Additional expression of T-cell engager in clinically tested oncolytic adeno-immunotherapy redirects tumor-infiltrated, irrelevant T cells against cancer cells to enhance antitumor immunity

Abstract

Background

Oncolytic adenoviruses (OAds) are the most clinically tested viral vectors for solid tumors. However, most clinically tested “Armed” OAds show limited antitumor effects in patients with various solid tumors even with increased dosages and multiple injections. We developed a binary oncolytic/helper-dependent adenovirus system (CAdVEC), in which tumors are coinfected with an OAd and a non-replicating helper-dependent Ad (HDAd). We recently demonstrated that a single low-dose CAdVEC expressing interleukin-12, programmed death-ligand 1 blocker, and HSV thymidine kinase safety switch (CAdTrio) induces significant antitumor effects in patients, including complete response. Similar to previous OAd studies, all patients primarily amplified Ad-specific T cells after treatment however, CAdVEC was still able to induce clinical responses even given at a 100-fold lower dose.

Methods

To address the mechanisms of CAdTrio-mediated antitumor effect in patients, we analyzed patients’ samples using Enzyme-linked immunosorbent spot (ELISpot) to measure T-cell specificity and quantitative polymerase chain reaction (qPCR) to measure CAdVEC viral genome copies at tumor sites. We then evaluated potential mechanisms of CAdVEC efficacy in vitro using live-cell imaging. Based on those results, we developed a new CAdVEC additionally expressing a T-cell engager molecule targeting CD44v6 to redirect tumor-infiltrating irrelevant T cells against cancer stem cell populations (CAdTetra) for further improvement of local CAdVEC treatment. We tested its efficacy against different cancer types both in vitro and in vivo including Ad pre-immunized humanized mice.

Results

We found that HDAd-infected cells escape Ad-specific T-cell recognition with enhanced tumor-specific T-cell activity through immunomodulatory transgenes. Since CAdVEC treatment initially amplified Ad-specific T cells in patients, we re-direct these virus-specific T cells to target tumor cells by additionally expressing CD44v6.BiTE from CAdTetra. CAdTetra significantly controlled tumor growth, repolarizing local and systemic responses against cancer cells in both immunologically “hot” and “cold” tumor models and also induced immunologic memory against rechallenged tumors.

Conclusions

Our results indicate that CAdTetra effectively induces adaptive T-cell responses against cancer cells by using tumor-infiltrating irrelevant T cells.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Oncolytic adenoviruses (OAds) are the most clinically tested viral vectors for solid tumors. Previous OAd clinical studies have evaluated adaptive immune responses after OAd treatment and found that Ad-neutralizing antibody titer has little/no impact on local (intratumoral) OAd treatment. However, even though OAd treatment amplifies Ad-specific T cells, the contribution of these T cells is not well understood.

WHAT THIS STUDY ADDS

• We investigated how low-dose CAdVEC (binary oncolytic/helper-dependent adenovirus system) led to durable responses in patients and found that helper-dependent Ad (HDAd)-infected cells attenuate Ad-specific T-cell recognition and elimination, unlike OAd-infected cells. In addition, immunomodulatory molecules derived from HDAd enhance the activity of tumor-specific T cells. Since CAdVEC treatment initially amplified Ad-specific (irrelevant) T cells in patients, we re-direct these virus-specific T cells to tumor cells by additionally expressing a bi-specific T-cell engager molecule (BiTE) from CAdVEC. Since patients have memory immune responses against Ad due to natural Ad infection, we evaluated the antitumor effects and mechanisms of the new CAdVEC in Ad-immunized humanized mouse models and confirmed that additional BiTE expression enhances antitumor effects and development of tumor-specific T cells in both immunologically “hot” and “cold” tumor models.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our data suggest the importance of Ad-specific T-cell contribution to OAd-mediated antitumor effects in patients. Since immunodeficient mouse models cannot address host immune contribution to OAd treatment, and human Ad-based vectors show limited infectivity and replication capacity in rodent cell lines, humanized mouse models may be clinically relevant models for OAd preclinical studies. Our data indicate that additional BiTE expression induces tumor-specific T-cell development regardless of tumor type without systemic toxicity unlike recombinant BiTE treatment, warranting clinical investigation of new CAdVEC.

Background

Oncolytic viruses (OVs) selectively replicate within infected cancer cells, causing direct tumor cell lysis with subsequent release of newly produced daughter virions which can infect neighboring malignant cells, creating a virtuous lytic cycle.¹ Beyond direct tumor cell lysis, it is well established that OV-mediated oncolysis can induce immunogenic cell death (ICD) and systemic immune stimulation. OVs have also shown clinical benefit, including the Food and Drug Administration (FDA) approval of an oncolytic herpes simplex virus, talimogene laherparepvec, for metastatic melanoma.² Due to their well-characterized virology, including intrinsic immune sensing mechanisms (eg, toll-like receptors^{3 4}), and ease of manufacturing, human oncolytic adenoviruses (OAds) are the most clinically tested viral vectors for solid tumors.⁵ However, most clinically tested “Armed” OAds show limited antitumor effects in patients with various solid tumors (except for brain tumors⁶) even with increased dosages and multiple injections. One potential reason for these unremarkable responses is suboptimal immune stimulation against cancer cells because most clinically tested “Armed” OAds encode only a single immunomodulatory molecule due to limited transgene capacity.⁷ To overcome this limitation, we developed a binary oncolytic/helper-dependent adenovirus system (CAdVEC), in which tumors are coinfected with an OAd and a non-replicating helper-dependent Ad (HDAd), which is devoid of all viral genes and has a cargo capacity of up to 34 kb.⁸ In this way, HDAds are able to express multiple immunomodulatory molecules in a single vector, and are replicated by co-infected OAd within malignant cells. We have previously demonstrated that CAdVEC expressing interleukin (IL)-12, programmed death-ligand 1 (PD-L1) blocker and HSV thymidine kinase (HSVtk) safety switch (CAdTrio) can achieve significant antitumor effects in preclinical and clinical studies.9,11

In a first-in-man clinical trial evaluating the safety of CAdVEC (NCT03740256), we found that a single low-dose intratumoral administration of CAdTrio is sufficient to control injected tumors in some patients despite a dose 100-fold lower than other trials with “Armed” OAds.¹¹ As in other OAd trials, all patients treated with CAdTrio immediately amplified humoral and cellular responses against Ad likely due to memory anti-viral responses as most adults have had natural Ad infections.¹² We confirmed that patients had pre-existing neutralizing adenoviral antibodies in their serum prior to treatment. Although the level of humoral response (neutralizing antibody) has shown limited impact on clinical outcomes of local OAd treatment,¹³ the contribution of cellular (T cell) responses is not well studied, partly due to limited preclinical animal models for human Ad-based OVs. Although Cotton rat and Syrian hamster models are semi-permissive for group C Ads (serotypes 1, 2, 5, and 6), group C Ads still show limited replication capacity in rat and hamster cancer cells compared with human cancer cells.¹⁴ The limited replication capacity of OAds in these semi-permissive models may mitigate viral antigen presentation, thereby affecting the evaluation of cellular immune responses against intratumoral injected OAds.

Based on our early clinical data, we hypothesized that pre-existing Ad-specific T cells are immediately amplified after OAd administration leading to the elimination of OAd-infected cells, thus attenuating OAd-mediated lytic effect and OAd-derived transgene expression as seen in previous OAd trials. In contrast, due to the lack of any viral genes in vector DNA, HDAd-infected cells escape Ad-specific T-cell recognition thereby allowing continued expression of therapeutic transgenes encoded within the HDAd. This hypothesis is supported by previous HDAd studies in large animal models which demonstrate long-lasting HDAd-mediated transgene expression.^{15 16}

Since natural Ad infection is common and most adults have pre-existing immunity against Ad, Ad-immunized immunocompetent animal models may be needed for preclinical evaluation of new OAds. Advantageously, we found that our humanized mouse model developed adaptive immune responses against Ad after local CAdVEC treatment in preclinical studies.^{10 11} Taking advantage of this model, we evaluated how pre-existing Ad-specific T cells influence CAdVEC antitumor effects and whether re-directing virus-specific “irrelevant” T cells against cancer cells using a bi-specific T-cell engager molecule (BiTE) enhances the development of adaptive T-cell responses against cancer cells using Ad-immunized humanized mice.

