ICAM1-targeted costimulation extends HER2 CAR T cytotoxicity to HER2-low and HER2-negative tumors through synergy with TCR signaling Authors: Yogindra Vedvyas, Yanping Yang, Aroshi Mitra, Yoo-Shin Kim, Nate R. Fredette, Ricardo de Azevedo, Daniel J. Kota, Irene M. Min, Moonsoo M. Jin Published: 2026 ABSTRACT Autologous T cells engineered to express a HER2 chimeric antigen receptor (CAR) have shown limited clinical efficacy. To broaden antitumor activity, we co-expressed a chimeric costimulatory receptor (CCR) targeting ICAM1 (ICCR) in HER2 CAR T cells. ICAM1 is upregulated by inflammatory cytokines, reinforces immune-synapse formation, and is elevated in aggressive or dedifferentiated tumors, including those with heterogeneous or low HER2 expression. HER2 CAR/ICCR T cells preserved potent cytotoxicity against HER2-high targets and showed significantly enhanced killing of HER2-low cell lines compared with HER2 CAR T cells alone. ICCR augmented NF-κB activation particularly under low HER2 conditions. Upon repeated stimulation with HER2-negative tumor cells, HER2 CAR/ICCR T cells underwent nearly 10-fold expansion with superior cytotoxicity, whereas HER2 CAR T cells failed to expand. In vivo, both constructs rapidly eliminated HER2-high tumors, but only HER2 CAR/ICCR T cells achieved complete clearance of HER2-low tumors, with delayed kinetics consistent with clonally expanded, tumor-specific T cells. TCR sequencing confirmed substantially greater clonal expansion in HER2 CAR/ICCR T cells across in vivo and in vitro studies. This dual-receptor platform (HMJ01) has supported IND authorization for a first-in-human trial in patients with HER2-positive and HER2-low gastric cancer. Introduction Human epidermal growth factor receptor 2 (HER2) is a transmembrane tyrosine kinase receptor that is amplified or overexpressed in approximately 15%–30% of breast and gastrointestinal cancers, where it promotes aggressive tumor growth and metastasis through HER2-driven signaling pathways.1 HER2-targeted therapies, such as monoclonal antibodies (trastuzumab, pertuzumab), tyrosine kinase inhibitors (tucatinib, neratinib), and antibody drug conjugates (ADCs; T-DM1, T-DXd), have provided significant survival benefits for patients with HER2-expressing breast, gastric, and colorectal cancers. However, despite these advances, most patients ultimately experience disease progression or relapse, with few remaining therapeutic options.2,3 Chimeric antigen receptor (CAR) T cell therapy offers a promising approach to target tumor-associated surface antigens, enabling direct tumor cell killing independent of major histocompatibility complex (MHC) presentation. While its application in solid tumors has historically faced challenges, recent encouraging clinical results of CAR T cells targeting disialoganglioside (GD2) in neuroblastoma and Claudin18.2 in gastrointestinal cancers demonstrate its feasibility and growing potential for treating otherwise refractory solid tumors.4,5,6 HER2 has long been considered one of the top-ranked targets for CAR T cell development. Early HER2 CAR T cells utilized a trastuzumab-derived single-chain variable fragment (scFv), but a case of fatal pulmonary toxicity following high-dose CAR T cell infusion (1 × 1010 cells) and interleukin (IL)-2 administration after lymphodepleting chemotherapy raised serious safety concerns.7 To address this, later generations of HER2 CAR T cells were engineered using lower-affinity scFvs such as FRP5 and administered at lower doses (1 × 104/m2 to 1 × 108/m2) without IL-2.8,9,10 Clinical studies using FRP5-based CAR T cells demonstrated safety across multiple solid tumor types, including sarcoma and glioblastoma, but antitumor efficacy remained limited.8,9,11 Resistances mechanisms to HER2-targeted therapies remain incompletely understood and likely vary depending on the therapeutic modality. For HER2-targeted ADCs, resistance may arise from HER2 downregulation or loss, increased drug efflux, and the activation of alternative signaling pathways.12,13 For HER2-directed CAR T cells, limited in vivo expansion and persistence—together with impaired functionality arising from suboptimal antigen engagement or insufficient costimulatory signaling—remain major barriers to therapeutic efficacy.8,11 Chronic antigen exposure may further promote T cell dysfunction through tolerance mechanisms.14 To enhance CAR T cell expansion and activation, we engineered HER2-specific CAR T cells that co-express a chimeric costimulatory receptor (CCR)15,16 targeting intercellular adhesion molecule 1 (ICAM1). The HER2 CAR incorporates the FRP5 scFv fused to a 4-1BB co-stimulatory module and a CD3ζ signaling domain, whereas the ICAM1-specific CCR (termed ICCR) consists of the R6.5 scFv17,18 linked to CD28 and OX40 intracellular domains. ICAM1 was strategically selected as the CCR target based on our prior studies demonstrating strong induction by T cell-derived cytokines such as IL-1 and interferon-γ (IFN-γ),19 as shown in our gastric20,21 and thyroid cancer18 studies, association with poor prognosis in gastric cancer20 and elevation in tumors that undergo dedifferentiation, including those that lose markers like HER2.22,23,24 This dual-receptor configuration is designed to deliver primary activation through HER2 recognition and complementary costimulatory signals through ICAM1 engagement. The ICCR is intended to augment NF-κB-mediated signaling,25 particularly in contexts where HER2 expression is low or heterogeneous and CAR signaling alone may be suboptimal. The CAR architecture leverages 4-1BB to support sustained anti-tumor activity and persistence,26 while CD28 and OX40 incorporated into the CCR promote central memory differentiation and long-term functional durability.27,28,29,30 Here, we show that this dual-receptor strategy enhances CAR T cell function and improves tumor control in preclinical models of HER2-positive and HER2-low gastric and breast cancers. ICCR has also shown, even in the absence of HER2, the ability to drive clonal expansion of tumor-specific T cells through endogenous TCR-mediated antigen engagement. Results ICCR enhances HER2 CAR T cell activity against HER2-low target cells in vitro To provide