Engineering a drug-inducible pyroptosis platform enables precise tumor suppression in colorectal cancer.

1
Introduction
Colorectal cancer (CRC) ranks among the most prevalent and lethal malignancies worldwide, accounting for substantial morbidity and mortality despite continuous advances in surgical resection, radiotherapy, and chemotherapy [1,2]. Unfortunately, the five-year survival rate for patients with advanced-stage CRC remains below 15 % [3,4], largely due to the emergence of chemoresistance and the immunosuppressive tumor microenvironment that blunts antitumor immunity. These limitations underscore the urgent need for novel therapeutic modalities that can overcome existing barriers to effective treatment [5].
Gene-based antitumor strategies—such as oncolytic viruses, RNA interference constructs, and gene-encoded antibodies—have shown promise owing to their high specificity and versatility [[6], [7], [8], [9], [10], [11], [12]]. However, challenges including off-target effects, inefficient in vivo delivery, and the risk of systemic cytokine release syndrome have thus far impeded their clinical translation [11,[13], [14], [15], [16], [17]]. To address these issues, synthetic biology has turned to inducible gene circuits that offer tight, externally controlled activation of therapeutic payloads, with applications spanning metabolic disorders and oncology [18].
Among these, drug-inducible switches are especially attractive because they leverage small molecules with established safety profiles to achieve precise spatial and temporal regulation of transgene expression and protein activity [[19], [20], [21]]. For instance, Yin et al. developed a protocatechuic acid–responsive switch in which a transcriptional repressor dissociates upon ligand binding, enabling tightly controlled insulin expression for blood-glucose management in diabetic mice and non-human primates [22]. Similarly, a switchable CAR-T platform was engineered so that rapamycin induces dimerization of the receptor's antigen-binding and signaling domains to activate the cells; in its absence, these domains remain separated, keeping CAR-T cells inactive in vivo [23].
Pyroptosis is an inflammatory form of programmed necrosis driven by gasdermin proteins, which oligomerize into transmembrane pores to rupture the plasma membrane and release intracellular contents, provoking robust immune responses [24,25]. Recent studies have shown that pyroptosis can enhance tumor immunogenicity and promote T-cell infiltration, positioning it as a potential immunotherapeutic modality [[26], [27], [28]]. GSDME—one of the six gasdermins and a key member of the pore-forming executioners in pyroptosis—consists of an N-terminal pore-forming domain (nGSDME) held inactive by a C-terminal autoinhibitory domain (cGSDME) [29,30]. Upon cleavage of the interdomain linker by caspase-3 or granzyme B, nGSDME is freed to oligomerize at the plasma membrane, compromise membrane integrity, and induce osmotic lysis [[31], [32], [33]]. This pore formation unleashes damage-associated molecular patterns (DAMPs) and proinflammatory cytokines such as IL-1β and IL-18, amplifying local inflammation [34,35]. Harnessing GSDME's potent pore-forming activity via an inducible switch could thus enable precise, on-demand execution of pyroptosis in tumor cells [36,37].
We designed and constructed a chemically inducible GSDME (CiGSDME) platform, incorporating small-molecule–responsive gene switches for precise control of GSDME-mediated pyroptosis. By replacing the native protease recognition sequence of GSDME with the TEV protease (TEVp) cleavage site [38], we generated an inactivated GSDME variant. Upon addition of the small-molecule inducer, TEVp is activated to cleave this variant, removing the C-terminal inhibitory domain and releasing the N-terminal fragment to oligomerize into membrane pores and trigger pyroptosis. We therefore assembled three small-molecule–inducible TEVp variants within the CiGSDME platform—namely CiGSDMEABA, CiGSDMEGZV, and CiGSDMEDNV—and identified CiGSDMEDNV as the most potent activator of GSDME-mediated pyroptosis. Delivery of this Danoprevir (DNV)-inducible GSDME circuit into patient-derived CRC organoids elicited robust, dose-dependent cell death. Moreover, oral administration of DNV in a mouse RKO xenograft model produced significant tumor growth inhibition. Together, these findings establish a promising strategy for CRC treatment—leveraging the safety of an FDA-approved small molecule to harness the potent antitumor activity of GSDME-driven pyroptosis—and lay the groundwork for future applications of drug-inducible cell-death switches in cancer therapy.