We and others¹⁷ have previously investigated the OVs expressing BiTE molecules against different antigens. We previously developed a CAdVEC expressing a BiTE molecule to overcome tumor antigen escape from single antigen-specific chimeric antigen receptor (CAR)-T cell therapy.¹⁸ However, the contribution of endogenous T-cell activity in response to the BiTE molecule were not evaluated. Cancer stem cells (CSCs) are highly tumorigenic and metastatic, making them a crucial target to eliminate cancer. CD44 has been a helpful marker to detect CSCs for various epithelial tumors.^{19 20} However, CD44 is also expressed by various non-malignant epithelial cells and immune cells.²¹ Among 10 splice variants (CD44v1-CD44v10), CD44 variant 6 (CD44v6) was the most frequently recovered marker of CSCs in solid tumors,^{22 23} and CD44v6 knock-down induced the loss of metastatic capacity in different solid tumor models.24,26 We have demonstrated that we can incorporate multiple different transgene expression cassettes into a single HDAd vector.²⁷ Our first cohort of patients in our clinical trial did not receive combination treatment with CAR-T cells. A single low-dose injection of CAdTrio induced T-cell infiltration into treated tumors and robust Ad-specific T-cell responses at early time point, suggesting that these tumor-infiltrated T cells are Ad-specific T cells. We therefore hypothesized that the addition of the CD44v6-targeting BiTE expression cassette into the clinical CAdTrio would allow these newly expanded “irrelevant” T cells to target malignant cells.

Here, we show that in our binary OAd/HDAd (CAdVEC) system, OAd-infected cells are eliminated by Ad-specific T cells while HDAd-infected cells are spared. Given the extended transgene expression by HDAd and their large transgene capacity, we incorporated the CD44v6.BiTE expression cassette into the clinically evaluated HDTrio creating HDTetra. The addition of the BiTE within clinical CAdVEC redirected tumor-infiltrating Ad-specific (irrelevant) T cells against cancer cells to induce further tumor antigen spread with increased development of tumor-specific T cells for better control of CAdVEC-untreated, distant tumors in Ad-immunized humanized mouse models.

Methods

Adenoviral vectors (HDAds and OAds)

The human adenoviral vectors OAd5/3Ad2E1AΔ24, OAdRFP, HDAd0 (no transgene), HDAdEGFP and HDAdTrio were previously developed and characterized in detail.^{8 10 11 28} To generate the CD44v6.BiTE, the anti-human CD44v6 single-chain variable region of BIWA-4 was fused with the anti-human CD3ε single-chain variable region with a sequence encoding a flexible glycine-serine linker.²⁷ The CD44v6.BiTE, IL-12p70, HSVtk, and PD-L1 blocking mini-antibody expression cassettes (CD44v6.BiTE driven by GAPDH promoter, IL-12p70 driven by EF1 promoter, HSVtk driven by hamster GRP78 promoter, and PD-L1 mini-antibody driven by human GRP94 promoter) were cloned into pHDΔ21E4 (HDAdTetra vector). After confirming the sequence, HDAdTetra were rescued with chimeric helper virus 5/3 (knob replacing Ad serotype 3 from Ad serotype 5) as previously described.²⁹

Primary cells

Human peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll-Paque Plus according to the manufacturer’s instructions (Axis-Shield) from patients (NCT03740256) or healthy donor whole blood (approved by the Baylor College of Medicine IRB Committee).

Activated T cells (ATCs) were generated as previously described.³⁰ Briefly, PBMCs were activated with OKT3 (Ortho Biotech, Bridgewater, New Jersey, USA) and CD28 antibodies (Becton Dickinson, Franklin Lakes, New Jersey, USA) and fed 10 ng/mL recombinant IL-7 and 5 ng/mL recombinant IL-15 every 2 days. For adenovirus-specific T cell (AdVST) generation, PBMCs pulsed with hexon/penton pepmix with subsequent stimulation with irradiated autologous ATCs pulsed with hexon/penton pepmix supplemented with 10 ng/mL IL-7 and 100 ng/mL IL-15. CD8⁺ T cells were isolated from bulk AdVSTs using a CD8 isolation column (Miltenyi Biotec). We used T cells that pass through the column as CD4⁺ T cells after confirming high CD4-positivity by flow cytometry. Ad-specificity of CD4+ and CD8+ T cells were determined by interferon (IFN-γ) Enzyme-linked immunosorbent spot (ELISpot) assay. For monitoring AdVST proliferation and distribution, retrovirus coding fusion protein enhanced green fluorescent protein (EGFP)-ffLuc was infected to expanded AdVSTs.

Multi-tumor-associated antigen-specific T cells (mTAA-Ts) were generated as previously described.³¹ Briefly, monocyte-derived dendritic cells (DCs) were generated from donor peripheral blood, loaded with pep-mixes spanning the tumor-associated antigens Survivin, SSX2, MAGEA4, PRAME, WT-1 and NY-ESO-1 (JPT Peptide Technologies), and co-cultured with PBMCs in the presence of a Th1-polarizing cytokine cocktail (IL-7 (10 ng/mL), IL-12 (10 ng/mL), IL-15 (5 ng/mL), and IL-6 (10 ng/mL)). From day 10, T cells were restimulated weekly with irradiated pepmix-pulsed DCs in the presence of IL-15 (5 ng/mL) or IL-2 (50–100 U/mL). The manufactured T-cell line was then cryopreserved before use.

Cell lines

Human non-small cell lung carcinoma cell line A549, human head and neck squamous cell carcinoma lines FaDu and SCC90, human breast cancer cell line SUM-159, human pancreatic line PANC-1, and human colorectal carcinoma cell line HCT-116 were obtained from the American Type Culture Collection (ATCC) (Manassas, Virginia, USA). Cell lines were authenticated using short tandem repeat profiling by ATCC. Cells were cultured under the conditions recommended. To generate the beta-2 microglobulin (B2M)-deficient or CD44-deficient cell lines, we removed the B2M or CD44 gene using CRISPR/Cas9. The gRNAs used for the knockout of B2M, (ttaatacgactcactataGGGTAGCGCGAGCACAGCTAgttttagagctagaaatagc) and CD44 (TtaatacgactcactataGGGCCACCAAAACTTGTCCAgttttagagctagaaatagc) were generated using the HiScribe T7 High Yield RNA Synthesis Kit (NEB) as previously described.^{18 32}

Following electroporation using Neon transfection system (Thermo Fisher Scientific) and expansion, B2M−/− and CD44 −/− cells were sorted using an SH800 cell sorter (Sony) after staining cells with anti-human HLA-ABC or CD44 antibody, respectively (BD Biosciences, BioLegend). To generate cell lines expressing the fusion protein EGFP-ffLuc, we infected cells with retrovirus encoding EGFP-ffLuc as previously described.^{18 30} EGFP+cells were sorted after three passages post-infection of retrovirus.

Co-culture experiments

Cancer cell lines, including knockout lines, were seeded in 48-well plates and infected with OAds and/or HDAds 24 hours later as described in the figure legends. AdVSTs or mTAA-Ts were added to the wells 48 hours post-infection at ratios described in the figure legends. In experiments assessing the impact of AdVSTs on OAd-infected cells and HDAd-infected cells, we used OAdRFP and HDAdEGFP. In experiments evaluating AdVSTs or mTAA-Ts killing ability of HDAdTrio-infected or HDAdTetra-infected cancer cells, we used GFP-expressing cancer cells. GFP-expressing AdVSTs were used in experiments to evaluate the proliferation of AdVSTs. Culture plates were analyzed using Incucyte live cell imaging and analysis system (Sartorius) and residual live cells were measured (as a total area of red fluorescent protein (RFP) or green fluorescent protein (GFP) signals).

