental details (Supplementary Data 4) are provided. ERAD-level gene switches for AAV-mediated gene therapy. a Application potential. ERAD-level gene switches can be installed onto single vector AAV systems, allowing delivery of POI cargos of close to 4 kb into target tissues through flexible choice of AAV serotypes. b Validation of TMP-regulated protein secretion in vivo. Mice were hydrodynamically injected with ecDHFRN18T, A19V-tagged SEAP (pGT68) and fed with TMP 3 times per day (total daily dose indicated). SEAP levels in bloodstream were measured at 24 h after the first TMP administration. Numbers above bars are fold changes relative to mice receiving saline. c Activation kinetics. 24 h after hydrodynamic injection of pGT68, mice received a single oral dose of TMP (4 mg/kg) before SEAP levels were tracked for 2 h. Numbers above bars are fold changes relative to mice receiving saline. d Compatibility of the TMP-regulated ERAD gene switch with various therapeutically relevant AAV serotypes. A constitutive expression unit for ecDHFRN18T, A19V-tagged SEAP (from pGT68) was transferred into AAV2 particles with tropisms for the liver (AAV2/8), muscle (AAV2/Myo) or central nervous system CNS (AAV2/9n). 14 days after intravenous (AAV2/8, AAV2/Myo) or intrathecal injection (AAV2/9n) of AAV particles, SEAP levels in the bloodstream or cerebral fluid of mice receiving or not receiving TMP were measured. e Long-term control of TMP-stimulated protein secretion in mice. TMP-stimulated SEAP production from mice treated with AAV2/8 particles was monitored for 1 year by recurrent TMP feeding at different timepoints. Numbers above bars are fold changes relative to mice receiving saline. f Potential impact of ERAD-level gene switches on ER stress in target tissues. One year after AAV2/8-particles injection, relative mRNA expression levels in liver were profiled by qRT-PCR using primers listed in Supplementary Table 1. Untreated mice of equal age were used as controls. Data are shown as mean ± SEM; n = 7 (b, c), n = 6 or 8 (d), or n = 5 or 6 (e, f) mice per group; circles show individual results. Statistical analysis was performed using one-way ANOVA with Tukey’s multiple comparison test (f). ns, not significant. Source data (Source Data file) and experimental details (Supplementary Data 4) are provided. On-demand secretion of cardioprotective therapeutics in mammalian cells and in mice Although we showed that ERAD-level gene switches can be engineered to regulate diverse types of therapeutic protein outputs, such as monoclonal antibodies (Fig. 2e), anorectic hormones (Fig. 2e, Supplementary Fig. 7c, d and Supplementary Fig. 13e) or growth factors (Supplementary Fig. 6e), rapid and on-demand therapeutic protein secretion may only be suitable for diseases that suffer from acute conditions. For this reason, we developed potential gene therapy approaches for cardiovascular diseases and chronic pain - two types of pathologies that may most benefit from the key characteristics of ERAD-level gene switches. We first engineered a Type-C gene switch for TMP-inducible secretion of a stabilized human tenecteplase variant created through fusion of an hIgG-Fc immunoglobulin domain to recombinant human TNK (ssTNK; hIgG-Fc-TNK) (Fig. 4a). Using an in vitro thrombolysis assay51, we confirm that ssTNK produced with this Type-C secretion strategy (Fig. 4b) was indeed bioactive in terms of clot lysis capability (Fig. 4c). Furthermore, TMP-dependent ssTNK production was observed within 30 min of stimulation either on a cellular level (Fig. 4d) or in mice (Fig. 4e), thus corroborating the rapid activation kinetics of ERAD-level gene switches. We also show that TMP-dependent ssTNK production levels in vivo (Fig. 4f) could be precisely titrated to reach specific efficacy windows of interest (Fig. 4g). This may form a promising starting point to develop future therapeutic regimens for on-demand thrombolysis.Fig. 4Potential gene therapy applications of intramuscular expression of ERAD-level gene switches.a On-demand secretion of cardioprotective therapeutics by AAV2/Myo-mediated gene delivery. AAV2/Myo particles containing gene switches for Type-C TMP-inducible tenecteplase (ssTNK) or Type-N/C TMP-inducible atrial natriuretic peptide secretion (ssANP) were intravenously injected to allow muscle tissues to mediate trigger-inducible release of ssTNK or ssANP. b TMP concentration-dependent ssTNK expression. HEK-293 cells were transfected with Type-C TMP-inducible ssTNK expression vector (pIU210) and cultured with different TMP concentrations. ssTNK levels were measured. c Validation of fibrinolysis by TMP-stimulated ssTNK production in vitro. Conditioned medium from HEK-293 cells transfected with pIU210 (0.4633 ng/mL ssTNK from DMSO-treated or 5.2158 ng/mL ssTNK from TMP-treated) was added to freshly formed fibrin clots. Time-dependent fibrinolysis was recorded by photographic analysis. d Activation kinetics of TMP-stimulated ssTNK production. HEK-293 cells were transfected with pIU210. ssTNK levels were monitored over 4 h after TMP exposure. e Activation kinetics of TMP-inducible ssTNK secretion in mice. 24 h after hydrodynamic injection of pIU210, mice received a single oral dose of TMP and ssTNK levels in blood were tracked for 2 h. f AAV2/Myo-mediated delivery of Type-C TMP-inducible ssTNK switch into mice. 14 days after intravenous injection of AAV2/Myo-ssTNK(TMP-C) particles, ssTNK levels in the bloodstream of mice receiving or not receiving TMP (12 mg/kg/day) were measured. g Estimation of efficacy window for ssTNK-mediated fibrinolysis. Conditioned medium from TMP-stimulated HEK-293 cells transfected with pIU210 was serially diluted and added to fibrin clots. After 15 h, dose-dependent fibrinolytic effect was recorded. Data are shown as mean ± SD; n = 3 independent experiments (b, d, g); mean ± SEM; n = 7 (e) or n = 6 or 8 (f) mice per group; circles show individual results. Statistical analysis was performed using one-way ANOVA with Dunnett’s multiple comparison test (b, g), two-way ANOVA with Bonferroni’s multiple comparison test (d), or two-tailed unpaired t test (f). The