Background

Autologous chimeric antigen receptor (CAR) T-cell therapy has revolutionized treatment for patients with B-cell leukemia, and multiple myeloma. Over one-third of all CAR T-cell patients treated to date with commercial CAR T cells products targeting CD19 or B-cell maturation antigen (BCMA), respectively, achieve complete and durable remissions.1 However, despite wide-scale efforts to tackle solid tumors, which account for 90% of all cancer types, they have yet to demonstrate similarly high therapeutic efficacy to that observed in hematologic malignancies. The immunosuppressive tumor microenvironment (TME) has been identified as one major challenge to the success of solid tumor CAR T-cell therapies in solid tumors,2 and will need to be addressed.

Receptor tyrosine kinase-like orphan receptor 1 (ROR1) is an oncofetal protein and an attractive target for immunotherapy of solid and hematologic tumors. ROR1 plays an important role during early embryonic development but remains absent from vital adult human tissues, except for expression in a subset of immature B-cell precursors in adult bone marrow, and low-level expression in adipocytes.3 4 By contract, ROR1 is overexpressed on the surface of a large array of hematologic tumors, including B-ALL, B-CLL, MCL, FL, MZL, DLBCL, and a subset of solid tumors, including ovarian, pancreatic, lung, skin, breast, and colon cancers.5–7 Monoclonal antibody moieties targeting ROR1 in the form of a naked antibody (cirmtuzumab), or as antibody-drug conjugate (zilovertamab vedotin) demonstrated an acceptable safety profile and showed antitumor activity in some subsets of treated patients,8 suggesting that targeting ROR1 for cancer therapy can be both safe and efficacious. However, to our knowledge, no study to date has demonstrated robust efficacy of ROR1-taregting CAR T-cell immunotherapy in solid tumors. In one phase I basket clinical trial including hematologic and patients who had a solid tumor, in which a number of patients with triple-negative breast cancer and non-small cell lung cancer were treated, ROR1-targeting CAR T cells exhibited an acceptable safety showed modest efficacy towards solid tumors. Post-treatment biopsy in one patient with non-small cell lung cancer with partial response revealed that CAR T cells infiltrated tumor poorly and became dysfunctional due to defective homing, low persistence, and diminished function at tumor sites.9 10 All of these characteristics are known to be precipitated by the immunosuppressive TME in solid tumors.

Transforming growth factor beta (TGFβ) is a master regulator of TME, and is known to be secreted by tumor cells, stromal fibroblasts, and other cells in many solid cancers, creating an immunosuppressive environment, inhibiting T-cell effector function, cytokine response, proliferation, and memory formation, promoting neoangiogenesis and metastasis.11 There are three isoforms of TGFβ in mammals: TGFβ1, 2, and 3. TGFβ signals by binding to TGFβ receptors one and two (RI and RII) on the cell surface, leading to phosphorylation and activation of transcription factor Smad2/3, which in turn activates responsive genes that inhibit T-cell proliferation and differentiation into helper T cells and cytotoxic T lymphocytes (CTLs).12 Overcoming the immunosuppressive effects of TGFβ in TME may therefore offer a unique opportunity to simultaneously improve multiple CAR T-cell attributes. Modulating the anti-tumor inhibitory effect of TGFβ has been studied by other groups and ours, including armoring CAR T with dominant-negative TGFβRII targeting Prostate-specific membrane antigen (PSMA) in prostate cancer13 and BCMA in multiple myeloma models,14 or knocking out TGFβRII in CAR T cells.15 Clinical trial employing PSMA-CAR T armored with a dominant negative (DN) form of TGFβRII showed promising results in patients with prostate cancer when administered at a safe dose.16

Here we report a novel, fully human ROR1-CAR-1 employing the single chain variable fragment (scFv)9 targeting domain, which effectively eliminated hematologic and solid tumor xenografts in mice. Furthermore, CAR-1 T cells armored with the TGFβRIIDN element overcame the inhibitory effect of TGFβ in vivo in pancreatic ductal adenocarcinoma(PDAC) xenograft models with constitutive or treatment-induced expression of TGFβ1. These findings support the application of TGFβRIIDN armor as a tool to ameliorate immunosuppressive TME for better treatment of patients with both hematologic and solid malignancies.

