Background Diffuse intrinsic pontine glioma (DIPG) and glioblastoma (GBM) are two highly aggressive and generally incurable gliomas with little therapeutic advancements made in the past several decades. Despite immense initial success of chimeric antigen receptor (CAR) T cells for the treatment of leukemia and lymphoma, significant headway into the application of CAR-T cells against solid tumors, including gliomas, is still forthcoming. The integrin complex alphav beta3 (αvβ3) is present on multiple and diverse solid tumor types and tumor vasculature with limited expression throughout most normal tissues, qualifying it as an appealing target for CAR-T cell-mediated immunotherapy.
Methods Patient-derived DIPG and GBM cell lines were evaluated by flow cytometry for surface expression of αvβ3. Second-generation CAR-T cells expressing an anti-αvβ3 single-chain variable fragment were generated by retroviral transduction containing either a CD28 or 4-1BB costimulatory domain and CD3zeta. CAR-T cells were evaluated by flow cytometry for CAR expression, memory phenotype distribution, and inhibitory receptor profile. DIPG and GBM cell lines were orthotopically implanted into NSG mice via stereotactic injection and monitored with bioluminescent imaging to evaluate αvβ3 CAR-T cell-mediated antitumor responses.
Results We found that patient-derived DIPG cells and GBM cell lines express high levels of surface αvβ3 by flow cytometry, while αvβ3 is minimally expressed on normal tissues by RNA sequencing and protein microarray. The manufactured CAR-T cells consisted of a substantial frequency of favorable early memory cells and a low inhibitory receptor profile. αvβ3 CAR-T cells demonstrated efficient, antigen-specific tumor cell killing in both cytotoxicity assays and in in vivo models of orthotopically and stereotactically implanted DIPG and GBM tumors into relevant locations in the brain of NSG mice. Tumor responses were rapid and robust with systemic CAR-T cell proliferation and long-lived persistence associated with long-term survival. Following tumor clearance, TCF-1+αvβ3 CAR-T cells were detectable, underscoring their ability to persist and undergo self-renewal.
Conclusions These results highlight the potential of αvβ3 CAR-T cells for immunotherapeutic treatment of aggressive brain tumors with reduced risk of on-target, off-tumor mediated toxicity due to the restricted nature of αvβ3 expression in normal tissues.
- chimeric antigen
- immunologic memory
Data availability statement
Data are available on reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
While chimeric antigen receptor (CAR)-T cells are changing the treatment paradigm for many hematologic cancers,1–6 CAR-T cells as a viable therapeutic strategy against solid tumors has been met with challenges resulting in a lack of major advancement. They include heterogeneic expression and downregulation of target antigen, unacceptable levels of antigen expression in normal tissues, impaired CAR-T cell trafficking, and immunosuppressive tumor microenvironments7 8. Common CAR-T cell targets receiving attention include B7-H3, EGFRvIII, GD2, HER2, and mesothelin.9–12 Finding additional targets broadly expressed on many tumors but not on normal tissues would greatly advance the field as clinical development of just one therapeutic could be widely applicable to many cancers.
Integrins are heterodimeric adhesion receptors important for cell migration, tissue invasion, and in some cases survival of tumors themselves. Additionally, they are critical for angiogenesis, being expressed on neovascular endothelium feeding growing tumors, and via this role can contribute to tumor progression and metastasis.13 Integrin alphav beta3 (αvβ3) expression is typically limited to newly forming endothelial cells14 15; however, its presence has been associated with a number of tumors, particularly within the central nervous system (CNS).16–18 Previous studies of αvβ3 as a therapy target focused on inhibiting its angiogenic role in tumorigenesis due to its expression on endothelial cells within tumor vasculature. The small molecule, cilengitide, an αvβ3 and αvβ5 antagonist was safe in phase I/II trials in children and adults with high-grade, αvβ3-expressing gliomas but was not effective.19 Similarly, antibodies targeting αvβ3 did not demonstrate a therapeutic benefit in patients with melanoma, leiomyosarcoma, and other solid tumors.20–23 Despite the lack of efficacy, disrupting αvβ3 did not have any significant toxicities, making it attractive as a potential CAR-T cell therapeutic. In addition, since αvβ3 is upregulated on non-malignant vascular endothelium feeding tumors,24 αvβ3 CAR-T cells may still be effective even if tumors themselves downregulate the target, a frequent phenomenon driving CAR-T cell resistance. To date, no other CAR-T cell has been developed that targets both tumor and surrounding vasculature.