p values indicate the significance of differences in the mean values versus the non-induced group (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); ns, not significant (p > 0.05). Experimental details are shown in Supplementary Data 4. Source data are provided as a Source Data file. Potential gene therapy applications of intramuscular expression of ERAD-level gene switches. a On-demand secretion of cardioprotective therapeutics by AAV2/Myo-mediated gene delivery. AAV2/Myo particles containing gene switches for Type-C TMP-inducible tenecteplase (ssTNK) or Type-N/C TMP-inducible atrial natriuretic peptide secretion (ssANP) were intravenously injected to allow muscle tissues to mediate trigger-inducible release of ssTNK or ssANP. b TMP concentration-dependent ssTNK expression. HEK-293 cells were transfected with Type-C TMP-inducible ssTNK expression vector (pIU210) and cultured with different TMP concentrations. ssTNK levels were measured. c Validation of fibrinolysis by TMP-stimulated ssTNK production in vitro. Conditioned medium from HEK-293 cells transfected with pIU210 (0.4633 ng/mL ssTNK from DMSO-treated or 5.2158 ng/mL ssTNK from TMP-treated) was added to freshly formed fibrin clots. Time-dependent fibrinolysis was recorded by photographic analysis. d Activation kinetics of TMP-stimulated ssTNK production. HEK-293 cells were transfected with pIU210. ssTNK levels were monitored over 4 h after TMP exposure. e Activation kinetics of TMP-inducible ssTNK secretion in mice. 24 h after hydrodynamic injection of pIU210, mice received a single oral dose of TMP and ssTNK levels in blood were tracked for 2 h. f AAV2/Myo-mediated delivery of Type-C TMP-inducible ssTNK switch into mice. 14 days after intravenous injection of AAV2/Myo-ssTNK(TMP-C) particles, ssTNK levels in the bloodstream of mice receiving or not receiving TMP (12 mg/kg/day) were measured. g Estimation of efficacy window for ssTNK-mediated fibrinolysis. Conditioned medium from TMP-stimulated HEK-293 cells transfected with pIU210 was serially diluted and added to fibrin clots. After 15 h, dose-dependent fibrinolytic effect was recorded. Data are shown as mean ± SD; n = 3 independent experiments (b, d, g); mean ± SEM; n = 7 (e) or n = 6 or 8 (f) mice per group; circles show individual results. Statistical analysis was performed using one-way ANOVA with Dunnett’s multiple comparison test (b, g), two-way ANOVA with Bonferroni’s multiple comparison test (d), or two-tailed unpaired t test (f). The p values indicate the significance of differences in the mean values versus the non-induced group (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); ns, not significant (p > 0.05). Experimental details are shown in Supplementary Data 4. Source data are provided as a Source Data file. In contrast to ssTNK, regulation of a similarly designed variant of atrial natriuretic peptide (ssANP; ANP-hIgG-Fc) using Type-C secretion strategies resulted in relatively poor fold-changes of TMP-inducible ssANP-secretion, demonstrating high basal expression (Fig. 5a, b) and slow activation kinetics (Fig. 5c). One reason for this poor degradation efficiency could be the relatively low cytosolic exposure area of the degron when applying a Type-C secretion strategy to ssANP regulation, in spite of its small molecular weight (Supplementary Data 2). To improve gene switch performance, we inserted the Type-C TMP-inducible ssANP-expression unit into the framework of a Type-N secretion system, thus forming a combined “Type-N/C” configuration (Fig. 4a). As expected, degradation efficiencies of such Type-N/C TMP-inducible ssANP-secretion system was markedly enhanced, resulting in improved dose-dependent regulation of absolute (Fig. 5a) and bioactive ANP levels (Fig. 5b). Using such improved gene switch, we were again able to rapidly secrete ssANP from mammalian cells (Fig. 5c) and in vivo within less than one hour after TMP exposure (Fig. 5d). Thus, following AAV2/Myo-mediated delivery of such Type-N/C gene switch regulating ssANP secretion into hypertensive mice, systolic blood pressure levels could be effectively regulated by oral TMP administration (Fig. 5e, f), indicating the potential of using such gene therapy strategy to provide on-demand medication for acute complications such as hypertension.Fig. 5On-demand regulation of blood pressure by oral TMP administration.a TMP concentration-dependent ssANP expression. HEK-293 cells were transfected with either Type-C (pIU177) or Type-N/C (pIU180) TMP-inducible ssANP expression vector and cultured with different TMP concentrations. ssANP levels were measured at 48 h post transfection. b Dose-dependent quantification of bioactive ANP. Receiver cells were co-transfected with an ANP-responsive plasmid system and incubated with conditioned medium from TMP-stimulated sender cells transfected with pIU177 or pIU180. SEAP levels were quantified at 24 h after medium transfer. c Activation kinetics of TMP-stimulated ssANP production in mammalian cells. HEK-293 cells were transfected with pIU177 or pIU180. ssANP levels in culture supernatants were monitored over 4 h after exposing the cells to 40 μM TMP. d Activation kinetics of TMP-inducible ssANP secretion in mice. 24 h after hydrodynamic injection of pIU180, mice received a single oral dose of TMP (4 mg/kg) and ssANP levels in the bloodstream were tracked for 2 h. e, f AAV2/Myo-mediated delivery and therapeutic efficacy of a Type-N/C TMP-inducible gene switch regulating ssANP secretion. 