MethodsGeneration of CAR constructs, lentiviral vector production and titration

The scFv sequence scFv9 or comparator sequence scFv R12,17 targeting the extracellular domain of human ROR1 was linked in frame to a human IgG4 hinge (aa 99–110 UniProt sequence ID P01861), human CD8+transmembrane domain (aa 183–206, UniProt sequence ID P01732), human 4-1BB co-stimulatory domain (aa 214–255, UniProt sequence ID Q07011), and human CD3-ζ activating domain (aa 52–163, UniProt sequence ID P20963), to generate CAR-1 and CAR-2, respectively. The bicistronic armored anti-ROR1 CAR was designed by combining ROR1-CAR-1 with the TGFβRIIDN armor via P2A ribosomal skip sequence. CAR sequences were cloned into a lentiviral vector (LV) expression cassette under the control of the human EF-1α promoter (Lentigen Technology, Gaithersburg, Maryland, USA). Lentiviral particles were generated by transient transfection of HEK 293 T cells, pelleted by centrifugation and stored at −80°C until transduction. LV titers were determined by the serial transduction of SUP-T1 cell line and real-time quantitative polymerase chain reaction (qPCR) analysis of gag and pol expression.

Primary human T cells

Whole blood was collected from healthy volunteers at Oklahoma Blood Institute (OBI) with donors’ written consent. Processed buffy coats were purchased from OBI (Oklahoma City, Oklahoma, USA). The human CD4+ and CD8+T cells were purified from buffy coats via positive selection using a 1:1 ratio of CD4-MicroBeads and CD8-MicroBeads (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s protocol resulting in a mixture of enriched CD4+ and CD8+T cells. CAR-T cells transduction and culture were performed as previously described.18

Cytotoxicity assay was performed as previously described18 with effector to target (E:T) ratio calculated in reference to CAR+T cells. Absolute potency (EC50) and relative potency of effector T cells were calculated using Prism software with a 4-parameter parallel-line analysis approach.

Impedance-based cytotoxicity assay

The assay was performed employing xCELLigence RTCA MP analyzer (Agilent Technologies, Santa Clara, California, USA) following the manufacturer’s instructions. Briefly, 40,000 AsPC-1 cells were co-cultured with 80,000 effector cells (ie, E:T ratio=2:1) and the cytolysis was monitored for 3 days. Data were analyzed by RTCA Software Pro (Agilent Technologies, Santa Clara, California, USA).

In vivo studies

In vivo studies were performed at LabCorp (Ann Arbor, Michigan, USA), as described in the online supplemental materials and methods, and are reported according to the ARRIVE guidelines.19 The animal studies were reviewed and approved by LabCorp Drug Development (Formerly Covance) Institutional Animal Care and Use Committee, ID: D16-00871 (A4671-01).

MACSima imaging cycling staining multiplex immunofluorescence analysis

Formalin-fixed paraffin-embedded (FFPE) xenograft tissues were prepared with deparaffinization (Xylene and Ethanol; Sigma-Aldrich, St. Louis, Missouri, USA) and antigen retrieval (TEC Buffer with pH 9) prior to immunofluorescence (IF) staining. We used MACSima imaging cycling staining (MICS) system (Miltenyi Biotec, Bergisch Gladbach, Germany) to acquire multiplex IF images. All antibodies were purchased from Miltenyi Biotec (Bergisch Gladbach, Germany) except for alpha-smooth muscle actin (34105S; Cell Signaling Technology, Danvers, Massachusetts, USA), TGFß1 (EPR21143; Abcam, Cambridge, UK). Multiplex images were processed using MACS iQ View software (Miltenyi Biotec, Bergisch Gladbach, Germany).

Statistical analysis

Statistical analysis was performed using Prism V.9.3.1 software (GraphPad, San Diego, California, USA). Measurements of bioluminescence, cytotoxicity of target cells, and expansion of T cells, were log-transformed prior to analysis using parametric tests. Data normality was determined by Kolmogorov-Smirnov test. For normally distributed data sets, statistical significance was determined by Student’s t-test, or one-way or two-way analysis of variance, followed by Tukey’s or Sidak’s multiple comparison test, respectively. For non-normally distributed data sets, non-parametric Mann-Whitney test, Kruskal-Wallis test with Dunn’s post hoc test, or Wilcoxon matched-pairs signed-rank test was used, as appropriate. Survival was evaluated by the Kaplan-Meier test; ns (non-significant) p>0.05, ∗p<0.05, ∗∗p<0.01, ∗∗∗p<0.001, and ∗∗∗∗p<0.0001. Error bars represent SEM.

Additional methods are provided in the online supplemental materials and methods.

