TEM8/ANTXR1-Specific CAR T Cells as a Targeted Therapy for Triple-Negative Breast Cancer

Triple-negative breast cancers (TNBC) are estrogen and progesterone receptor-negative and lack amplification of the human epidermal growth factor receptor 2 (HER2) gene. Accounting for 15% to 20% of cases, TNBC is associated with a particularly aggressive phenotype and a higher incidence of recurrence (1, 2). To date, there are no approved targeted therapies for TNBC, new therapies are thus desperately needed.

In 2000, St. Croix identified tumor endothelial marker 8 (TEM8) as a marker of tumor-associated versus normal endothelium in colorectal cancer (3). Also known as anthrax toxin receptor 1 (ANTXR1), TEM8 is an integrin-like cell surface protein and has been shown to play a role in endothelial cell migration and invasion (3–5). TEM8 is conserved in mice and humans, sharing an overall 98% amino acid sequence homology (6). Blocking and knocking out TEM8 resulted in a decline in tumor growth in several preclinical cancer models (5–10). Importantly, deletion of the TEM8 gene in mice did not result in defects in angiogenesis associated with wound healing or normal development (9, 10).

Mounting evidence implicates a role for TEM8 in TNBC pathogenesis. TEM8 is elevated in invasive breast cancer tissue relative to normal tissue and correlates with a higher incidence of recurrence in basal breast cancer (11, 12). Overexpressing TEM8 in a preclinical breast cancer model increased tumor growth and metastasis (13). In 2013, Chen and colleagues identified TEM8 as a marker of breast cancer stem-like cells (BCSC; ref. 14). Taken together, these data support TEM8 as a potential target not only of TNBC tumor-associated vessels, but also of TNBC tumor cells and potentially BCSCs.

In this study, we sought to determine whether T cells rendered specific to TEM8 using chimeric antigen receptors (CAR) could serve as a novel targeted therapy for TNBC. Specifically, we explored the effector functions of TEM8 CAR redirected T cells against TNBC cells as well as against the tumor-associated vasculature.

Frozen tissue microarrays (BRF404b) containing primary breast cancer cases and normal adjacent breast tissue controls were obtained from US Biomax, Inc.

Frozen tissue microarrays were stained as previously described (9). L2 in antibody form (mouse–human reverse chimeric antibody with human variable domains and mouse IgG2a constant domains) was not suitable for immunofluorescence with high background observed in the negative control (secondary only), specifically when staining mouse tissues.

Total cell lysates were prepared using Pierce IP Lysis Buffer supplemented with protease and phosphatase inhibitors (Thermo Scientific) and quantified with the BCA Protein Assay Kit (Thermo Scientific). The protein samples were separated via SDS–PAGE and then transferred to polyvinylidene difluoride membranes. Blots were incubated in Odyssey Blocking Buffer (LI-COR) followed by overnight incubation with mouse anti-SB5 antibody diluted in blocking buffer. Blots were then exposed to the IRDye secondary antibodies (LI-COR) for 60 minutes at room temperature and washed again. Blots were detected using (LI-COR) Odyssey Infrared Imaging System and analyzed using the ImageJ software.

The monoclonal antibody L2 has previously been described (5, 7, 9). The L2 scFv was generated using the sequence for a single light chain of L2 antibody connected by a flexible glycine serine linker, followed by a single heavy chain of the L2 antibody. L2-based constructs were assembled, synthesized, and sequence-verified as previously described (15–17). A retroviral vector encoding the fusion protein eGFP-Firefly Luciferase (eGFP.FFluc) previously described (18) was used to generate firefly luciferase expressing T cells, MDA-MB-468 and LMD231 cells (in vivo experiments).

To produce retroviral supernatant, human embryonic kidney (HEK) 293T cells were cotransfected with either the L2 2G or L2 3G encoding SFG retroviral plasmids, Peg-Pam-e plasmid encoding MoMLV gag-pol and the plasmid containing the sequence for the RD114 envelope (19). T cells were transduced with retroviral vectors containing their respective CARs as described (20). In order to generate GFP-ffluc CAR T cells, equal parts of eGFP-firefly reporter gene and CAR containing retroviral supernatants were used to cotransduce primary human T cells. T cells were then normalized for GFP expression and CAR density. Evidence of light output was confirmed prior to use using a luminometer following the addition of d-luciferin substrate.