complementary co-stimulation to CAR T cells, we constructed a lentiviral vector encoding a second-generation HER2 CAR (FRP5.4-1BB.CD3ζ) co-expressed with an ICAM1-specific CCR. The ICCR is comprised of an R6.5-derived scFv targeting ICAM1,17,18 fused to CD28 and OX40 costimulatory domains (Figures 1A and 1B). Both receptors were expressed from a single transcript via a self-cleaving T2A ribosome-skipping sequence. We first compared the cytolytic activity of HER2 CAR/ICCR T cells with the second-generation HER2 CAR T cells against a panel of target cell lines with varying levels of HER2 and ICAM1 expression. Flow cytometry revealed that HER2-positive cell lines, BT-474 (breast cancer), SK-BR-3 (breast cancer), and NCI-N87 (gastric cancer), expressed low levels of ICAM1 (1+), while HER2-low or negative cell lines, including MDA-MB-468 (breast cancer), SW1990 (pancreatic cancer), 8505C (thyroid cancer), MDA-MB-231 (breast cancer), Capan-2 (breast cancer), SNU-638 (gastric cancer), and HT-29 (colorectal cancer), exhibited moderate to high ICAM1 expression (2+ or 3+) (Figure 1C; Figure S1). These findings are consistent with prior reports of ICAM1 overexpression in HER2-low patient tumor samples,22 supporting its suitability as a CCR target, particularly against tumors with low or heterogeneous HER2 expression. In the cytotoxicity assays, incorporation of the ICCR significantly increased the lytic activity of HER2 CAR T cells against HER2-low target cells (SW1990, 8505C, MDA-MB-231, Capan-2, MKN-45, SNU-638, and HT-29) (Figure 1C; Figure S2). For HER2-high targets (BT-474, SK-BR-3, and NCI-N87), HER2 CAR and HER2 CAR/ICCR T cells exhibited similarly high cytotoxicity. Notably, HER2 CAR/ICCR T cells also mediated higher killing of HER2-negative MDA-MB-468 cells relative to HER2 CAR T cells alone, although lysis remained modest and below 15% (Figure 1C).Figure 1ICCR enhances HER2 CAR T cell activity against HER2-low/negative target cells(A) Lentiviral vectors encoding HER2 CAR (top) and HER2 CAR/ICCR (bottom) constructs. (B) Schematic representation of the HER2 CAR and HER2 CAR/ICCR expression in T cells. (C) Cytolytic activity of CAR T cells against target cells with varying levels of HER2 expression. CAR T cells were co-incubated with target cells at an E:T ratio of 2.5:1, and percentage of specific lysis was analyzed at 24 h later. Data represent mean ± SD (n = 4). The levels of cell surface HER2 and ICAM1 expression were characterized as negative (−), 1+, 2+, and 3+ based on flow cytometric analysis. Statistical significance was determined by two-way ANOVA with Šídák’s multiple comparisons test. (D) NF-κB activity of NF-κB-fLuc-expressing Jurkat cells that were engineered to express HER2 CAR, HER2 CAR-ICCR or left unmodified (NT) after a 16-h stimulation with HER2-negative (MDA-MB-468), HER2-low (293 T, SW1990, 8505C, MDA-MB-231, HT-29, and Capan-2) and HER2-high (SK-BR-3 and BT-474) target cells at an E:T ratio of 1:1. Each data point represents mean of triplicate wells against a single cell line. Statistical significance was determined by unpaired, two-tailed t test. (E) Expansion of NT, HER2 CAR T, and HER2 CAR/ICCR T cells following repeated stimulation with HER2-negative MDA-MB-468 cells every 3–5 days. Data represent mean ± SD (n = 3). Statistical significance was determined by two-way ANOVA with Tukey’s multiple comparisons test. (F) Frequency of CAR- or CAR/ICCR-expressing cells within HER2 CAR T or HER2 CAR/ICCR T cell populations before and after serial rounds of MDA-MB-468 stimulation, respectively. Data represent individual values from triplicate analyses. (G) Specific lysis of MDA-MB-468 cells by MDA-MB-468-stimulated HER2 CAR T or HER2 CAR/ICCR T cells at an E:T ratio of 2.5:1. Cytotoxicity was assessed 72 h after target cell exposure. Data represent mean ± SD (n = 3). Statistical significance was determined by unpaired, two-tailed t test. ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. (H) The proportions of the 20 most abundant TCR beta clonotypes from T cells collected on days 0 and 7 (HER2 CAR T) or 0, 7, and 19 days (HER2 CAR/ICCR T) after coincubation with target cells are displayed as pie charts. Each segment reflects the frequency of an individual clonotype, calculated as its clone size divided by the total number of unique CDR3 sequences. The cumulative percentage accounted for by the top 20 clones is indicated at the center of each chart. ICCR enhances HER2 CAR T cell activity against HER2-low/negative target cells (A) Lentiviral vectors encoding HER2 CAR (top) and HER2 CAR/ICCR (bottom) constructs. (B) Schematic representation of the HER2 CAR and HER2 CAR/ICCR expression in T cells. (C) Cytolytic activity of CAR T cells against target cells with varying levels of HER2 expression. CAR T cells were co-incubated with target cells at an E:T ratio of 2.5:1, and percentage of specific lysis was analyzed at 24 h later. Data represent mean ± SD (n = 4). The levels of cell surface HER2 and ICAM1 expression were characterized as negative (−), 1+, 2+, and 3+ based on flow cytometric analysis. Statistical significance was determined by two-way ANOVA with Šídák’s multiple comparisons test. (D) NF-κB activity of NF-κB-fLuc-expressing Jurkat cells that were engineered to express HER2 CAR, HER2 CAR-ICCR or left unmodified (NT) after a 16-h stimulation with HER2-negative (MDA-MB-468), HER2-low (293 T, SW1990, 8505C, MDA-MB-231, HT-29, and Capan-2) and HER2-high (SK-BR-3 and BT-474) target cells at an E:T ratio of 1:1. Each data point represents mean of triplicate wells against a single cell line. Statistical significance was determined by unpaired, two-tailed t test. (E) Expansion of NT, HER2 CAR T, and HER2 CAR/ICCR T cells following repeated stimulation with HER2-negative MDA-MB-468 cells every 3–5 days. Data represent mean ± SD (n = 3). Statistical significance was determined by two-way ANOVA with Tukey’s multiple comparisons test. (F) Frequency of CAR- or CAR/ICCR-expressing cells within HER2 CAR T or HER2 CAR/ICCR T cell populations before and after serial rounds of MDA-MB-468 stimulation, respectively. Data represent individual values from triplicate analyses. (G) Specific lysis of MDA-MB-468 cells by MDA-MB-468-stimulated HER2 CAR T or HER2 CAR/ICCR T cells at an E:T ratio of 2.5:1. Cytotoxicity was assessed 72 h after target cell exposure. Data represent mean ± SD (n = 3). Statistical significance was determined by unpaired, two-tailed t test. ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. (H) The proportions of the 20 most abundant TCR beta clonotypes from T cells collected on days 0 and 7 (HER2 CAR T) or 0, 7, and 19 days (HER2 CAR/ICCR T) after coincubation with target cells are displayed as pie charts. Each segment reflects the frequency of an individual clonotype, calculated as its clone size divided by the total number of unique CDR3 sequences. The cumulative percentage accounted for by the top 20 clones is indicated at the center of each chart. To evaluate downstream signaling, we measured NF-κB activity in HER2 CAR- and HER2 CAR/ICCR-expressing Jurkat T cells stimulated with target cells expressing varying levels of HER2 and ICAM1 (Figure 1D). These Jurkat T cells were transduced with an NF-κB-firefly luciferase (fLuc) reporter construct to enable quantification of NF-κB activation. Upon stimulation with HER2-low targets, HER2 CAR/ICCR T cells exhibited significantly higher NF-κB activity than HER2 CAR T cells. Both HER2 CAR and HER2 CAR/ICCR T cells showed robust NF-κB activation when exposed to HER2-high targets and minimal activation with HER2-negative target cells. Jurkat non-transduced (NT) cells without CAR expression did not exhibit NF-κB response against any targets, supporting that the response was specific to CAR and CCR engagement. ICCR promotes expansion and function of tumor-reactive T cells Building on our observation that HER2 CAR/ICCR T cells displayed enhanced killing of HER2-negative MDA-MB-468 cells (Figure 1C), we hypothesized that this activity reflected alloreactive T cells whose function was further potentiated by ICCR signaling. To further confirm this hypothesis, HER2 CAR T and HER2 CAR/ICCR T cells were repeatedly stimulated with MDA-MB-468 cells at 3–5-day intervals. Over 16 days, HER2 CAR/ICCR T cells expanded by nearly 10-fold, whereas HER2 CAR T cells failed to expand without HER2 antigen stimuli and remained static, similar to NT controls (Figure 1E). Selective enrichment of HER2 CAR/ICCR-positive cells was evident by increasing from ∼50% at baseline to nearly 90% after repeated stimulation, while no such enrichment occurred in HER2 CAR T cell cultures (Figure 1F). Moreover, HER2 CAR/ICCR T cells conditioned through serial exposure to MDA-MB-468 cells exhibited enhanced cytotoxic activity against MDA-MB-468 targets relative to HER2 CAR T cells (Figure 1G). Next, to determine whether expansion of HER2 CAR/ICCR T cells resulted in clonal bias, we examined TCR clonotypes in serially stimulated T cell cultures using TCR sequencing (TCR-seq). The frequency of specific clonotypes within the HER2 CAR/ICCR T cells increased progressively during co-incubation, rising from day 0 to day 2 and further expanding by day 19 (Figure 1H). By contrast, HER2 CAR T cells did not exhibit increased number of TCR clonotypes by day 2 and failed to expand sufficiently for analysis by day 19. Collectively, these data suggest that ICCR delivers robust co-stimulatory support for TCR-driven activation, enhancing the expansion and function of tumor-reactive T cells even in the absence of cognate CAR antigen. ICCR co-expression extends HER2 CAR T cell activity to HER2-low tumors We next evaluated the antitumor efficacy of HER2 CAR/ICCR T cells in metastatic breast cancer models using HER2-low MDA-MB-231 tumors and MDA-MB-231 tumors engineered to overexpress HER2 (MDA-MB-231-hHER2) via lentiviral transduction (Figure 2). Intravenous (i.v.) injection of either MDA-MB-231 or MDA-MB-231-hHER2 cells reliably established pulmonary metastases (Figures 2C–2H). In the HER2-high MDA-MB-231-hHER2 model, HER2 CAR/ICCR T cells exhibited antitumor activity comparable to that of HER2 CAR T cells, with both treatments achieving complete tumor remission within 2 weeks following a single injection of 1 × 107 CAR T cells (Figures 2A–2D). In contrast, in the HER2-low MDA-MB-231 model, HER2 CAR T cells delayed tumor progression but failed to induce tumor regression. Strikingly, treatment with HER2 CAR/ICCR T cells resulted in complete tumor eradication in three of four mice (Figures 2F–2I). All treated mice maintained stable body weight trajectories, and no overt signs of toxicity were observed (Figures 2E–2J). Notably, tumor clearance in the HER2-low model occurred with delayed kinetics, emerging approximately 4–5 weeks after T cell injection (Figure 2H), in contrast to the rapid responses observed in the HER2-high setting (Figure 2C). This delayed antitumor activity suggests a mechanism distinct from direct CAR-mediated cytotoxicity, likely reflecting ICCR-mediated costimulatory support for endogenous TCR-driven antitumor responses by tumor-reactive T cells.Figure 2HER2 CAR/ICCR T cells exhibit robust antitumor activity in both HER2-high and HER2-low breast tumor models(A) Flow cytometric histogram showing HER2 expression in MDA-MB-231-hHER2 cells engineered to overexpress HER2. (B) Experimental schematic for evaluating HER2 CAR and HER2 CAR/ICCR T cells in the HER2-high MDA-MB-231-hHER2 model. NSG mice were intravenously injected with 5 × 105 MDA-MB-231-hHER2 cells and treated 5 days later with CAR T cells (1 × 107 per mouse) or left untreated (No T). (C) Whole-body bioluminescence images (BLI). (D) Quantification of whole-body BLI over time. Data represent mean ± SD (n = 5 biologically independent mice). (E) Body weight changes over time (n = 5 biologically independent mice). (F) Flow cytometric histogram showing HER2 expression in parental MDA-MB-231 cells. (G) Experimental schematic for evaluating HER2 CAR and HER2 CAR/ICCR T cells in the HER2-low MDA-MB-231 model. NSG mice were intravenously injected with 5 × 105 MDA-MB-231 cells and treated 5 days later with CAR T cells (1 × 107 per mouse) or left untreated (No T). (H) Whole-body bioluminescence images. (I) Quantification of whole-body BLI over time. Data represent mean ± SD (n = 2–4 biologically independent mice). (J) Body weight changes over time (n = 2–4 biologically