2
Methods and material
2.1
Ethics
All animal studies were conducted in compliance with institutional ethical standards, having received approval from the Institutional Animal Care and Use Committee (ZJCLA-IACUC-20011255) and adhering to the national guidelines for animal welfare established by the Ministry of Science and Technology of China. Human tissue sample collection and associated clinical data usage were authorized by the Ethics Committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine (KY2025-089-B).

2.2
Plasmid construction
Supplementary Table 1 outlines the complete design and assembly specifications for each expression plasmid, while Supplementary Table 2 contains the full DNA sequences of CiGSDME elements. Selected constructs were generated with the MultiS One Step Cloning kit (Vazyme, C113-01), following the supplier's protocol. Sequence validation of all plasmids was performed by Shanghai Saiheng Technology.

2.3
Cell culture and transfection
The following cell lines were maintained in DMEM (Gibco, C11995500BT) containing 10 % FBS (AusGeneX, FBSSA500-S) and 1 % penicillin-streptomycin (Beyotime, ST488-1/2) at 37 °C with 5 % CO2: HEK-293T (ATCC, CRL-11268), MDA-MB-231 (ATCC, HTB-26) and colorectal cancer lines: RKO (ATCC, CRL-2577), HT29 (ATCC, HTB-38), SW480 (ATCC, CCL-228), and SW620 (ATCC, CCL-227). B16F10 (ATCC, CRL-6475) cell line was maintained in RPMI-1640 (31800022) containing 10 % FBS (AusGeneX, FBSSA500-S) and 1 % penicillin-streptomycin (Beyotime, ST488-1/2) at 37 °C with 5 % CO2. Hepa1-6 (ATCC, CRL-1830) cell line was maintained in DMEM (Gibco, C11995500BT) supplemented with 10 % FBS (AusGeneX, FBSSA500-S), 1 % Sodium Pyruvate (Gibco, 11360070), 1 % Glutamax (Gibco, 35050061), and 1 % penicillin-streptomycin (Beyotime, ST488-1/2) at 37 °C with 5 % CO2 in a humidified incubator.
For transfection experiments, HEK-293T cells were transfected with an optimized polyethyleneimine (PEI)-based protocol [39]. Briefly, cells were seeded at 6 × 104 cells/well in 24-well plates. Following 18 h of incubation, transfection was performed using PEI (Polysciences, 24765; MW 40,000) at a 3:1 PEI:DNA ratio in 50 μL solution for 6 h. After replacing with fresh medium, solutions of different concentrations of small molecules were introduced to the supernatant 18 h post-transfection. B16F10 and Hepa1-6 cells were transfected with Lip8000™ reagent (Beyotime, C0533) according to the manufacturer's protocol.

2.4
Lentivirus production and transduction
HEK-293T cells (6.5 × 106) were seeded in 15 cm dishes and cultured for 18 h (∼70 % confluence) prior to transfection. For lentivirus production, cells were cotransfected with: Target gene expression vector (15 μg), packaging plasmid psPAX2 (15 μg; Addgene #12260) and VSV-G envelope plasmid pMD2.G (7.5 μg; Addgene #12259). Transfection was performed using a PEI-based method. After 6 h, the medium was replaced with fresh DMEM. Viral supernatants were harvested at 48 h post-transfection, filtered (0.45 μm; Pall #4654), and concentrated by centrifugation (25,000×g, 1.5 h, 4 °C). Pellets were resuspended in 200 μL PBS (Sangon Biotech #A600100-001) and stored at −80 °C.
For transduction, CRC cells (RKO, HT29, SW480, SW620; 2 × 106 cells/well in 6-well plates) and MDA-MB-231 cells were incubated with viral concentrate (MOI = 3/10/10/30/25) and polybrene (8 μg/mL; Sigma #H9268). Medium was replaced 24 h post-infection.

2.5
Lactate dehydrogenase (LDH) release assay
Cytotoxicity was assessed using the CytoTox 96 Assay Kit (Promega) according to the manufacturer's protocol. Absorbance readings at 490 nm were obtained using a BioTek microplate reader with Gen5 software (v2.04).

