Peripheral blood was collected from healthy donors following written informed consent, and all animal experiments were conducted according to the institutional review board and research ethics committees of the Huazhong University of Science and Technology, PR China (2019-S1010 and 2019-IEC-S213) to ensure the ethical and welfare of human participants and animals involved in the research process.

Cell lines in this study, including A375 (human melanoma cell line), A549 (human lung carcinoma cell line), BGC823 (human gastric carcinoma cell line), Hep2 (human head and neck squamous cell carcinoma cell line), HuH-7 (human hepatocellular carcinoma cell line), LoVo (human colon carcinoma cell line), PANC-1 (human pancreatic carcinoma cell line), U87MG (human glioblastoma cell line), 5637 (human bladder carcinoma cell line), MRC-5 (human embryonic lung fibroblasts line), and HEK293 (human embryonic kidney cell line) were obtained from the Cell Bank of Chinese Academy of Sciences (SGST, Shanghai, China). The composition of the culture medium is DME/F12 (Invitrogen, Carlsbad, CA) medium supplemented with 10% fetal bovine serum (HyClone, Logan, UT), 1% L-glutamine (Invitrogen, Carlsbad, CA), and 1% penicillin-streptomycin (Invitrogen, Carlsbad, CA) in a CO₂ incubator at 37°C for most of these cell lines, while it will undergo appropriate modifications for the other lines. FAP5 cDNA was amplified using FAP-specific primers in 2× qPCR BIO SyGreen Blue Mix Hi-ROX Master Mix (PCR BioSystems) and then transduced into various cancer cells through the well-constructed plasmids to obtain the cancer cells with high expression of FAP5.

Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll separation method (Stem Cell Technologies, Vancouver, BC, Canada). PBMCs were cultured in RPMI 1640 medium supplemented with 200 IU/ml human recombinant interleukin (IL)-2 and activated by Dynabeads™ CD3/CD28 Human T-activator (Cat. 11131D, Thermo Fisher, Waltham, MA) for 3 days according to the manufacturer’s instruction. Alternatively, they can be directly isolated using a blood cell separator (Fresenius CEM TEC., Germany).

The activated PBMCs were directly used in various experiments in vivo or in vitro or transfected with the lentivirus expressing CARs to obtain CAR-T^HN3 targeting GPC3 by us (15). In addition, the immunotoxins (scFv fused with PE24, the 24-kDa cytotoxic domain of Pseudomonas exotoxin A) targeting the membrane-distal N-lobe of GPC3 previously developed by us, named as J80A-PE24, could suppress tumor growth much greater than naked HN3-PE24 in a xenograft mouse model (27).

The DSTE^FAP5/CD3 was produced by fusing the single-chain variable fragments (scFvs) of anti-FAP 28H1 from patent EP3333194A1-1 and anti-CD3 OKT3 to construct the FAP5/CD3-DSTE expression fragment.

The final fragment VH_FAP5-VL_FAP5-VH_CD3-VL_CD3 was constructed by Nanjing Kingsley Company and assembled into the pHG52d34.5-CMV shuttle vector (also referred to as pGFP) established in our laboratory for subsequent homologous recombination. The parental virus OHSV2-GFP and OHSV2-DSTE^CD19/CD3 derived from laboratory were utilized as controls. The genome of OHSV2-DSTE^FAP5/CD3 was obtained by recombineering and subsequently sequenced by the Tsingke Biotech (Beijing, China). The sequencing results were aligned to the original FAP/CD3-DSTE plasmid sequences by SnapGene software. OHSV2-DSTE^FAP5/CD3, OHSV2-DSTE^CD19/CD3, and OHSV2-GFP were used to infect Vero cells for 48–72 h, and the supernatant was collected after the addition of a virus-releasing solution. Afterwards, virus was purified and titrated.

OHSV2-DSTE^FAP5/CD3 was used to infect Vero cells for 48 h at multiplicity of infection (MOI) of 0.01. After infection, the viral supernatants were collected and purified using High-Affinity Ni-Charged Resin (GenScript). Subsequently, the DSTE^FAP5/CD3 protein released by infected cells was eluted with the appropriate buffer. The resultant protein solution was purified through 20K Slide-A-Lyzer Dialysis Cassettes according to the manufacturer’s instructions (Thermo Fisher).