Though αvβ3 is known to be expressed on glioblastoma (GBM) and melanoma, studies investigating the breadth of targeting it with CAR-T cells are lacking. Screening additional tumor types for αvβ3, correlating functional assays including xenograft models to rigorously test the efficacy and safety of αvβ3 CAR-T cells, and deeper analysis of αvβ3 CAR-T cell antitumor response characteristics are necessary. In this study, we sought to evaluate the ability of αvβ3 CAR-T cells to target the histone H3K27M mutant diffuse midline glioma, diffuse intrinsic pontine glioma (DIPG), and GBM brain tumors, two highly lethal malignancies currently lacking effective therapeutics. DIPG is a highly aggressive tumor of the pons during early childhood and is universally fatal within 1 year on average even with aggressive interventions.25 26 Likewise, GBM, the most common primary brain tumor in adults, is characterized by very poor survival and limited effective therapeutics.27 We designed and tested new αvβ3 CAR-T cells in separate orthotopic brain tumor xenograft models representing DIPG and GBM. Our results provide a comprehensive evaluation of CAR-T cell products and analyses of the antitumor responses of αvβ3 CAR-T cells including efficacy, persistence, and memory cell characteristics in vivo following treatment of these CNS tumors.
Materials and methods
Tumor cell lines were obtained from the following sources: patient-derived DIPG cells were kindly provided by M. Monje (Stanford University), U87 from L. Lum (University of Virginia), and U251 from D. Brautigan (University of Virginia). Maintenance and culture of DIPG lines was performed according to an established protocol.28 U87, U251, and 293GP from C. Mackall (Stanford University) were maintained in DMEM containing 10% FBS. Tumor cells were lentivirally transduced with eGFP/firefly luciferase (eGFP/ffLuc). SU-DIPG-36 and U87 were single-cell cloned and evaluated for GFP by flow cytometry. Peripheral blood mononuclear cells (PBMCs) were obtained from healthy and consenting donors by apheresis under an Institutional Review Board approved protocol (UVA-IRB#18842).
CAR-T cell construct design, synthesis, and cloning
The scFv sequence for targeting αvβ3 was derived from the monoclonal antibody clone LM609 with assistance from the University of Virginia Antibody Engineering and Technology Core. The variable regions of the heavy and light chains were assembled with a (GGGGSx3) linker. The scFv sequence was preceded by the human GM-CSF-R signal sequence and followed by human CD8α hinge and transmembrane domain. This was linked to either a CD28 or 4-1BB intracellular signaling domain followed by the CD3ζ signaling chain. The entire construct was generated by gene synthesis (ThermoFisher). CAR constructs were then cloned into a retroviral vector. After sub-cloning, sequences were confirmed by DNA sequencing.
Retrovirus production and T cell transduction
Retroviruses were produced via transient transfection of 293GP cells. Cells were seeded on poly-D-lysine-coated plates (10mg/cm2) and transfected via Lipofectamine2000 (ThermoFisher) with plasmids encoding the CARs and RD114. PBMCs were stimulated with T cell activation and expansion beads (ThermoFisher) at a 1:1 ratio with 40 IU/mL IL-2 for 3 days. Retroviral supernatants were added to Retronectin-coated plates (Takara) and centrifuged for 2 hours. T cells were transduced on days 3 and 4 and cultured in AIM-V medium (ThermoFisher) containing 5% fetal bovine serum (FBS), 1% L-glutamine, 1% HEPES and 1% Pen/Strep with 300 IU/mL IL-2. Beads were removed on day 5, and T cells were expanded with 300 IU/mL IL-2 until harvest.
Surface expression of αvβ3 on tumor cells was evaluated by staining with PE-conjugated anti-αvβ3 (LM609; Millipore Sigma) or isotype control. Analysis of CAR-T cells was performed using antibodies against CD3(HIT3a), CD4(OKT4), CD8(SK1), CD45RO(UCH1.1), CD45RA(H100), CCR7(G043H7), CD127(A019D5), PD-1(EH12.2H7), LAG-3(11C3C65), TIM-3(F38-2E2), CD45(2D1), and TCF-1(7F11A10) (Biolegend). Expression of αvβ3 CAR was detected using biotinylated protein-L (ThermoFisher) followed by streptavidin-PE (Biolegend). Intracellular staining for TCF-1 was performed using Foxp3/Transcription Factor Staining Buffer Set (eBioscience). All samples were stained with Fixable Viability dye (Biolegend). Data were acquired on a NovoCyte3005 Flow Cytometer (ACEA Biosciences) and analyzed using FlowJov10.