14 days after intravenous injection of AAV2/Myo-ssANP (TMP-N/C) particles into hypertensive mice, systolic blood pressure (SBP) (e) and ssANP levels (f) in the bloodstream of mice receiving or not receiving TMP (12 mg/kg/day) were measured. Hypertensive mice not receiving AAV treatments were used as controls. Data are shown as mean ± SD; n = 3 (a–c); mean ± SEM; n = 6 (d) or 7 (e, f); circles show individual results. Statistical analysis was performed using one-way ANOVA with Dunnett’s test (a) or two-way ANOVA with Bonferroni’s test (b–f). The p values indicate the significance of differences in the mean values versus the non-induced group (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); ns, not significant (p > 0.05). For the Type-N/C group in (a): p = 0.0012 (0.01 vs. 0 μM); all other groups vs. 0 μM: p < 0.0001. Experimental details are shown in Supplementary Data 4. Source data are provided as a Source Data file. On-demand regulation of blood pressure by oral TMP administration. a TMP concentration-dependent ssANP expression. HEK-293 cells were transfected with either Type-C (pIU177) or Type-N/C (pIU180) TMP-inducible ssANP expression vector and cultured with different TMP concentrations. ssANP levels were measured at 48 h post transfection. b Dose-dependent quantification of bioactive ANP. Receiver cells were co-transfected with an ANP-responsive plasmid system and incubated with conditioned medium from TMP-stimulated sender cells transfected with pIU177 or pIU180. SEAP levels were quantified at 24 h after medium transfer. c Activation kinetics of TMP-stimulated ssANP production in mammalian cells. HEK-293 cells were transfected with pIU177 or pIU180. ssANP levels in culture supernatants were monitored over 4 h after exposing the cells to 40 μM TMP. d Activation kinetics of TMP-inducible ssANP secretion in mice. 24 h after hydrodynamic injection of pIU180, mice received a single oral dose of TMP (4 mg/kg) and ssANP levels in the bloodstream were tracked for 2 h. e, f AAV2/Myo-mediated delivery and therapeutic efficacy of a Type-N/C TMP-inducible gene switch regulating ssANP secretion. 14 days after intravenous injection of AAV2/Myo-ssANP (TMP-N/C) particles into hypertensive mice, systolic blood pressure (SBP) (e) and ssANP levels (f) in the bloodstream of mice receiving or not receiving TMP (12 mg/kg/day) were measured. Hypertensive mice not receiving AAV treatments were used as controls. Data are shown as mean ± SD; n = 3 (a–c); mean ± SEM; n = 6 (d) or 7 (e, f); circles show individual results. Statistical analysis was performed using one-way ANOVA with Dunnett’s test (a) or two-way ANOVA with Bonferroni’s test (b–f). The p values indicate the significance of differences in the mean values versus the non-induced group (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); ns, not significant (p > 0.05). For the Type-N/C group in (a): p = 0.0012 (0.01 vs. 0 μM); all other groups vs. 0 μM: p < 0.0001. Experimental details are shown in Supplementary Data 4. Source data are provided as a Source Data file. Rapid control of TMP-stimulated painkiller release in the CNS providing on-demand analgesia Apart from being compatible with regulation of protein secretion in the context of ERAD-level gene switches, another key advantage of Trimethoprim (TMP) as a clinically-established trigger molecule is the ability of TMP to cross the blood-brain barrier and manipulate gene activities within the central nervous system (CNS)41. To this end, chronic pain may be a disorder that may highly benefit from a CNS-specific gene therapy providing on-demand and remote-controlled drug actions. Thus, to provide a gene therapy approach capable of providing on-demand analgesia in its strictest sense, we sought to create an ERAD-level gene switch for TMP-simulated ssHwTx secretion, followed by AAV2/9n-mediated delivery into the dorsal root ganglia (DRG) at the subarachnoid space, ultimately allowing oral TMP administration to mediate spontaneous painkiller release right at the main body site of pain signal transmission (Fig. 6a). Interestingly, effective TMP-inducible ssHwTx production was achieved with either Type-C or Type-N secretion strategies (Fig. 6b, c), but therapeutic efficacy in vivo was only observed with the Type-N secretion strategy (Fig. 6d–f). One plausible explanation may be the presence of additional residues added to the ssHwTx protein during Type-C furin-processing (Fig. 2e), as the active site of the analgesic peptide Huwentoxin-IV is known to be at the C-terminus52. Thus, degradation efficiency does not appear to be the bottleneck for therapeutic activity of ssHwTx, as both Type-C and Type-N secretion accounted for similar TMP-dependent production levels (Fig. 6c, f). Instead, ssHwTx produced from Type-N TMP-inducible secretion may not contain any foreign amino acid residues (Fig. 2e), and could therefore account for the analgesic effects produced in mice as soon as ssHwTx-secretion levels reached a typical efficacy window of >100 ng/mL in the bloodstream (Fig. 6d–f and Supplementary Fig. 14). Thus, after confirming that improvement of pain sensation was achieved within one hour after oral administration of a single TMP dose (Fig. 6g–j), we transferred the Type-N TMP-inducible gene switch for ssHwTx secretion into CNS-selective AAV2/9n vectors (Fig. 6a). As expected, suppression of thermal (Fig. 6k) or mechanical pain (Fig. 6l) was mediated upon oral TMP-intake over prolonged timespans, thus offering the perspective of achieving patient-compliant, side-effect free, on-demand and injection-free pain management in the long term. We also showed that ssHwTx-secretion triggered by one single TMP stimulation event could account for pain suppression of at least 3 days in duration (Fig. 6m), which may facilitate precise dosing of the therapeutic regimen. At the same time, mice receiving oral administrations of TMP exhibited substantially reduced chronic pain behavior following formalin injection on the plantar surface of a hind paw (Fig. 6n). In the future, gene therapies enabling rapid pain suppression upon stimulation using a patient-compliant and CNS-compatible trigger signal such as Trimethoprim may pave the way to effective pain treatment strategies in humans.Fig. 6A regulated gene therapy approach for chronic pain treatment.a On-demand analgesia by AAV2/9n-mediated TMP-inducible painkiller release. AAV2/9n particles containing Type-N TMP-inducible ssHwTx secretion were intrathecally injected for long-term pain relief by oral TMP stimulation. b TMP concentration-dependent ssHwTx expression. HEK-293 cells were transfected with Type-C (pLYL110) or Type-N TMP-inducible ssHwTx vector (pLYL225) and cultured with different TMP concentrations, ssHwTx was measured. c