Driving CARs on the Highway to Solid Cancer: Some Considerations on the Adoptive Therapy with CAR T Cells

The choice of a safe antigen

While it is clear that T cells can be redirected toward any target, the difficulty lies in the choice of a targetable antigen that is specific for the cancer cells while ideally absent on healthy tissues. Such truly “tumor-specific” antigens suitable for selective tumor targeting are rare. For instance, EGFR-variant three (EGFRvIII) is a tumor-specific mutant protein; targeting of EGFRvIII by CAR T cells is not expected to harm, since healthy tissues with the unmutated EGFR are not recognized. On this basis, at least three trials targeting EGFRvIII by CAR T cells were initiated at different centers (NCT02209376, NCT01454596, and NCT02664363). Another group of unique antigens are characterized by post-translational modifications. For instance, Muc1 is glycosylated differently in various tumors compared to healthy tissues. ³⁴ Targeting antigens with aberrant glycosylation patterns or with mutations may be a strategy for increasing selectivity in targeting solid tumors, albeit with the frequent limitation that not all cancer cells in a tumor lesion express the mutant antigen.

Most antigens targeted by CAR T cells are not tumor selective but are expressed by healthy cells. Consequently, most CAR T cells not only target the tumor lesion but also attack healthy tissues. While in the case of CD19 targeting in the treatment of lymphoma/leukemia, depletion of healthy B cells is clinically manageable, targeting solid cancer antigens may provoke severe adverse events when expressed by essential tissues. The situation is less severe when the tumor-associated antigens are shared by non-essential tissues, such as the prostate-associated antigen PSMA (NCT01140373, NCT00664196, and NCT01929239).

The targeted antigen can also be displayed by the tumor stroma like fibroblast activation protein (FAP), ³⁵ which is also expressed by fibroblasts in healing wounds. Vascular endothelial growth factor receptor-2 (VEGF-R2) is an alternative target in the tumor stroma. ³⁶

Self-antigens with a restricted surface expression are of particular interest in this context. For instance, CEA is expressed on the apical surface of epithelia in the gastrointestinal tract facing the lumen, thereby invisible to immune cells. Accordingly, CAR T cell therapy targeting CEA did not produce severe adverse events in preclinical models ³⁷ or in a recent trial to treat metastatic colorectal cancer. ¹⁸ The situation is different when CEA is targeted by TCR T cells, which recognize MHC class I presented CEA displayed by healthy gastrointestinal epithelia cells in a non-polarized fashion. Accordingly, a trial using anti-CEA TCR T cells caused severe colitis, with the destruction of healthy colon epithelia. ³⁸ CEA is also an example of antigens that are released by cancer cells, and serum CEA levels can be used as a biomarker for tumor progression. In order to target CEA, a binding domain is required that is not blocked by the serum antigen, which is the case for the BW431/26 scFv, as reported by Hombach et al. ³⁹

Self-antigens are also worthwhile targets for CAR T cell therapy when the antigen level on cancer cells is abnormally high compared to healthy cells, providing a “therapeutic window.” Examples of such shared self-antigens are Her2, cMet, mesothelin, GD2, EphA2, CD171, and CD133. The discrimination between healthy and malignant cells in this situation depends on a number of parameters. Binding affinity is a major factor, as previously shown in experimental models. ^37,40,41

The application of third-generation CAR T cells with two co-stimulatory domains and targeting Her2 for the treatment of metastatic melanoma induced fatal side effects in one case; within minutes after infusion, the patient suffered from severe distress, massive pulmonary cell infiltrates, and tissue destruction, leading to death 5 days later, despite intensive care intervention. ²¹ The adverse event was unexpected, since the targeting domain was derived from the well-established anti-ErB2 antibody Herceptin, which is applied in adjuvant therapy with a low rate of toxicity. Targeting Her2 in metastatic sarcoma in another trial proved to be safe, with some evidence of efficacy with a dose of CAR T cells (i.e., 10⁸ cells), ²⁵ which was three logs lower than the fatal case of Her2 CAR T cell treatment. The experience from this case indicates that the CAR T cells have to be thoroughly evaluated in a careful escalation trial. No prediction can be deduced from experiences from antibody therapies.