Cytotoxicity and cytokine production assays
Cytotoxicity was evaluated using a luciferase-based assay measuring bioluminescence. Tumor cells expressing eGFP/ffLuc were coincubated with non-transduced T cells, CD19.28ζ, αvβ3.28ζ, or αvβ3.BBζ CAR-T cells for 18 hours at effector-to-target (E:T) ratios ranging from 10:1 to 2.5:1. Tumor cell lysis was determined by measuring residual luciferase activity and calculated as follows: percent lysis=100−(((average signal from T cell treated tumor wells)/(average signal from untreated target tumor wells))×100). In vitro cytokine production by CAR-T cells was evaluated through coincubation of non-transduced T cells, CD19.28ζ, αvβ3.28ζ, or αvβ3.BBζ CAR-T cells with tumor cells at a 1:1 ratio. Effector cell numbers were normalized across each group with transduction efficiency of CAR+ T cells. Coculture supernatants were collected after 24 hours and tested for interferon gamma (IFN-γ), interleukin-2 (IL-2), and tumor necrosis factor alpha (TNF-α) by ELISA (Biolegend).
Orthotopic xenogeneic mouse models
For DIPG, 5×105 patient-derived SU-DIPG-36 cells expressing eGFP/ffLuc were implanted by stereotactic injection into the pons of 6–8 weeks old NOD-scid IL2Rg− (NSG) mice (coordinates from lambda: M/L: +1mm; A/P: −0.8mm; D/V: −5mm). For GBM, 5×105 U87 cells expressing eGFP/ffLuc were implanted by stereotactic injection into the pons of NSG mice (coordinates from bregma: M/L: +3mm; A/P: +1mm; D/V: −3.5mm). On day 21, 2×106 CAR-T cells were delivered intratumorally using the same coordinates for tumor implantation. Tumor burden was assessed at least weekly via bioluminescent imaging (BLI) following intraperitoneal injection of D-luciferin (IVIS Spectrum; PerkinElmer). Progression-free survival was determined as the number of days from CAR-T cell treatment to the first day of disease progression or death.
A 20-organ normal tissue microarray (TMA; BN1002b; US BioMax) was stained with anti-αvβ3 antibody (23C6; R&D Systems) (performed by the Biorepository and Tissue Research Facility core; University of Virginia). For staining of DIPG tumor cells in mouse brain tissue, mice were perfused and brains harvested and fixed in paraformaldehyde. Brain sections were stained with anti-αvβ3 antibody (LM609). Slides were digitally scanned and evaluated for αvβ3 staining in QuPath, an open-source software application for digital histopathology analysis.28 Positive cell detection was performed on normal tissue TMA and mouse brain tissues at three threshold levels (1+, 2+, and 3+), and H-scores were calculated by using the following formula: [1×(%cells 1+) + 2×(%cells 2+) + 3×(%cells 3+)].
All statistical analyses were performed using GraphPad Prism. Data are presented as means±SEM unless otherwise noted in figure legends. Results were analyzed by one-way or two-way analysis of variance (ANOVA) with Tukey’s multiple comparisons post-test or unpaired non-parametric Mann-Whitney test for determining significant differences between experimental groups. Kaplan-Meier survival analysis was performed with Log-rank (Mantel-Cox) test for comparison between treatment groups. Bioluminescence images and data were analyzed with Living Image Software (PerkinElmer). Analysis of tumor growth curves was performed by first transforming bioluminescent values of individual mice at each time point to log form and then confirming Gaussian distribution using the Shapiro-Wilk test for normality. Unless noted in figure legends, means of each treatment group were compared at each time point by two-way ANOVA with Tukey’s multiple comparisons post-test.
αvβ3 expression on DIPG, GBM and normal tissues
Given that αvβ3 expression has been reported across some cancers, we sought to evaluate expression on DIPG and GBM for the purpose of CAR-T cell-directed therapy. αvβ3 expression was detected on 6/6 patient-derived DIPG lines and 2/2 GBM lines screened (figure 1A). In particular, αvβ3 was highly expressed on SU-DIPG-36, SU-DIPG-IV, SU-DIPG-21 (H3.1K27M mutants) and exhibited moderate expression on SU-DIPG-13, SU-DIPG-17, and SU-DIPG-27 (H3.3K27M mutants) (figure 1B). αvβ3 was also highly expressed on GBM (U87 and U251). The cancer cell lines H358 (lung), BT-20 (breast), and Caco-2 (colon) were negative for αvβ3 expression. These data are supported by publicly available gene expression data in DIPG and pediatric GBM patient specimens, particularly for integrin αv (ITGAV), relative to normal tissues (online supplemental figure 1A). Additionally, we found within the Cancer Cell Line Encyclopedia project database that glioma cell lines as a whole were elevated in expression of ITGAV and ITGB3 relative to other lines (including B-ALL, T-ALL, AML, and neuroblastomas) and normal organ tissues (online supplemental figure 1B). These results support the potential for CAR-T cell-directed targeting in this setting.