Activation kinetics of TMP-stimulated ssHwTx production. HEK-293 cells were transfected with pLYL110 or pLYL225. ssHwTx levels were monitored after TMP exposure. d–f Analgesic efficacy of TMP-inducible ssHwTx secretion. After hydrodynamic injection of pLYL110 or pLYL225, mice were fed 3 times of TMP. Analgesic scores against thermal (d), mechanical pain (e) and ssHwTx levels (f) were profiled. g Activation kinetics of TMP-inducible ssHwTx secretion in mice. After hydrodynamic injection of pLYL225, mice received a single oral dose of TMP and ssHwTx levels were tracked for 2 h. h–j Rapid analgesia by TMP-stimulated ssHwTx secretion. Analgesic scores against thermal (h), mechanical (i) and ssHwTx-levels (j) were profiled in mice received a single dose of TMP after hydrodynamic injection of pLYL225. k–m Long-term pain management by AAV2/9n-delivered Type-N TMP-inducible gene switch regulating ssHwTx secretion. Mice received intrathecal injection of AAV2/9n particles. Analgesia scores against thermal (k) or mechanical (l) pain were profiled over several weeks. (m) In week 6, analgesia scores after a single TMP dose were followed for 5 days. n Analgesic efficacy in a chronic inflammatory pain model. 6 weeks after AAV2/9n injection, mice were fed with TMP. Pain behavior was analyzed by scoring licking and biting times. Data are shown as mean ± SD; n = 3 (b, c); mean ± SEM; n = 5–8 (d–n); circles show individual results. Statistical analysis was performed using one-way ANOVA with Dunnett’s test (b), two-way ANOVA with Bonferroni’s test (c–j), one-way ANOVA with Tukey’s test (k, l), or two-tailed unpaired t test (m, n). The p values indicate the significance of differences in the mean values versus the non-induced group (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); ns, not significant (p > 0.05). Experimental details are shown in Supplementary Data 4. Source data are provided. A regulated gene therapy approach for chronic pain treatment. a On-demand analgesia by AAV2/9n-mediated TMP-inducible painkiller release. AAV2/9n particles containing Type-N TMP-inducible ssHwTx secretion were intrathecally injected for long-term pain relief by oral TMP stimulation. b TMP concentration-dependent ssHwTx expression. HEK-293 cells were transfected with Type-C (pLYL110) or Type-N TMP-inducible ssHwTx vector (pLYL225) and cultured with different TMP concentrations, ssHwTx was measured. c Activation kinetics of TMP-stimulated ssHwTx production. HEK-293 cells were transfected with pLYL110 or pLYL225. ssHwTx levels were monitored after TMP exposure. d–f Analgesic efficacy of TMP-inducible ssHwTx secretion. After hydrodynamic injection of pLYL110 or pLYL225, mice were fed 3 times of TMP. Analgesic scores against thermal (d), mechanical pain (e) and ssHwTx levels (f) were profiled. g Activation kinetics of TMP-inducible ssHwTx secretion in mice. After hydrodynamic injection of pLYL225, mice received a single oral dose of TMP and ssHwTx levels were tracked for 2 h. h–j Rapid analgesia by TMP-stimulated ssHwTx secretion. Analgesic scores against thermal (h), mechanical (i) and ssHwTx-levels (j) were profiled in mice received a single dose of TMP after hydrodynamic injection of pLYL225. k–m Long-term pain management by AAV2/9n-delivered Type-N TMP-inducible gene switch regulating ssHwTx secretion. Mice received intrathecal injection of AAV2/9n particles. Analgesia scores against thermal (k) or mechanical (l) pain were profiled over several weeks. (m) In week 6, analgesia scores after a single TMP dose were followed for 5 days. n Analgesic efficacy in a chronic inflammatory pain model. 6 weeks after AAV2/9n injection, mice were fed with TMP. Pain behavior was analyzed by scoring licking and biting times. Data are shown as mean ± SD; n = 3 (b, c); mean ± SEM; n = 5–8 (d–n); circles show individual results. Statistical analysis was performed using one-way ANOVA with Dunnett’s test (b), two-way ANOVA with Bonferroni’s test (c–j), one-way ANOVA with Tukey’s test (k, l), or two-tailed unpaired t test (m, n). The p values indicate the significance of differences in the mean values versus the non-induced group (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001); ns, not significant (p > 0.05). Experimental details are shown in Supplementary Data 4. Source data are provided. Discussion Gene therapy has returned to the forefront of modern personalized medicine, showing considerable advantages over classical drug injections by enabling in situ production of difficult-to-express therapeutic proteins with high bioavailability1. However, a critical limitation of currently approved gene therapy products is the lack of temporal regulation in vivo; once delivered, the therapeutic transgene is permanently (over)expressed in the body and cannot be re-adjusted or interrupted using specific control signals12. Synthetic biology, an engineering principle of biology focusing on the design of sophisticated gene regulation strategies in mammalian cells, may provide attractive solutions to increase the spatiotemporal precision of therapeutic transgene regulation53. Though, only a few synthetic gene switches may qualify for therapeutic usage. First, the regulation system must be simple in design and shall comprise a minimal number of expression units and/or xenogenic parts, which are important factors not only for efficient gene delivery, but also to minimize adverse effects related to transgene immunogenicity48. Also, the gene regulation system should not depend on too many cell type-specific factors to ensure that the synthetic gene switch may operate in many target tissues in vivo26. Third, the trigger signal used to regulate therapeutic transgene expression should be safe, patient-compliant and bioavailable when administered into patients - features that may be warranted by FDA-approved small molecules54. So far, the Branaplam (LMI070)-responsive XON-system operating at the level of RNA splicing may be a system that fulfills all these criteria, demonstrating compatibility with AAV-mediated delivery of a diverse set of meaningful therapeutic target proteins26. However, as