Flow cytometry was performed on Accuri C6 cytometer (BD Biosciences, Franklin Lakes, NJ). Human-derived FAP expression was detected with anti-FAP antibody (phycoerythrin, accession number: MIH1, BD Biosciences). T cells were analyzed using the following antibodies: anti-CD3 (APC, accession number: HIT3a, BD Biosciences), anti-CD4 (FITC, accession number: RPA-T4, BD Biosciences), anti-CD8 (FITC, accession number: RPA-T8, BD Biosciences), anti-CD25 (PE, accession number: MA251, BD Biosciences), and anti-CD69 (APC, accession number: FN50, BD Biosciences). Cytokine, including IL2, IL4, IL6, IL10, TNF, and IFN-γ in culture supernatants or peripheral blood, was detected with BD™ Cytometric Bead Array (CBA) Human Th1/Th2 Cytokine Kit II. All flow cytometry data were processed with FlowJo v7.6.5 and FCAP Array v3.0.

To investigate the cytotoxicity of free DSTE, oncolytic virus, and T-cell-mediated killing, an average of 1.5 × 10⁴ cancer cells per well was seeded into E-Plate 16 (ACEA Biosciences Inc, San Diego, CA) and co-cultured with pre-activated PBMCs (cancer cells, PBMCs = 1:2) or/and fibroblasts treated by OHSV2-DSTE^FAP5/CD3 (MOI = 0.1) alone or the other agents and OHSV2 or OHSV2-DSTE^CD19/CD3 (MOI = 0.1) as control virus. Cellular vitality of cancer cells or others was continuously monitored by the xCELLigence Real-Time Cell Analyzer (ACEA Biosciences Inc, San Diego, CA), following the manufacturer’s protocol. The half-maximal inhibitory concentration (IC50) was calculated by a dose–response inhibition (variable slope) curve with GraphPad Prism V8.0 (GraphPad Software, Inc, La Jolla, CA).

We selected engineering high-expressing FAP5 cancer cells, BGC823-GFP-FAP5 cell lines with three different levels of FAP5 expression to investigate free DSTE or OHSV2-DSTE^FAP5/CD3 impairment on the cell proliferation in vitro continuously in the above-mentioned cell proliferation experiment.

RNA was extracted from cell lines, mouse tumor tissue, or rabbit spleen, or other components using RNA Simple Total RNA Kit (Tiangen Biotech, Beijing, China), and reverse transcription was carried out to detect levels of IL2RA, GZMB, and PRF1, which are T-cell activation markers by RT-qPCR using 5 × HiScript II QRTSuperMix II (Vazyme Biotech, Nanjing, China) and iTaq Universal SYBR Green Supermix (Bio-Rad, Redmond, WA). Specific primers were designed through the software of Invitrogen’s Vector NTI^® Advance 11.5.1. The RT-qPCR procedures briefly were as follows: pre-denaturation at 95°C for 1 min, followed by 40 cycles of 95°C for 5 s, 61°C for 31 s, 95°C for 15 s, and 60°C for 60 s, ending up with heating to 95°C. The relative expression level of the target gene was calculated by the 2^−△△CT method with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene as an internal reference for three independent biological replicates.

6-week-old female BALB/c nude mice were purchased from the Animal Center of Huazhong Agricultural University. Three million HuH-7 cells were subcutaneously injected into the right flank of nude mice feeding under specific pathogen-free conditions. After tumor mass reached the size of approximately 100–200 mm³, mice were randomized into seven groups listed in Table 1 for details. Tumor volume was calculated from the formula: tumor volume (mm³) = (length × width × width)/2 by digital calipers. We first carried out dose-found research of OHSV2-DSTE^FAP5/CD3 in vivo in a HuH-7 subcutaneous mouse model intratumorally injected with three different doses of OHSV2-DSTE^FAP5/CD3 (CCID₅₀ = 1 × 10⁶, 1 × 10⁷, and 1 × 10⁸, respectively) or control OHSV2-GFP, once a week, for a total of 4 weeks. Subsequently, formal high-order combination research will be conducted according to the specified schedule and dosages shown in Table 1 . Mice in the control group and treatment groups were sacrificed for analysis at day 42 after tumor inoculation or at any time due to the oversized tumors.