Given the high-level of CAR-T cell activity in patients following antigen exposure, limiting ‘on-target, off-tumor’ toxicities due to normal tissue expression of antigen is imperative for safe clinical usage. While αvβ3 expression is well characterized and has been shown to be primarily restricted to endothelial cells in newly forming blood vessels and absent in resting endothelial cells in normal organ systems, we performed additional screening for its presence in normal tissues. First, we performed gene expression analysis of ITGAV and ITGB3 using the expO Normal Tissue Expression array in the NCI Oncogenomics Database. Neither gene was highly expressed in represented normal organ tissues, except ITGB3 in the thyroid (figure 1C). Second, we assessed expression of αvβ3 by immunohistochemistry in a normal human tissue microarray consisting of 20 organ types, each from five individual donors (figure 1D). αvβ3 staining was absent in nearly all or exhibited minimal staining in others as determined by H-score (figure 1E). Of note, skin (3/5 donors) and ovary (2/4 donors) exhibited detectable but modest staining. These results provide further validation of the limited expression of αvβ3 in normal healthy tissues, supporting the targeting of this integrin by CAR-T cells.
Design and phenotypic characterization of CAR-T cells redirected against αvβ3
The scFv domain used in the design of the CARs targeting αvβ3 was based on the well-characterized monoclonal antibody clone LM609. LM609 has been previously evaluated in various clinical studies and found to be safe but with no reported treatment efficacy.23 29 For the purposes of our study, we designed two αvβ3-CAR constructs with humanized and optimized LM609 VH and VL chains, followed by the CD8α transmembrane domain, and CD28 or 4-1BB costimulation with CD3ζ (figure 2A). αvβ3.28ζ and αvβ3.BBζ CARs were highly expressed by T cells, exhibiting an average of 76% and 66% positivity, respectively, across multiple normal donors (figure 2B). Positivity and surface expression levels (eg, MFI) of αvβ3.28ζ and αvβ3.BBζ CARs were comparable to the CD19.28ζ CAR,30 31 which was included throughout this study as a negative control and for comparison purposes. αvβ3 CAR-T cells exhibited vigorous ex vivo expansion ranging from 100 to 1000-fold by day 7–9 post-transduction (figure 2C).
Successful clinical outcomes from CAR-T cell therapy occur when cell persistence and memory formation are achieved. Specifically, less-differentiated CAR-T cells and memory populations have been shown to exhibit superior antitumor responses.32 Therefore, we assessed the memory cell populations of αvβ3 CAR-T cells following ex vivo expansion. Evaluation by flow cytometry revealed a spectrum of memory and effector subset differentiation. Expanded αvβ3 CAR-T cells were comprised of effector and central memory subsets as well as stem cell memory T-like cells (figure 2D). There were no significant differences in the overall percentage of CD45RO+ populations in CD4+ T cells between non-transduced T cells, αvβ3.28ζ, αvβ3.BBζ, or CD19.28ζ CAR-T cells; however, CD8+ T cells exhibited increased frequencies of CD45RO+ cells in αvβ3.28ζ and αvβ3.BBζ CAR-T cells relative to controls (figure 2E). Both CD8 and CD4 αvβ3.BBζ CAR-T cell populations trend towards a less differentiated profile compared with αvβ3.28ζ CAR-T cells, although, there were no statistically significant differences between the two among profiled subsets except in the frequency of CD8+CCR7−CD127− effector memory cells (figure 2F). The modest differences in CAR-T cell differentiation may be attributed to some form of low-level activation as a result of CAR expression.
αvβ3 CAR-T cells efficiently eliminate DIPG and GBM in vitro
The antitumor activity of αvβ3 CAR-T cells was evaluated by determining cytotoxicity and effector cytokine production against DIPG and GBM tumor cells in vitro. CAR-T cell-mediated cytotoxicity was evaluated against patient-derived DIPG lines, SU-DIPG-36, SU-DIPG-XIII, SU-DIPG-IV, and SU-DIPG-17 by exposing eGFP/ffLuc-expressing DIPG cells to non-transduced, CD19.28ζ, αvβ3.28ζ, or αvβ3.BBζ CAR-T cells at E:T ratios ranging from 10:1 to 2.5:1. Both αvβ3 CAR-T cells exhibited substantial tumor cell killing against each DIPG line at all E:T ratios, with αvβ3.BBζ CAR-T cells showing a more modest yet still high level of killing at the lowest E:T ratio (figure 3A,B). Little to no cytotoxicity of DIPG cells was observed when cocultured with non-transduced or CD19 CAR-T cells. DIPG tumor cells also induced significant levels of effector cytokine production by αvβ3 CAR-T cells, but not controls, as determined by measurements of secreted IFN-γ, IL-2, and TNF-α (figure 3C). These data show the specific and robust antitumor responses of αvβ3 CAR-T cells against DIPG. Next, we evaluated efficacy of αvβ3 CAR-T cells against GBM. Antitumor activity of αvβ3 CAR-T cells was evaluated against eGFP/ffLuc-expressing U87 and U251. Both αvβ3.28ζ and αvβ3.BBζ CAR-T cells exhibited a high degree of tumor cell killing at all E:T ratios (figure 3D,E). Furtherore, αvβ3 CAR-T cells produced effector cytokines on GBM tumor cell exposure (figure 3F). The αvβ3-negative cell lines H358 and BT-20 (figure 1A) were used as target cells to validate specificity of antitumor activity against αvβ3. Importantly, there was an absence of cytotoxicity of αvβ3.28ζ and αvβ3.BBζ CAR-T cells against BT-20 and H358 (figure 3G), indicating that expression of αvβ3 on target cells is necessary for antitumor activity. Lastly, αvβ3.28ζ and αvβ3.BBζ CAR-T cell killing of U87 tumor cells was abolished following target blockade via anti-αvβ3 antibody (LM609), demonstrating specificity of the CAR for αvβ3 (figure 3H). These results demonstrate that αvβ3.28ζ and αvβ3.BBζ CAR-T cells are highly effective at killing DIPG and GBM brain tumor cells expressing αvβ3.