we have experimentally shown, the activation kinetics of this gene switch may lie in a similar range as conventional transcriptional (Supplementary Fig. 5a) and translational regulation systems (Supplementary Fig. 5c) - requiring at least 4 hours to observe significant trigger-inducible protein secretion (Supplementary Fig. 5b). Thus, the TMP-inducible ERAD-level gene switch developed in this work may also offer the same advantages of circuit simplicity (Fig. 3a), low transgene immunogenicity (Fig. 3e), applicability to various cell types (Fig. 2c) and tissues (Fig. 3d) and responsiveness to clinically licensed drugs (Supplementary Fig. 1f), but provides markedly faster activation kinetics both at a cellular level (Fig. 2b) and in vivo (Fig. 3c). Using histopathological analysis by H&E staining, we showed that all the cumulative immune reactions observed in the livers of mice may result from the typical transgene delivery process as previously characterized for AAV-vectors in general50, showing no direct correlation to whether an ERAD-level gene switch is permanently expressed and/or activated (Supplementary Fig. 12). Though similar and even faster secretion rates can be achieved at a molecular level with other cell-based approaches based on ER retention (Supplementary Fig. 5d), conditional protein aggregation (Supplementary Fig. 13) or exocytosis (Fig. 1a), the TMP-regulated ERAD-level gene switch may launch a fastest gene therapy approach developed to date showing minimal safety constraints. In vivo, therapeutically active levels of transgene expression can be consistently achieved within the first hour after oral drug intake (Figs. 4e, 5d and 6g), which forms an important basis for future gene therapies that seek on-demand drug action. With cardiovascular disorders and chronic pain, we define two medical scenarios that may most benefit from regulated gene therapies with rapid drug actions (Figs. 4–6). For pain treatment in particular, gene therapies designed for on-demand secretion of painkillers may indeed become a clinically competitive treatment strategy. Whereas treatment of hypertension may not necessarily urge gene therapy approaches for similar reasons as for type-1 diabetes - where conventional therapies may keep practical benefits in the longer term55, current standards of pain management are limited to morphine injection, which undisputedly calls for better and safer therapies56. Recombinant analgesic peptides such as Huwentoxin-IV (blocking Nav1.7) or ω-conotoxin (blocking Cav2.2) are promising side-effect-free alternatives to morphine, but repeated intrathecal injection of these peptides into the dorsal root ganglia (DRG) of the CNS is technically demanding and may also raise safety concerns57. Thus, in situ production of analgesic peptides capable of inhibiting these clinically accepted pain targets (Nav1.7 and/or Cav2.2) may result in effective pain management strategies58. Recently, an AAV-mediated gene therapy approach was described for permanent silencing of Nav1.7 genes upon intrathecal expression of a dCas9-based synthetic transcription factor59. However, potential consequences of such permanent gene silencing may cause uncertainties for the patients’ compliance and safety in the longer term. Regulated control over Nav1.7 suppression was achieved by a cell therapy approach allowing for spearmint-triggered secretion of Huwentoxin-IV into the bloodstream of mice, but the slow activation kinetics of the underlying transcription-based gene regulation system can only mediate analgesia at almost 24 h after exposure to spearmint aroma52. Here, the TMP-stimulated gene switch developed in this work may combine the best of all recent attempts. First, it allows the use of CNS-specific AAV vectors to “pre-install” the therapeutic gene switch into the DRG following a single intrathecal injection. Because TMP can naturally cross the blood-brain barrier upon oral intake, in situ secretion of Huwentoxin-IV in the longer term can then be triggered at optimal bioavailability, and in a genuinely on-demand, patient-compliant and non-invasive manner. Second, AAV-based transgene delivery may last for several years in vivo, which may be the therapeutic regimen of choice for patients suffering from life-long and intractable pain syndromes. In contrast to morphine treatments, neither TMP nor Huwentoxin-IV are expected to elicit addiction-related side effects. Because the therapeutic action of Huwentoxin-IV secretion can be accurately dosed both in strength (Supplementary Fig. 14) and duration of analgesia (Fig. 6k–n), complications potentially arising from drug resistance may also be reduced. Lastly, both the dynamics (i.e. speed) and the molecular components of the potential gene therapy product (i.e. the type of AAV vector, transgene payloads and therapeutic outputs) were designed to comply with a high clinical standard. Thus, this work may lay the foundation for a long-sought solution towards a severe medical complication that is afflicting more than 10% of the human population and causing more economic burden than cancer, cardiovascular diseases and diabetes together60. From a synthetic biology perspective, the Trimethoprim (TMP)-responsive ecDHFRN18T, A19V mutant is a well-established degron motif used to regulate a multitude of different target proteins41. However, the majority of studies capitalize on fusing this conditional degron to various intracellular proteins, such as fluorescent reporters41, chimeric antigen receptors61, translational inhibitors62, light-sensitive domains63, anti-VEGF proteins64, thus achieving TMP-dependent control over the stability of these protein targets. While a similar Shield-1 responsive degron system was applied to regulate interleukin secretion in cell therapy settings65, this degron motif was ineffective in engaging the ERAD pathway and could not account for trigger-inducible gene regulation according to the aim of our studies (Fig. 1c and Supplementary Fig. 1f). In contrast, we show that the strategy of designing ERAD-level gene switches involving ecDHFRN18T, A19V may become a new standard to enable