Blood samples were obtained from healthy donors on a protocol approved by the institutional review board of Baylor College of Medicine. Written informed consent was obtained from all donors. All parental cell lines were used less than 6 months after receipt or resuscitation. Breast cancer cell lines ((Hs578T, MDA-MB-231, MDA-MB-436, MDA-MB-468, and SK-BR-3) were purchased from the ATCC. The lung metastasis-derived LMD231 cell line was a gracious gift from Dr. Harikrishna Nakshatri (Indiana University, Bloomington, IN; ref. 21). Breast cancer lines were grown in DMEM (Invitrogen) with 10% fetal calf serum (Hyclone) and 2 mmol/L GlutaMAX (Invitrogen). Endothelial cell lines: HMMEC (ScienCELL) and HC6020 (CELL biologics) were cultured in Endothelial Cell Medium EGM Complete Medium (CC-3024; Lonza), 10% FBS, Endothelial Cell Growth Supplement (ECGS), 90 Mg/mL, Na heparin, 30 Mg/mL Endothelial Cell Growth Supplement, 10 ng/mL epidermal growth factor (EGF), vascular endothelial growth factor (VEGF; 0.5ng/mL), 0.5% bovine serum albumin (BSA) and ascorbic acid (1 μg/mL). Raji and T cells were cultured in RPMI-1640, 10% FCS and 2 mmol/L GlutaMAX (Invitrogen).

Samples were run on either the Gallios Flow Cytometer: 3 lasers, 10-color configuration (Beckman Coulter) or the BD Accuri C6 Flow Cytometer (Becton Dickinson). Data analysis was done on ≥10,000 events using the Kaluza (Beckman Coulter) and FlowJo (TreeStar) data analysis software, respectively. Cells were washed once with PBS containing 1% FBS (FACS buffer) prior to addition of antibodies. After 30 minutes to 1 hour of incubation at 4°C in the dark, the cells were washed for analysis.

Cytotoxicity assays were performed as previously described (22). Nontransduced T cells were used to normalize the percentage of CAR-positive cells. The mean percentage of specific lysis of triplicate wells was calculated according to the following formula: (test release – spontaneous release)/(maximal release – spontaneous release) × 100.

Effector T cells (CAR expressing T cells or nontransduced T cells) from healthy donors were cocultured with TEM8-positive and TEM8-negative cell lines at a 1:1 effector to target ratio in a 96-well plate. After 24 to 48 hours of incubation, culture supernatants were harvested and ELISA determined the presence of IFNγ and IL2 as per the manufacturer's instructions (R&D Systems).

All animal experiments were conducted on a protocol approved by the Baylor College of Medicine Institutional Animal Care and Use Committee (IACUC). Animals were regularly examined for any signs of stress and euthanized according to preset criteria.

Subcutaneous model: Six- to, 10-week female athymic nude mice were purchased from taconic (NCRNU-F Homozygous CrTac: NCr-Foxn1^nu; Taconic) for animal studies. Mice received a subcutaneous injection of 2 × 10⁵ MDA-MB-468 eGFP.FFluc labeled cells in 100 μL phenol-red free Matrigel on day 0 (Corning). Nine days after tumor cell injection, animals received a single dose of 25 × 10⁶ T cells for antitumor studies. Animals were imaged using the IVIS Lumina II as previously described (17). For bioluminescence experiments, intensity signals were log-transformed and summarized using mean at baseline and multiple subsequent time points for each mouse group.

Human breast cancer patient-derived xenograft models: Fresh Patient-derived xenograft (PDX) tumor fragments were transplanted in the epithelium-free, right inguinal fat pad of 4-week-old SCID/Bg mice, and allowed to engraft as previously described (23). The slower growing TNBC PDX BCM-2665 was allowed to engraft for 21 days (∼140 mm³) prior to treatment. The more aggressive claudin-low TNBC PDX WHIM12 was allowed to engraft for 13 days (∼165 mm³). To account for the difference in two models’ growth rate, BCM-2665 mice received 3 doses of 5 × 10⁶ L2 3G CAR T cells after transplant via intratumoral injection, while WHIM12 mice received 5 doses. Irrelevant CAR (CD19) T cells, nontransduced T cells, and untreated tumors served as controls. Body weight and tumor volume measurements were taken twice a week for the duration of the studies. Tumors were harvested when they reached the maximum allowable size as per IACUC protocols.