αvβ3 CAR-T cells induce rapid tumor regression of well-established orthotopic DIPG tumors
In order to validate antitumor activity of αvβ3 CAR-T cells and to demonstrate in vivo efficacy, we used an orthotopic xenograft model for DIPG. First, NOD-scid-IL2Rgammanull (NSG) mice were implanted with SU-DIPG-36 cells expressing GFP/ffLuc into the pons region of the brainstem using previously established stereotactic coordinates33 34 (figure 4A). Tumor cell engraftment within the pons was confirmed by H&E staining and immunohistochemical staining for human αvβ3 (figure 4B). At 3 weeks postimplantation, pretreatment tumor burden was assessed using bioluminescent imaging indicating that DIPG tumors were large and well established. Tumor-bearing mice were then administered 2×106 control CD19.28ζ, αvβ3.28ζ, or αvβ3.BBζ CAR-T cells by intratumoral (IT) injection using the same coordinates as tumor implantation.
In three independent experiments, using CAR-T cells generated from three different normal human donors, we observed rapid tumor regression within the first week of treatment in mice administered αvβ3 CAR-T cells (figure 4C,D). In contrast, tumors in mice receiving CD19.28ζ CAR-T cells continued to progressively grow. By day 14 post-treatment, tumors in αvβ3.28ζ CAR-T cell treated mice were reduced to background bioluminescent levels. Similarly, tumors in mice treated with αvβ3.BBζ CAR-T cells were also significantly reduced, but remained detectable. Tumors did eventually recur; however, longitudinal monitoring showed that αvβ3 CAR-T cells slowed or partially inhibited DIPG growth for an extended period of time as evidenced by the length of progression-free survival (figure 4E). Of note, tumor recurrence in CAR-treated mice was slow over time and remained at or below initial pretreatment levels for the duration of the experiment (8 weeks post-treatment). Mice surviving long term required euthanasia due to development of xenogeneic graft-versus-host disease, so additional survival and tumor monitoring could not be completed.
Across three separate and individual experiment replicates, disease resulted in lethality in 42% of mice given CD19 CAR-T cells, whereas 100% of mice that received αvβ3.BBζ CAR-T cells survived the duration of the experiment (figure 4F). Unexpectedly, 39% of mice that were treated with αvβ3.28ζ CAR-T cells died, of which most occurred within the first 11 days after injection and following significant reductions in tumor size. This could potentially be due to treatment-induced toxicity and/or trauma associated with repeat injections into the brainstem area. In summary, these results demonstrate that αvβ3 CAR-T cells can act rapidly to significantly reduce DIPG tumors in mice, prolong tumor growth on relapse, and extend survival.
No evidence of T cell exhaustion or antigen escape leading to DIPG recurrence
Tumor recurrence following CAR-T cell treatment can be attributed to various factors including inefficient T cell dose or expansion, exhaustion, lack of CAR-T cell persistence or loss of target antigen expression on tumor cells. In order to determine why DIPG tumors slowly recurred over time following an initial response to αvβ3 CAR-T cell treatment, we investigated whether any of these factors may have contributed to relapse. As expected, reduction in tumor burden coincided with a significant expansion of αvβ3.28ζ and αvβ3.BBζ CAR-T cells within the circulation compared with CD19 CAR-T cells by day 14 as these cells were easily detectable following IT administration (figure 5A).
