Optimizing CARs for ocular delivery

Abstract

Effective methods for treating retinoblastoma while preserving vision are an unmet clinical need. Subretinal delivery of a hydrogel containing T cells that secrete the cytokine IL-15 and express a chimeric antigen receptor directed at the ganglioside protein GD2 completely controls retinoblastoma in immunocompromised mice, with no obvious damage to the surrounding retina.

Retinoblastoma is the most common ocular tumor of childhood and causes substantial morbidity in very young children. 40% of all cases of retinoblastoma occur due to a germline cancer predisposition, and patients with this predisposition are more likely to develop bilateral retinoblastoma. Treatment can require enucleation (surgical removal) of the eye and subsequent loss of vision, particularly in advanced disease. Ophthalmic-artery chemosurgery, a localized treatment that profoundly changed retinoblastoma therapy, has dramatically decreased the rate of enucleation in unilateral and bilateral retinoblastoma, saving the majority of affected eyes without compromising survival¹. Although outcomes are excellent in developed countries due to early diagnosis, patients with metastatic and recurrent disease (a more common scenario in developing countries) fare less well. Thus, novel therapeutic strategies to effectively control retinoblastoma and preserve the globe of the eye and a child’s vision are sought after, especially therapies with minimal short- and long-term side effects. T cells that express a genetically engineered chimeric antigen receptor (CAR T cells) have been highly successful in the treatment of refractory hematologic cancers, and while they have been less effective in solid cancers thus far, there are ongoing efforts to address this through the optimization of CAR T cell products and delivery strategies. Relevant to the setting of retinoblastoma, CAR T cell immunotherapy should not have the same risks of secondary malignancies as those of chemotherapy and radiation, which makes this approach particularly attractive in these typically young patients. In this issue of Nature Cancer, Wang et al. demonstrate the therapeutic efficacy of locally delivered CAR T cells designed to target tumor cells expressing the ganglioside protein GD2 (‘GD2 CAR T cells’), in preclinical models of retinoblastoma².

Wang et al.² test a previously described CAR construct that incorporates the single-chain variable fragment derived from the monoclonal antibody dinutuximab (14g2a), an endodomain of the co-receptor CD28, and the T cell antigen receptor CD3ζ chain, in two orthotopic retinoblastoma models³. GD2 is a ganglioside that is expressed on retinoblastoma and other pediatric cancers, has limited expression in normal tissue (and no expression on normal retina), and has been validated as an immunotherapeutic target, which makes it an attractive candidate antigen in retinoblastoma. The authors demonstrate robust antigen-specific tumor-cell killing in two retinoblastoma cell lines by GD2 CAR T cells, relative to that of control CAR T cells targeting an antigen (CD19) not expressed on retinoblastoma. GD2 CAR T cells injected subretinally and intra-tumorally transiently control retinoblastoma xenografts, but cytokine IL-15–secreting GD2 CAR T cells encapsulated in chitosan–polyethylene glycol hydrogel mediate complete control (Fig. 1). The authors demonstrate that IL-15 increases the proliferation and persistence of GD2 CAR T cells, and that delivery of CAR T cells within a hydrogel improves their distribution within the tumor. The translational impact of these promising results is elevated by the observations that the surrounding retina is not damaged in treated mice, and that this treatment results in improved function compared with that of retinae in mice treated with the control cells.

Fig. 1 |. Treatment of retinoblastoma with GD2 CAR T cells.

a, Untreated retinoblasoma. b, GD2 CAR T cells injected subretinally and intra-tumorally transiently control retinoblastoma growth. c, Secretion of IL-15 by GD2 CAR T cells enhances survival of the T cells and results in improved retinoblastoma control. d, IL-15-secreting GD2 CAR T cells encapsulated in chitosan–polyethylene glycol hydrogel demonstrate complete retinoblastoma control.