2.6
Chemical-inducible pyroptosis assay in mammalian cells
HEK-293T cells (6 × 104) were transfected with different CiGSDME variants (ABA, GZV, or DNV forms). Following a 6-h incubation, the transfected cells were exposed to corresponding inducers: ABA (test range: 0.01–100 μM; standard dose: 10 μM), GZV or DNV (test range: 0.01–10 μM; standard dose: 1 μM). Pyroptotic cell death was evaluated by measuring LDH release 48 h post-treatment.

2.7
Hematoxylin-eosin staining
Paraffin-embedded tumor specimens were sectioned at 4–5 μm thickness and sequentially processed through dewaxing in xylene I and II (10 min each), rehydration in a graded ethanol series (100 %→95 %→80 %→70 %, 5 min each step) with a final water rinse, nuclear staining with Harris hematoxylin (5 min) followed by water rinse, acid-alcohol differentiation (1 %, 10–15 s), lithium carbonate bluing (1 min), cytoplasmic counterstaining with 1 % eosin Y (1–2 min), gentle washing, dehydration through 95 % ethanol (twice) and 100 % ethanol (twice, 2 min each), clearing in xylene I and II (3 min each), and finally mounted with neutral balsam.

2.8
Construction of RKO xenograft model
RKO cells (5 × 106) were suspended in Matrigel (1:1 ratio) and subcutaneously implanted into the flanks of 6-week-old BALB/c-nu mice (100 μL/mouse). Tumors were allowed to grow for 25 days until reaching 50–100 mm3 before initiating interventions. The treatment regimen consisted of: Intratumoral lentiviral injection (5 × 106 TU) and oral DNV administration (50 mg/kg, Q2D) for 14 days. Tumor dimensions were monitored every 4 days, with volumes calculated as (length × width2)/2. Humane endpoints were strictly enforced: animals were euthanized by trained personnel when tumors exceeded 1500 mm3 or showed signs of distress. In rare instances where tumors surpassed this threshold between measurements, immediate euthanasia was performed.

2.9
Cell death detection with flow cytometry
Cell viability was evaluated using the Annexin V-FITC/PI apoptosis detection kit (Beyotime, C1062L) according to manufacturer's instructions. At specified time points, culture supernatant was collected and adherent cells were washed with PBS before being detached using EDTA-free trypsin. The cell suspension was then neutralized with conditioned medium and centrifuged at 300×g for 5 min. The resulting pellet was resuspended in binding buffer and stained with 5 μL Annexin V-FITC and 10 μL propidium iodide (PI) for 15 min at room temperature in the dark. Finally, the stained cells were immediately analyzed by flow cytometry using a Sony SA3800 instrument (Sony, Japan), with subsequent data analysis performed using FlowJo software (Tree Star, USA).

2.10
Organoid isolation and transfection
Human colorectal cancer organoids were established from surgical specimens by mincing tumor tissue into 1–2 mm3 fragments followed by digestion in Advanced DMEM/F12 medium supplemented with 1 mg/mL collagenase IV, 10 μM Y-27632, and 1 % penicillin-streptomycin at 37 °C for 1–2 h with gentle agitation. After digestion, the cell suspension was filtered through a 70 μm strainer, washed with PBS, and centrifuged at 300×g for 5 min before being resuspended in Matrigel and plated as 20 μL domes in 48-well plates. Organoids were maintained in complete culture medium containing B27, N2, EGF, Noggin, and R-spondin at 37 °C with 5 % CO2. For lentiviral transduction, organoids at 60–70 % confluence were dissociated into 10–20 cell clusters, incubated with lentivirus (MOI = 10) in medium containing 8 μg/mL polybrene for 24 h, and then re-embedded in Matrigel as 20 μL droplets in 48-well plates for continued culture.

2.11
Organoid death detection
Organoid viability was evaluated 48 h post-DNV treatment (1 μM) using a CAM/PI dual-staining assay (Solarbio, CA1630). Following lentiviral transduction and drug exposure, samples were washed with cold PBS and incubated in 1 × detection buffer containing 1 μM CAM and 2 μM PI for 30 min (37 °C, dark). After removal of excess dye, fluorescence imaging was performed with filter sets optimized for CAM (488/515 nm) and PI (561/617 nm).