To examine whether the functions of major organs and systems such as liver, kidney, bone marrow, myocardium, and endocrine system have been impaired by different schedules of administration, we collected peripheral venous blood of mice at the end of the experiment and analyzed levels of various serum enzymes (such as alanine aminotransferase, ALT; aspartate aminotransferase, AST; and creatine kinase, CK), blood glucose, blood lipids, albumin, total bilirubin, creatinine, electrolytes, and cytokine.

Once the mice were sacrificed by cervical dislocation, major organs and partial tumor tissue were collected and immersed in 10% neutral-buffered formalin for fixation. Subsequently, the samples were embedded in paraffin, sectioned (5 μm), and hematoxylin–eosin stained. Additionally, the levels of FAP5 expression were detected by immunohistochemical staining, and grading was assessed by a digital camera (Leica ICC50 HD, Germany) to gather the area and density of the dyed region and calculate the integrated optical density (IOD) value, from five randomly selected fields (Image-Pro Plus 6.0).

These analyses are independently evaluated by two pathologists, and in case of inconsistency, a discussion will be held to make the decision.

For the high-order combined therapy group, we conducted a preliminary analysis of the mechanism by single-cell RNA sequencing, in addition to T-cell activation characterization and cytokine levels. Tumor tissues were harvested from mice in different treatment groups at the end of the experiments for sample preparation and single-cell RNA sequencing. In short, after digested into cell suspension, they were filtered by a 40-μm strainer and resuspended in the PBS solution to obtain single cells for single-cell sequencing. Single-cell capture was performed by the BD Rhapsody Single-Cell Analysis System (BD Biosciences, Franklin Lakes, NJ) for library construction. Upon preparation of the libraries, they were quantified using the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA) and the Qubit 4.0 (Thermo Fisher Scientific, USA). Finally, the libraries were sequenced on the Illumina NovaSeq 6000 (Illumina, USA), and 300-bp reads (including 150-bp paired-end reads) were generated. Single-cell RNA sequencing data were subjected to multiple analyses including quality control, alignment, clustering, marker gene identification, annotation of clusters, and other analyses.

The clustering results were visualized using t-distributed stochastic neighbor embedding (tSNE) and uniform manifold approximation and projection (UMAP). The marker genes for each cluster were identified using the default parameters through the “FindAllMarkers” function in Seurat. The original clusters were annotated based on the MouseRNAseqData dataset via SingleR (v.1.0.1). T-cell clusters were extracted for further sub-analysis using the subset function. To enhance the distinction between cell types, the ImmGenData dataset was utilized for cluster annotation. Differentially expressed genes (DEGs) related to interleukin (IL) and interferon (IFN) family were screened from the T-cell cluster with |Log2-fold change |  > 0.5 and p-value < 0.05 as thresholds and visualized through violin plot by Seurat. Furthermore, genes related to CAFs and T-cell exhaustion were also analyzed.

Quantitative data are usually displayed as mean ± SD from at least three biological replicates and related to the control. Two-tailed unpaired Student’s t-tests were performed for comparisons of two independent data sets. Comparisons among three or more groups were performed using two-way ANOVA with Tukey’s multiple comparison test. The statistical significance was denoted as follows: n.s, non-significant; *p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001. Statistical analysis was conducted by GraphPad Prism 8.0.

DSTE^FAP5/CD3 was constructed using the FAP5 monoclonal antibody 28H1, paired with the CD3 monoclonal antibody OKT3. In combination with peripheral blood mononuclear cells (PBMCs), DSTE^FAP5/CD3 significantly inhibited the proliferation of BGC823 cells overexpressing FAP5, and fibroblasts, in a time-dependent manner, with maximal effect at 60 h (95% tumor cell apoptosis) monitored by CellInsight CX5 high-content system ( Figures 1A, B ), compared with DSTE^CD19/CD3 (preserved in our laboratory only utilized as a control).