Flow cytometry

We used the following fluorochrome-conjugated monoclonal antibodies: anti-human CD3 (clone UCHT1), CD4 (SK3), CD8 (RPA-T8), CD25 (2A3), TIM-3 (7D3), CD278 (DX29), LAG-3 (T47-530), CTLA-4 (BNI3), CD134 (ACT35), CD137 (4B4-1), CCR7 (3DI2), CD45RO (UCHL1), PD-1 (MIH4), CD39 (TU66) PD-L1 (29E.2A3), CD20 (B9E9), CD56 (N901), CD14 (MϕP9), CD33 (HIM3-4), CD11b (M1/70), CD11c (B-ly6), CD16 (3G8), CD80 (L307.4), CD206 (19.2), CD44v6 (2F10), and HLA-ABC (G46-2.6) (BD Biosciences, Beckman Coulter, BioLegend). Cells were stained with these antibodies or the appropriate isotype control antibodies for 30 min at 4°C. We determined live/dead discrimination via the exclusion of 7-aminoactinomycin D+ cells (BD Pharmingen) or Fixable Viability Stain 780-positive cells (BD Biosciences). Stained cells were analyzed by Gallios flow cytometer (Beckman Coulter) or BD FACSymphony (BD Biosciences) at the Flow Cytometry Core at Texas Children’s Hospital. We analyzed the data with Kaluza software V.2.2.1 (BD Bioscience) or Cytobank (Cytobank) according to the manufacturer’s instructions.

IFN-γ ELISpot assay

For evaluating patients’ samples (NCT03740256), PBMCs (5×10⁵) in 200 µL of complete medium were seeded in a 96-well Millipore ELISpot plate (BD Biosciences) with Ad-pepmix (hexon/penton), PRAME, SSX2, MAGE-A4, NYESO-1, Survivin, MART1, or WT1. Plates were incubated for 12–18 hours in a 37°C incubator and then developed according to the manufacturer’s protocol. Plates were scanned using a Mabtech IRIS ELISpot reader (Mabtech).

For quantifying Ad-specificity of AdVSTs, the bulk, CD4⁺, and CD8⁺ AdVSTs (1×10⁴) were seeded with Ad peptide mix (hexon and penton) or EBV peptide mix (EBNA1, BZLF1, LAMP1, and LAMP2). Plates were incubated for 12–18 hours in a 37°C incubator and then developed. Plates were scanned using a Mabtech IRIS ELISpot reader.

Tumor sphere formation assay

80 µL of Matrigel (Corning, Lowell, Massachusetts, USA) were added to a 48-well culture plate and incubated at 37°C for 1 hour. 2,000 single cells, including knockout lines, resuspended in 300 µL of culture media were plated in the Matrigel-covered wells and cultured in an Incucyte live cell imaging system. Half of the culture media was replaced every 3 days. A total number of tumor spheres with a size of 10,000 µm² or more were measured and counted by the Incucyte system.

Animal experiments

All animal experiments were approved by the Baylor College of Medicine Institutional Animal Care and Use Committee (AN-6144).

For the xenograft models, 2×10⁶ PANC-1 cells, 1×10⁶ HCT-116 cells or 4×10⁶ SCC90 cells were resuspended in 100 µL of phosphate-buffered saline (PBS) and subcutaneously injected into the right flank of 7–8-week-old NSG male or female mice. 2 million SUM-159 cells were resuspended in 100 µL of Matrigel and orthotopically injected into the mammary fat pad of 7–8-week-old NSG female mice. After tumor volumes reached >50 mm³, mice received 5×10⁶ AdVSTs expressing ffluc intravenously. 7 days post-injection of AdVSTs, CAdVEC (OAd:HDAd=1:1) were injected into each tumor in the total amount of viral particle described in the figure legends.

Humanized mice were generated as previously described.^{10 11} Briefly, newborn (1–2 days from birth) male and female NSGA2_SGM3 mice were sublethally irradiated (100 cGy) and intrahepatically injected with 5×10⁴ human CBU-derived CD34+cells. 8 weeks post-injection, PBMCs were collected from these mice and stained them with human CD45 flow antibody (BD Biosciences) to confirm humanization. Mice showing more than 15% reconstitution of human immune cells were used for experiments.

To generate Ad-immunized humanized mice, a total of 1×10⁸ viral particles (vp) of first-generation Ad (FGAd) were injected intramuscularly. The same dose of FGAd was injected again as a booster 3 weeks post-injection. Blood samples were collected 1-week post-second injection, and the levels of Ad-neutralizing IgG were determined as described previously.¹¹ Humanized mice showing Ad-neutralizing IgG were used for experiments.

For evaluation of antitumor effects of CAdVEC in Ad-immunized humanized mice with orthotopic breast tumor, 2×10⁶ SUM-159 cells expressing ffLuc resuspended in 100 µL Matrigel were injected into the mammary fat pad. After tumor volumes reached >100 mm³, a total of 1×10⁶ vp of CAdVEC (OAd:HDAd=1:1) were injected intratumorally in a volume of 20 µL. For evaluation of antitumor effects of CAdVEC in Ad-immunized humanized mice with a pancreatic tumor, 2×10⁶ PANC-1 cells were resuspended in 100 µL of PBS and subcutaneously injected into the right flank. After tumor volumes reached >100 mm³, a total of 1×10⁷ vp of CAdVEC (OAd:HDAd=1:1) were injected into the tumor. For the tumor rechallenge model, 2×10⁶ PANC-1 cells expressing ffLuc resuspended in 100 µL PBS were injected into the left flank of the mice, which had controlled initial PANC-1 tumor growth and survived over 10 weeks. AdVSTs or tumors expressing ffluc were assessed using an in vivo imaging system (Xenogen).³⁰ The endpoint was established at a tumor volume of >1,500 mm³ or a body weight of <80%. For humanized mice, we also set an endpoint at 28 weeks post-humanization due to xenoreaction.

Isolation of tumor-infiltrating immune cells and co-culture of tumor-infiltrating T-cell in vitro

After rinsing harvested tumors with PBS, tumors were minced and incubated in Roswell Park Memorial Institute Medium (RPMI) containing human tumor dissociation reagents (Miltenyi Biotec) at 37°C for 1 hour. Cells were passed through a 70 µm cell strainer (BD Pharmingen), and murine stroma cells were removed using a Mouse Cell Depletion Kit (Miltenyi Biotec). Human cells were stained with the antibodies described in figure legends. For tumor-infiltrating T cell (TIL) co-culture, human CD8+T cells were isolated using a CD8 isolation column (Miltenyi Biotec) from tumor-infiltrating immune cells and cultured in complete media supplemented with 6,000 ng/mL rIL-2 overnight. The human CD8+TIL and ffLuc expressing SUM-159, A549 or FaDu plated on a 96-well plate 1-day earlier, were co-cultured at an effector to target ratio of 1:1. After a 24 hours co-culture, we measured residual live cancer cells (ffLuc activity) using a luciferase assay system (Promega) and measured by a plate reader (BMG Labtech).

Cytokine analysis

We collected blood samples from xenograft and humanized mouse models at the time points indicated in the figure legends. We ran Multiplex Cytokine Immunoassay (EMD Millipore) from these plasma samples according to the manufacturer’s protocol.¹¹

Immunohistochemistry

The Human Tissue Acquisition and Pathology Core at Baylor College of Medicine stained tumor biopsies from patients in our Phase I clinical trial (NCT03740256) with anti-human CD44v6 antibody (VFF-7; Thermo Fisher Scientific) or anti-human CD8 antibody (C8/144B: Dako).

Ad-neutralizing IgG analysis

Ad-neutralizing IgG levels in the blood of humanized mouse samples were measured as described previously.¹¹ Briefly, heat-inactivated serum was diluted and mixed with HDAd5/3EGFP, then used to infect A549 cells. EGFP positivity was measured 24 hours post-infection with Incucyte (Sartorius) to determine neutralizing antibody titers in the serum samples using a standard curve regression analysis.