CAR T cells infiltrating the tumor lesion

Trafficking and infiltration of systemically applied CAR T cells are major issues in solid cancer. Some tumors are highly fibrotic and difficult to penetrate physically; others have low or no chemokine signaling preventing T cell trafficking to the lesion. In order to ensure accumulation within the malignant lesion, CAR T cells were locally applied by endoscopy or puncture into or in the near vicinity ¹⁹ or by installation of biodegradable pads with CAR T cells. ⁴² The strategy is only applicable for accessible tumors, which is not the case for the majority of solid cancer lesions. Preclinical models indicate that systemically applied CAR T cells can traffic to and accumulate at the tumor tissue in the immune competent host with the targeted antigen expressed by healthy tissues. ³⁷ Accordingly, CEA CAR T cells preferentially accumulated in the mesenchyme but less in the parenchyma of the tumor in a recent trial. ¹⁸

Trafficking is guided by a balanced mixture of cell bound factors such as adhesion molecules and soluble factors, including chemokines, which is frequently different in solid-tumor lesions compared to healthy tissues. Sensing these factors through chemokine receptors guides T cells to the tumor lesion. While freshly isolated T cells display the entire panel of receptors, prolonged ex vivo amplification after genetic engineering likely goes along with an altered panel of chemokine receptors on the engineered T cells. Transgenic expression of a specific chemokine receptor may render the T cell sensitive for a particular tumor-secreted chemokine. For instance, transgenic CXCR2, the receptor for CXCL1, expressed by transgenic T cells improved trafficking to melanoma ⁴³ ; transgenic expression of CCR2b improved trafficking to neuroblastoma ⁴⁴ ; and expression of CCR2 enhanced targeting and antitumor activity against mesothelioma. ⁴⁵

Despite accumulation, T cell extravasation and migration into the tumor lesion is frequently suppressed on various levels, including the loss of adhesion molecules on vascular endothelial cells required for transmigration. Vascular evasion of CAR T cells and finally infiltration into the tumor lesion can be improved by targeting vascular endothelial growth factor receptor-2 (VEGF-R2), which is overexpressed by tumor endothelial cells. ⁴⁶ Blocking of the endothelin-B receptor, a migration inhibitory receptor, improves tumor infiltration, ⁴⁷ indicating that T cells are actively prevented from infiltration by some tumor-associated endothelial cells. Normalization of vasculature by low-dose angiogenesis inhibitors may facilitate vasculature extravasation and thereby sustain T cell therapy in the long term. ⁴⁸

T cell migration through tissues occurs more often than previously thought. T cells can penetrate through the blood–brain barrier and enter the central nervous system (CNS) ⁴⁹ and other immune-privileged sites such as the testes and eyes. ⁵⁰ The profound migratory capacity of T cells allows the treatment of tumors that are otherwise difficult to access such as brain tumors and prostate cancer. The impaired capability to migrate through solid tumors may also be due to the loss of heparanase, responsible for degrading heparan sulfate proteoglycans in the stroma, by CAR T cells during the manufacturing process. ⁵¹ Re-expression of heparanase reconstitutes the ability to degrade the extracellular matrix and improves T cell migration capacities.

T cell proliferation, persistence, and exhaustion

In contrast to first-generation CAR T cells, T cells with a second-generation CAR persist in the peripheral blood of treated patients for weeks to months. T cell amplification after administration seems to be required to achieve an effective T cell-to-tumor cell ratio. T cell amplification is the best predictor of clinical efficacy. Improving CAR T cell amplification is currently a major focus of research. Based on experiences in the treatment of leukemia, incorporation of the CD28 or 4-1BB co-stimulatory domain clearly improves T cell persistence. CD28 co-stimulation more rapidly induces tumor regression compared to 4-1BB, which predominantly improves T cell persistence in the long term. Thereby, CD28 CAR T cells seem to be more suitable for the early induction of clinical remission, while 4-1BB CARs may be more suitable for long-term control beyond primary tumor elimination. Other co-stimulatory domains as well as combinations thereof are beneficial for some cell types ⁹ or induce maturation in a specific fashion. Extended chronic antigen exposure induces T cell exhaustion with the loss of proliferative capacities. Tumorigenesis frequently goes along with T cell exhaustion, which is finally irreversible. ⁵² Programmed cell death-1 (PD-1) on T cells seems to be crucial in this process. Since PD-1 in exhausting cells is regulated on the genetic and epigenetic level, ⁵³ modulating these regulators may provide the opportunity to prevent exhaustion. Accordingly, PD-1-deficient T cells are resistant to exhaustion in an experimental model. ⁵⁴