independent mice). Statistical significance in (D and I) was determined by two-way ANOVA with Tukey’s multiple comparisons test. ns, not significant; ∗p < 0.05. (K) The proportions of the 20 most abundant TCR beta clonotypes of T cell products and T cells collected from mouse spleen are displayed as pie charts. Each segment reflects the frequency of an individual clonotype, calculated as its clone size divided by the total number of unique CDR3 sequences. The cumulative percentage accounted for by the top 20 clones is indicated at the center of each chart. HER2 CAR/ICCR T cells exhibit robust antitumor activity in both HER2-high and HER2-low breast tumor models (A) Flow cytometric histogram showing HER2 expression in MDA-MB-231-hHER2 cells engineered to overexpress HER2. (B) Experimental schematic for evaluating HER2 CAR and HER2 CAR/ICCR T cells in the HER2-high MDA-MB-231-hHER2 model. NSG mice were intravenously injected with 5 × 105 MDA-MB-231-hHER2 cells and treated 5 days later with CAR T cells (1 × 107 per mouse) or left untreated (No T). (C) Whole-body bioluminescence images (BLI). (D) Quantification of whole-body BLI over time. Data represent mean ± SD (n = 5 biologically independent mice). (E) Body weight changes over time (n = 5 biologically independent mice). (F) Flow cytometric histogram showing HER2 expression in parental MDA-MB-231 cells. (G) Experimental schematic for evaluating HER2 CAR and HER2 CAR/ICCR T cells in the HER2-low MDA-MB-231 model. NSG mice were intravenously injected with 5 × 105 MDA-MB-231 cells and treated 5 days later with CAR T cells (1 × 107 per mouse) or left untreated (No T). (H) Whole-body bioluminescence images. (I) Quantification of whole-body BLI over time. Data represent mean ± SD (n = 2–4 biologically independent mice). (J) Body weight changes over time (n = 2–4 biologically independent mice). Statistical significance in (D and I) was determined by two-way ANOVA with Tukey’s multiple comparisons test. ns, not significant; ∗p < 0.05. (K) The proportions of the 20 most abundant TCR beta clonotypes of T cell products and T cells collected from mouse spleen are displayed as pie charts. Each segment reflects the frequency of an individual clonotype, calculated as its clone size divided by the total number of unique CDR3 sequences. The cumulative percentage accounted for by the top 20 clones is indicated at the center of each chart. To determine whether late-onset tumor regression observed in MDA-MB-231 model with HER2 CAR/ICCR T cell treatment was due to clonal expansion of tumor-reactive T cells, we analyzed TCR clonotype changes by sequencing TCR repertoires in peripheral blood and spleen. Both HER2 CAR T and HER2 CAR/ICCR T cells underwent significant clonal expansion, with the top ten TCRβ chains comprising 25% and 63% of the total TCR repertoire pools, respectively. The most dominant two clones accounted for 50% of the overall HER2 CAR/ICCR T cell clonotypes, compared with 15% in HER2-only CAR T cells. We speculate that a subset of clonally expanded HER2 CAR and HER2 CAR/ICCR T cells represent cells activated by recognition of mouse tissues, a well-described phenomenon when human T cells are introduced into murine hosts.31,32 In HER2 CAR/ICCR T cells, ICCR signaling likely contributed to the expansion of both tumor-reactive clones and T cells responding to murine tissues. Together, they underscore the therapeutic potential of HER2 CAR/ICCR T cells across a broad spectrum of HER2 expression and establish a clear rationale for advancing this dual-receptor platform toward clinical translation. Manufacturing of HER2 CAR/ICCR T cells under cGMP-compliant conditions Based on the preclinical efficacy shown with HER2 CAR/ICCR T cells, we established a current good manufacturing practice (cGMP)-compliant process for producing HER2 CAR/ICCR T cells for potential clinical application in HER2-positive and HER2-low solid tumors. Four lots of HER2 CAR/ICCR T cells (“HMJ01”) were successfully manufactured using the CliniMACS Prodigy platform, starting from cryopreserved peripheral blood leukopak material obtained from two healthy donors. Production closely adhered to our previously established clinical manufacturing protocol33 and yielded consistent, high-quality cell products across all runs. Starting from 30 million CD4+/CD8+-enriched T cells, an average of 3.5 ± 0.6 billion viable cells were obtained after a 10-day manufacturing process, representing a 117 ± 21-fold expansion (Figure 3A). Final products demonstrated high cell viability (98.0% ± 0.8%) and CD3+ T cell purity (99.3% ± 1.0%) (Figure 3B). Transduction efficiency was consistent across all lots, with 45.7% ± 9.0% HER2 CAR and 46.9% ± 4.0% ICCR expression in final product cells as assessed by flow cytometry (Figure 3B). Four lots of HMJ01 cells were evaluated for cytotoxicity against the panel of target cells (Figure S1). Their specificity and killing capacity were comparable to HER2 CAR/ICCR T cells produced manually using G-Rex 6 M well plate.Figure 3HER2 CAR/ICCR T cells effectively eliminate s.c. gastric tumors(A) Fold expansion of HMJ01 T cells over a 10-day manufacturing process. (B) Viability, CD3+ T cell purity, and surface expression of HER2 CAR and ICCR in HMJ01 T cell products manufactured with CliniMACS Prodigy system. (C) Expansion of HMJ01 T cells cultured in media with or without exogenous cytokine support. Data in (A)–(C) represent four manufacturing lots generated from leukopaks from two healthy donors (two lots per donor). (D) Experimental schematic for evaluating three batches of HMJ01 T cells in a subcutaneous NCI-N87 gastric tumor model. NSG mice were subcutaneously implanted with 1 × 106 NCI-N87 cells and treated 7 days later with non-transduced T cells (NT, 1 × 107 per mouse), HMJ01 T cells (1 or 2 × 107 per mouse) or left untreated (No T). (E) Whole-body bioluminescence images. (F) Quantification of whole-body BLI over time. Data represent mean ± SD. (G) Average tumor volume measurements over time. Data represent mean ± SD. Significance was determined by two-way ANOVA with Tukey’s multiple comparisons test. ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. (H) Tumor growth curves for individual mice. (I) Body weight changes over time. (E–I) No T, n = 5 mice examined over two independent experiments; NT, n = 4 mice examined over one independent experiment; HMJ01, n = 12 mice examined over two independent experiments using 3 different lots of CAR T cells manufactured with T cells from 2 healthy donors. HER2 CAR/ICCR T cells effectively eliminate s.c. gastric tumors (A) Fold expansion of HMJ01 T cells over a 10-day manufacturing process. (B) Viability, CD3+ T cell purity, and surface expression of HER2 CAR and ICCR in HMJ01 T cell products manufactured with CliniMACS Prodigy system. (C) Expansion of HMJ01 T cells cultured in media with or without exogenous cytokine support. Data in (A)–(C) represent four manufacturing lots generated from leukopaks from two healthy donors (two lots per donor). (D) Experimental schematic for evaluating three batches of HMJ01 T cells in a subcutaneous NCI-N87 gastric tumor model. NSG mice were subcutaneously implanted with 1 × 106 NCI-N87 cells and treated 7 days later with non-transduced T cells (NT, 1 × 107 per mouse), HMJ01 T cells (1 or 2 × 107 per mouse) or left untreated (No T). (E) Whole-body bioluminescence images. (F) Quantification of whole-body BLI over time. Data represent mean ± SD. (G) Average tumor volume measurements over time. Data represent mean ± SD. Significance was determined by two-way ANOVA with Tukey’s multiple comparisons test. ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. (H) Tumor growth curves for individual mice. (I) Body weight changes over time. (E–I) No T, n = 5 mice examined over two independent experiments; NT, n = 4 mice examined over one independent experiment; HMJ01, n = 12 mice examined over two independent experiments using 3 different lots of CAR T cells manufactured with T cells from 2 healthy donors. To assess the risk of malignant transformation, a concern given rare cases of secondary T cell malignancies in CAR T cell-treated patients,34,35 we evaluated the long-term growth potential of HMJ01 cells in T cell growth media without exogenous cytokine supplementation for 2 months. None of the four lots of HMJ01 cells (from two healthy donors) exhibited sustained expansion under these conditions, indicating an absence of autonomous growth or immortalization (Figure 3C). In contrast, all cultures expanded robustly in cytokine-supplemented complete T cell growth media for approximately 2 weeks before plateauing. These findings support the safety of HMJ01 cells with no evidence of cytokine-independent proliferation. HMJ01 CAR T cells effectively eliminate HER2-positive gastric tumors We next evaluated the efficacy of three independently manufactured lots of HMJ01 products (n = 2 donor leukopaks) against subcutaneous (s.c.) xenografts of HER2-positive gastric cancer cell line NCI-N87 (Figure 3D). In contrast to the continued tumor progression observed in left untreated (No T) and NT groups, which reached tumor volumes of approximately 1,000 mm3 a single i.v. dose of HMJ01 cells induced significant tumor killing. Eight out of twelve HMJ01-treated mice achieved complete responses, with tumors becoming non-palpable by the study endpoint (Figures 3E–3H). Across all HMJ01-treated mice, tumor volumes remained below 150 mm3. Additionally, these mice maintained stable body weight and exhibited no signs of treatment-related toxicity (Figure 3I). Gastric cancer patients with peritoneal carcinomatosis have a median survival of only 6 months due to limited treatment options.36 Peritoneal dissemination is the most common pattern of recurrence and metastasis in gastric cancer and is associated with poor prognosis.37 To test the antitumor activity of HMJ01 cells in a setting that more closely reflects clinical manifestations, we established an intraperitoneal (i.p.) gastric cancer model by xenografting HER2-high NCI-N87 cells (1 × 106) in the peritoneal cavity of NSG mice (Figure 4A). Five days later, mice were treated with a single dose of HMJ01 cells (1 × 107 cells) or left untreated. While control mice exhibited progressive tumor growth in the peritoneal cavity, all mice treated with HMJ01 cells showed tumor regression, resulting in significantly prolonged survival (Figures 4B–4E). Necropsy and ex vivo bioluminescence imaging confirmed the extensive peritoneal tumor burden and metastatic dissemination to the liver, spleen, lungs, and intestines in untreated mice (Figure 4F and 4G). In contrast, HMJ01-treated mice showed minimal residual disease, with greatly reduced or absent tumor signals in major organs (Figure 4F and 4G). Taken together, these studies highlight the strong translational potential of HMJ01 cells, even within the immunosuppressive, poorly penetrable tumor environment of the peritoneum, which typically presents significant challenges for CAR T cell therapies.Figure 4Antitumor activity of HER2 CAR/ICCR T cells in an intraperitoneal gastric cancer xenograft model(A) Schematic of the intraperitoneal NCI-N87 tumor model. NSG mice were implanted with 1 × 106 fLuc-expressing NCI-N87 cells via intraperitoneal injection and treated on day 5 post-xenograft with 1 × 107 HER2 CAR/ICCR T cells (HMJ01) via tail vein injection or left untreated (No T). (B) Whole-body bioluminescence images of NCI-N87-engrafted NSG mice. (C) Quantification of total body bioluminescence intensity. Data represent mean ± SD (n = 4–5 biologically independent mice). Significance was determined by two-way ANOVA with Tukey’s multiple comparisons test. ∗p < 0.05. (D) Summary of body weight changes over time. (E) Kaplan-Meier survival curves. p value was determined by log rank (Mantel-Cox) test. (F) Necropsy images obtained at 7 weeks post-treatment, with tumors outlined by blue dashed circles (n = 1). (G) Ex vivo bioluminescence imaging of tumors and major organs harvested from No T or HMJ01-treated mice at 9- and 13-weeks post-treatment, respectively (n = 3). Antitumor activity of HER2 CAR/ICCR T cells in an intraperitoneal gastric cancer xenograft model (A) Schematic of the intraperitoneal NCI-N87 tumor model. NSG mice were implanted with 1 × 106 fLuc-expressing NCI-N87 cells via intraperitoneal injection and treated on day 5 post-xenograft with 1 × 107 HER2 CAR/ICCR T cells (HMJ01) via tail vein