ResultsThe novel fully-human ROR1-CAR-1 demonstrates high cell surface density, similar cytotoxic activity, and superior cytokine production as compared with the comparator ROR1-CAR-2 in vitro

In the present study, we employed a novel ROR1-targeting CAR construct containing a fully-human binder scFv9, termed CAR-1, and a comparator CAR based on the clinically-characterized construct using the chimeric rabbit/human scFv-R12,9 17 termed CAR-2 (figure 1A). Both CARs were comprised of a short IgG4-Fc hinge, 4-1BB co-stimulatory domain, and CD3 zeta signaling domain. The affinity of CAR binding domains was characterized by biolayer interferometry. The scFv9 and scFv R12 showed a comparable binding affinity towards human ROR1 protein, whereas no binding to mouse ROR1 protein was observed for either binder, confirming their specificity to human ROR-1 (online supplemental figure 1A). Lentiviral transduction of human primary T cells with CAR-1 or CAR-2 at multiplicity of infection 40 yielded viral copy number integration within the clinically acceptable range (<5 copies per cell, online supplemental figure 4). Moreover, ROR1 CAR-1 exhibited a greater T-cell surface density (mean fluorescence intensity, average MFI 164 vs 43, p<0.01) as compared with CAR-2, with a similar percentage of total CAR-positive cells (figure 1B,C). ROR1 CAR-1 and CAR-2 products had a comparable percentage of CD8+CAR T cells (p>0.05), whereas CAR-2 had a slightly greater percentage of CD4+CAR T cells (p<0.05), however, CAR-1 MFI was higher in both CD4+and CD8+ CAR-positive populations (p<0.001); (online supplemental figure 1B).

Figure 1

Characterization of in vitro and in vivo function of the novel ROR-1 CAR-1, and the comparator CAR-2, against hematologic tumors. (A) Schematic diagram of CAR-1 and CAR-2 ROR1-1 targeting constructs. (B) Representative flow cytometry plots for ROR1 CAR expression on CAR-1 and CAR-2-transduced normal donor T cells, and UTD control. (C) Pooled CAR expression percentage CAR+cells (left) and MFI (right) for transduction experiments in T cells from three healthy donors, and un-transduced T cells (UTD) control, as measured by flow cytometry at day 8 post transduction; mean±SEM, Student’s t-test. (D) Quantification of ROR1 cell surface density, expressed in molecules per cell, in different hematologic cell lines, as quantified by flow cytometric ABC assay, n=2. (E) In vitro cytotoxic activity of CAR T cells when co-cultured for 18 hours with Jeko-1, RPMI-8226, or HL-60 tumor cell lines; one representative donor of three is shown. (F) Quantification of cytokines secreted in 18 hours co-culture of CAR T cells with Jeko-1 cell line by ELISA, pooled results from three donors, mean±SEM, mixed effect model. (G) NSG mice were implanted with Jeko-1 cells (i.v., 0.5e6 cells/mouse; 6 mice/group) at day #−6, followed by staging at day #−1, CAR T cells were administered (i.v., 3e6 CAR+T cells/mouse) at day #0; tumor progression was quantified by bioluminescence imaging (H, I). Tumor progression in study groups was compared by Kruskal-Wallis analysis, with Dunn’s post hoc test. Body weight change was monitored (J) peripheral blood was sampled at the indicated time points and the tumor cells (K) and human T cells (L) were quantified by flow cytometry. Student’s t-test. All data are shown as mean±SEM; statistical significance is denoted as *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001, ns-non-significant. CAR, chimeric antigen receptor; IFN, interferon; IL, interleukin; i.v., intravenous; MFI, mean fluorescence intensity; ROR1, receptor tyrosine kinase-like orphan receptor 1; scFv, single chain variable fragment; TNF. tumor necrosis factor.

ROR1 is a 120 kDa protein containing extracellular immunoglobulin-like, Frizzled, and Kringle domains.3 To evaluate the domain specificity of CAR-1 and CAR-2, we incubated CAR T cells with protein fragments containing the sequence of Ig-like, Frizzled, or Kringle regions of the ROR1 receptor ectodomain, and quantified CAR-bound ROR1 fragments by flow cytometry. ROR1 CAR-2 is associated with both the Ig-like and the Frizzled region of ROR1, in agreement with a previous report,17 and thus validating our approach. By comparison, CAR-1 bound to the Ig-like region fragment of ROR1, but not to the Frizzled domain fragment. Therefore, CAR-1 targets ROR1 ectodomain region that is different from, but partially overlapping with that of CAR-2 (online supplemental figure 1C).