Tumor-associated immune suppression

Tumors establish a potent immune suppression on various levels, for example through repressor cells such as regulatory T (Treg) cells or myeloid derived suppressor cells (MDSCs), through suppressive cytokines such as IL-10 or transforming growth factor (TGF)-β, or through metabolic processes. TGF-β repression can be overcome by CD28 co-stimulation through the CAR. ⁵⁵ In the presence of tumor-infiltrating Treg cells, the antitumor activity of CAR T cells is increasingly suppressed upon CD28 CAR T cell activation with the release of IL-2. The CAR T cell antitumor response is improved by the abrogation of IL-2 release through a modified CD28 co-stimulatory CAR domain, ⁵⁶ avoiding that IL-2 released upon CAR T cell activation feeds the Treg cells, which in turn suppress the antitumor T cells. An alternative strategy to neutralize TGF-β repression is the transgenic expression of a dominant negative form of the TGF-β receptor, as shown in a preclinical melanoma model. ⁵⁷ The suppressive effect of IL-4, frequently released by prostate cancer, can be converted into a stimulatory effect to T cells by a recombinant “switch” receptor, which binds IL-4 and transmits a stimulatory IL-7 signal. ⁵⁸ Co-expressed with an anti-PSCA, CAR T cells increased activation against prostate cancer cells and showed improved antitumor activity.

Activated T cells are highly susceptible to receptor-mediated repression. They furthermore increase their sensitivity to suppression by increasing the level of repressive receptors during an effector response, including PD-1, cytotoxic T lymphocyte-associated antigen-4 (CTLA-4), or Fas. T cells with 4-1BB CAR seem to be more resistant to PD-1-mediated repression and exhaustion than T cells with CD28 CAR. ⁵⁹ However, various strategies are currently being explored to make CAR T cells more resistant to repression. For instance, blocking PD-1 by an antibody improved the antitumor activity of Her2-specific CAR T cells ⁶⁰ and mesothelin-specific CAR T cells in preclinical models. ⁶¹ As a first example, blocking anti-PD1 by pembrolizumab along with CAR T cell therapy is currently explored in the treatment of neuroblastoma (NCT01822652). Cell-intrinsic PD-1 shRNA blockade or a PD-1 dominant negative receptor also improves CAR T cell activity in preclinical models. ⁵⁹ The dominant negative PD-1, which consisted only of the extracellular binding domain, improved the persistence and antitumor activity in a mesothelioma model likely through competition in PD-L1 binding with the endogenous PD-1. ⁵⁹ A PD-1:CD28 “switch receptor” converted the negative signal into a CD28 stimulatory signal upon binding to PD-L1. ^62,63 CAR T cells with such PD-1:CD28 receptor showed improved accumulation and antitumor activity in large established mesothelioma and prostate cancer and executed an improved antitumor response in the presence of myeloid derived suppressor cells (MDSCs), which express high levels of PD-L1. Kobold et al. reported an increased accumulation of IFN-γ-producing T cells and a beneficial ratio of CD8⁺ T cells to MDSCs in the tumor tissue upon switch receptor activation. ⁶⁴ Co-expressed with the tumor targeting CD3ζ CAR, the switch receptor can also be used to make the CAR T cell antitumor attack more selective by sustaining full T cell activation when the CD3ζ CAR binds to the tumor target. Engagement of PD-L1 outside the tumor tissue without engagement of the tumor-associated antigen by the CAR would repress the CAR T cell.

On the other hand, CAR T cells show tremendous flexibility in responding to activation, for example by upregulating a panel of suppressive receptors. For instance, antibody mediated PD-1 blockade is frequently counteracted by the upregulation of T cell immunoglobulin mucin-3 (TIM-3) and lymphocyte-activated gene-3 (LAG-3) in CAR T cells upon engagement of target. ⁶⁵ Accordingly, CAR T cells with the PD-1:CD28 “switch receptor” prevented upregulation of both LAG-3 and TIM-3/CEACAM1. Despite PD-1, a number of other suppressive factors are still in place, including T cell intrinsic inhibitory enzymes such as SHP-1 and SHP-2 and surface inhibitory receptors such as LAG-3 or TIM-3, which are also upregulated upon T cell activation. ⁶⁵ As a consequence, the blockade of more than one inhibitory pathway is required to make the CAR T cells resistant to repression. Moreover, the blockade needs to be restricted to the CAR T cell itself in order to avoid a generalized “de-blockade” of circulating blood T cells and finally auto-immunity.