10x Visium analysis

We prepared sections from post-CAdVEC formalin-fixed paraffin-embedded) SUM159 tumors from humanized mice. Total RNA was extracted from sections using the RNeasy Plus Mini Kit and quantified using NanoDrop 2000. The Baylor Genomic and RNA Profiling and Histopathology Cores performed RNA expression profiling with the Visium Spatial Gene Expression system (10x Genomics). Data quality control, normalization, and analyses were performed using spaceranger count (V.2.1.0) following Visium Spatial Gene Expression guidelines and using slide layout files. Spatial plots were generated with the Seurat R package for spatial gene expression analysis (V.4.3.0). To perform differential gene expression analysis, variable features were generated for each sample using SCTransform and the samples were then merged using the union of variable features for principal component analysis (PCA), clustering and uniform manifold approximation and projection (UMAP). Plots were generated using the function SpatialDimPlot and differentially expressed genes were identified with the likelihood-ratio test for single spot gene expression, using the Seurat function FindMarkers for a range of parameter values. Preranked Gene Set Enrichment Analysis was performed in Python using the software gseapy and pathways were ranked by their negative enrichment score. Only two pathways had p values<0.01 and the top 5 out of 18 were plotted.

Statistics and reproducibility

Results are represented as means of two or more independent experiments (biological replicates). Data were analyzed as described in figure legends using GraphPad Prism V.9 (Dotmatics, Boston, Massachusetts, USA). Differences in means were considered significant at p<0.05.

Results

Low-dose CAdTrio treatment initially induced Ad-specific T cells in patients, but HDAd-infected cells attenuate Ad-specific T-cell recognition

As part of our ongoing Phase I clinical Trial (NCT03740256), we have treated the first seven patients in two escalating dose levels (DL1: 5×10⁹ vp and DL2: 1×10¹⁰ vp) with CAdTrio alone (OAd:HDTrio=1:1),¹¹ and found that a single intratumoral injection of CAdTrio mediates durable antitumor responses in injected lesions in our initial cohort despite limited patient numbers (one patient: complete response, four patients: partial response and two patients: stable disease) (online supplemental figure S1).¹¹ These responses are promising, especially given that these dosages of CAdTrio (OAd amount: 2.5×10⁹ vp (DL1), 5×10⁹ vp (DL2)) are 100-fold lower than other OAd clinical trials.33,35 To address antitumor mechanisms of low-dose CAdTrio treatment in these patients, we first evaluated antigen specificity of T cells in pretreatment and 1-week post-treatment PBMCs (figure 1A). We found that most patients induced Ad-specific T cells compared with common cancer/testis antigens (CTAs) in a dose-dependent manner 1-week post-injection, similar to other OAd clinical trials.^{33 36 37} We found a transient reduction of peripheral CD8+T cells¹¹ and CD8+T cell infiltration in CAdTrio-injected tumors at this time point (figure 1B, online supplemental figure S2). These data together suggest that tumor-infiltrating CD8+T cells 1-week post-CAdTrio injection may be Ad-specific (irrelevant) rather than tumor-specific T cells. We also quantified Ad vector copies in tumor biopsies 1-week post-CAdTrio injection (online supplemental figure S3). Although there was no difference in pre-existing Ad immunity of patients enrolled in DL1 or DL2 prior to CAdTrio treatment, and patients in DL2 received a half-log higher amount of CAdTrio, these DL2 patients showed lower vector copies in tumor biopsy samples than patients in DL1. One possible explanation for these data is that OAd-infected cells are immediately cleared by dose-dependent Ad-specific T cells. This could be why OAd clinical trials typically require repeated injections of OAds. However, our patients responded well to a single CAdTrio injection even at reduced doses compared with previous OAd trials. Since HDAd lacks all Ad genes within the vector DNA²⁹ and HDAd-derived transgene expression was detected for more than 5 years in non-human primates after a single administration,¹⁵ we hypothesized that HDAd-infected cells evade Ad-specific T-cell recognition, resulting in constitutive transgene expression at the treated tumor site. To address this, we generated Ad-specific T cells and co-cultured them with partially HLA-matched cancer cell lines (online supplemental figure S4). We first infected CAdVEC (OAd expressing RFP: HDAd expressing EGFP=1:1 mix) then added Ad-specific T cells 2 days post-infection. We monitored RFP and EGFP signals in infected cancer cells in the presence or absence of Ad-specific T cells (figure 1C, online supplemental figure S5). In this setting, we used a low dose (10 vp/cell) that we know from previous studies9,1118 will not cause tumor cell lysis nor spread propagated viruses to neighboring cells. Thus, there are four groups of cells in this experimental context; uninfected cells, cells infected with OAd only (RFP expressing), cells infected with HDAd only (GFP expressing), and cells co-infected with both vectors (RFP and GFP expressing which appear yellow online supplemental figure S5). We found that only OAd-derived RFP signals were significantly reduced in the presence of Ad-specific T cells in all cell lines while the GFP signal from HDAd was maintained. While there is no significant difference in GFP expression with or without Ad-specific T cells, there is a trend for reduced GFP when infected cells are co-cultured with Ad-specific T cells, suggesting that the population of cells co-infected with HDAd and OAd are still targeted and eliminated due to the presence of the OAd. We confirmed that OAd-derived RFP elimination is dependent on major histocompatibility complex class I presentation as this effect was diminished in β2 microglobulin knock out cell lines (online supplemental figure S6). These data support that, although local CAdTrio treatment stimulates patients’ memory T cells against Ad and reduces OAd activity, HDAd-infected cells, in contrast, attenuate the elimination by Ad-specific T cells, leading to long-lasting expression of HDAd-encoded immunomodulatory molecules (IL-12 and PD-L1 blocker) at the treated tumor site.

Figure 1. Patients initially amplify Ad-specific T cells after CAdVEC treatment, but HDAd-infected cells escape Ad-specific T cells recognition. (A) PBMCs were isolated from patients’ blood pre and 1-week post CAdVEC treatment in our ongoing Phase I clinical trial (NCT03740256). We performed an IFN-γ ELISpot assay with adenoviral peptides or peptides of cancer/testis antigens. (B) Matched pretreatment and 1-week post-treatment tumor biopsies from patient #7 were stained with human CD8 IgG for immunohistochemistry (IHC). Representative images presented. (C) HLA-A2⁺ cancer cells were infected with total 10 viral particles (vp)/cell of OAdRFP and HDAdEGFP (OAd:HD=1:1). Infected cancer cells were co-cultured with HLA-A2⁺ AdVSTs 48 hours post-infection at effector to target (E:T) ratio of 1:10 for SUM-159 or SCC-90, and 1:4 for PANC-1 or HCT-116. Live infected cells were measured as a total area of RFP or GFP signals using Incucyte live cell imaging and analysis system. Data are presented as means±SD (n=6). P values were determined by two-tailed t-test. (D) PBMCs were isolated from patients’ blood pre and 2–4-week post CAdVEC treatment in our ongoing Phase I clinical trial (NCT03740256). We performed an IFN-γ ELISpot assay with peptides of cancer/testis antigens. (E) EGFP-expressing HLA-A2⁺ cancer cells were infected with 100 vp/cell of HDAd0 (no transgene) or HDAdTrio. Infected cancer cells were co-cultured with multi-tumor associated antigen-specific T cells (mTAA-T cell) from an HLA-A2⁺ donor 48 hours post-infection at an E:T ratio of 5:1. Residual live cancer cells were measured as a total area of GFP signals using an Incucyte. Data are presented as means±SD (n=6). P values were determined by one-way analysis of variance. Ad, adenovirus; AdVST, adenovirus specific T cell; CAdVEC, binary oncolytic/helper-dependent adenovirus system; CR, complete response; DL, clinical trial dose level; EGFP, enhanced green fluorescent protein, ELISpot, enzyme linked immunosorbent spot; GFP, green fluorescent protein; HDAd, helper-dependent Ad; IFN, interferon; OAds, oncolytic adenoviruses; PBMC, peripheral blood mononuclear cell; PDAC, pancreatic ductal adenocarcinoma; PR, partial response; RFP, red fluorescent protein; SD, stable disease.