Two major challenges of CAR T cell therapy for solid tumors are ensuring adequate proliferation of the cells in vivo to mount an effective immune response, and efficient trafficking to the tumor site. Wang et al. equip the genetically modified CAR T cells with the ability to secrete IL-15, an immunostimulatory cytokine known to enhance proliferation². Furthermore, they facilitate T cell trafficking by injecting CAR T cells directly into the tumor and ‘packaging’ the cells within a chitosan–polyethylene glycol hydrogel, which both improves distribution of the cell suspension and retains localized stimulation with IL-15. Indeed, the addition of these two features leads to complete control of retinoblastoma in the subretinal tumor models and supports the persistence of viable CAR T cells in the tumor site 60 days after injection.

Additional factors that impair CAR T cell efficacy include antigen heterogeneity on the surface of the tumor, as well as an immunosuppressive tumor microenvironment. For CAR T cells designed to recognize a single antigen, the optimal target is one that is uniformly and highly expressed by all tumor cells. However, antigen downregulation or loss is a common mechanism employed by tumors to evade CAR T cell therapy⁴. One strategy for addressing this well-recognized problem is targeting two or three antigens simultaneously by CAR T cells⁴. In addition to GD2, HER2 and CD171 (L1CAM) have also been identified in retinoblastoma and are validated immunotherapeutic antigens that could be simultaneously targeted^5,6. Wang et al. do not determine whether antigen loss contributes to CAR T cell failure in their experiments that do not incorporate IL-15 and hydrogel² (Fig. 1a,b). Nonetheless, the durability of tumor clearance in experiments utilizing GD2 CAR T cells that secrete IL-15 encapsulated in a hydrogel is encouraging. Immunosuppressive myeloid cells respond to inflammation and cancer, which potently inhibits effector T cell function and potentially contributes to the resistance of solid tumors to CAR T cells. Indeed, targeting myeloid-derived suppressor cells improved GD2 CAR T cell efficacy in preclinical sarcoma models⁷. Immunosuppressive myeloid cells have also been found within samples from patients with retinoblastoma and in transgenic retinoblastoma mouse tumors^8,9. Thus, combinatorial strategies for addressing known barriers in immunotherapy for solid tumors will probably facilitate more effective CAR T cell therapy in retinoblastoma.

GD2 CAR T cells were among the first CAR T cell products tested in patients and have shown preclinical efficacy in neuroblastoma, sarcoma, melanoma, breast cancer and, most recently, H3K27M-mutant diffuse midline glioma^3,7,10–12. The most extensive clinical experience with GD2 CAR T cells has been in patients with neuroblastoma, in whom they have demonstrated safety, feasibility and objective responses. However, clinically meaningful responses have been limited. All successful clinical trials for the treatment of hematologic cancers with CAR T cells (and all trials of GD2 CAR T cells in humans thus far) have evaluated only systemic delivery, which makes direct delivery into the subretinal space, as tested by Wang et al.², especially unique^10,11,13. The preference for avoiding systemic delivery of CAR T cells after lymphodepletion is mainly to reduce toxicity from cytokine-release syndrome and/or on-target/off-tumor binding of CAR T cells. For these reasons, local or regional delivery of CAR T cells is being evaluated in multiple trials treating solid tumors of the central nervous system. The eye is considered to be immunologically privileged to preserve vision and prevent damage by infiltrating immune cells¹⁴. Thus, in addition to the eye’s being well suited for localized delivery of CAR T cells for accessibility reasons, ocular delivery may be a scenario in which allogeneic products may not be rejected and, therefore, may be more feasible than delivery at other sites^14,15. Use of such ‘off-the-shelf’, allogeneic CAR T cells increases the likelihood that this therapy will be accessible to patients in low-, middle- and high-income countries¹⁵. In summary, Wang et al. convincingly show that IL-15 secretion and hydrogel encapsulation enhance the in vivo functionality of GD2 CAR T cells², in support of not only further investigation of such a product for children with retinoblastoma, but also the potential applicability of similar designs for the loco-regional delivery of CAR T cells in other solid tumors.