TNBC lung metastasis model: 5 × 10⁵ LMD231.eGFPffluc cells were systemically administered to 10-week-old athymic nude mice. Starting on day 8, mice received 3 consecutive doses of 3 million L2 3G CAR (90% transduction) or nontransduced T cells. Evidence of disease was measured by bioluminescence imaging.

For cytokine release and cytotoxicity Student t test was used to compare two groups at a time. For bioluminescence data, signal intensity changes from baseline at each time point were calculated and compared using paired t tests or Wilcoxon signed-rank test. Log-rank test was used to compare the survival distribution between treatment groups.

To validate TEM8 as a target, we stained 6 primary TNBC samples for TEM8 and compared the staining pattern to normal adjacent breast tissue. TEM8 (green) was detected in all TNBC cases, while little to no staining was observed in control tissue (Fig. 1A). To determine the location of TEM8 expression we stained primary TNBC vessels with anti-CD31 (platelet endothelial cell adhesion molecule 1 [PECAM-1]) and plasmalemma vesicle-associated protein (PV1/MECA-32) antibodies as previously described (9, 24). We observed two patterns of TEM8 expression in primary TNBC, a perivascular stromal signature (TNBC-1, 5 and 6) as well as diffuse staining throughout the tumor tissue (TNBC-2, 3, and 4; Fig. 1A). Normal adjacent breast tissues did not express TEM8 (n = 3). Based on these results, we decided to test TNBC cell lines for TEM8 expression (5, 25). As determined by Western blot, TNBC cell lines Hs578T, MDA-MB-231, MDA-MB-436, and MDA-MB-468 express TEM8 (564 aa, 62-kDa band) (Fig. 1B). Normalized to the housekeeping gene GAPDH, Hs578T had the highest level of expression, followed by MDA-MB-231, MDA-MB-436, and MDA-MB-468 (Fig. 1C). The HER2-amplified luminal breast cancer cell line SKBr3 did not express appreciable levels of TEM8 (Fig. 1B and C).

We used flow cytometry to determine the surface expression of TEM8, by TNBC and tumor endothelial cell lines (Fig. 1D and E). Of the TNBC cell lines MDA-MB-468 expressed the highest level of TEM8 on the surface, followed by MDA-MB-231, MDA-MB-436, and Hs578T. We compared tumor-derived endothelial cells to normal endothelial cells. Breast tumor-associated endothelial cells (HC 6020) and murine tumor-associated endothelial cells (2H11) expressed TEM8 whereas normal human microvascular mammary endothelial cells (HMMEC) and human umbilical vein endothelial cells did not (26). The murine brain tumor endothelial cell line bEND.3 expressed TEM8, albeit to a much lesser extent than the other tumor endothelial cell lines (Fig. 1E). Assessment of TEM8 expression in a panel of normal tissues was negative for TEM8 (Supplementary Fig. S1A).

CARs consist of the single-chain variable fragment (scFv) of a monoclonal antibody acting as a target-recognition extracellular domain fused with signaling domains traditionally derived from the T-cell receptor complex (27). T cells can be genetically modified to express CAR molecules on their surface, so that upon binding of the target antigen killing is induced (19, 27). Use of a CD28 CD3-ζ signaling domain has been shown to mediate both the killing and proliferative capacity of CAR T cells (28); however, addition of a TNFα receptor-like costimulatory moiety, specifically 41BB, has demonstrated increased in vivo persistence (29). We created second-generation (CD28.CD3-ζ) and third-generation (CD28.41BB.CD3-ζ) TEM8-specific CAR molecules derived from the scFv of the TEM8 antibody L2, which we hereafter refer to as L2 2G and L2 3G, respectively (Fig. 2A; ref. 9). Primary human T cells from three healthy donors were transduced with either L2 2G or L2 3G CAR transgenes with similar transduction rates (Fig. 2B).

In cytotoxicity assays, L2 2G and 3G CAR T cells effectively killed TNBC lines Hs578T, MDA-MB-231, MDA-MB-436, and MDA-MB-468, human breast tumor endothelial line HC 6020 and murine tumor endothelial cell lines 2H11 and bEND.3 (Fig. 2C). TEM8-negative Raji cells were not killed.