2.12
Statistical analysis
All in vitro results are presented as mean ± SD (n = 3 biological replicates), while in vivo data show mean ± SEM (n = 5 mice/group). Researchers conducting histopathology and animal studies were blinded during sample processing and data collection, though other experiments involved unblinded analysis. Statistical comparisons employed two-tailed unpaired t-tests (two groups) or one-way ANOVA (multiple groups) using GraphPad Prism 8.0.1, with significance thresholds set at ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001. No exclusions were applied to experimental subjects or data points.

3
Results
3.1
Development of a CiGSDME for precise pyroptosis
Pyroptosis is executed by pore-forming proteins such as GSDME, which is activated through proteolytic cleavage to release its N-terminal effector domain [[29], [30], [31]]. To achieve inducible and orthogonal control of pyroptosis in tumor cells, we engineered a variant of gasdermin E (designated TEV-GSDME), in which the endogenous caspase-3 cleavage site was replaced with a tobacco etch virus protease (TEVp)-specific recognition sequence. This strategy reduces interference from native caspase activity, thereby limiting off-target pyroptosis in healthy cells. Upon recognition and cleavage by TEVp, the inhibitory C-terminal domain of the engineered GSDME is cleaved off, releasing the N-terminal fragment to oligomerize and form membrane pores that trigger pyroptosis (Fig. 1A). This re-engineered GSDME construct provides the foundation for a modular, chemical-inducible pyroptosis platform—CiGSDME.
To regulate TEVp activity, we employed an abscisic acid (ABA)-inducible dimerization system [40]. ABA is a non-toxic and cost-effective plant hormone commonly used in synthetic biology. TEVp was split into two inactive fragments, each fused to ABA-responsive dimerization domains (ABI or PYL1). Following ABA treatment, the fragments heterodimerize, reconstituting an active TEVp enzyme that cleaves TEV-GSDME and initiates pyroptosis (Fig. 1B).
We next assessed ABA-induced pyroptosis in vitro by co-transfecting HEK-293T cells with the split TEVp constructs (ABI-nTEVp and PYL1-cTEVp) [41] along with TEV-GSDME. In the presence of ABA, cells exhibited characteristic pyroptotic morphology, including cytoplasmic swelling and plasma membrane rupture (Fig. 1C). Furthermore, Annexin V-FITC/PI staining revealed a significant increase in the proportion of cells with compromised membrane integrity compared to the ABA-free control group (Fig. 1D and E). To further quantify cytolytic cell death, we performed a lactate dehydrogenase (LDH) release assay. In line with the Annexin V/PI results, ABA treatment resulted in a marked elevation in LDH release, confirming pyroptosis-induced cytotoxicity (Fig. 1F).
Given that ABA is not clinically approved, we sought to develop clinically relevant gene switches using two FDA-approved antiviral drugs, grazoprevir (GZV) and Danoprevir (DNV), which have been shown to induce dimerization of NS3a/GNCR1 and NS3a/DNCR2, respectively [42]. Treatment with GZV or DNV resulted in a significant increase in LDH release, indicating successful activation of pyroptosis by these clinically compatible switches (Fig. 1G and H). Importantly, LDH assays indicated that treatment with ABA, GZV, or DNV alone did not result in a significant increase in LDH release compared to untreated controls (Fig. S1), confirming that the observed cell death was attributable to drug-induced pyroptosis rather than toxicity of the small molecules themselves.