Previous studies have demonstrated that CAR-T cells can become exhausted during ex vivo manufacturing as in the case of GD2.28ζ CAR-T cells, despite strong in vitro cytotoxicity.35 Thus, expression profiles for the inhibitory receptors PD-1, LAG-3, and TIM-3 on αvβ3.28ζ and αvβ3.BBζ CAR-T cells following ex vivo expansion were analyzed (figure 5B). Although αvβ3.28ζ CAR-T cells did exhibit significantly higher LAG-3 expression on CD8 T cells, inhibitory receptor expression on αvβ3 CAR-T cells was generally comparable with CD19.28ζ CARs in both CD4 and CD8 T cells (figure 5C). Thus, αvβ3.28ζ and αvβ3.BBζ CAR-T cells exhibit inhibitory receptor expression that would be expected as a result of recent T cell activation and are not elevated relative to CD19.28ζ CAR-T cells following ex vivo production. Given these results, in addition to efficient proliferation during ex vivo expansion (figure 2C), we conclude that αvβ3 CAR-T cells are not exhausted prior to administration into mice bearing DIPG tumors.
To assess the impact of antigen exposure on αvβ3 CAR-T cells in vivo, inhibitory receptor expression was analyzed on CAR-T cells in blood on day 7 and day 14 post-treatment. PD-1 levels increased slightly from day 7 to day 14 in both αvβ3CAR groups (figure 5D). LAG-3 and TIM-3 expression levels were low and did not increase from day 7 to day 14 on either CAR-T cell group (figure 5E,F). Importantly, expression levels of PD-1, LAG-3, and TIM-3 on peripheral CAR-T cells were substantially lower relative to their expression induced during initial activation and expansion prior to being transferred into DIPG xenografts. In conclusion, αvβ3 CAR-T cells responding to DIPG tumors in vivo do not appear to show signs of exhaustion at the time when tumors began to regrow, as there was not an overt nor widespread increase in multiple inhibitory receptors, a characteristic indication of the development of T cell exhaustion.36 37
Tumor escape through target antigen loss could have been a cause for DIPG recurrence following αvβ3 CAR-T cell treatment. To address this possibility, brains were harvested at the experimental end-point (day 56), processed for microscopic analysis, and human αvβ3 expression evaluated by immunohistochemistry. αvβ3 expression was detected throughout multiple areas of brain sections indicating significant DIPG tumor infiltration in mice treated with CD19 CAR-T cells. Importantly, strong αvβ3 expression in the pons and subventricular zone from mice treated with either αvβ3 CAR demonstrates that DIPG tumors retain αvβ3 expression following CAR treatment (figure 5G,H). Although αvβ3 expression was detectable in the subventricular zone, a common area of invasion by DIPG,38 tumor infiltration into this region was substantially diminished in αvβ3 CAR-T cell treated mice. Given that αvβ3 expression is retained, these results demonstrate that tumor recurrence following αvβ3 CAR-T cell treatment is not due to target antigen loss.
αvβ3 CAR-T cells eradicate established glioblastoma tumors in mice
To test the in vivo antitumor activity of αvβ3 CAR-T cells against GBM, luciferase-expressing U87 tumor cells were orthotopically implanted into the right forebrain of NSG mice via stereotactic injection, and tumors were allowed to grow for 3 weeks. GBM tumor-bearing mice were administered 2×106 CD19.28ζ, αvβ3.28ζ, or αvβ3.BBζ CAR-T cells by IT injection using the same coordinates as tumor implantation. Reduction in tumor burden was evident by day 7 post-treatment in both αvβ3 but not CD19 CAR groups and progressively declined until it was completely eliminated by day 21 (figure 6A,B, online supplemental figure 3A). Not surprisingly, CD28 containing CAR-T cells eliminated tumors more rapidly than 4-1BB CAR-T cells. This is consistent with known effects of CD28 vs 4-1BB signaling on CAR-T cell phenotype, in which CD28 typically provides stronger signaling.39 40 Meanwhile, GBM tumors treated with CD19 CAR-T cells continued to grow aggressively.
In contrast to treatment of DIPG tumors, IT treatment of GBM with αvβ3.28ζ CAR-T cells did not result in any deaths from treatment related toxicity. This suggests that potential toxicity from CAR-T cells may manifest differently in the pons compared with other regions of the brain. Mice treated with αvβ3 CAR-T cells were completely protected from lethality as 100% of mice survived long term, whereas 80% of control mice died by day 70 (figure 6C). Importantly, all αvβ3 CAR-T cell treated mice remained tumor free for the duration of the experiment. These data show that αvβ3 CAR-T cell treatment was highly effective at completely eliminating GBM tumors in vivo without any observable evidence of toxicity or tumor recurrence.
Systemic administration of αvβ3 CAR-T cells leads to complete regression of glioblastoma
The previous results demonstrated effective antitumor activity of αvβ3 CAR-T cells injected intratumorally. We next aimed to determine if CAR-T cells administered in a systemic manner could migrate to orthotopic GBM tumors. Mice were injected intravenously with 10×106 CD19 or αvβ3.BBζ CAR-T cells, and tumor burden was monitored by BLI (figure 6D, online supplemental figure 3B). By day 14, significant tumor regression was evident in mice treated with αvβ3.BBζ CAR-T cells (figure 6D–E). Mice treated with αvβ3.BBζ CAR-T cells remained disease-free for the duration of the experiment and exhibited superior survival compared with mice treated with control CAR-T cells (figure 6F). These results demonstrate that αvβ3 CAR-T cells can be administered systemically to effectively treat GBM tumors.