Next, we assessed the CAR T cells’ cytokine release against TNBC and tumor endothelial cell lines (Fig. 2D). L2 3G CAR T cells released significantly higher amounts of IFNγ and IL2 in response to all four TNBC cell lines and 2H11 when compared with L2 2G CAR T cells. In coculture with bEND.3 IFNγ only but not IL2 was higher with L2 3G, and only L2 2G released IL2. The modest expression of TEM8 by bEND.3 cells coupled with the lower threshold of activation required for L2 2G cells could explain why L2 2G, but not L2 3G, cells were activated by bEND.3. NT T cells did not release detectable amounts of cytokines.

BCSC are associated with an aggressive phenotype, therapy resistance, and poor prognosis in TNBC (30, 31). As such, we tested TNBC cell lines for expression of BCSC markers CD44 and CD24 (Supplementary Fig. S2A). In particular, independent of CD24 expression, CD44 has been identified as a stem cell marker associated with a poor prognosis in TNBC (32). TNBC cell line MDA-MB-436 cells contained sufficient populations of CD44⁺ and CD44⁻ cells to analyze TEM8 surface expression. Relative to the CD44⁻ fraction CD44⁺ cells had significantly higher levels of TEM8 as determined by flow cytometry (Fig. 2E). To test whether L2 CAR T cells could potentially target BCSC and inhibit mammosphere formation, a hallmark of BCSCs, Hs578T cells expressing a GFP reporter gene were grown in nonadherent nondifferentiating conditions and cultured with either L2 CAR T cells or nontransduced control cells. Hs578T cells are enriched for stem-like cells and were >99 % CD44⁺ (Fig. 2F). L2 CAR T cells significantly reduced the number of mammospheres compared with NT T cells (Fig. 2G and H; P < 0.05), and when compared with irrelevant CAR T cells (CD19), we saw the same results (Supplementary Fig. S2B).

To compare L2 2G and L2 3G CAR T cells functionally both in vitro and in vivo, we normalized for GFP fluorescence and CAR expression (Supplementary Figs. S3A and S3B). In long-term killing assays at a 1:1 and 1:5 effector: target ratios, L2 2G CAR T cells killed TNBC cell lines MDA-MB-468 and HS578T more quickly than L2 3G CAR T cells (Fig. 3A). When testing proliferation against TEM8⁺ cell lines, L2 2G and L2 3G CAR T cells proliferated similarly as determined by flow cytometry using CellTrace Violet staining. We did detect a discernable difference however when using both low (0.2 μg/mL) and high (1.0 μg/mL) concentrations of recombinant TEM8 protein as a source of stimulation (Fig. 3B). L2 2G CAR T cells were able to proliferate more against lower levels of antigen. Nontransduced T cells served as a control, only proliferating in the presence of IL2. To distinguish how these differences affect the T-cell subset phenotype, specifically naïve, central memory and effector memory T cell subsets, we assayed L2 2G and L2 3G CAR T cells for expression of CCR7 and CD45RA (Supplementary Fig. S3C). Adding a 41BB-signaling moiety (L2 3G CAR) resulted in higher preservation of the central memory phenotype (CCR7⁺/CD45RA⁻) upon stimulation with Hs578T and MDA-MB-468 cells (Fig. 3C). This difference was nearing significance with a P value of 0.06 (n = 3 donors).

Lastly, we compared L2 2G and L2 3G CAR T cells in vivo. To test the ability of these cells to persist upon chronic antigen exposure, L2 2G or L2 3G CAR T cells were injected into MDA-MB-468 tumor bearing mice. Labeled with firefly luciferase, T cells were monitored over time. L2 3G CAR T cells, relative to L2 2G CAR T cells, persisted longer in vivo (n = 5). Nontransduced T cells served as a control (Fig. 3D).