Mechanistically, our results showed a significant upregulation in the proliferation of both CD4+ and CD8+ T cells in the DSTEFAP5/CD3 plus PBMC group (2,557 ± 76.59 and 9,362 ± 101.8, respectively), compared to the control group (1,141 ± 79.23 and 6,215 ± 120.7, respectively, p < 0.05) ( Figures 1C, D ). In addition, the proliferation amplitude between CD4+ and CD8+ T cells was approximately similar in DSTE^FAP5/CD3 plus PBMC group. Furthermore, multiple cytokine levels in the supernatant were found to be significantly elevated, particularly IFN-γ, IL-6, and TNF ( Figures 1E, F ).

Next, DSTE^FAP5/CD3 was transduced into OHSV2 to generate a novel virus, named OHSV2-DSTE^FAP5/CD3 ( Figure 2A ), which was used to infect Vero cells. At 60 h post-infection, the concentration of DSTE^FAP5/CD3 reached 0.12 ng/ml, despite lower than that achieved by recombination methods, indicating that insertion of DSTE gene fragments did not impair replication of OHSV2 relative. Subsequently, we collected the supernatant and found that DSTE^FAP5/CD3 also exhibited killing activity ( Figures 2B, C ). Moreover, the levels of DSTE^FAP5/CD3-induced cytokines, including IL-2, IL-4, IL-6, IL-10, TNF, and IFN-γ were markedly elevated in vitro, indicating the activation of T cells and immune function enhancement in BGC823 cells overexpressing FAP5 cells and MRC-5 cells ( Figures 2D–G ), suggesting that OHSV2-DSTE^FAP5/CD3 is successfully constructed; thus, in-depth exploration is necessary to achieve our therapeutic objectives.

To further explore the lytic function of OHSV2-DSTE^FAP5/CD3, we initially examined human melanoma cells, which demonstrate relatively higher sensitivity to most immunotherapy drugs. Our findings showed that OHSV2-DSTE^FAP5/CD3 can exhibit uncompromising lytic activity compared to its parental virus ( Figures 3A, B ). Furthermore, the in vitro anti-tumor efficacy of OHSV2-DSTE^FAP5/CD3 against HuH-7 cells was significantly greater than that observed in CT-26, MC-38, 4T1, and BGC823 cell lines lacking FAP5 expression, and analogous to control virus, such as OHSV2-DSTE^CD19/CD3 or OHSV2, albeit weaker than its effect on melanoma cells ( Table 2 ). These results imply that HuH-7 cells possess an inherent susceptibility to immune-mediated killing induced by OHSV2-DSTE^FAP5/CD3, accompanied by significant T-cell activation ( Figures 3C, D ), prompting our team to select this cell line for further investigation.

OHSV2 demonstrates limited replication capacity and consequently fails to directly eliminate fibroblasts or other nonepithelial stromal cells with normal antiviral pathways, such as interferon secretion. Furthermore, co-culturing cancer cells and fibroblasts may affect their respective survival rates, although these effects can vary depending on multiple factors. Hence, co-cultivation of HuH-7 and FAP-expressing stromal fibroblasts was lastly carried out. The results suggested that OHSV2-DSTE^FAP5/CD3 exhibited direct HuH-7 cell killing and T-cell-mediated fibroblast elimination ( Figure 3E ). However, the potential enhancement of oncolytic effects through CD3-induced T-cell clustering activation remains to be elucidated. Moreover, OHSV2-DSTE^FAP5/CD3 exhibits negligible cytotoxicity toward fibroblasts in the absence of cancer cells or PBMCs, indicating that its function is constrained by environmental conditions, which may confer potential safety advantages for further exploration and even clinical development.