αvβ3 CAR-T cells develop long-lived memory following successful GBM treatment
Lastly, we sought to determine whether long-lived memory CAR-T cells develop following eradication of GBM tumors. Both αvβ3.28ζ and αvβ3.BBζ CAR-T cells demonstrated exceptional persistence following tumor clearance as intratumorally administered T cells were easily detectable day 42 post-treatment (figure 7A). A large proportion of αvβ3.28ζ and αvβ3.BBζ CAR-T cells expressed the memory marker CD127 but were also PD-1+ (figure 7B,C). Interestingly, mice treated with αvβ3.28ζ CAR-T cells had a significantly higher percentage of CD127+ cells compared with mice treated with αvβ3.BBζ CARs.
Some antigen-experienced T cells that exhibit properties of memory and PD-1 expression, yet retain the ability to self-renew in the absence of antigen, are responsive to checkpoint inhibition41 42 and express the transcription factor TCF-1. TCF-1 regulates stem-like functions of pre-exhausted T cells (PD-1Int), and TCF-1+ tumor-infiltrating lymphocytes were shown to be elevated in melanoma patients who had responded to checkpoint receptor inhibition.43 Thus, we sought to determine if αvβ3 CAR-T cells express elevated levels of TCF-1 following successful treatment of GBM tumors. We observed that CD8+CD127+PD-1+ populations in αvβ3.28ζ and αvβ3.BBζ CAR-treated mice expressed significantly higher levels of TCF-1 relative to the more terminally differentiated CD127-PD-1+ populations (figure 7D,E). As expected, TCF-1 expression was not increased in CD8+CD127+PD-1+ cells in mice treated with CD19 CARs. These data demonstrate that antigen-experienced αvβ3 CAR-T cells with self-renewal capacity develop following eradication of tumors.
Finally, persistence of αvβ3 CAR-T cells was evaluated 70 days following treatment of GBM tumors. We observed that αvβ3 CAR-T cells were still present at day 70 in blood, spleen, and brain (figure 7F). Analysis of memory cell subsets by staining for memory markers CD127 and CCR7 on CAR-T cells was performed. The spectrum of memory T cell differentiation appeared similar in blood, spleen, and brain. Although the highest proportion of CAR-T cells were the more differentiated CCR7-CD127- subset, substantial frequencies of CCR7+CD127+, CCR7+CD127−, and CCR7-CD127+ subsets were detected in all tissues in which T cells were recovered (figure 7G). The results of these experiments show that αvβ3 CAR-T cells develop memory and persist long term in vivo.
We report that CAR-T cells targeting αvβ3 represent an attractive candidate for successful therapeutic treatment of malignant and fatal brain tumors. In this study, we determined that αvβ3 on the H3K27M mutant diffuse midline glioma, DIPG, and GBM tumors can be targeted by CAR-T cells. Importantly, αvβ3 CAR-T cells were highly effective in vivo, mediating significant regression of orthotopic tumors and displaying persistence and memory development. Despite eventual recurrence of DIPG tumors in mice, we were able to rule out specific factors known to be significant limitations of CAR-T cell therapy that typically contribute to relapse: poor CAR expansion, T cell exhaustion, and target antigen escape. Of note, DIPG tumors were large and very well established at the initiation of CAR-T treatment, therefore, a higher dose or repetitive dosing may improve these outcomes. Given the dire prognoses of these high-grade gliomas and absence of effective therapies to treat them, αvβ3 CAR-T cell therapy may offer a promising avenue for these patients.
DIPG tumors are notoriously infiltrative.38 44 In our orthotopic DIPG experiments, we used a DIPG line that exhibits aggressive growth features (in vitro and in vivo observations), which likely resulted in dissemination and almost certainly contributes to the challenge of achieving successful treatment. Invasion of other areas and niches of the brain by infiltrating DIPG cells occurred as has been demonstrated to happen in human disease and in DIPG mouse xenografts.33 Despite this, we were able to significantly lower tumor burden, extend progression-free survival, and overall survival through treatment with αvβ3 CAR-T cells. Of note is the possible toxicity driven by αvβ3 CAR-T cells that use CD28 costimulation in some mice with DIPG tumors. CAR-T cell-induced toxicity was observed in a subset of mice with DIPG following treatment with GD2 CARs containing 4-1BB costimulation.45 This study and ours suggests that the anatomical location of DIPG in the pons may be of particular concern, given the critical functions performed there. It is also interesting to speculate on the role of costimulatory domains in CAR signaling and how these impact toxicities. In the case of orthotopic GBM tumors, we observed complete clearance by αvβ3 CAR-T cells without any ensuing tumor recurrence or toxicity. These observations suggest potential differences in the development of treatment-related toxicity based on anatomical location of gliomas and call for incorporation of control or suicide mechanisms for CAR-T cells prior to clinical trials.