Next, we evaluated the cytotoxicity of ROR1-CARs in vitro against ROR-1 positive cell lines Jeko-1 (MCL), and RPMI-8226 (multiple myeloma), and the acute promyelocytic leukemia HL-60 ROR-1 negative cell line, in an overnight co-culture assay. CAR-1 and CAR-2 showed similar killing potency towards Jeko-1, and RPMI-8226 cell lines which express ROR1 at high density (figure 1D,E, online supplemental figure 1D,E). No cell lysis of the ROR-1 negative HL-60 cell line was detected (figure 1E); demonstrating the specificity of CAR T cells against ROR1. Further, Jeko-1 cells elicited significantly higher production of cytokines interferon (IFN)-γ, tumor necrosis factor (TNF)-α and interleukin (IL)-2 from CAR-1 as compared with CAR-2 (figure 1F), suggesting a more efficient antitumor response of CAR-1.

CAR-1 and CAR-2 were equally effective in eradicating hematologic Jeko-1 MCL disseminated xenografts in vivo

We next evaluated the antitumor activity of CAR-1 and CAR-2 in Jeko-1 xenograft model (figure 1G).CAR-1-transduced and CAR-2-transduced T-cell products showed similar CD4 and CD8 composition before infusion (online supplemental figure 1F), and mediated comparable tumor regression. Although CAR-2, as compared with CAR-1, appeared to mediate somewhat shallower level of remission on day 27, comparable remissions on day 34, and marginally deeper remissions on day 41, the differences between the experimental groups were not statistically significant. Furthermore, both CARs improved survival in Jeko-1-bearing NSG mice, whereas tumor alone and un-transduced T cells (UTD) control groups exhibited rapid tumor progression (figure 1H,I). No significant body weight loss was observed in mice treated with either CAR during this study (figure 1J). Peripheral blood analysis revealed efficient clearance of Jeko-1 cells post CAR-1 or CAR-2, whereas the number of Jeko-1 cell increased 1000-fold in untreated mice by study day 10 and remained high thereafter (figure 1K), indicating tumor progression. By contrast, starting from day 3 post CAR therapy, peripheral blood CAR-1+CD8+ and CAR-1+CD4+ T cells both expanded faster than the respective CAR-2+populations (figure 1L), demonstrating that CAR-1 reacted to tumors more vigorously.

CAR-1 showed enhanced cytokine response and equal or greater cytotoxicity against ROR1+ OVCAR-3, Capan-2 and NCI-H226 solid tumor cell lines in vitro, as compared with CAR-2

We further investigated the antitumor reactivity of CAR-1 and CAR-2 ROR1 T cells against solid cancer cell lines in vitro. Flow cytometric quantification of ROR1 surface density on solid tumor cell lines NCI-H226, Capan-2, and OVCAR-3 (lung, pancreatic, ovarian, respectively) revealed a range of ROR1 expression densities (figure 2A, online supplemental figure 2A). In overnight co-culture assays, CAR-1 T cells, as compared with CAR-2, exhibited comparable cytotoxic potency against OVCAR-3 and NCI-H226 tumor lines, but greater potency against Capan-2, which may reflect overcoming the intrinsic resistance of pancreatic tumors to T-cell therapy by CAR-1 (figure 2B, online supplemental figure 2B). Additionally, greater amounts of IFN-γ, TNF-α, and IL-2 were produced by CAR-1 as compared with CAR-2 T cells against the ovarian carcinoma OVCAR-3 as well as NCI-H226 and Capan-2 cell lines (figure 2C), consistently with greater elaboration of cytokines by CAR-1 in response to hematologic tumor lines (figure 1F). These results demonstrate a universal heightened cytokine response of CAR-1, irrespective of tumor type.

Figure 2

CAR-1, but not CAR-2, suppressed tumor progression in the orthotopic OVCAR-3 mouse xenograft model of ovarian cancer, and exhibited higher cytokine response to solid tumor lines in vitro. (A) Quantification of ROR1 expression on the surface of various solid tumor cancer cell lines; the experiment was performed in duplicates employing anti-ROR1 Ab from BD Biosciences; a separate experiment was also performed in duplicates using anti-ROR-1 Abs from Miltenyi Biotec and R&D Systems with similar results. (B) Representative killing curves of CAR T cells against various solid cancer cell lines (OVCAR-3, Capan-2, and NCI-H226) in an 18 hours co-culture. (C) Cytokine production from the experiments in (B) was quantified by ELISA, pooled results from three independent donors are shown, mean±SEM. Experimental groups were compared by mixed effect model (D–I): Efficacy of CAR T cells in in vivo ovarian cancer OVCAR-3 xenograft model: NSG mice (five mice/group) were implanted (i.p.) with OVCAR-3 cell line (10e6 cells/mouse) at day −7, followed by staging at day −1; CAR T cells (5e6 CAR+T cells/mouse) were administered (i.v.) at day 0 (D); tumor progression was quantified by bioluminescence imaging; groups were compared by mixed effect analysis with Tukey’s post hoc test. (E, F) Body weight was monitored (G); blood was sampled at the indicated time points to quantify CAR+T cells in both CD8+ and CD4+ subpopulations (H) as well as memory T cells (I). Groups were compared by two way analysis of variance with Sidak’s post hoc test (H) or Student’s t-test (I). All results are presented as mean±SEM; statistical significance is denoted as *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001, ns-non-significant. Ab, antibody; BLI, bioluminescence; CAR, chimeric antigen receptor; IFN, interferon; IL, interleukin; i.p., intraperitoneal; i.v., intravenous; ROR1, receptor tyrosine kinase-like orphan receptor 1; TEM, efector memeory T cells; TCM - central memory T cells; TNF, tumor necrosis factor; UTD, un-transduced T cells.