3.2
Characterization of the CiGSDME for inducing cell pyroptosis
To gain a comprehensive understanding of the performance and responsiveness of our CiGSDME, we systematically characterized its activation profile under the control of three small-molecule inducers: abscisic acid (ABA), grazoprevir (GZV), and Danoprevir (DNV). HEK-293T cells were transfected with the corresponding CiGSDME constructs and subsequently treated with a range of concentrations of ABA, GZV, or DNV. The level of pyroptosis induction was quantitatively assessed by measuring lactate dehydrogenase (LDH) release into the culture supernatant (Fig. 2A). Across all three systems, we observed a clear concentration-dependent increase in LDH release, indicating that inducer dose directly modulates the extent of TEVp activation and subsequent GSDME cleavage. Notably, even at submicromolar concentrations as low as 0.01 μM, all three inducers elicited a statistically significant elevation in LDH release compared to untreated controls, underscoring the high sensitivity of the CiGSDME platform to small-molecule inputs (Fig. 2B–D).
Further dose–response analysis revealed that each inducer exhibited a distinct threshold and saturation point for pyroptosis activation. Specifically, ABA-induced LDH release plateaued at 10 μM, while both GZV and DNV reached maximal induction at 1 μM. Among the three systems, DNV consistently produced the highest LDH levels at saturating doses, suggesting superior efficiency of the DNV-responsive switch in mediating pyroptotic cell death (Fig. 2B–D). These findings highlight the tunable nature of the CiGSDME platform and suggest that inducer choice can be strategically selected to balance sensitivity, dynamic range, and translational potential.
To further investigate the temporal dynamics of pyroptosis induction, we monitored LDH release over time following exposure to the saturating concentrations of each inducer. Kinetic analysis showed that LDH release began to increase at approximately 12 h post-induction, with statistically significant differences emerging between treated and control groups at later time points (Fig. 2E–G). This time-resolved data indicates that CiGSDME-mediated pyroptosis progresses in a controlled, time-dependent manner, providing a useful window for downstream functional studies or therapeutic intervention. Overall, these data show that CiGSDME is tunable, time-controlled, and responsive to different inducers, making it a flexible system for regulated pyroptosis in mammalian cells.
To test its relevance in cancer models, we introduced the CiGSDMEDNV into four human colorectal cancer (CRC) cell lines—RKO, HT29, SW480, and SW620—and measured LDH release after DNV treatment. All four cell lines showed strong, DNV-dependent increases in LDH levels, indicating that the system worked well across different CRC backgrounds (Fig. 2H–K). These findings suggest that CiGSDME is broadly compatible with colorectal cancer cells and may be useful for targeted colorectal cancer therapies.
To further evaluate its versatility, we next tested the system in three additional cancer cell lines representing other tumor types: Hepa1-6 (murine hepatocellular carcinoma), B16F10 (murine melanoma), and MDA-MB-231 (human triple-negative breast cancer). In all cases, CiGSDMEDNV induced marked, drug-dependent pyroptosis, underscoring its applicability beyond CRC and highlighting its potential for broader oncological applications (Fig. S2).

3.3
CiGSDMEDNV-mediated tumor inhibition in Patient‐Derived colorectal cancer organoids
To evaluate the translational potential of CiGSDMEDNV in a clinically relevant model, we established patient-derived organoids (PDOs) from colorectal cancer (CRC) tissues and tested the efficacy of CiGSDMEDNV in this ex vivo setting (Fig. 3A).
Tumor tissues were obtained from two CRC patients, and their malignant characteristics were histologically confirmed by hematoxylin and eosin (H&E) staining prior to organoid derivation (Fig. 3B). Detailed clinicopathological features of the enrolled patients are summarized in Supplementary Table S3. Fresh tumor specimens were mechanically minced, enzymatically digested, and filtered to obtain single-cell suspensions, which were then embedded in Matrigel for three-dimensional (3D) culture. The organoids were maintained in a specialized culture medium, refreshed every three days, and their growth was continuously monitored under a microscope. Microscopic examination confirmed that the organoids exhibited characteristic vacuolar morphology, consistent with established CRC organoid features (Fig. 3B).
The resulting CRC PDOs were transduced with lentiviral vectors encoding the DNV-responsive CiGSDMEDNV construct. Following transduction, organoids were treated with Danoprevir (DNV) to initiate TEVp activation and subsequent GSDME-mediated pyroptosis. Cell death was assessed using a Calcein AM/propidium iodide (PI) dual-staining assay, which simultaneously labels live cells with intact membranes (Calcein AM+, green fluorescence) and dead or dying cells with compromised membrane integrity (PI+, red fluorescence) [43]. DNV treatment led to a significant increase in PI-positive cells within the organoids, indicating effective induction of pyroptotic cell death. In contrast, organoids transduced with the same construct but not exposed to DNV showed minimal PI signal, suggesting low basal activation and minimal background toxicity (Fig. 3C). This observation was consistently reproduced across both patient-derived organoid lines, demonstrating the high inducibility and minimal leakiness of CiGSDMEDNV under non-induced conditions.
These findings clearly illustrate that the CiGSDMEDNV enables potent and selective induction of cell death in human CRC organoids upon small-molecule stimulation. The ability to trigger pyroptosis in PDOs—a model that retains the structural, genetic, and phenotypic features of primary tumors—underscores the potential of this approach for therapeutic applications in colorectal cancer.