OAds are known to develop and/or enhance tumor-specific T cells through ICD of tumor cells.⁷ As such, we confirmed new CTA-specific T-cell circulation in responders at 2–4 weeks post injection in patients’ PBMCs (figure 1D), but Ad-specific T cells are still the majority at these time points (online supplemental figure S7). Since we previously demonstrated that IL-12 and PD-L1 blocker derived from CAdTrio enhanced the activity and persistence of adoptively transferred T cells in various solid tumor preclinical models,^{9 10 18} we hypothesized that these transgenes also enhance the antitumor activity of tumor-specific T cells in patients. We co-cultured HDAd-infected cancer cell lines with HLA-partially matched cognate CTA-specific T cells and monitored antitumor activity (figure 1E and online supplemental figure S8). CTA-specific T cells significantly increased their antitumor activity in the presence of HDTrio compared with cells infected with control HDAd (HD0: no transgene). To address how IL-12 and PD-L1 blocker enhanced tumor-specific T-cell activity, we phenotyped CD8+T cells pre-co-culture and post-co-culture (online supplemental figure S9). Phenotypically exhausted effector CD8+T cells in which existed pre-co-culture mostly disappeared post-co-culture in both HD0 and HDTrio conditions regardless of target cell line. However, CD8+T cells co-cultured with HDTrio-infected cells had increased activated central-memory-like T-cell phenotype compared with co-culture with HD0-infected cells (online supplemental figure S9). Since these CD8+T cells are still CD39 positive, and CD39 is a marker for tumor-specific CD8+ TILs,³⁸ these results suggest that HDTrio-derived immunomodulators support tumor-specific CD8 T-cell persistence. These results also suggest that, although low-dose CAdTrio injection initially leads to Ad-specific T-cell amplification, HDTrio-infected cells evade Ad-specific T-cell recognition and HDTrio-derived immunomodulatory molecules enhance the activity and/or persistence of tumor-specific T cells induced by CAdTrio treatment. These results validate why a single injection of low-dose CAdTrio induces durable antitumor activity in patients.

HDTetra additionally expressing a T-cell engager molecule induces Ad-specific T-cell killing of cancer cells in vitro and in vivo

Although current CAdTrio (OAd plus HDTrio) controlled CAdTrio-injected tumor growth in patients, distant (CAdTrio untreated) tumors were still incurable,¹¹ indicating that for local CAdTrio treatment to have a broader systemic antitumor effect, more tumor-specific T cells must be generated. HDTrio has an additional 21 kb transgene capacity, and our data suggest that at early time points, TILs were Ad-specific (irrelevant) T cells. We thus hypothesized that HDTrio additionally expressing a T-cell engager molecule (BiTE) would re-target tumor-infiltrating irrelevant (non-tumor specific) and virus-specific T cells against cancer cells, leading to enhanced tumor cell killing and additional tumor antigen spread with adaptive immune development against cancer cells, especially for metastatic tumor cells.

Our strategy is to target CD44, specifically CD44v6 which is a known marker for CSCs, thereby reducing tumor growth and metastasis.39,41 We confirmed that CD44 knock-out cancer cell lines have significantly reduced spheroid formation compared with unmodified cell lines (online supplemental figure S10). We confirmed that our patient samples are CD44v6 positive regardless of tumor type (figure 2A). Interestingly, one patient (Pt #1) initially responded to CAdTrio injection but relapsed in a juxtaposed region with high expression of CD44v6 (re-enrolled as Pt #5), suggesting that CD44v6 positivity may associate with cancer metastasis and/or recurrence. Since we previously developed a CD44v6-targeting CD3-engaging T-cell molecule (CD44v6.BiTE),¹⁸ we hypothesized that the addition of CD44v6.BiTE expression to our clinically tested HDTrio would re-direct irrelevant T cells against CSC populations at CAdVEC-injected tumor sites, resulting in CSC antigen spread and the development of adaptive immune responses against CSC-derived metastatic tumors. We constructed HDTrio additionally expressing CD44v6.BiTE^{18 27} (HDTetra) and confirmed transgene expression (figure 2B). To evaluate CD44v6.BiTE function, we infected CD44v6 positive cancer cell lines with HDAds (online supplemental figure S11) and co-cultured them with Ad-specific T cells 48 hours post-HDAd infection. We monitored both cancer cell viability and T-cell proliferation (figure 2C). Although HDTrio induced Ad-specific T-cell proliferation in some cell lines, only HDTetra led to significant cancer cell killing by Ad-specific T cells regardless of CD44v6 expression levels. In addition, Ad-specific T cells showed significantly better proliferation in the presence of HDTetra compared with those in control HDAds. These results indicate that CD44v6.BiTE induces Ad-specific T-cell killing of cancer cells, and other immunomodulatory molecules (IL-12 and PD-L1 blocker) enhance their activity and proliferation. To address whether we can have similar effects in vivo, we transplanted CD44v6 positive cancer cell lines into NSG mice and systemically administered Ad-specific T cells after confirming tumor formation. After the infiltration of adoptively transferred Ad-specific T cells was confirmed at the tumor sites (7 days post-administration), we intratumorally injected CAdTrio (OAd plus HDTrio=1:1 ratio) or CAdTetra (OAd plus HDTetra=1:1 ratio) at that time. We monitored T-cell proliferation and tumor growth (figure 3). Although Ad-specific T cells significantly proliferate in the presence of HDTetra in all tested cell lines in vitro, we found that in vivo Ad-specific T-cell proliferation is cell line-dependent. However, we found that CAdTetra-treated groups induced significant tumor growth control compared with control mice. This tumor control also correlates with increased circulating IFN-γ levels in the blood of mice. In contrast, AdVST plus CAdTrio is insufficient to control tumor growth. These results indicate that the CD44v6.BiTE molecule can re-target tumor-infiltrating Ad-specific (irrelevant) T cells against cancer cells, and other immunomodulatory molecules (IL-12 and PD-L1 blocker) support their antitumor activity both in vitro and in vivo.

Figure 2. HDTetra additionally expressing CD44v6.BiTE leads to Ad-specific T-cell killing of cancer cells in vitro. (A) CD44v6 immunohistochemistry of tumor biopsies from patients in our ongoing Phase I clinical trial (NCT03740256). Representative images presented. (B) Schematic structure of HDAdTrio and HDAdTetra. A549 cells were infected with 500 viral particles (vp)/cell of HDAdTrio or HDAdTetra. Media were collected 48 hours post-infection. Medium samples were subjected to western blotting for PD-L1 mini-antibody, which is detected by anti-HA antibody, and assessed for human IL-12p70 by ELISA assay. Data are presented as means±SD (n=4). A549 cells were infected with 500 vp/cell of HDAdTrio or HDAdTetra. Media containing ganciclovir (GCV) were added to denoted wells 24 hours post-infection and refreshed daily. After 5 days viable cells were fixed and stained with crystal violet. Representative image presented. (C) (Left) EGFP expressing HLA-A2⁺ cancer cells were infected with 100 vp/cell of HDAd0, HDAdTrio or HDAdTetra. Infected cancer cells were co-cultured with HLA-A2⁺ AdVSTs 48 hours post-infection at an E:T ratio of 1:10. Residual live cancer cells were measured as a total area of GFP signals using an Incucyte. Data are presented as means±SD (n=6). P values were determined by one-way ANOVA. (Right) HLA-A2⁺ cancer cells were infected with 100 vp/cell of HDAd0, HDAdTrio or HDAdTetra. Infected cancer cells were co-cultured with EGFP expressing HLA-A2⁺ AdVSTs 48 hours post-infection at an E:T ratio of 1:10. Live AdVSTs were measured as a total area of GFP signals using an Incucyte. Data are presented as means±SD (n=6). P values were determined by one-way ANOVA. AdVST, adenovirus specific T cell; ANOVA, analysis of variance; BiTE, bi-specific T-cell engager molecule; Colon, colorectal carcinoma; EGFP, enhanced green fluorescent protein; ET, effector to target; ER/PR, estrogen/progesterone positive breast cancer; HDAd, helper-dependent Ad; HNC, head and neck cancer, HSVtk, HSV thymidine kinase; IL, interleukin; PD-L1, programmed death-ligand 1; PDAC, pancreatic ductal adenocarcinoma; TNBC, triple-negative breast cancer.