Only CAR-1, but not CAR-2, mediated tumor regression in in vivo ovarian cancer model

To investigate the antitumor response of CARs against solid tumor in in vivo, we chose the OVCAR-3 ovarian xenograft cancer model (figure 2D). Although both CARs showed similar cytotoxic activity against OVCAR-3 in vitro, only CAR-1 mediated OVCAR-3 tumor regression in vivo (figure 2E,F), it is worth noting that CAR-1 and CAR-2 CAR-T products had similar CD4+ and CD8+ CAR+ T-cell composition before infusion (online supplemental figure 2C). Additionally, mice administered CAR-1+T cells did not lose weight as compared with mice administered either CAR-2, UTD T cells, or tumor alone (figure 2G). The rapid expansion of CD8+T and memory T cells after CAR T-cell administration predicts positive clinical outcomes.20 Analysis of peripheral blood samples from mice treated with either CAR-1 or CAR-2+ T cells revealed a more rapid expansion of CAR-1 in both CD8+ and CD4+cell compartments (figure 2H). The percentage of CD8+T cells was significantly elevated in CAR-1 treatment group as compared with CAR-2 on days 3–17, and the percentage of CD4+T cells was elevated on days 3–10, respectively, and was similar to CAR-1 expansion kinetics in the hematologic JeKo-1 xenograft (figure 1L). Additionally, from day 3 to day 10 post administration, CAR-1-transduced T cells showed a rapid expansion of effector memory T cells (TEM) in both CD8 (from 0.53% to 23%) and CD4 (from 4.2% to 17.6%) compartments, indicative of prompt effector CAR T-cell activity, as compared with CAR-2 (from 5% to 17% for CD8+TEM, and no increase in % of CD4+TEM). Similarly, CD8+ and CD4+CAR-1+central memory T cells (TCM) cells expanded faster than the respective CAR-2 populations (figure 2I), demonstrating the effective formation of immune cell memory and reserve for durable CAR-1+T-cell function. In summary, CAR-1 exhibited superior antitumor efficacy in in vivo xenograft model of ovarian cancer, which is attributed to the timely expansion of CAR+T cells, and enrichment for TEM and TCM phenotypes in both CD8+ and CD4+T-cell populations.

TGFβRIIDN-armored CAR-1 attenuated the inhibitory effect of TGFβ1 on CAR T-cell cytotoxic activity in vitro

Having demonstrated high in vitro and in vivo potency of ROR-1 CAR T cells, we proceeded to protect CAR-1+T cells from the inhibitory effects of TGFβ. We constructed LV co-expressing ROR1-CAR and the TGFβRIIDN armor element, separated by ribosomal skip site, to facilitate co-expression of the two polypeptides (figure 3A). TGFβRIIDN is a truncated form of TGFβ receptor II, capable of TGFβ binding, but devoid of intracellular signaling activity,14 thus attenuating the TGFβ-induced suppression of T cells. The armored ROR1-CAR was expressed robustly on healthy donor T cells with comparable enriched CAR+TN and TCM phenotypes in both the CD8+ and CD4+T-cell fraction, similarly ROR-1 CAR alone (figure 3B). The overexpression of TGFβRIIDN element on the surface of armored CAR-1+T cells was visualized by flow cytometry using an anti-TGFβRII antibody (figure 3C). Canonical TGFβ signaling through TGFβRII leads to phosphorylation of transcription factors Smad2 and Smad3 (pSmad 2/3). We observed a time-dependent reduction of pSmad2/3 in TGFβRIIDN-armored CAR-1+T cells treated with TGFβ1 compared with non-armored CAR-1+T cells (figure 3D), which validates the functional effect of the DN TGFβRII on TGFβ1 signal transduction.