3.4
CiGSDMEDNV-mediated tumor inhibition in an RKO mouse model
To assess the in vivo therapeutic efficacy of the DNV-inducible CiGSDME system, we employed a colorectal cancer xenograft model using RKO cells, which harbor an MSI-high/CIMP-high genomic profile associated with sensitivity to immunotherapy and reliably form tumors in vivo [44]. We assessed tumor growth inhibition mediated by CiGSDMEDNV-induced pyroptosis in this xenograft model (Fig. 4A). RKO cells were subcutaneously implanted into the dorsal flank of mice, and once the tumor volume reached 50–100 mm3, the tumors were subsequently transduced with lentiviruses carrying the DNV-responsive CiGSDMEDNV construct. To improve infection efficiency, we combined three vectors of CiGSDMEDNV into two vectors before viral transduction (Fig. S3). Mice were then randomly assigned to one of four treatment groups: PBS (G1), pLL3.7 (G2), CiGSDMEDNV DNV (0 mg/kg) (G3), and CiGSDMEDNV DNV (50 mg/kg) (G4). On days 0, tumors were injected with the respective lentiviral vectors. DNV was administered orally at a dose of 50 mg/kg every two days for two weeks, and tumor size was monitored over a period of one month (Fig. 4B).
Compared to the other groups, the G4 group showed a significant reduction in tumor weight (Fig. 4C) and marked inhibition of tumor growth (Fig. 4D). Importantly, tumor volume in the G3 group remained comparable to those in the PBS and vector-only controls, suggesting that the tumor suppression observed in the G4 group was specifically induced by DNV-activated CiGSDMEDNV-mediated pyroptosis. These findings demonstrate that the CiGSDMEDNV platform enables efficient induction of tumor cell pyroptosis and robust suppression of tumor progression, supporting its therapeutic potential in colorectal cancer.