rapid secretion of target proteins (Fig. 1 and Supplementary Fig. 1). Moreover, TMP was shown to exhibit a > 10,000-fold selective inhibition for bacterial DHFRs over mammalian analogues66, thereby achieving antibacterial activities by means of blocking tetrahydrofolate biosynthesis and interfering with bacterial DNA synthesis67. Thus, whether or not bacteria will ultimately develop resistance mechanisms against TMP, the ecDHFRN18T, A19V-dependent system is expected to remain largely orthogonal to the human metabolism, as we confirmed that TMP had no significant impact on endogenous folate metabolism of human cells (Supplementary Fig. 2c). Also, the 12 mg/kg/day dose consistently used throughout our study in mice would fall into a safe and clinically-approved therapeutic window for humans68, which would correlate to a human dose of <100 mg/day as estimated using an FDA-certified conversion rule based on body surface area normalization (U.S. Food and Drug Administration (2005), Guidance for Industry). Therefore, although the Trimethoprim (TMP)-responsive gene switch may already fulfill many criteria of a clinically eligible gene therapy product, we anticipate that other ERAD-level gene switches responding to other types of trigger signals may also be developed in a user-defined and systematic manner. First, conditional degron systems may be selected using technologies such as random mutagenesis and directed evolution. Then, selected degron signals may be fused to a secretory protein of interest to test the involvement of the ERAD machinery when regulating protein expression. Lastly, ERAD-level gene switches may be characterized and selected for fastest activation kinetics and potential compatibility with Type-C and/or Type-N secretion. Because the cellular entry process of small molecules—including tissue/organ distribution, cell membrane penetration, and binding to the designed regulatory elements could remain a key limiting factor that determines the ultimate speed of the current gene switch, instant or near‑instant (second‑ to minute‑scale) response rates in the future could be achieved by ERAD-level gene switches developed to respond to physical stimuli such as light or ultrasound. Nevertheless, this work uses a simple, pragmatic and yet rational approach to achieve a longstanding goal of enabling rapid protein secretion in mammalian cells, introduces a future class of gene therapies providing on-demand therapeutic transgene activity, and may increase the clinical utility of synthetic biology-based gene regulation systems. Methods Ethics This study was conducted in strict accordance with all relevant ethical regulations and animal welfare legislation in China. The animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Westlake University (Protocol ID: 24-088-XMQ) and performed in accordance with the Animal Care Guidelines of the Ministry of Science and Technology of the People’s Republic of China. Mice were housed under a 12-h light–dark cycle, typically with five mice per cage. The ambient temperature was maintained at 21 ± 1 °C, with a humidity level of 50 ± 10%. Vector design Construction and reference details for all expression vectors are provided in Supplementary Data 3. Some expression vectors were constructed by in-fusion cloning using the Seamless Cloning Kit (Beyotime Biotechnology, Shanghai, China; cat. no. D7010M). PCR-amplification reactions were performed using KOD One PCR Master Mix (Toyobo Inc., Osaka, Japan; cat. no. KMM-201). Ligation reactions were performed using T4 DNA ligase (New England Biolabs, Beverly, MA; cat. no. M0202L). Restriction endonucleases were purchased from New England Biolabs. The primer synthesis and gene synthesis were completed by Suzhou Genewiz Biotech Co., Ltd. Plasmid DNA was extracted using the TIANprep Mini Plasmid Kit (TIANGEN Biotech, Beijing, China; cat. no. GDP103-03). DNA sequencing service was provided by Tsingke Biotechnology (Beijing, China). Cell culture and transfection Human embryonic kidney cells (HEK-293T, ATCC: CRL-3216), murine hepatoma cells (Hepa1-6, ATCC: CRL-1830), KRASG12D-positive pancreatic ductal adenocarcinoma cells (KPC, Shanghai Model Organisms Center: NM-YD04), murine neuroblasts (N2a, ATCC: CRL-131), malignant human glioma cells (U-87 MG, ATCC: HTB-14), human glioblastoma astrocytoma cells (U251, Shanghai Model Organisms Center: NM-N03) and adipose-derived mesenchymal stromal cells (hADSC, ATCC: PCS-500-011) were cultivated in Dulbecco’s modified Eagle’s medium (DMEM; Thermo Fisher Scientific, Waltham, MA; cat. no. 12100046) supplemented with 10% (v/v) fetal bovine serum (Gibco FBS, Australia; Thermo Fisher Scientific, Waltham, MA; cat. no. 10099141, lot no. 2177370) and 1% (v/v) penicillin/streptomycin solution (PenStrep; Beyotime Biotechnology, Shanghai, China; cat. no. ST488). ElectroBeta18 were cultivated in Roswell Park Memorial Institute 1640 Medium (RPMI medium; VivaCell Biosciences, Shanghai, China; cat. no. C3010) supplemented with 10% FBS and 1% PenStrep. For passaging, cells of pre-confluent cultures were detached by incubation in 0.05% trypsin-EDTA (Sangon Biotech, Shanghai, China; cat no. A610629-0050; lot no. F319BA0030) for 3 min at 37 °C, collected in 10 ml of cell culture medium, centrifuged for 2 min at 200 × g, and resuspended in fresh culture medium at 1.5 × 105 cells/mL, and then seeded into new tissue culture plates (Wuxi NEST Biotechnology, Wuxi, China). All cells were cultured at 37 °C in a humidified atmosphere of 5% CO2 in air. Cell number and viability were quantified using an Invitrogen Countess II AMQAX1000 Cell Counter (Thermo Fisher Scientific). All consumables, stock solutions, buffers, and media used are listed in Supplementary Data 1, and detailed information regarding cell transfection experiments is available in Supplementary Data 4. AAV production AAV2/9n-ssHwTx(TMP), AAV2/Myo-ssANP(TMP), AAV2/Myo-ssTNK(TMP), AAV2/9n-SEAP(TMP), AAV2/Myo-SEAP(TMP), AV2/8-SEAP(TMP) and AAV2/8-SEAP(Grazoprevir) were produced by PackGene Biotech (Guangzhou, China) using the transfer plasmids pSL1047, pIU202, pIU221, pSL660 or