Figure 3

Dominant negative TGFβRII (DN) obstructed TGFβ1 signaling in T cells transduced with CAR-1 and reduced the inhibitory effect of TGFβ1 on CAR T cells’ cytotoxic activity against pancreatic cancer cell line AsPC-1 in vitro. (A) schematic diagram of constructs of CAR-1 alone and CAR-1 armored with DN. (B) At day 8 of transduction, CAR expression (left: flow plots, center: graph from the flow plots) and memory phenotype (right) of both CD8+ and CD4+T cells transduced with CAR-1 or CAR-1+DN were analyzed by flow cytometry; three independent experiments were performed, employing three donors, with similar results. (C) Expression of TGFβRII in T cells transduced with CAR-1 or CAR-1+DN was assessed by flow cytometry; three independent experiments were performed, employing three donors, with similar results. (D) CAR T cells were IL-2 starved for 22 hours to synchronize the cells followed by treatment with TGFβ1 (10 ng/mL) for 5, 15, or 30 min; cells were then stained with pSmad2/3 and subject to flow cytometry analysis; left panel: flow plots: right panel: graph from the plots in the left panel (results are the mean±SE of three donors). (E) Expression of ROR1 on AsPC-1 cell line was assessed by flow cytometry. (F) AsPC-1 was co-cultured with CAR T cells without or with TGFβ1 (1 or 10 ng/mL); tumor cell lysis was measured by xCELLigence; left: % cytolysis (results are representative of three donors); right: cytotoxic relative potency of CAR T cells treated with TGFβ1 versus non-treatment (results are the mean±SE of three donors). Mean±SE of three donors. Two-way analysis of variance with Sidak’s post hoc test *p<0.05; ***p<0.001; **p<0.01; ***p<0.001. (G) Cytokine production from the experiments in (F) was quantified by ELISA; results are the mean±SE of three donors. (H,I) Production of TGFβ1 either in active or latent form by various solid tumor cell lines (H) or by AsPC-1 ectopically overexpressing TGFβ1 (I) was assessed by ELISA; data are representative of two independent experiments with similar results. (J) AsPC-1 overexpressing TGFβ1 (AsPC-1/TGFβ1) or AsPC-1 control was co-cultured with CAR T cells, % cytolysis of tumor cells was shown, results are the representative of three donors. (K) Cytokine production from the experiments in (J) was quantified by ELISA. Mean±SE of three donors; Student’s t-tests; statistical significance is denoted as *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001, ns-non-significant. CAR, chimeric antigen receptor; IFN, interferon; IL, interleukin; TGF, transforming growth factor; TNF, tumor necrosis factor; UTD, un-transduced T cells.

TGFβ is known for its negative effect on cytotoxic T cells, including inhibition of the expression of multiple effector molecules (granzyme A, granzyme B, perforin, IFN-γ and TNF-α).21 To demonstrate the functional effect of TGFβRIIDN on antitumor activity of CAR-transduced T cells in vitro, we co-cultured CAR T cells with pancreatic adenocarcinoma AsPC-1 cells,22 highly positive for ROR1 (figure 3E), in the presence of TGFβ1. TGFβ1 reduced the cytotoxic activity of CAR-1+T cells, and decreased the production of IFN-γ, TNF-α, and IL-2 in the co-culture supernatant, all of which were restored in the armored CAR-1+T cells (figure 3F,G). AsPC-1 cells naturally express low levels of latent form of TGFβ1 in cell culture which was detected by ELISA on activation by acidic treatment of culture supernatants (figure 3H). AsPC-1 cell line engineered to stably overexpress TGFβ1 (AsPC-1/TGFβ)23 was therefore employed to investigate the effect of the TGFβRIIDN armor in vitro and in vivo. AsPC-1/TGFβcells produced high amounts of both active (approximately 16,000 pg/mL) and latent (pg/mL) forms of TGFβ1 when cultured overnight (figure 3I). Cytotoxicity of CAR-1+T as well as its production of cytokines (ie, IFN-γ, TNF-α, and IL-2) were dramatically reduced when co-cultured with AsPC-1/TGFβ in comparison to AsPC-1 control cell; however, this effect was attenuated for CAR-1+T armored with TGFβRIIDN (figure 3J,K). Thus, the TGFβRIIDN armor lessened the inhibitory effect of TGFβ1 on the cytotoxicity of CAR T cells in vitro.