Figure 3. HDTetra additionally expressing CD44v6.BiTE leads to Ad-specific T-cell killing of cancer cells in vivo. Cancer cells were transplanted into the right flank of NSG mice (SUM-159 cells were transplanted into the mammary fat pad) (n=5 per group). A total of 5×10⁶ AdVSTs expressing ffLuc were systemically administered after tumor volumes reached >50 mm³. We injected a total of 1×10⁶ vp of CAdVEC (Onc:HD=1:1) into the SUM-159 tumor, 1×10⁷ vp into the PANC-1 or SCC-90 tumor, and 1×10⁸ vp into the HCT-116 tumor 7 days post-injection of AdVSTs. Tumor volumes and bioluminescence of AdVSTs were monitored at different time points. Representative animal images presented. Serum samples were collected pre and 1-week post-CAdVEC treatment, and the serum IFN-γ levels were analyzed by multiplex cytokine immunoassay as means±SD (n=5). P values were determined by one-way analysis of variance. AdVST, adenovirus specific T cell; CAdVEC, binary oncolytic/helper-dependent adenovirus system; HNC, head and neck cancer; IFN, interferon; vp, viral particles.

CAdTetra controls tumor growth and elicits host antitumor immunity in Ad-immunized humanized mice

Due to the limited infectivity and replication capacity of human Ad-based OV in rodent cell lines,^{14 42} we have established humanized mice reconstituting human innate and adaptive immune cells, and demonstrated that these mice can develop adaptive immune responses against Ad and cancer cells.^{10 11} Although we evaluated the safety and antitumor efficacy of clinically tested CAdTrio in these humanized mouse models (IND19439), these mice were Ad-naïve mice, and our patients have memory immune responses against Ad due to previous natural Ad infection.¹¹ To mimic patients’ pre-existing Ad immunity, we immunized mice with Ad and used them to evaluate how memory immune responses against Ad, especially pre-existing Ad-specific T cells, impacts local CAdVEC antitumor effects (online supplemental figure S12). In addition, we tested whether CAdTetra (additional CD44v6.BiTE) re-directs tumor-infiltrating Ad-specific (irrelevant) T cells against cancer cells, leading to tumor antigen spread and subsequently diversified adaptive immune development against cancer cells. We first used SUM-159, an immunologically “hot” tumor model because we previously confirmed high immune cell infiltration in this tumor compared with a “cold” breast tumor model.¹¹ Although we previously demonstrated that CAdTrio significantly controlled SUM-159 tumor growth in naïve humanized mice,¹¹ we found that there was no significant difference between control and CAdTrio in Ad-immunized mice (figure 4A, online supplemental figure S13). In contrast, CAdTetra-treated mice showed significant antitumor effects compared with other groups. These results indicate that CD44v6.BiTE derived from CAdTetra re-directs tumor-infiltrating/resident T cells against cancer cells. To address how additional CD44v6.BiTE changes the tumor immune microenvironment, we phenotyped tumor-infiltrating immune cells 4 weeks after CAdVEC treatment (figure 4B). Although there was no significant difference in the composition of innate immune cell populations, CAdTetra treatment significantly increased activated CD8+T cell (CD8+, CD56+T cell) infiltration compared with other groups. Further analysis of T-cell subsets revealed that CAdTetra significantly increased terminally differentiated CD8+T cells (Tim-3, CD39, 4-1BB positive) compared with other groups (figure 4C). Since these CD8+T cell markers are associated with tumor-specific T cells,⁴³ we repeated the above experiment and isolated CD8+T cells from tumors 4 weeks post-CAdVEC injection (online supplemental figure S14A). We co-cultured these TILs with SUM-159 or irrelevant cancer cells and measured lytic effects in vitro (figure 4D). As irrelevant cancer cells, we used A549 for potential residual OAd and FaDu for potential residual CD44v6.BiTE in isolated CD8+T cell samples (online supplemental figure S14B). Although CD8+T cells infiltrated untreated control tumors, there was minimal target cell killing by these TILs. In contrast, CD8+T cells derived from CAdTetra-treated tumors showed specific and significant target cell lysis. CAdTrio treatment increased the infiltration of activated CD8+T cells, and these TILs showed specific target cell lysis, however, the lytic effect was significantly lower than that seen in TILs from CAdTetra-treated mice. These data indicate that additional CD44v6.BiTE expression through CAdTetra induced tumor-specific T-cell development and infiltration compared with CAdTrio in Ad-immunized humanized mice. Since there was no significant difference in circulating human Th1 cytokine levels between CAdTrio and CAdTetra (online supplemental figure S15A), these data also suggest that additional local expression of BiTE through CAdTetra does not induce systemic toxicity (eg, cytokine release syndrome) unlike systemic BiTE treatment. To address whether tumor-specific CD8+T cell infiltration affects tumor gene signatures, we performed 10x Visium spatial transcriptomics of tumors 4 weeks post-CAdVEC in a repeated experiment (figure 5, online supplemental table 2Suzuki_Supplemental Tables (zenodo.org)). Although SUM-159 is a homogenous cell line, the untreated control tumor sample contained spatially variable genes that coincided with different gene expression clusters within the tumor, likely due, in part, to different levels of immune cell infiltration within different regions of the tumor. We found that only 1–2% of spots were shared between control and CAdVEC-treated samples, indicating that both CAdTrio and CAdTetra treatments alter overall tumor gene signatures similar to that seen in one of our patients.¹¹ There is a region of mitochondrial messenger RNA transcript enriched spots, a marker of dead cells (cluster 0), in both CAdTrio and CAdTetra treated tumors, indicating that this tumor cell death was primarily induced by oncolysis. The majority of spots were from clusters that were unique to CAdTrio or CAdTetra, although 14% of spots were from clusters that were shared between CAdTrio and CAdTetra. These results suggest that additional BiTE expression from CAdTetra impacts overall tumor gene signatures. CAdTetra-cluster 5, which surrounds cluster 0, showed significant enrichment of genes associated with “cytokine activity” when differential expression was performed on genes detected in a small fraction of spots. Since we found limited overlapping spots across samples, we next examined genes enriched in unshared spots (green squared spots in blue Venn diagram) (online supplemental figure S16, online supplemental table 2Suzuki_Supplemental Tables (zenodo.org)). Most genes with positive fold change in unshared clusters were unique to the untreated control (73%). For instance, PFKP, rate-limiting enzyme of glycolysis, and MTHFD2, folate-coupled mitochondrial metabolic enzyme, highly express in in vitro-cultured SUM-159 cells and are associated with poor prognosis in patients with breast cancer (online supplemental table 3Suzuki_Supplemental Tables (zenodo.org)).44,46 9% and 3% of upregulated genes in unshared clusters overlapped with the control, in CAdTrio and CAdTetra, respectively. These data correlate with CAdVEC-dependent antitumor outcomes. Genes uniquely upregulated in response to CAdTrio treatment include inflammation-related genes (eg, IL-1A, NFKB), suggesting that CAdTrio treatment constitutively induces inflammation within the tumor. In contrast, genes upregulated in CAdTetra treatment include immune cell markers (eg, CD74, TNFRSF14), suggesting that additional BiTE expression accelerates immune cell infiltration into the tumor. Although the only difference between CAdTrio and CAdTetra is additional BiTE expression, these data suggest that the additional tumor killing via the BiTE molecule not only enhances tumor-specific T-cell development but also changes the overall tumor microenvironment, including immune cell signatures in SUM-159 tumors.