injection or left untreated (No T). (B) Whole-body bioluminescence images of NCI-N87-engrafted NSG mice. (C) Quantification of total body bioluminescence intensity. Data represent mean ± SD (n = 4–5 biologically independent mice). Significance was determined by two-way ANOVA with Tukey’s multiple comparisons test. ∗p < 0.05. (D) Summary of body weight changes over time. (E) Kaplan-Meier survival curves. p value was determined by log rank (Mantel-Cox) test. (F) Necropsy images obtained at 7 weeks post-treatment, with tumors outlined by blue dashed circles (n = 1). (G) Ex vivo bioluminescence imaging of tumors and major organs harvested from No T or HMJ01-treated mice at 9- and 13-weeks post-treatment, respectively (n = 3). Discussion The present study demonstrates that integrating an ICCR into HER2-directed CAR T cells can significantly elevate antitumor activity by harnessing both HER2-directed CAR activity and endogenous TCR signaling. Although HER2-targeted therapies have transformed the management of HER2-expressing cancers, most patients ultimately experience disease progression or relapse despite sequential lines of HER2-directed agents. HER2 CAR T cells built with FRP5 scFv have been proven safe against indications of advanced sarcoma and brain cancers but have shown limited clinical response.8,9,11 This clinical reality underscores the need for more potent yet safe immunotherapies, particularly capable of functioning across a broad spectrum of HER2 expression, including tumors with heterogeneous, low, or negligible levels. Our findings show that HER2 CAR/ICCR T cells may possess such activity by coupling CAR-mediated recognition with potent, antigen-agnostic co-stimulation that amplifies endogenous TCR-driven immunity. A central insight from this work is that ICCR enables T cells to sustain productive signaling even when CAR antigen density is insufficient to support robust activation. In particular, when HER2 antigen density was low, ICCR augmented NF-κB signaling significantly in the presence of HER2 CAR. The ICCR function we describe here is consistent with prior studies showing that CCR targeting antigens such as GD2,38 CD38,15 CLEC12A,16 and CD8039 can augment either TCR- or CAR-mediated responses by increasing sensitivity to low antigen density and supporting T cell expansion and persistence. The distinguishing feature of ICCR relative to CCRs against other targets is its choice of ligand, ICAM1, which is inducible by inflammatory cytokines and upregulated in aggressive subtypes including HER2-enriched and triple-negative breast cancers.23 By analogy to the ICAM1-LFA-1 interaction in the periphery, which stabilizes TCR-peptide-MHC engagement at the immune synapse, ICCR may provide superior co-stimulatory support for the clonal expansion of T cells bearing tumor-reactive TCRs. Furthermore, in a separate study, ICCR engagement was shown to substantially enhance intercellular adhesion strength, surpassing the stability provided by native TCR-peptide/MHC and coreceptor interactions.25 This mechanical reinforcement conferred by ICAM1-targeting in the ICCR likely facilitates more efficient recruitment of signaling machinery and enhanced integration of NF-κB-biased co-stimulation, collectively lowering the activation threshold for both CAR- and TCR-mediated responses. The functional consequences of this synapse strengthening were evident in HER2-negative tumor models. HER2 CAR/ICCR T cells displayed enhanced killing of MDA-MB-468 cells, prompting the hypothesis that ICCR potentiates alloreactive or tumor-reactive T cells through endogenous TCR signaling. Repeated stimulation experiments confirmed this model: HER2 CAR/ICCR T cells underwent nearly 10-fold expansion over 16 days, whereas HER2 CAR T cells failed to expand and remained static. This proliferative advantage was accompanied by preferential enrichment of transgene-positive cells and superior cytotoxicity following serial conditioning. Although we have not directly demonstrated that this ICCR-driven expansion and cytotoxicity requires TCR recognition of peptide-MHC on target cells, our previous work showed that ICCR-mediated T cell proliferation and killing were completely lost when TCR was knocked out in T cells or MHC was ablated in target cells.25 A similar pattern emerged in vivo using HER2-low xenografts. Whereas HER2 CAR and HER2 CAR/ICCR T cells were equally effective against HER2-high tumors, HER2 CAR/ICCR T cells achieved significantly greater tumor control in HER2-low models, with a pronounced therapeutic advantage that became evident after approximately 30 days of treatment. This delayed yet abrupt tumor regression suggests the involvement of TCR-driven responses distinct from CAR-mediated signaling and is consistent with TCR recognition of alloantigens and the clonal expansion of tumor-reactive T cells. TCR-seq further supported this model, revealing progressive clonal accumulation in HER2 CAR/ICCR T cells, with the top two clones comprising half of the repertoire—more than 3-fold higher than in HER2-only CAR T cells. Collectively, these results show that embedding ICCR within the CAR architecture delivers strong costimulatory reinforcement to TCR-driven activation, a phenomenon we previously observed more directly in tumor-infiltrating T cells engineered to express ICCR.25 To support FDA investigational new drug (IND) submission, we developed a good manufacturing practice (GMP)-compliant manufacturing process using the automated, closed CliniMACS Prodigy platform, closely mirroring the procedures used in our prior clinical manufacturing protocol.33 HMJ01 T cells generated under these conditions were subsequently tested in vivo in HER2-positive gastric cancer models, including both s.c. and i.p. xenografts. In each model, HMJ01 achieved robust tumor control, demonstrating both feasibility and therapeutic potential for clinical translation. Based on the strength of the clinical manufacturing process and the supporting preclinical pharmacology and safety data, we received IND authorization to evaluate HER2 CAR/ICCR T (labeled HMJ01) in patients with HER2-positive or HER2-low gastric cancer. The limitations of our preclinical safety assessment, however, must