Tumor-associated hypoxia and nutrient deprivation

A hallmark of solid-tumor lesions is hypoxia associated with an altered metabolism and starvation from nutrients. In particular, indoleamine-2,3-dioxygenase (IDO) of cancer cells and MDSCs produce kynurenine through degradation of trypthophan, which blocks CAR T cell activation, amplification, and cytolytic activity. ⁶⁶ Cyclophosphamide and fludarabine, used for lymphodepletion before CAR T cell therapy, decreases IDO production, providing a strong rationale for preconditioning or for the use of IDO inhibitors. Prostaglandin E2 (PGE2) and adenosine, both produced by protein kinase-A (PKA), also inhibit T cell activity; PKA disruption through disrupting PKA anchorage within the membrane improved infiltration, persistence, and antitumor activity of anti-mesothelin CAR T cells. ⁶⁷ In response to hypoxia, T cells alter their metabolism. Redirecting the metabolism by co-stimulation provides a path to prevent exhaustion and to prolong survival. ¹⁰ On the other hand, adenosine and potassium ions together with other metabolic waste products are highly enriched in tumor tissues, preventing productive T cell activation. The situation demands strategies to improve the T cell metabolic fitness through the overall control of metabolism in order to allow T cell survival in the hostile tumor environment.

Tumor-associated stroma

Solid tumors often display a strong mechanical barrier composed of stroma fibroblasts, which hamper CAR T cells from penetration. Moreover, these stromal cells strongly support tumor progression by secreting growth factors, cytokines, and chemokines. Preclinical models demonstrate that successful treatment of advanced tumors requires the destruction of tumor stroma cells; IFN-γ is crucial in this process. ⁶⁸ The level of IFN-γ released by CAR T cells into the attacked tumor tissue can be increased by providing co-stimulation, in particular through CD28. Apart thereof, the stroma fibroblasts of nearly all epithelial cancers express fibroblast activating protein-1 (FAP-1), which can be targeted by CAR T cells. Together with cancer cell targeting CAR T cells, anti-FAP CAR T cells improved the antitumor response. ³⁵ Anti-FAP CAR T cells inhibited the formation of stroma in a preclinical model, reduced the density in tumor vasculature, and disrupted the spatial orientation of ductal pancreatic adenocarcinoma, resulting in reduced tumor progression. ⁶⁹ Taken together, cancer cell targeting combined with the destruction of tumor stroma improves eliminating solid-tumor lesions.

Activating the local endogenous immune response

Solid-tumor lesions display a tremendous heterogeneity in cancer cells and antigen expression, which demands the recruitment and activation of the endogenous immune system in order to eliminate the tumor lesion. The host immune response can be recruited to the tumor lesion by CD40L expressed by the CAR T cells in the treatment of CD40⁺ tumors. The antigen presentation by dendritic cells is improved and PD-1 expression decreased. ⁷⁰ Similarly, activation of the STING pathway activates dendritic cells, sustains T cell activation, and mediates innate immune cell recognition of tumors. ⁷¹ Alternatively, the local cytokine network can be modulated in order to activate innate immune cells (e.g., by the release of transgenic IL-12). In particular, CAR T cells with the inducible release of transgenic IL-12, so-called “fourth-generation” CAR T cells or TRUCKs (T cells redirected for universal cytokine mediated killing), ⁷² activate macrophages for killing of cancer cells not recognized by CAR T cells. ⁷³ In this context, NFAT inducible IL-12 release upon CAR signaling allows local IL-12 deposition without an increase in systemic IL-12 levels, which has a high risk of toxicity. The inducible release of a “payload” by CAR T cells upon engagement of targeted antigen provides a safety strategy for toxic compounds that accumulate, as long as the CAR target is present. Loss of CAR signaling results in switch-off of the IL-12 release. Other strategies were subsequently reported, including local administration of anti-Muc-16(ecto) CAR T cells with constitutive IL-12 release. ⁷⁴

The efficacious CAR T cell dose

CAR T cell dosing does not follow classical pharmaceutical roles, since the cells amplify in the patient after application. In the treatment of hematologic malignancies, doses of 10⁹–10¹¹ CAR T cells were usually applied. However, a single dose of 1.42 × 10⁷ CAR T cells was sufficient to induce a sustained antitumor response in the treatment of chronic lymphocytic leukemia. ⁷⁵ The increase in CAR T cell number in the peripheral blood is a good indicator for clinical efficacy in the treatment of hematologic malignancies. CAR T cell loss in peripheral blood results in relapse of the disease. ^76,77 However, in the case of solid tumors, peripheral blood is not the compartment of therapeutic action. CAR T cell accumulation in the tumor tissue is hard to record. So far, the effective CAR T cell dose in the treatment of solid cancer still needs to be thoroughly defined. Some efficacy was obtained with 10⁸ CAR T cells per kilogram. ¹⁸ Even a smaller dose of CAR T cells may be sufficient, assuming that the CAR T cells amplify when engaging cognate antigen in the infiltrated tumor lesion. In a recent trial, anti-CEA CAR T cells expanded in the peripheral blood and tumor tissue of those patients who received high doses of CAR T cells. However, the effect was only transient. A second infusion re-increased the CAR T cell number in the short term. ¹⁸ A low number of CAR T cells persisted in the cerebrospinal fluid after successful treatment of glioblastoma, ²⁰ as did the Her2-specific CAR T cells in the therapy of sarcoma. ²⁵ Persistence and antitumor activity may be improved by the co-expression of telomerase. ⁷⁸