Figure 4. CAdTetra shows better antitumor effects and tumor-specific T-cell development than CAdTrio in Ad-immunized humanized mouse model with triple-negative breast cancer. (A) FfLuc expressing SUM-159 cells were orthotopically transplanted into the mammary fat pad of Ad-preimmunized humanized female mice (n=5 per group). After tumor volume reached >100 mm³, a total of 1×10⁶ viral particles of CAdVEC (OAd:HD=1:1) were injected intratumorally. Tumor volume and bioluminescence of cancer cells were monitored at the indicated time points. Data are presented as means±SD (n=5). P values were determined by one-way ANOVA. Tumor samples were collected at 4 weeks post CAdVEC. Freshly isolated tumor-infiltrating immune cells were phenotyped using different immune cell markers (B) or T-cell markers (C). Representative FlowSOM data are presented from each group. %Totals are presented as means±SD (n=5). P values were determined by one-way ANOVA. (D) We isolated the tumor-infiltrating human CD8⁺T cells from the SUM-159 tumors at 4 weeks post CAdVEC. We co-cultured them with ffLuc expressing SUM-159, A549, or FaDu cells at an effector-to-target ratio of 1:1. After 24 hours co-culture, we measured residual live cancer cells using a luciferase assay system. % Lysis are presented as means±SD (n=5). P values were determined by one-way ANOVA. ANOVA, analysis of variance; CAdVEC, binary oncolytic/helper-dependent adenovirus system; NK, natural killer; OAd, oncolytic adenovirus.

Figure 5. CAdVEC treatment changes gene signatures in triple-negative breast cancer tumors of adenovirus-immunized humanized mouse model. We performed Visium spatial gene expression with the SUM-159 tumors collected 4 weeks post CAdVEC. Spots on the cytassist slides are colored by their cluster labels (top panels). The contribution of spots to each cluster for all samples is shown in the bar chart and the percentage of spots shared between clusters in each treatment is shown in the Venn diagram. The top five pathways from Gene Set Enrichment Analysis of cluster 5, that is, shared between CAdTrio and CAdTetra are shown. Only receptor ligand activity and cytokine activity were significant. CAdVEC, binary oncolytic/helper-dependent adenovirus system.

Since SUM-159 is an immunologically “hot” tumor filled with immune cells, including T cells, we also evaluated the effect of additional CD44v6.BiTE in pancreatic ductal adenocarcinoma (PDAC), an immunologically “cold” tumor model. PDAC tumors show little immune cell activation/infiltration within the immunosuppressive tumor microenvironment, and patients with PDAC rarely benefit from immunotherapies.⁴⁷ PANC-1 tumors from Ad-immunized mice show immune cell infiltration, but the level was significantly lower than that in SUM-159 tumor (online supplemental figure S17). Although these mice were pre-immunized with Ad, the same as in the SUM-159 model (online supplemental figure S13), CAdTrio treatment significantly controlled tumor growth compared with control mice (figure 6A). These results indicate that oncolysis, IL-12 and PD-L1 blocker can control tumor growth of “cold” tumors even with pre-existing Ad immunity. However, unlike the CAdTrio treatment, all CAdTetra-treated mice had controlled tumor growth over time. On histological analysis, we found only stromal cells in these residual tumors treated with either CAdTrio or CAdTetra (online supplemental figure S18). Since CAdTetra-treated mice showed long-term tumor control, we rechallenged these mice and remaining CAdTrio-treated mice with the same cell line 10 weeks post initial CAdVEC injection to address whether this tumor control is associated with the development of tumor-specific T cells and immunological memory (figure 6B). Although CAdTrio controlled treated tumor growth in some mice initially, the rechallenged tumor grew similarly to control naive mice. In contrast, CAdTetra-treated mice significantly controlled rechallenged tumor growth compared with control mice. These data suggest that CD44v6.BiTE-mediated tumor killing by tumor-infiltrating T cells in PANC-1 tumors led to the development of tumor-specific T cells, resulting in tumor growth control of rechallenged cells. To address whether this control can be attributed to CD8+T cells, we harvested tumors day 35 post-rechallenge and phenotyped tumor-infiltrating immune cells after combining tumor samples per treatment group due to limited amounts of TILs (figure 6C). Although mice treated with CAdTrio showed limited T-cell infiltration, CAdTetra-treated mice showed T-cell infiltration with predominantly CD8+T cells, unlike control mice with predominantly CD4+T cells. While the control mice still showed more CD8+T cells than CAdTetra-treated mice, these cells were unable to inhibit tumor growth. Additionally, CAdTetra-treated mice showed an increase in the natural killer cell population, which may reflect CAdVECs overall immunostimulatory activity due to the CD44v6.BiTE from CAdTetra. These data support that CD44v6.BiTE-mediated tumor lysis accelerates the development of tumor-specific T cells, leading to immunologic memory and effective control of rechallenged tumors.

Figure 6. CAdTetra shows better antitumor effects and tumor-specific T-cell development than CAdTrio in Ad-immunized humanized mice model with PDAC. (A) PANC-1 cells were transplanted into the right flank of Ad-preimmunized humanized mice (n=5 per group). After tumor volume reached >100 mm³, a total of 1×10⁷ vp of CAdVEC (OAd:HD=1:1) were injected intratumorally. Tumor volumes were monitored at the indicated time points. Data are presented as means±SD (n=5). P values were determined by one-way ANOVA. Kaplan-Meier survival curve after CAdTrio or CAdTetra administration in mice (n=5). P values were determined using the log-rank Mantel-Cox test (df=3). (B) FfLuc expressing PANC-1 cells were injected into the left flank of the mice, which had controlled initial PANC-1 tumor growth and survived over 10 weeks, as a rechallenge model. Tumor volume and bioluminescence of cancer cells were monitored at the indicated time points. Representative images presented. P values were determined by one-way ANOVA. (C) Tumor samples were collected 35 days post-rechallenge. Due to the limited number of tumor-infiltrating immune cells in this “cold” tumor model, samples were combined per treatment to have sufficient cell numbers and phenotyped using different immune cell markers. ANOVA, analysis of variance; CAdVEC, binary oncolytic/helper-dependent adenovirus; HD, helper-dependent adenovirus; NK, natural killer; OAd, oncolytic adenovirus; PDAC, pancreatic ductal adenocarcinoma.

These data demonstrate that the additional CD44v6.BiTE expression from CAdTetra enhances adaptive immune development against cancer cells, resulting in better tumor control than CAdTrio in both immunologically “hot” and “cold” tumors in clinically relevant Ad-immunized humanized mouse models.

Discussion

While OAds have been clinically studied for decades, the contribution of cellular immune responses against OAds with clinical outcomes are not well studied. Here, we find that Ad-specific T cells are immediately amplified after our binary oncolytic/helper-dependent adenoviral immunotherapy CAdVEC administration in patients similar to other OAd trials. However, we also find that Ad-specific T cells target OAd-infected cells whereas HDAd-infected cells avoid Ad-specific T-cell recognition.