Furthermore, the levels of multiple cytokines, including IL-2, IL-4, IL-6, IL-10, TNF, and IFN-γ in the supernatant of the OHSV2-DSTE^FAP5/CD3 group co-cultured with cancer cells and fibroblasts, showed a significant increase compared to groups with either cell type alone ( Figure 3F ). This suggests that T cells derived from PBMCs are primarily activated by OHSV2, with DSTE playing a secondary regulatory role. It should be noted that our interpretation is based solely on a correlation study and requires further validation.

Subsequently, the FAP5-targeted function of OHSV2-DSTE^FAP5/CD3 in high FAP5 expression setting was explored. Initial assessment of FAP5 expression levels across various tumor cell lines revealed that only MRC-5 and U87MG cell lines exhibited high expression levels of FAP5, with an expression of 45% and 18%, respectively ( Figures 4A, B ). Conversely, FAP5 expression was nearly undetectable in other malignant tumor cells such as A375 and A549, consistent with its established role as a marker for tumor stroma. Following 48-h co-incubation, OHSV2-DSTE^FAP5/CD3 can perform greater cancer cell lysis in a FAP5 level-dependent manner, through comparing three cell lines of BGC823 with 90%, 53%, and 17% of FAP5 expression levels to parental BGC823 ( Figures 4C, D ) and HuH-7 (p < 0.05), suggesting that the additive antitumor activity of OHSV2-DSTE^FAP5/CD3 is dependent on the direct on-target effect of DSTE^FAP5/CD3. Moreover, T cells were significantly activated by OHSV2-DSTE^FAP5/CD3, supported by the upregulation of T-cell activation-related genes, including GZMB, IL-2RA, and PRF1, and increased secretion of cytokines (IL2, IL4, IL6, IL10, TNF, and IFN-γ) ( Figures 4E, F ).

These results collectively demonstrated that OHSV2 is an outstanding platform to express DSTE^FAP5/CD3, and HCC is a preferred prey for OHSV2-DSTE^FAP5/CD3 because of at least triple effects of promoting proliferation and activation of T cells, along with a direct oncolytic effect observed in vitro.

Based on the aforementioned findings, OHSV2-DSTE^FAP5/CD3 has been characterized as possessing robust oncolytic activity, moderate stromal elimination capability, significant immune activation potential, and a favorable safety profile. In addition, our previous research demonstrated that the CAR-T cells (CAR-T^HN3) and immunotoxins (J80A-PE24) targeting GPC3 developed in our laboratory exhibited significantly greater cytotoxicity than single-agent treatments, although further optimization is warranted. Therefore, we propose a high-order combination regimen comprising OHSV2-DSTE^FAP5/CD3, CAR-T^HN3, and J80A-PE24. This combination is anticipated to exhibit synergistic anti-tumor effects and is theoretically feasible while circumventing challenges associated with integrating therapeutics from different teams in future clinical studies.

First, we found that the combination of OHSV2-DSTE^FAP5/CD3 with GPC3-targeting immunotoxins exhibited significantly enhanced cytotoxicity against HuH-7 cells (p<0.05) compared to monotherapies ( Figure 5A ). It is worth mentioning that OHSV2-DSTE^FAP5/CD3 plus CAR-T^HN3 induced higher cytokine levels than single-agent treatment ( Figure 5B ), which is a major contributor to severe toxicity in CAR-T cell therapy, although showing significant synergistic anti-tumor effects. Thus, we reduced the CAR-T cells dosage by 50% and found that cytokine levels were similar to those in the other treatments but with a 30% decrease in efficacy. Therefore, this adjusted dosing strategy was employed in subsequent in vivo experiments to balance safety and efficacy. Lastly, what exceeded our expectations is that adding GPC3-targeting immunotoxins into OHSV2-DSTE^FAP5/CD3 and GPC3-targeting CAR T cells group resulted in a significant increase in the killing effect on tumor cells ( Figure 5C ). We speculated that this may be due to a saturation of cytotoxic potential for the latter combination in vitro, where no major barriers hinder effector cell infiltration and activation. However, this scenario is difficult to replicate in vivo, especially in solid tumors. Therefore, this strategy remains valuable to be explored in HuH-7 mice models.