pSL552, respectively. AAV2/9n-EGFP and AAV2/Myo-EGFP were kindly gifted by PackGene Biotech (Guangzhou, China). shRNA knockdown The shRNA target sequences for human SYVN1 were designed using the Sigma-Aldrich Predesigned shRNA portal according to the RNAi Consortium (TRC) guidelines. The corresponding oligonucleotides (sense and antisense) were synthesized by Tsingke Biotechnology (Beijing, China), annealed to form double-stranded fragments, and inserted into the pLKO.1 vector (Addgene, plasmid #10878) between the AgeⅠ and EcoRⅠ restriction sites via T4 DNA ligase-mediated ligation. Positive clones were confirmed by Sanger sequencing. The target sequences and the full oligonucleotide sequences used for cloning are listed in Supplementary Table 2. Fluorescence imaging Tissues from mice injected with AAV2/Myo-EGFP (intravenous) or AAV2/9n-EGFP (intrathecal) were harvested and fixed with 4% PFA solution (4°C, overnight) and immersed in 30% sucrose until submersion. After being embedded in Tissue-Tek® O.C.T. Compound (Sakura Finetek Inc., Torrance, CA; cat. no. cat. no. 4583), the tissues were cut into 15 μm slices and stained with DAPI Staining Solution (Beyotime Biotechnology; cat. no. C1006). Images were visualized under a Nikon ECLIPSE Ts2 Fluorescence Microscope with respective excitation and emission filter sets (EGFP: 488/509 nm). Images were acquired by OPLENIC software (version x64, 10.1.14643.20190511). Confocal imaging was performed on a Zeiss LSM 980 Airyscan confocal microscope (Carl Zeiss Microscopy; Jena, Germany) using a ×63 objective lens and excitation spectral laser lines set at 405 nm (DAPI) or 594 nm (ER-Tracker). Reverse transcription-polymerase chain reaction (RT-PCR) Total RNA was extracted from the liver of mice using a Trizol-based RNA extraction method. First-strand cDNA was synthesized using the HiScript II Q RT SuperMix for qPCR (Vazyme Biotech, Nanjing, China; cat. no. R223-01). For quantitative analysis, the PCR reaction was carried out with an initial step of 95 °C for 30 s followed by 40 cycles of 95 °C for 10 s and 60 °C for 30 s on the Applied Biosystems QuantStudio 1 Real-Time PCR System (ThermoFisher Scientific) using the ChamQ Universal SYBR qPCR Master Mix (Vazyme Biotech; cat. no. Q711-02) and the primers listed in Supplementary Table 1. The relative cycle threshold (CT) was determined and normalized against the expression level of endogenous human (ACTB) or mouse actin beta (Actb) genes. RNA sequencing analysis Libraries of extracted RNA were sequenced by Suzhou Genewiz Biotech Co., Ltd. (Suzhou, China) for RNA-seq analysis. Raw data were processed following a standard analysis pipeline. Gene set enrichment analysis was performed by GSEA (version 4.3.3)69 using the C5 collection of MSigDB70. Enrichment bubble plots were generated using GraphPad Prism (version 10.6.1). Western Blot analysis Protein samples were denatured in 5× SDS-PAGE loading buffer at 98 °C for 10 min and separated by SDS-PAGE using 8% or 10% polyacrylamide gels prepared with an SDS-PAGE kit. Proteins were subsequently transferred onto PVDF membranes. Membranes were probed with primary antibodies followed by incubation with HRP-conjugated goat anti-rabbit IgG secondary antibody or HRP-conjugated goat anti-mouse IgG secondary antibody. Signals were detected using BeyoECL Moon reagent and imaged with an Amersham Imager 600 system (GE Healthcare). Primary antibodies used in this study include anti-His-Tag (mouse monoclonal, Abclonal, Cat# AE003, Lot: 3600009978, dilution: 1:10000), anti-PIGK (Rabbit monoclonal, Abcam, Cat# ab201693, Lot: 1020725-33, dilution: 1:1000), anti-α-Tubulin (Rabbit monoclonal, Abclonal, Cat# A6830, Lot: 3600014101, dilution: 1:10000) and anti-GAPDH (mouse monoclonal, Santa Cruz Biotechnology, Cat# sc-32233, Lot: L2019, dilution: 1:1000). Secondary antibody include HRP-conjugated goat anti-rabbit IgG (Beyotime Biotechnology, Cat# A0208, dilution: 1:1000) and HRP-conjugated goat anti-mouse IgG (Beyotime Biotechnology, Cat# A0216, dilution: 1:1000). Thrombolysis assay Human blood from healthy donors was collected into vacutainers containing 3.2% trisodium citrate at a 9:1 blood-to-anticoagulant ratio (KWS Medical Instrument, Shijiazhuang, China), with 0.1 µg/mL PGE1 added to the collection tubes. After gentle inversion mixing, the blood was centrifuged at 330 × g for 10 min to obtain platelet-rich plasma (PRP) as the supernatant. The PRP was then subjected to a second centrifugation at 1000 × g for 10 min at 22 °C to yield platelet-poor plasma (PPP). Clot formation was triggered by adding 10 mM CaCl₂ to 400 µL of PPP in a Chrono-log cuvette (ThermoFisher Scientific), followed by parafilm sealing, gentle inversion, and incubation at 37 °C for 2 h. For clot lysis assessment, 20 µL of culture medium containing tissue plasminogen activator (tPA) was introduced 2 h post-induction, and clot size was recorded photographically over time. Experiments involving human blood samples were performed according to the directives of the Ethics Committee of Westlake University (study number: 20240311XMQ001). Quantification of target gene expression Expression levels of human placental secreted alkaline phosphatase (SEAP) in culture supernatants were quantified based on p-nitrophenyl phosphate-derived light absorbance at 415 nm. In brief, 80 µL of heat-inactivated supernatants (30 min at 65 °C) were transferred to a 96-well plate containing 100 µL of 2x SEAP assay buffer [20 mM L-homoarginine, 1 mM MgCl2, 21% (v/v) diethanolamine, pH 9.8] per well. Immediately after addition of 20 µL pNPP solution [120 mM stock solution in 2x SEAP assay buffer] per well, time-dependent increase of light absorbance at 405 nm was followed over 30 min using a Multiskan SkyHigh Microplate Spectrophotometer (ThermoFisher Scientific) set at 37 °C. SEAP levels in mouse serum were profiled using a SEAP chemiluminescence assay kit (Abcam, Cambridge, UK; cat no. AB133077). NanoLuc levels were profiled using the Nano-Glo® Luciferase Assay System (Promega, Madison, WI; cat. no. N1120). Bevacizumab, Rituximab, B6H12, ssANP, ssHwTx, and ssADPN levels in culture supernatants