To investigate the impact of TEM8 CAR T cells on the growth of established and vascularized TNBC xenografts, eGFP.FFLuc-expressing TEM8⁺ MDA-MB-468 cells were injected subcutaneously into the flanks of 6- to 8-week-old female athymic nude mice. On day 9, mice received a single dose of L2 2G or 3G CAR T cells, NT T cells, or T cells expressing an irrelevant CAR (n = 10 per group) intratumorally. Both L2 2G and L2 3G induced regression of MDA-MB-468 xenografts compared with NT T cells and mice that received no treatment (P = 0.02, 36 days after treatment; P = 0.01, 50 days after treatment; Fig. 3E). L2 3G cells were significantly better at controlling the tumor at day 59 (P = 0.04). Furthermore, L2 3G but not L2 2G significantly improved survival relative to NT T cells (P = 0.003; log-rank/Mantel–Cox test; Fig. 3F). Tumors were explanted to analyze the effect of L2 CAR T cells on vessel density. We examined the effects of the CAR T cells on the vascular component of the tumor. H&E staining of tumor explants revealed dense tumor tissue in control groups, with nests of tumor cells and collapse of the tumor infrastructure in L2-treated tumors (Supplementary Fig. S4A). Immunofluorescence staining for CD31/MECA-32 revealed that L2 2G and L2 3G CAR T cell–treated groups had significantly decreased vascularization (P = 0.0001 and 0.001, respectively), whereas tumors that received no treatment, NT T cells or T cells expressing an irrelevant CAR (HER2 2G) had an intact vascular supply (Fig. 3G; Supplementary Fig. S4B). TEM8 staining (green) revealed incomplete tumor targeting (Supplementary Fig. S4B). Further, while tumors recurred in mice treated with L2 CAR T cells as determined by bioluminescence imaging of the tumor cells, the supporting tumor microenvironment was affected as indicated by a smaller tumor volume as determined by external caliper measurements (Supplementary Fig. S4C).

PDX models have been shown to mimic much of the microenvironment of the tumor of origin (23). To test the ability of TEM8 CAR T cells to target patient-derived tumors, we evaluated the effect of treatment on two TNBC PDX models: BCM-2665 (generated by Dr. Michael Lewis at Baylor College of Medicine from a patient with locally advanced TNBC) and WHIM12 (generated by Dr. Matthew Ellis then at Washington University from an especially aggressive claudin-low breast cancer; Horizon; Cambridge, MA). Similar to the two patterns, we observed in primary TNBC samples the BCM-2665 PDX induced host TEM8 expression and displayed a predominately perivascular expression, while TEM8 expression by WHIM12 PDX was expressed diffusely throughout the tissue, expressed by both the tumor cells and perivascularly (Fig. 4A).

As L2 3G CAR T cells demonstrated superior efficacy over L2 2G CAR T cells in vivo, we proceeded with L2 3G CAR T cells for testing in orthotopic patient-derived xenograft models. Nontransduced T cells and sham injected (PBS-treated) served as a control. Body weight and tumor volume measurements were taken twice a week for the duration of the study. L2 3G CAR T cells induced regression in 6 of 7 BCM-2665 PDX tumor–bearing mice and significantly extended the median survival from 21 days to 60 days. One of 7 mice displayed no evidence of disease on day 60 (Fig. 4B and C). In the WHIM12 PDX model, L2 CAR T cells induced tumor stabilization relative to control groups and significantly extended median survival from 11 to 26.5 days, with 2 of 10 mice still alive at day 60 (Figs. 4D and E). Taking advantage of the stromal TEM8 signature of the BCM-2665 PDX, we tested whether TEM8 CAR T cells could specifically target the tumor microenvironment and decrease vascularization. As indicated by CD31 and MECA-32 staining (red), TEM8 CAR T cell–treated tumors displayed evidence of vascular targeting with necrotic pockets devoid of vessels (Fig. 4F).

The TNBC lung metastasis-derived cell line LMD231 was used to test our CAR T cells in a systemic model. Similar to the parental line MDA-MB-231, LMD231 cells were predominately CD44⁺/CD24⁻ and TEM8 positive (Fig. 5A and B). To determine whether this level of expression was detectable by our CAR T cells, LMD231 cells were incubated at a 1:1 ratio with L2 CAR T cells overnight. L2 CAR T cells recognized and released high levels (>2,000 pg/mL) of IFNγ and IL2 as determined by ELISA (Fig. 5C). LMD231.eGFPfffluc cells were injected systemically via tail vein into athymic nude mice. Tumors were allowed to engraft and vascularize for eight days, after which they received 3 consecutive doses of 3 × 10⁶ L2 3G CAR T cells or NT T cells. Tumor burden was measured using bioluminescence imaging. L2 3G CAR T cells induced significant tumor regression in mice such that no mice developed detectable metastasis to the bones and brain of mice as observed in the control group (Fig. 5D and E). Lastly, this provided a significant survival advantage (n = 4, P < 0.05; Fig. 5F).