4
Discussion
Pyroptosis has emerged as a promising strategy for cancer therapy due to its ability to induce immunogenic cell death and enhance antitumor immune responses [45]. However, the lack of precise and tunable control over pyroptosis limits its utility in targeted applications. Although GSDME can be activated by caspase-3 to induce pyroptosis, this endogenous mechanism is embedded within the apoptotic pathway and cannot be externally regulated. As a result, therapeutic strategies relying on native caspase activation may suffer from off-target effects, limited timing control, and insufficient tissue specificity.
To overcome these limitations, we developed CiGSDME—a synthetic biology–based platform for programmable pyroptosis induction through small-molecule control. Mechanistically, this system decouples pyroptotic execution from the endogenous caspase network by engineering a TEV protease recognition sequence into the GSDME linker region, allowing GSDME activation only upon orthogonal protease input. Furthermore, the protease itself is split and fused to chemically inducible dimerization domains, forming a two-layer regulatory logic that minimizes leaky activation and enhances temporal resolution. Importantly, this design ensures that GSDME variant remains inert in the absence of exogenous input, preventing unintended pyroptosis and enhancing safety.
Compared with conventional transcriptional or optogenetic switches, which often require hours to days for response or involve specialized hardware [[46], [47], [48]], CiGSDME enables rapid, post-translational control of cell death using pharmacologically tractable inputs. To facilitate clinical translation, we reconfigured the system to respond to two FDA-approved antivirals, grazoprevir (GZV) and Danoprevir (DNV), which induce protease reconstitution via established small-molecule–dependent dimerization modules. Notably, DNV demonstrated superior sensitivity and potency, likely due to its pharmacokinetic properties or its interaction dynamics with the protease domains. Time-course studies revealed a clear initiation of pyroptotic signaling at 12 h post-treatment, confirming the system's controllability and tunability.
The use of patient-derived colorectal cancer organoids further supports the clinical relevance of CiGSDMEDNV. These organoids, which preserve intratumoral heterogeneity and drug responses, offer a closer approximation to human disease than traditional cell lines. The fact that DNV-induced activation triggered robust pyroptotic death only in the presence of the system highlights its specificity and low background activity. This opens the door for combinatorial strategies, such as pairing CiGSDMEDNV with immune checkpoint inhibitors, where controlled pyroptosis may improve antigen presentation and immune infiltration. Compared with apoptosis-based killing (e.g., Bax overexpression) or purely cytotoxic effectors (e.g., DTA, HSV-TK), CiGSDME-mediated pyroptosis uniquely couples efficient tumor cell lysis with potent pro-inflammatory and immunogenic effects, enabling the conversion of “cold” tumors into “hot” tumors and bypassing apoptosis resistance, thereby offering superior translational potential in both apoptosis-refractory and immune-desert cancers [49]. In our BALB/c-nu mouse model, which lacks functional T cells, CiGSDME-mediated tumor suppression is likely driven by both direct tumor cell lysis and robust innate immune activation (e.g., NK cells, macrophages, dendritic cells), indicating that both mechanistic arms remain operative even in the absence of adaptive immunity.
In vivo validation in the RKO xenograft model demonstrated that oral delivery of DNV was sufficient to activate the system and suppress tumor growth, supporting the translational potential of CiGSDMEDNV. However, systemic delivery of small-molecule inducers may still lead to non-specific activation in non-tumor tissues, especially in systems with leaky expression or widespread promoter activity. This emphasizes the importance of developing delivery systems that enable localized induction of the circuit. Potential scalable and clinically relevant strategies could include systemic administration via adeno-associated virus (AAV) vectors, lipid nanoparticles, or the incorporation of tumor-specific promoters, each of which may improve targeting efficiency and reduce off-target effects. Future work should focus on developing spatially gated versions of the GSDME variant. For example, pairing drug-inducible switches with tumor-specific promoters, light-activated control, or nanoparticle-based delivery vehicles may further enhance specificity. Alternatively, integrating dual-input logic gates requiring both chemical and tissue-specific signals could increase safety. Beyond DNV, other clinically relevant small-molecule inducers—such as rapamycin/rapalogs for FKBP–FRB chemically inducible dimerization systems [50]—could be adapted to drive CiGSDME activation, providing additional flexibility for translational applications.
Future work should focus on developing spatially gated versions of GSDME variant. For example, pairing drug-inducible switches with tumor-specific promoters, light-activated control, or nanoparticle-based delivery vehicles may further enhance specificity. Alternatively, integrating dual-input logic gates requiring both chemical and tissue-specific signals could increase safety. From a systems perspective, the modularity of CiGSDME makes it broadly adaptable—not only to other solid tumors beyond colorectal cancer, but also to diverse effectors beyond pyroptosis, such as necroptosis or ferroptosis, depending on therapeutic needs.
In summary, CiGSDME offers a versatile platform for externally regulated, clinically compatible pyroptosis induction. Its tight control, drug responsiveness, and compatibility with human-relevant models position it as a promising tool for precision cancer therapy. With further optimization of delivery and activation specificity, this strategy may contribute to the next generation of programmable cell-death–based therapeutics.

CRediT authorship contribution statement
Xiang Yao: Writing – original draft, Validation, Investigation, Data curation. Yu Wei: Investigation, Data curation. Yuan Gao: Validation, Investigation. Lei Li: Validation, Data curation. Junchi Liu: Investigation. Wenmin Zhou: Data curation. Tao Yan: Investigation. Letian Gong: Investigation. Yang Zhou: Writing – review & editing, Supervision, Funding acquisition, Conceptualization. Ganglong Gao: Writing – review & editing, Supervision, Resources, Conceptualization.

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