and animal serum were quantified with a human IgG Fc ELISA kit (Immunology Consultants Laboratory Inc., Portland, OR; cat. no. E-80G). shGLP1 levels in culture supernatants were quantified with a Bioactive GLP-1 ELISA kit (Merck Millipore, Darmstadt, Germany; cat. no. EGLP-35K). mTSLP levels in culture supernatants were quantified with a mouse TSLP ELISA kit (Solarbio Life Sciences, Beijing, China; cat. no. SEKM-0139). ssTNK levels in culture supernatants and animal serum were quantified with an Invitrogen human tPA ELISA kit (ThermoFisher Scientific, USA; cat. no. BMS258-2). Human Erythropoietin levels in culture supernatants were quantified with a Human Erythropoietin ELISA kit (Abcam, Cambridge, UK; cat no. ab274397). Quantification of murine ALT, AST, TBIL, ALB, ALP and TBA levels in animal serum was performed by Wuhan Servicebio Technology Co., Ltd. (Wuhan, China). Animal experiments 8-week-old and 60-week-old male C57BL/6 J and NSG mice were ordered through the Laboratory Animal Resources Center (LARC) of Westlake University. All mice were housed in Laboratory Animal Resources Center of Westlake University at a constant temperature and humidity, with ad libitum access to standard rodent food and water. The detailed information on AAV or plasmid injection experiments is provided in Supplementary Data 4. Hypertensive mouse model. The hypertensive mouse model was established by constant administration of 1 mg/mL L-arginine methyl ester hydrochloride (L-NAME) to the drinking water of C57BL/6 J mice over one month. Systolic blood pressure (SBP) of the animals was monitored using a tail-cuff plethysmography device (ZS-BPIV; Beijing Zhongshi Dichuang Technology Development Co., Ltd, Beijing, China). Mice with SBP levels exceeding 125 mmHg were selected into the hypertensive group for subsequent experiments. Hydrodynamic tail vein injection. Endotoxin-free plasmids were diluted in Ringer’s solution (consisting of 147 mM NaCl, 4 mM KCl, 1.13 mM CaCl2) to achieve a final injection volume of 100 µL per gram body weight and injected into tail veins of > 6-week-old mice using 5 mL syringes. Drug administration. TMP (0-4 mg/kg dissolved in saline) was administered by oral gavage. Grazoprevir (3 mg/kg dissolved in saline) was administered by intraperitoneal (i.p.) injection or oral gavage. Blood sampling. Blood was obtained from the submandibular vein of mice and allowed to clot by incubating at 4 °C for 2 h. Following this, the serum was separated from the clotted blood by centrifuging at 8000 × g for 8 min. Histopathological analysis. Liver tissues were collected and fixed in 4% Paraformaldehyde overnight, and dehydrated using an Ethanol gradient. Following clearing in xylene, tissues were embedded in paraffin and sectioned into 5 µm slices using a Leica RM2235 Microtome. Sections were stained with hematoxylin and eosin at Jingfei Biotech (Hanghzou, China), where tissue histology was observed and analyzed by a pathologist who was blinded to the treatment. Thermal pain assay. 3 days after acclimatization of mice to the testing room, animals were gently held by the experimenter wearing a cotton glove and blinded to group assignments, allowing one-third of the protruding mouse tail to immerse into a water bath set to 52 °C. The latency of the tail flick response was immediately recorded using a conventional stopwatch by setting a maximum time limit of 10 seconds. Mechanical pain assay. The mechanical sensitivity threshold of mice was measured by using a BIO-EVF4 electronic von Frey instrument (BIOSEB Inc., Vitrolles, France). In brief, mice were acclimatized to the testing room for 2–3 days before a punctuate stimulus was delivered to the mid-plantar area of each anterior paw from the mesh floor through a plastic tip, and the withdrawal threshold was automatically recorded. The final value was obtained by averaging 5 repeated measurements. Chronic inflammatory pain assay. 2–3 days after acclimatization to the testing room, mice were kept in individual chambers and injected with 10 µL of 5% buffered formalin solution on the plantar surface of a hind paw using 30 G Hamilton syringes (catalogue no. HA-80508-13) and video-captured for 1 h. The total licking and biting time of the formalin-treated paw over the course of 60 min (phase I, acute pain behaviour, 0–10 min; phase II, chronic pain behaviour, 10–60 min) was counted by an investigator who was blinded to the group assignments. Statistics and reproducibility Two-tailed unpaired Student’s t-tests or one-way ANOVA were used to evaluate differences between two groups or more than three groups, respectively, and two-way ANOVA was used to assess the effects of two independent variables and their interaction. All statistical analyses were performed using GraphPad Prism 10 (GraphPad Software, San Diego, CA, USA), and band intensities in western blot assays were analyzed with ImageJ (National Institutes of Health, Bethesda, USA). Unless otherwise stated, all experiments were conducted with at least two independent biological replicates, yielding consistent results. Sample sizes were determined based on sufficient statistical power, experimental feasibility, and convenience, with no statistical methods used to predetermine sample size; no data were excluded from the analyses. Data are presented as mean ± SD/SEM with *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 considered statistically significant; p > 0.05 was considered not significant. Detailed statistical parameters and corresponding p values are provided in the Source Data xls-file, figure legends, or the figures as indicated. For all experiments, investigators were not blinded to group allocation during experiments but were blinded during data analysis. All figures were originally created by the authors using Illustrator (V28.2, Adobe, USA), and no previously created elements were used. 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 Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Reporting Summary Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Reporting Summary Transparent Peer Review file Source data Source Data Source Data