ROR1 CAR-1 with the dominant negative TGFβ receptor II armor overcame the inhibitory effect of TGFβ on T cells in the AsPC-1 pancreatic cancer xenograft model overexpressing TGFβ1

To better understand the effect of the TGFβRIIDN armor on the antitumor activity of ROR1-CAR T cells in the TGFβ-rich tumor environment in vivo, a common feature of many solid tumors, NSG mice were implanted subcutaneously with AsPC-1/TGFβ1 pancreatic tumor cells modified to stably overexpress and secrete TGFβ1. The AsPC-1/TGFβ1 tumor-bearing mice were treated with ROR1-CAR-1+ T cells with or without armor, and evaluated for 49 days (figure 4A). Mice treated with the armored CAR T cells rapidly and efficiently rejected tumors, whereas the non-armored CAR T cells administration resulted in stable disease (figure 4B,C). Body weights remained normal for both armored and non-armored CAR-treated mice (online supplemental figure 3A). Peripheral blood armored CAR T cells expanded faster in response to tumor between study days 2 and 12 (figure 4D), due to greater expansion of both CD8+ and CD4+CAR-T+ populations, then contracted similarly to CAR-1 alone (figure 4E). Serum pro-inflammatory cytokines IFN-γ and granulocyte-macrophage colony-stimulating factor (GM-CSF) were only moderately elevated after either CAR on day 5, and subsided to baseline levels by day 15, whereas TNF-α, IL-2 and IL-6 remained low (online supplemental figure 3B). No cytokine was significantly induced in the armored CAR-1 treatment group, as compared with CAR-1 alone (online supplemental figure 3B). Notably, on day 5 post CAR T, there was a significant reduction of the circulating TGFβ1 in both armored and non-armored CAR T-cell groups, as compared with non-treated or UTD-treated mice; which was attributed to the cytotoxic activity of CAR-Ts against TGFβ-producing AsPC-1/TGFβ tumors (figure 4F). However, in the non-armored CAR group, peripheral blood TGFβ1 resurged by day 15 to match the level of TGFβ1 in the non-treated and UTD-treated control mice. By contrast, in mice treated with the armored CAR T cells, serum TGFβ1 levels remained low, suggesting that the expanded armored CAR-1+T-cell population may function as a TGFβ cytokine sink, in addition to the direct tumor killing (figure 4F).

Figure 4

TGFβRIIDN attenuated the inhibitory effect of TGFβ1 in AsPC-1 xenograft model of pancreatic cancer overexpressing TGFβ1. NSG mice (five mice/group) were implanted subcutaneously with AsPC-1/TGFβ1 cells (1e6 cells/mouse) at day −15, followed by staging and CAR T-cell infusion (i.v., 5e6 CAR+T cells/mouse) at day 0 (A) and tumor volumes were monitored (B–C). Statistical significance determined by Mann-Whitney test (B) and Wilcoxon matched—pairs signed-rank test (C) T cells isolated from peripheral blood at the indicated time points were quantified by flow cytometry for total cell number (D) or CAR+ components in both CD8+ and CD4+ subpopulations (E) results were analyzed by Mann-Whitney test. Blood from mice sampled at day 5 and day 15 post T-cell infusion was quantified for TGFβ1, groups were compared by Student’s t-test (F). Tumor tissues collected at day 7 post T-cell infusion and at study termination were subjected to immunohistochemistry staining for TGFβ1, CD3, and IgG control (G); the number of ROR1+cells was quantified by MACSima imaging cycling staining high content immunofluorescent microscopy (H–J); tissues were probed for CD8a (T cells), ROR1 (CAR-targeted tumor antigen), TGFβ and alpha-smooth muscle actin (myofibroblasts, stroma) with DAPI counterstaining for nuclei (H); T cell-rich, T cell-intermediate and T cell-low regions of tumors were evaluated; an example of higher magnification of these regions is shown in (I); and the quantification of ROR1+cells from each of these regions (n=3 each) is shown in violin plots representing cell number distribution in each group (J); statistical analysis was performed by two-way analysis of variance with Sidak’s post hoc test. All results are presented as mean±SEM; statistical significance is denoted as *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. BLI, bioluminescene; CAR, chimeric antigen receptor; DAPI, 4',6-diamidino-2-phenylindole; DN, dominant negative; i.v., intravenous; ROR1, receptor tyrosine kinase-like orphan receptor 1; s.c., subcutaneous; SMA, smooth muscle actin; TGF, transforming growth factor; UTD, un-transduced T cells.

Immunohistochemical staining of tumor tissues harvested at day 7 post T-cell infusion revealed a high level of TGFβ being produced by the AsPC-1/TGFβ tumor cells, and CAR-1 dependent infiltration of CD3+T cells into the tumors (figure 4G), which was not observed in the UTD control group. At study termination, intense TGFβ staining was still observed within the tumors in tumor alone, UTD and the non-armored CAR-1+ T-cell group, in agreement with the observed incomplete tumor regression. By contrast, in the armored CAR-1 group, by study termination most tumors resolved completely, and we only succeeded in harvesting tissue from the tumor site in one mouse, which was TGFβ-negative (figure 4G).