Because murine and human TEM8 have a 98% amino acid sequence homology and L2 CAR T cells lyse murine tumor endothelial lines 2H11 and to a lesser extent bEND.3 cells in vitro, it is important that the mice be monitored for any signs of distress. We, nor the veterinary staff, detected any clinical signs of distress from the use of systemic administration of CAR T cells to these mice. Further, to test for potential “on-target/off tumor toxicity,” a dose of 10 × 10⁶ L2 3G CAR T cells was administered to both female athymic nude mice and male SCID mice and tested for evidence of vascular injury using cytokine multiplex analysis and H&E staining of harvested major organs. Relative to L2 2G CAR T cells, L2 3G CAR T cells required higher levels of antigen to proliferate (Fig. 3B). However, we chose L2 3G CAR cells to test in toxicity studies for two reasons: these cells release on average higher levels of IFNγ, a key player in cytokine release syndrome (33), but also for their increased antitumor activity in vivo making them the most viable for clinical translation. Analysis of vascular injury markers and of harvested organs showed no evidence of vascular injury when compared with control mice (Supplementary Figs. S5A, S5B, and S6).

Here, we report on the first TEM8-specific CAR molecule that is able to cotarget both TNBC breast cancer stem cells and tumor-associated vasculature. TEM8 CAR T cells killed TNBC tumor cells, inhibited progenitor-mediated mammosphere formation, obliterated tumor neovasculature and induced regression of established PDXs in local and metastatic murine models of TNBC.

There is an increasing realization that solid tumors create or exploit complex multicellular niches for their growth and that understanding and targeting such niches is critical for improving therapeutic success. Expressed in targetable levels on TNBC tumor cells, BCSCs, as well as the tumor microenvironment, TEM8 represent a unique antigen for targeted cancer immunotherapy. Indeed, despite their single specificity, TEM8 CAR T cells function as a broad-spectrum multispecific biologic that simultaneously targets various elements of the TNBC tumor and its microenvironment.

Importantly, TEM8 is upregulated in the stem cell compartment of TNBC (14). This population of cells has been linked not only to tumor initiation and progression but also to metastasis and resistance to chemotherapy and radiotherapy, which explains at least in part the limited efficacy of standard-of-care measures for TNBC (34–36). After conventional therapy, breast cancers have been reported to selectively display mesenchymal and tumor-initiating cell features, a population we show is preferentially TEM8 positive (37). Therefore, we anticipate that TEM8-expressing TNBC cells represent a viable target for immunotherapeutic intervention that will abrogate the compartment responsible for failure of conventional therapies.

L2 CAR T cells readily released immunostimulatory cytokines IFNγ and IL2 upon encountering TEM8 and killed TEM8-positive TNBC and tumor endothelial cell lines, showing cross-reactivity against murine and human TEM8. Favorably, the L2 scFv antibody fragment was derived from a fully human Fab antibody, a clinically appealing characteristic avoiding the induction of human anti-mouse antibodies and clearance by the immune system in future clinical application. Collectively, these features provided a sound justification to focus further evaluation of TEM8-targeted T cells in our preclinical studies on cells expressing the L2 scFv-based CAR.

While demonstrating comparable cytolytic capabilities as L2 2G T cells in vitro, L2 3G CAR T cells were more efficacious than the former against established TNBC xenografts in vivo. Notably, L2 3G CAR T cells displayed increased cytokine release and preservation of a more favorable central memory phenotype when compared with L2 2G T cells in in vitro experiments, which could explain their enhanced in vivo activity. L2 3G demonstrated a significantly higher degree of in vivo expansion and persistence than L2 2G cells that correlates with increased percentage of central memory cells.

Treatment with a single dose of L2 CAR T cells did not completely eradicate the established TNBC xenografts. In both, L2 2G CAR T cell– and L2 3G CAR T cell–treated groups, tumors eventually progressed. Nevertheless, nearly 2 months after treatment, mice treated with L2 3G CAR T cells still had significantly smaller tumors than mice treated with NT T cells and staining for CD31 revealed that L2-treated tumors had significantly reduced vascularization.