In addition, our results showed that T-cell activation markers, including GZMB, IL2RA, and PRF1, were more significantly upregulated in the OHSV2-DSTE^FAP5/CD3 plus CAR-T cell group compared to the OHSV2-DSTE^FAP5/CD3 or CAR-T cell group, suggesting an enhancement in T-cell effector function, consistent with the findings from plenty of preclinical studies ( Figures 5D, E ).

Given that GPC3-targeting immunotoxins and GPC3-targeting CAR T cells have already been explored in vivo in our previous research, we conducted a dose-finding research of OHSV2-DSTE^FAP5/CD3 in HuH-7 subcutaneous mouse model, intratumorally injected with three different doses of OHSV2-DSTE^FAP5/CD3 (CCID₅₀ = 1×10⁶, 1×10⁷, and 1×10⁸) or control OHSV2-GFP. A relatively low dose of 1×10⁷ was selected for subsequent experiments, as it exhibited more significant tumor regression than the 1×10⁶ group or OHSV2-GFP group ( Supplementary Figure 1 ) while showing comparable efficacy to the 1×10⁸ group. However, a tendency of gradual acceleration in tumor growth was observed at a later stage, even though OHSV2-DSTE^FAP5/CD3 treatment effectively inhibited the tumor growth initially. These observations suggested that resistance to oncolytic viruses necessitates overcoming through the development of combination therapies.

Subsequently, we found that the innovative three-order combination by incorporating OHSV2-DSTE^FAP5/CD3, GPC3-targeting CAR-T cells (CAR-T^HN3), and immunotoxins (J80A-PE24) exerted the most potent anti-tumor effect in the mouse HCC model, as evidenced by the smallest tumor volumes at day 42 ( Figures 6A, B ). Unexpectedly, complete tumor eradication was observed in 40% of the mice (two out of five). Poor tumor growth could be ruled out, as the tumor volume increased to approximately 400 mm³ during the initial stage of treatment. Moreover, all other mice exhibited normal tumor growth kinetics before receiving this triple-agent combination therapy. In addition, hallmark of T-cell activation and PBMC counts were more pronounced in OHSV2-DSTE^FAP5/CD3-based combined therapy group ( Figures 6C, D ). Therefore, our conceptual design of a high-order combination therapy integrating OHSV2-DSTE^FAP5/CD3, CAR-T cells, and immunotoxins—three agents with distinct, non-overlapping mechanisms—was successfully transformed into reality. Our results also demonstrated that OHSV2-DSTE^FAP5/CD3-based treatment was superior to anti-PD-1 antibody-based treatment, although this was not a head-to-head comparison between single agents. Furthermore, the OHSV2-DSTE^FAP5/CD3-based combinations were more effective than single- or dual-agent regimens, including lenvatinib plus anti-PD-1 treatment, which has shown high tumor response rates in clinical trials but has not been approved by the FDA recently.

To evaluate the safety of OHSV2-DSTE^FAP5/CD3 in combination with GPC3-targeting CAR-T cells (CAR-T^HN3) and immunotoxins (J80A-PE24), a series of tests were conducted. Given the potential for increased toxicity due to the combined effects of these agents and the known risks of severe adverse events associated with CAR-T cell therapy (e.g., cytokine release syndrome or neurotoxicity), we selected a moderate dose of OVs and a half-dose of CAR-T cells.

First, mouse body weights showed no significant descending or differences across different groups through regular monitoring ( Figure 7A ). At the same time, the parameters reflecting organ function, including the glutathione transaminases, phosphocreatine kinases, myocardial enzymes, albumin, bilirubin, creatinine, hemocyte numbers, and glucose levels in peripheral blood, occasionally displayed, at most, grade 1 deterioration ( Supplementary Table S1 ), suggesting that all treatment schemes, especially our novel triple-agent combination, did not induce overlying toxic effects in major organs or tissues, such as the liver, kidney, myocardium, or bone marrow. Furthermore, these findings revealed a decoupling between therapeutic efficacy and toxicity, although the underlying mechanisms remain to be elucidated at present. Moreover, serum cytokine levels were not significantly increased in the triple-combination group at the end of the experiment on day 42 compared to controls ( Figure 7B ), suggesting that although the treatments activated immune responses within the tumor microenvironment, they did not lead to excessive systemic cytokine release. Additionally, histopathological analysis revealed no significant tissue damage in the monotherapy group (OHSV2-DSTE^FAP5/CD3) ( Figure 7C ) or in combination therapy groups compared to the blank control.