Another potential toxicity faced by all new CAR-T cell products in the preclinical phase of development is the risk of so called ‘on-target, off-tumor’ killing of normal cells that may express low-levels of the target antigen. The scFv used in our products does not recognize a murine counterpart. So, while our data suggests highly restrictive αvβ3 expression in normal human tissues, the absence of ‘on-target, off-tumor’ killing in our mouse studies does not rule out the possibility of such toxicity in humans. A carefully designed and implemented clinical trial of this product with this potential in mind is therefore warranted.
Others have studied αvβ3 targeting CAR-T cells in a limited capacity. Wallstabe et al46 suggested that CAR-T cells re-directed against αvβ3, were able to control disease in mice engrafted with a single melanoma cell line in limited experiments but did not examine other tumor types or effects on tumor vasculature. In a separate study, αvβ3-expressing melanoma was targeted in a syngeneic mouse model using a CAR-T that incorporated echistatin as the targeting domain, rather than an scFv.47 While murine-specific CAR-T was able to slow tumor growth, animals still succumbed to disease. Our study differs from these in three principal ways. First, our αvβ3 CAR-T cells use humanized and optimized scFv targeting, which has been shown to decrease the risk of anti-CAR-T cell immune responses as best exemplified by CD19 CARs using a humanized scFv rather than the relapse-prone murine-based FMC63 domain.48 49 Second, ours is the first to successfully target the most lethal high-grade gliomas in both children and adults using αvβ3 CAR-T cells and extends our earlier work of using CARs to target medulloblastoma.50 Third, we demonstrate that intravascular administration of αvβ3 CAR-T cells is sufficient to control well-established intracranial GBM in animals.
An unresolved area of interest in the use of CAR-T cells targeting αvβ3, is their potential to recognize cells of the tumor vasculature expressing αvβ3. A limitation of the xenograft models used here is that anti-αvβ3 scFv in the CAR does not bind to murine αvβ3. Thus, determining the contribution of tumor-associated endothelial cells being targeted by CAR-T cells is not possible. In the study using echistatin as the targeting domain against αvβ3, investigators observed destruction of tumor blood vessels in mice implanted with B16 melanoma, but not resting vessels within normal tissues, and significantly stalled tumor growth.47 Future studies are necessary to scrutinize the effects of αvβ3 CAR-T cells on tumor vasculature and how this may impact antitumor responses.
Our study demonstrates a renewed potential for targeting integrins, specifically αvβ3, for antitumor activity. Importantly, our results demonstrating robust responses in tumor-bearing mice, coupled with the restricted nature of αvβ3 expression in normal tissues, underscore the potential for both highly effective and safe use of αvβ3 CAR-T cells for treatment of DIPG and GBM in patients. Therefore, results of this study support development of Phase I clinical trials of αvβ3 CAR-T cells for adults and children with αvβ3-expressing malignancies.
Data availability statement
Data are available on reasonable request.
Patient consent for publication
Human PBMC collection from healthy donors was performed according to an Institutional Review Board approved protocol (UVA-IRB#18842). All animal studies were conducted under a protocol approved by and in compliance with policies set by the Institutional Animal Care and Use Committee.
We would like to acknowledge the following core facilities at the University of Virginia: The Research Histology Core, the Biorepository and Tissue Research Facility, and the Molecular Imaging Core, which are supported by the University of Virginia School of Medicine and through the University of Virginia Cancer Center National Cancer Institute P30 Center Grant. We would also like to acknowledge the University of Virginia Center for Comparative Medicine for providing animal care and services. The accompanying graphical abstract was created with BioRender.com.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Contributors Conceptualisation: DAC, JdR and DWL. Methodology: DAC, JdR and DWL. Investigation: DAC, JdR, LL and EA. Visualisation: DAC. Funding acquisition: DAC and DWL. Project administration: DWL. Supervision: DAC and DWL. Writing – original draft: DAC. Writing – review and editing: DAC and DWL. Guarantor: DWL.
Funding The Childhood Brain Tumor Foundation (DWL); The V Foundation for Cancer Research (DWL); UVA Strategic Investment Fund (DWL).
Competing interests DWL serves as a consultant to Juno Therapeutics/BMS, Harpoon Therapeutics, and Amgen, and his institution receives clinical trial funding from Kite Pharma/Gilead. His spouse is an employee of Karyopharm Therapeutics. DAC and DWL have a patent application pending based on the results presented in this study. All other authors declare that they have no competing interests.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.