be acknowledged. FRP5-based CAR T cells have previously shown an acceptable safety profile in patients, but adding a costimulatory signal could theoretically alter overall safety. Because FRP5 (anti-HER2) and R6.5 (anti-ICAM1) do not cross-react with murine orthologs, on-target, off-tumor toxicities cannot be evaluated in this mouse model, and safety monitoring was restricted to body weight, survival, and clinical appearance. Across these readouts, cohorts treated with HER2 CAR T cells versus HER2 CAR/ICCR T cells showed comparable body weight trajectories and no overt distress or abnormal behavior, indicating no evident increase in acute toxicity with ICCR co-expression. In the clinical setting, it will be important to understand not only how treatment responses align with the spectrum of HER2 expression—whether assessed by IHC or by HER2 copy number—but also how the immune system dynamically engages with the tumor. Accordingly, we will examine whether ICCR induces the emergence of tumor-specific clonal expansions and evaluate how these evolving T cell populations relate to clinical outcomes. In summary, these findings establish ICCR as a powerful co-stimulatory module that enhances NF-κB signaling downstream of antigen recognition, strengthens cellular contacts, promotes clonal expansion, and broadens the functional range of CAR T cells. By enabling effective tumor clearance across a continuum of HER2 expression, the HER2 CAR/ICCR platform addresses a major limitation of current antigen-restricted CAR therapies and provides a compelling rationale for clinical translation. Materials and methods Cell lines Human breast cancer cell lines (MDA-MB-468, MDA-MB-231, BT-474, SK-BR-3), gastric cancer cell line (NCI-N87), pancreatic cancer cell lines (SW1990, Capan-2), colorectal cancer cell line (HT-29), as well as 293 T and Jurkat T cell lines were purchased from the American Type Culture Collection (ATCC). Gastric cancer cell lines MKN-45 and SNU-638 were acquired from DSMZ and the Korean Cell Line Bank (Seoul National University, Seoul, Republic of South Korea), respectively. The human thyroid cancer cell line 8505C was obtained from Sigma. MDA-MB-468, MDA-MB-231, SW1990, and 293 T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Corning Life Sciences) supplemented with 10% fetal bovine serum (FBS, Gemini Bio). NCI-N87, BT-474, MKN-45, SNU-638, 8505C, and Jurkat T cells were maintained in RPMI-1640 medium (Corning Life Sciences) supplemented with 10% FBS. SK-BR-3, Capan-2, and HT-29 were cultured in McCoy’s 5a medium (ATCC) supplemented with 10% FBS. All tumor cell lines were transduced with a fLuc-F2A-GFP (FLuc-GFP) lentiviral vector (Biosettia) for bioluminescence-based cytotoxicity assays and in vivo imaging experiments. Cells were cultured in a humidified incubator at 37°C with 5% CO2 and routinely tested for mycoplasma using the MycoAlert detection kit (Lonza). CAR construct design The HER2 CAR/ICCR bicistronic construct was synthesized and cloned by GenScript into a lentiviral plasmid backbone under the regulation of a human elongation factor 1α (EF1α) promoter. The HER2 CAR consists of an scFv derived from the anti-HER2 mAb FRP5, followed by a CD8 hinge and transmembrane domain, a 4-1BB costimulatory domain, and the CD3ζ signaling domain. The ICCR includes an scFv from the R6.5 mAb, which binds human ICAM1 with approximately 10 nM Kd,40 a CD8 hinge, a CD28 transmembrane and costimulatory domain, and an OX40 costimulatory domain. The CAR and ICCR components are co-expressed from a single transcript via a self-cleaving T2A ribosome-skipping sequence. To enable the detection of expression by flow cytometry, a His tag and a FLAG tag were introduced at the N terminus of the CAR and ICCR, respectively. A second-generation HER2 CAR was used as a control and consists of a c-Myc tag, FRP5 scFv, CD8 hinge and transmembrane, and 4-1BB·CD3ζ signaling domains. CAR T cell manufacturing Lentiviral vectors suitable for IND-enabling studies were manufactured by GenScript using a third-generation lentiviral system. Viral titers were quantified by flow cytometry following transduction of Jurkat T cells with serial dilutions of the vector. HMJ01 CAR T cells were manufactured using the fully automated CliniMACS Prodigy system (Miltenyi Biotec) following a process previously described.33 Cryopreserved leukapheresis products (leukopak) from 2 healthy donors (AllCells) were thawed, diluted 1:2 with PlasmaLyte A (Baxter) containing 1% human serum albumin (HSA, Gemini Bio-Products), and sterile-docked into the Prodigy system. A single-use TS 520 tubing set (Miltenyi Biotec) was employed throughout the manufacturing process, beginning with immunomagnetic enrichment of CD4+ and CD8+ T cells. A total of 30 × 106 viable T cells were seeded into the Prodigy Culture Chamber at the end of day 0 and activated with MACS GMP T cell TransAct reagent (Miltenyi Biotec). Transduction was performed on days 1 and 2 using 1 mL of lentivirus vector each day in 10 mL of PlasmaLyte A with 1% HSA. Cells were expanded in TexMACS medium supplemented with 3% human AB serum, 12.5 ng/mL IL-7, and 12.5 ng/mL IL-15, and harvested on day 10. For non-GMP T cell manufacturing, T cells were similarly activated, transduced, and expanded in G-Rex 6 M well plate (Wilson Wolf Manufacturing) using identical complete T cell growth media. Final cell products were cryopreserved in a 1:2 mixture of PlasmaLyte A with 1% HSA or medium and CryoStor CS10 (STEMCELL Technologies). Flow cytometry Cells were washed with PBS containing 1% BSA and stained with fluorophore-conjugated antibodies at 4°C for 20 min. Subsequently, cells were washed twice and analyzed on Symphony A5SE or Symphony A1 flow cytometers (BD Biosciences). Data were analyzed using FlowJo software. The antibodies used in this study include antigen-presenting cell (APC) anti-human HER2 (BioLegend, Clone 24D2, Cat# 324408), APC anti-human ICAM1 (BioLegend, Clone HA58, Cat# 353112), AF647 anti-human CD3 (BioLegend, clone HIT3a, Cat# 300322), fluorescein isothiocyanate (FITC) anti-c-Myc tag (Miltenyi Biotec, clone SH1-26E7.1.3, Cat# 130-116-485), phycoerythrin (PE) anti-His tag (BioLegend, clone J095G46, Cat# 362603), APC