Taken together, these findings have largely dispelled our initial concerns regarding the potential for severe on-tumor and off-target side effects, such as cytokine storms, interstitial pneumonia, and hepatic impairment, resulting from combination therapy containing OHSV2-DSTE^FAP5/CD3, because of the relatively low specificity of FAP and GPC3 target antigens, of course, suggesting that this novel strategy integrating different immune therapeutic drugs is worth further clinical development.

Further exploration of the mechanism underlying the novel strategy combining OHSV2-DSTE^FAP5/CD3 with other immune therapeutic drugs is still necessary. Although not tremendously critical for clinical translation efforts, such insights can guide the iterative development of next-generation therapeutics. To this end, various cell populations, probably including CD3+ T cells, CAFs, and other immune cells T cells, were isolated from tumor tissues across different treatment groups and identified seven distinct cell clusters (types) based on gene marker expression profiles by single-cell sequencing ( Supplementary Figure S2 ).

Notably, the proportions of CD4+ and CD8+ T cells were significantly increased in the OHSV2-DSTE^FAP5/CD3-based triple-agent regimen groups (including OHSV2-DSTE^FAP5/CD3 + J80A-PE24 + CAR-T^HN3 group and OHSV2-DSTE^FAP5/CD3 + J80A-PE24 + anti-PD-1 group), compared to other treatment groups ( Supplementary Table S2 ), however, there is no significant difference between two triple-agent regimen groups or among other dual-agent regimen groups, besides blank control ( Figures 8A, B ). The gene amplification related to CD8+ T cell exhaustion was also observed, suggesting that our constructed OHSV2-DSTE^FAP5/CD3 effectively activates T cells in vivo.

Lastly, following the expression profiling of CAFs, we categorized them into myofibroblast CAFs (myCAFs) and inflammatory CAFs (iCAFs). Remarkably, the OHSV2-DSTE^FAP5/CD3 + J80A-PE24 + CAR-T^HN3 group demonstrated the most pronounced CAF depletion among all treatment groups ( Figure 8A ). Interestingly, the gene expression patterns of the OHSV2-DSTE^FAP5/CD3 +J80A-PE24+CAR-T^HN3 group did not significantly differ from those of other treatment groups ( Supplementary Figure S2 ), suggesting that distinct treatment modalities did not impact the overall gene expression profiles of CAFs. To further corroborate the depletion of CAFs in the tumor microenvironment, we conducted immunohistochemistry to evaluate the expression of FAPs in the OHSV2-DSTE^FAP5/CD3 + CAR-T cells + immunotoxin combination group and revealed a significant reduction in CAFs, contrasting with the effect observed in the OHSV2-DSTE^FAP5/CD3 monotherapy group ( Figure 8C ), where the impact on CAF elimination was different in vitro. This finding suggests that CAFs and cancer cells are interdependent, sharing a common fate for survival. We propose that this interdependence may stem from the requirement of direct or indirect support from cancer cells for the survival of CAFs or could be facilitated by bidirectional communication between CAFs and tumor cells. In addition, we cannot discount the possibility that non-specific bystander T-cell activation may be triggered by various factors, rather than the activation of CD3+ T cells through CAF-dependent pathways. Nonetheless, further insights into these mechanisms can be gained through T-cell receptor (TCR) sequencing.

Taken together, despite being derived from a limited set of assays, our findings provide evidence that, mechanistically, the synergistic anti-tumor effects of the high-order combination of OHSV2-DSTE^FAP5/CD3, CAR-T cells, and immunotoxins, are likely involved in the functions of OHSV2-DSTE^FAP5/CD3 to prime CD3+ T-cell activation and enhance proliferation of CD4+ and CD8+ T cells and even disrupt the entangled relationship between cancer cells and fibroblasts.