As part of our ongoing clinical trial evaluating the safety of our CAdVEC we were surprised to find that a single low-dose intratumoral injection of CAdTrio, which expresses the immunostimulatory molecules IL-12, PD-L1 blocker and safety switch HSVtk, elicited durable antitumor responses against CAdTrio-injected tumors, including a complete response.¹¹ Given that our initial dose was 100-fold lower than what has been used in previous OAd clinical trials, we sought to determine the mechanism of CAdVEC’s efficacy. We found that all patients treated with CAdTrio had increased Ad-specific T cells compared with CTA-specific T cells in their PBMCs, with increased T-cell infiltration of the treated tumor at early time points. Since DL2 patients showed higher Ad-specific T cells with lower Ad copies than DL1 patients in tumor biopsy samples, early after CAdTrio treatment, most TILs may be irrelevant (Ad-specific) T cells which eliminate OAd-infected cells. Responders in our trial also developed new CTA-specific T-cell circulation at later time points, suggesting that OV-dependent development of tumor antigen-specific T cells takes time. This also suggests that sample collection, including tumor biopsy, timing should be an important consideration for better evaluation of OV-dependent antigen spread following adaptive immune development against tumor cells in clinical studies. CAdTrio-dependent CTA-specific T cells were insufficient to cure distant (CAdVEC-untreated) tumors even in conjunction with checkpoint inhibitor in some patients, suggesting further tumor-associated antigen spread with the development of tumor-specific T cells is required to enhance the antitumor effect of local oncolytic adeno-immunotherapy, including CAdVEC, treatment.

Although OAds have produced significant antitumor effects in preclinical models, often immunodeficient xenograft mouse models, these results have not translated to clinical efficacy. However, our binary OAd/HDAd CAdVEC platform is unique in clinical outcome¹¹ and, as we show here, interaction with cellular immune responses against Ad. As other OAd clinical trials use replication competent Ads, transduced cells are likely rapidly cleared by re-stimulated and amplified memory Ad-specific T cells, leading to suboptimal clinical outcome. Whereas in our platform, the HDAd-transduced cells are spared from clearance by Ad-specific T cells that were reinvigorated by the presence of the OAd. Because the HDAd vector lacks all viral coding sequences, no viral proteins are produced in infected cells. In this way, single administration of HDAds confer long-lasting (>5 years) transgene expression in large animal models.^{15 16} Therefore, our binary platform provides immediate immune stimulation with the amplification of Ad-specific T cells by the OAd and the incorporation of an engager molecule into the HDAd allows the retargeting of these irrelevant T cells towards cancer cells. These important findings were made possible through our advanced Ad-immunized humanized mouse models which develop adaptive immune responses to both Ad and cancer cells.¹¹ Our humanization methodology allows for the generation of approximately 100 humanized mice from one cord blood donor, providing mice with consistent reconstitution of human innate and adaptive immune cell subsets. However, a limitation of this model is that results from different cord blood donors may be inconsistent. While syngeneic semi-permissive models allow for the consistent evaluation of treatment-mediated complete immune responses, human Ad replication is limited in these models and because the immune systems are murine, the resultant immune responses may not accurately predict human immune responses.⁴⁸ Our humanized mouse model provides the use of human cancer cell lines, or even patient-derived xenografts, with complete OAd replication potential and the ability to model pre-existing humoral and cellular immunity to CAdVEC, creating a more instructive preclinical model.

It is well understood that the infiltration of CD8+T cells has a direct impact on the efficacy of checkpoint inhibitors, however, recent reports showed that most TILs are irrelevant T cells (eg, virus-specific).^{38 49} We also confirmed that CD8+T cells infiltrate SUM-159 tumors but these CD8+T cells have limited reactivity against SUM-159 cells, suggesting that our humanized mice mimic some aspects of human tumors. Although IL-12 can promote the activation of CD8+T cells,⁵⁰ activated irrelevant CD8+T cells cannot contribute to antitumor activity or may accelerate the removal of OV-infected cells, if these T cells are OV-specific. Tumor-infiltrating CD4+T cells augment immune tolerance in the tumor microenvironment (TME) and enhance tumor growth.⁵¹ We also found that PANC-1 (PDAC) tumor is CD4+T cell dominant similar to that seen in PDAC patients.⁵² Since CD4+T cells are not a uniform cell population, and IL-12 can polarize CD4+T cells to T_H1 T cells,⁵³ tumor-infiltrating CD4+T cells may be repolarized to T_H1 T cells through IL-12 derived from CAdVEC. Since BiTE molecules can engage both CD8+ and CD4+ T cells against target cells,¹⁸ these activated/repolarized “irrelevant” T cells can be retargeted to cancer cells. As we show here, directing tumor-resident T cells with the CD44v6.BiTE enhances the development of tumor-specific CD8+T cells in both immunologically “hot” triple-negative breast cancer (TNBC) and “cold” PDAC tumors. In addition, we also found that BiTE-mediated tumor killing leads to favorable tumor gene signatures in SUM-159 tumors. However, we still do not know how and when additional BiTE-mediated tumor killing induced this signature. Monitoring tumor gene signatures with infiltrating immune cells at different time points will be necessary to address how the BiTE molecule combined with immunomodulatory molecules (IL-12 and PD-L1 blocker) modify the host TME, resulting in favorable outcomes in both immunologically “hot” and “cold” tumor models.

Even though the FDA has approved immune checkpoint inhibitors (ICIs) for patients with TNBC, the eligibility criteria require PD-L1 positive tumors. In a PDAC clinical trial, ICIs did not enhance antitumor activity.⁴⁷ Since CAdTetra significantly enhances tumor-specific T cell development compared with CAdTrio, and combination with CAdTrio and ICI showed long-term tumor control in our patients,¹¹ the use of ICIs with CAdTetra will be warranted in patients with epithelial tumors because these tumors have well characterized CD44v6 positivity.²² Although systemic administration of CD44v6 antibody led to dose-dependent toxicity in patients,⁵⁴ local CD44v6.BiTE expression via CAdTetra showed no toxicity (eg, cytokine release syndrome) in our advanced models due to limited systemic circulation. It should be noted that while the BiTE molecule may be immunogenic, results from the clinical study of the FDA-approved CD19.BiTE blinatumomab shows that neutralizing antibody development occurred in less than 1% of patients after systemic and frequent infusion.⁵⁵ We expect that is due to the intratumoral expression of our CD44v6.BiTE, the immunogenicity would be less compared with a systemically administered product. Our results suggest that the CAdTetra-derived BiTE mainly serves to redirect tumor resident/infiltrated T cells against cancer cells.

If we find limited antitumor effects in CAdTetra-untreated tumors in future clinical study, we could identify other surface markers on treatment-resistant residual tumors and incorporate additional BiTEs into the available 18 kb transgene cargo capacity of the HDTetra vector for further tumor antigen spread to broadly target this population. For instance, we found that the CAdVEC-treated SUM-159 tumor increased the expression of immune checkpoint molecule CD276 (B7-H3) compared with the control tumor. Since B7-H3 is highly expressed in different types of solid tumors including TNBC,⁵⁶ is associated with tumor metastasis,⁵⁷ and several clinical trials using antibody-drug conjugate or CAR-T cells against B7-H3 are ongoing, it will be enlightening to evaluate whether additional targeting of B7-H3 with a BiTE molecule enhances CAdVEC antitumor effects, including tumor-specific T-cell development. Solid tumors are complex structures and successful elimination may require targeting components other than malignant cells themselves. As such, we have developed various BiTE molecules targeting other tumor-associated antigens, including tumor stroma antigens,²⁷ like fibroblast activation protein to target the extensive stromal component common in PDAC tumors. Further studies are warranted to compare and/or add other BiTE molecule(s) targeting stromal component if these activities could be further augmented. Likewise, as the immunosuppressive TME often downregulates expression of co-stimulatory molecules resulting in suboptimal activation of TILs,⁵⁸ we have previously generated HDAd vectors encoding co-stimulatory molecules, and those co-stimulatory molecules enhanced tumor-specific T-cell activity.²⁷ Since CAdTetra provides only signal 1 T-cell receptor (TCR) engagement (BiTE) and signal 3 cytokine stimulation (IL-12), additional co-stimulation (signal 2) could convert “irrelevant” T cells to fully activated, tumor-targeting T cells, like CAR-T cells, at the tumor site. As these data presented here demonstrate, the CAdVEC platform is a flexible and potent cancer therapeutic agent that directly lyses tumor cells with simultaneous stimulation and retargeting of endogenous immune responses at the treated tumor site, thereby safely eliciting systemic antitumor responses in both immunologically “hot” and “cold” solid tumors.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.