PDX models are an invaluable resource for assessing the potential efficacy of a therapy in a preclinical setting. Our L2 CAR T cells effectively induced regression and stasis in TNBC PDX models BCM-2665 and claudin-low WHIM12, respectively. Increased expression of ANTXR1, the TEM8 gene, has been correlated with a poorer survival outcome in TNBC and induction of TEM8 expression in murine 4T1 breast cancer cells increased growth and metastasis (11, 13, 14). Our CAR strategy targets TEM8 on the surface; however, TEM8 can also be expressed in other isoforms internally. The WHIM12 PDX in particular expresses high levels of intracellular TEM8, which could explain at least in part the reduced efficacy of CAR T cells against the WHIM12 PDX versus BCM-2665.

In conclusion, our study presents TEM8 as a promising target antigen for CAR-based cancer immunotherapy of TNBC. We showed that TEM8-specific (L2) CAR T cells are capable of killing TNBC tumor parenchyma, tumor endothelial cells, as well as BCSCs, inducing regression of established TNBC xenografts. Thus, we propose TEM8 CAR T cells as an attractive targeted therapy for TNBC.

T.T. Byrd is a listed inventor on a patent pertaining to vascular-targeted T-cell therapy, publication number: WO 2014164544 A. M.T. Lewis is a Limited Partner at StemMed Ltd. and is manager at StemMed Holdings. N. Ahmed is a consultant/advisory board member for Adaptimmune. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of General Medical Sciences, National Heart, Lung and Blood Institute or National Institutes of Health.

Conception and design: T.T. Byrd, J. Hicks, M.L. Baker, M.T. Lewis, M. Hegde, B.S. Fletcher, B. St. Croix, N. Ahmed

Development of methodology: T.T. Byrd, A. Pignata, H. Samaha, S. Seaman, L. Dobrolecki, V.S. Salsman, K. Bielamowicz, J. Hicks, M.L. Baker, P.H. Sorensen, M.J. Ellis, M.T. Lewis, N. Ahmed

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.T. Byrd, K. Fousek, A. Pignata, C. Szot, L. Dobrolecki, V.S. Salsman, K. Bielamowicz, D. Landi, N. Rainusso, J. Hicks, M.L. Baker, W.S. Wels, J. Koch, M.J. Ellis, M.T. Lewis, N. Ahmed

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.T. Byrd, K. Fousek, H. Samaha, H.Z. Oo, D. Landi, J. Hicks, S. Powell, M.L. Baker, W.S. Wels, J. Koch, B. Deneen, M.T. Lewis, M. Hegde, B. St. Croix, N. Ahmed

Writing, review, and/or revision of the manuscript: T.T. Byrd, K. Fousek, A. Pignata, C. Szot, H. Samaha, L. Dobrolecki, H.Z. Oo, K. Bielamowicz, D. Landi, J. Hicks, S. Powell, W.S. Wels, J. Koch, M.J. Ellis, M.T. Lewis, M. Hegde, N. Ahmed

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Samaha, H.Z. Oo, J. Hicks, B.S. Fletcher, B. St. Croix, N. Ahmed

Study supervision: M. Hegde, N. Ahmed

Other (review of final manuscript draft): N. Rainusso

This research was supported in part by the Stand Up To Cancer - St. Baldrick's Pediatric Dream Team Translational Research Grant (SU2C-AACR-DT1113). Stand Up To Cancer is a program of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the Scientific Partner of SU2C. Authors directly supported by this grant include N. Ahmed and M. Hegde. This research was also funded by the following T32 training grants: award numbers T32GM088129 and 5T32HL092332 from the National Institute of General Medical Sciences and National Heart, Lung and Blood Institute, respectively. The authors directly supported by both T32 grants include T. Byrd and K. Fousek. The Cytometry and Cell Sorting Core at Baylor College of Medicine also supported this project with funding from the NIH (AI036211 and CA125123).

We would like to thank Drs. Malcolm Brenner, Motthafar Rimawi and Kent Osborne for their insight and professional advice and Mrs. Catherine Gillespie and Dr. Sujith Joseph for editing this manuscript. We would also like to thank summer students Daphine Mugayo, Yasmina Fayed, Jibwa Jakana-Lule and Jordan Clinton for their technical assistance.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.