To further investigate CAR T-cell effect in the mouse tumors, we performed multiplex IF MICS analysis of CAR-1 and the armored CAR-1-treated tumor tissues on day 7 post-treatment. FFPE tumor tissue sections were stained for TGFβ, CD8+T cells (cytotoxic CAR-T), ROR1 (targeted tumor antigen), αSMA (alpha-smooth muscle actin, a marker of myofibroblasts), and 4',6-diamidino-2-phenylindole (DAPI, live cell nuclei) (figure 4H). Tumor morphology consistent with human PDAC was revealed, with ROR1+ tumor lesion surrounded by a ring of infiltrating CD8+CAR T cells. ROR1+ tumor cells were highly prevalent in the live (DAPI-rich) peripheral tumor regions, and were supported by a network of mouse-derived alpha-smooth muscle-positive myofibroblasts, a component of desmoplastic tumor stroma,24 whereas tumor centers tended to be necrotic and lacked the nuclear DAPI stain. Intense TGFβ1 expression was observed in the tumor and the surrounding stroma regions, constituting a TGFβ1-rich TME (figure 4H). In order to investigate the impact of CAR-T cell infiltration on the tumor, triplicate tumor regions with high-T-cell, intermediate-T-cell and low-T-cell infiltration intensity were defined and marked for CAR-1 and armored CAR-1 tissues (figure 4I, one representative region at each intensity level is shown). We then quantified ROR-1+cells for each region type using MICS imaging (figure 4J)

In the CAR-1 treated tumor, a statistically significant increase in ROR1+ cell number was observed between T-cell poor-regions, intermediate-regions and rich-regions, respectively, indicating that T cells engage with tumor cells in accordance with ROR-1 expression levels, but no CAR-mediated reduction in ROR1+ tumor cells was detected at the time of tumor harvest, day 7. By contrast, in the armored CAR-treatment group, ROR1+ T-cell number increased between T-cell low-regions to T-cell intermediate-regions, demonstrating antigen-driven influx of CAR T cells, but was significantly reduced between the T-cell intermediate-regions and T-cell rich-regions, indicating an ongoing, CAR-mediated killing of ROR1+tumor cells in the T-cell-rich-regions. Furthermore, when comparing the T-cell rich-regions treated with the armored CAR-1+ T cells and CAR-1 alone, the total number of the remaining live ROR1+ T cells in the armored CAR-treated region was significantly diminished as compared with CAR-1 alone (figure 4J). Therefore, we have observed on an organism level as well as on a cellular level that the armored CAR-1+ T cells were more effective in clearing AsPC-1 tumors in mice than the non-armored CAR-1+ T cells, under comparable TGFβ1-high conditions.

Armored CAR-1 T cells accelerate tumor clearance in TGFβ-inducible model of pancreatic cancer

To evaluate the function of DN TGFβ armor under conditions where TGFβ is not overexpressed by tumor cells, we conducted an in vivo xenograft study in mice bearing the parental AsPC-1 PDAC tumors without overexpression of TGFβ. AsPC1 cells naturally produce modest levels of TGFβ1 in culture, as shown in figure 3H. Tumor-bearing NSG mice were treated with CAR-1, armored CAR-1, UTD negative control, or left untreated (tumor alone), (figure 5A). The armored CAR-1 DN T cells rejected tumors by study day 24, whereas CAR-1 alone-treated mice only became tumor-free after day 33 (figure 5B); and the rate of tumor regression was significantly more advanced in the group treated with the armored CAR-1 T cells compared with the non-armored CAR-1, (figure 5C); demonstrating greater efficacy of the armored CAR-1+ T cells. Unexpectedly, immunohistochemical images of tumor tissue harvested on day 7 post-treatment revealed TGFβ1 presence within the tumor lateral and medial regions only in CAR-treated groups, but not in the non-transduced CAR control UTD, or tumor alone groups (figure 5D). Moreover, TGFβ1 staining occurred selectively in regions infiltrated by T cells (CD3+), indicating that TGFβ1 production in this model was CAR-dependent. Using the multiplex IF MICS staining, we again analyzed T-cell rich-regions, T-cell intermediate-regions, and T-cell poor-regions, classified based on the fraction of infiltrating CD8+T cells (figure 5E–G representative images for CAR-1+DN are shown). Live tumor cells (ROR1+) were similarly present in all three regions, as did the desmoplastic stroma-forming myofibroblasts (αS