anti-FLAG tag (BioLegend, clone L5, Cat# 637308). Calcein Blue (Sigma-Aldrich, Cat# M1255) staining along with forward- and side-scatter gating was used to exclude dead cells. Cell surface expression levels of HER2 and ICAM1 in target cells were classified as follows: negative, mean fluorescence intensity (MFI) fold change ≤2; 1+, MFI fold change >2 and ≤20; 2+, MFI fold change >20 and ≤200; and 3+, MFI fold change >200. Cytotoxicity assays fLuc-expressing tumor target cells (MDA-MB-468, SW1990, MDA-MB-231, MKN-45, SNU-638, Capan-2, HT-29, BT-474, SK-BR-3, and NCI-N87) were seeded in tissue-culture-treated flat-bottom white 96-well plates (Corning Life Sciences) at a density of 5 × 103 cells per well. CAR T cells were co-cultured with target cells at an effector-to-target ratio of 2.5:1 in medium containing 150 μg/mL D-luciferin (Gold Biotechnology). Luminescence was quantified by a TECAN Spark multimode microplate reader at specified time points. The percentage of viability was calculated by dividing the relative light units (RLUs) by those of target cell only wells. The cytotoxicity (%) was calculated as [(target cell only RLU-test RLU)/target cell only RLU] × 100%. NF-κB activation assay Jurkat T cells were transduced with a lentiviral reporter construct encoding fLuc under the control of the NF-κB promoter (QIAGEN, Cat# 336841), along with lentiviral vectors encoding either HER2 CAR or HER2 CAR/ICCR. Jurkat reporter cells (1 × 105) were co-cultured in triplicate with stimulator cells (1 × 105) expressing varying levels of HER2 and ICAM1 in tissue-culture-treated flat-bottom 96-well plates in 100 μL culture media containing 150 μg/mL D-luciferin (Gold Biotechnology). After 16 h of incubation at 37°C, luminescence was measured using a microplate reader (TECAN Infinite M1000PRO). NF-κB activation was quantified as relative luminescence intensity and normalized to unstimulated controls for each Jurkat T cell group. Cytokine-independent growth assay HMJ01 HER2 CAR/ICCR T cells (four different lots from two donors) were thawed and plated in G-Rex 6M multi-well culture plates (Wilson Wolf Manufacturing) in T cell growth media (TexMACS medium +5% human AB serum) either without exogenous cytokine supplementation or supplemented with 12.5 ng/mL of IL-7 and IL-15. Viable cells were counted and passaged every 4–7 days by replacing half of the media with fresh culture media. In vivo mouse studies Six to eight week old male and female NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were purchased from The Jackson Laboratory (Stock# 005557) and housed under specific pathogen-free conditions at the animal core facility of Houston Methodist Research Institute (Houston, TX, USA). All animal procedures were approved by the Institutional Animal Care and Use Committee at Houston Methodist Research Institute. The s.c. gastric tumor model was established by injecting 1 × 106 fLuc-expressing NCI-N87 cells suspended in 50 μL RPMI-1640 medium (Corning Life Sciences) and mixed 1:1 with 50 μL of Matrigel (Corning Life Sciences) into the upper flank. For the i.p. gastric tumor model, 0.5 × 106 fLuc-expressing NCI-N87 cells were injected into the peritoneal cavity. Systemic metastatic models of breast cancer were generated via i.v. injection of fLuc-expressing MDA-MB-231-hHER2 (0.5 × 106 per mouse) or MDA-MB-231 (0.5 × 106 per mouse) cells. Freshly thawed CAR T and NT control T cells were administered i.v. at doses of 1 or 2 × 107 cells/mouse as indicated in the schematics and figure legends. Tumor growth was monitored weekly using the IVIS Spectrum in vivo imaging system (PerkinElmer). Mice were injected with 200 μL of 15 mg/mL D-luciferin (Gold Biotechnology) 15 min prior to imaging, i.p. for s.c. and systemic models and s.c. for i.p. tumor-bearing mice. Bioluminescence images were analyzed using the Living Image v.4.7.2 (PerkinElmer). For s.c. tumors, tumor volume (V) was measured with a caliper and calculated using the formula V = (length × [width]2)/2. Mice were randomized to ensure uniform tumor establishment before T cell treatment. For ex vivo imaging, mice were euthanized 15 min after D-luciferin injection, and major organs were collected and imaged using the IVIS Spectrum system. Peripheral blood samples were collected via lateral saphenous vein at indicated time points. Serum was isolated by centrifugation at 2,000 × g for 10 min at 4°C and stored at −80°C for cytokine analysis. Mice were monitored daily by the husbandry staff and euthanized using a CO2 chamber followed by cervical dislocation when humane endpoints were reached, such as hunched posture, rough coat, impaired mobility, paralysis, or >20% body weight loss. Spleens from mice treated with either HER2 CAR T or HER2 CAR/ICCR T cells were collected at study endpoint and flash-frozen. Splenocytes were isolated and mechanically dissociated for total RNA extraction prior to TCR-seq. TCR repertoire sequencing TCR β-chain variable region was amplified from cDNA using degenerate primer sets.41 Total RNAs were extracted from T cells using QIAGEN Mini RNA kit, and first-strand cDNA was synthesized using Superscript III (Invitrogen) with oligo(dT) primers. PCR-amplified Vβ products were purified and subjected to next-generation sequencing on an Illumina MiSeq platform (sequencing performed at the MD Anderson ATGC core facility). TCR β-chain variable region sequences (TRBV) were analyzed using IgBlast configured for TCR annotation. Sequences were aligned to the IMGT-curated germline database to determine V(D)J gene usage, CDR3 boundaries, and productive rearrangements. Statistics Statistical analyses were performed using Prism 10 (GraphPad, Inc.). An unpaired, two-tailed t test or two-way ANOVA with a multiple comparisons test was used to evaluate statistical significance between groups. Mouse survival curves were generated using the Kaplan-Meier method, and statistical significance was assessed using the log rank (Mantel-Cox) test. A p value <0.05 was considered statistically significant. Data and code availability All data relevant to the study are included in the article or uploaded as supplemental information. Requests for additional data should be made to the corresponding author.