Which antigen serves best in targeting solid tumors while avoiding off-tumor toxicities?

Tumor selective antigens are the preferred targets for a redirected T cell therapy, like the glycosylation variant of Muc1 for the treatment of adenocarcinoma ³⁴ or fetal antigens such as the fetal achetylcholin receptor for the treatment of rhabdomyosarcoma. ⁷⁹ The identification of more suitable antigens will broaden the application of CAR T cell therapy for solid tumors. However, their expression by healthy tissues demands additional strategies to obtain sufficient selectivity in tumor targeting. Such strategies include the combinatorial recognition of two antigens co-expressed by cancer cells but not by healthy cells. ⁸⁰ Two CARs, each recognizing one antigen, complement in signaling upon engaging the two antigens but do not activate the T cell upon engaging one antigen only (“and” logic combination). In an alternative setting, a co-expressed inhibitory CAR can block T cell activation when engaging an antigen on healthy tissues. ⁸¹

How should the CAR design be optimized for a specific target with a predictable functionality in the patient?

The CAR design so far relies on historical trial-and-error experience. The in vitro characterization of new CARs often fails to predict their in vivo functionality. There is no “one-fits-all” CAR; the various structures of targetable antigens require different structures of the extracellular CAR domain. High-throughput assays with higher predictive value in the characterization of T cell activation are demanding to make testing for new targets and diseases faster and more efficient. Beyond the classical CAR design, synthetic biology will help to create modular receptors that improve antitumor activities of engineered immune cells. As an example, a synthetic Notch-based receptor recognizes with the extracellular domain defined targets and with intracellular domain releases by proteolytic cleavage transcriptional activators or repressors that enter the nucleus for executing their function. ⁸² The universal applicability of “synNotch” receptors in the adoptive T cell therapy earns further exploration in the context of a redirected immune response toward solid tumors.

Which T cell subset performs best in the long term against solid tumors?

T cells in different maturation stages require different signals. On the other hand, co-stimulatory signals can be used to fix redirected T cells in a defined maturation stage. ¹⁰ Synthetic biology to redirect T cell maturation and metabolism by CARs will help to optimize the T cell performance after transfer to patients.

How can a high-quality T cell product be manufactured in a standardized process and for an increasing number of patients?

CAR T cell manufacturing is currently performed as a manual process in a few academic and commercial centers. In each center, a different production protocol with narrow and wide differences is applied. The exploration of the CAR T cell therapy in Phase II/III trials and the application to a broad variety of solid tumors requires a more robust and harmonized manufacturing process, which can be applied by non-specialists in various medical units.

Will “universal” CAR T cells outsmart “individualized” CAR T cells?

The current manufacturing of the individual T cell product for each patient is cost, time, and labor intensive. The limitation may be overcome by off-the-shelf treatment with pre-manufactured CAR T cells that are genetically engineered in order to avoid recognition by the patient's immune system. ⁸³

Will major CAR T cell therapy–associated adverse events be controlled in a standardized study design?

The currently accumulating clinical data and the management of adverse events are difficult to compare due to various differences in the trial protocols, the T cell product itself, the preconditioning of patients, and other relevant factors. The current situation demands a more rigorous standardization of the clinical exploration in the near future. Grading and management protocols are currently being explored in clinical practice in order to provide validated clinical procedures in case of CAR T cell toxicities. ⁷⁷

Will there be a specific preconditioning for each cancer?

Prior to T cell therapy, patients are subjected to non-myeloablative lymphodepletion in order to allow extensive CAR T cell amplification through providing space and growth factors and through attenuating suppressive factors. ⁸⁴ Lymphodepleting preconditioning in connection with adoptive cell therapy has also proven effective for the treatment of solid tumors, ⁸⁵ since the tumor milieu is modulated by depleting suppressor cells and by promoting the release of tumor antigens, which additionally activate a patient's endogenous T cells for an antitumor response. In this context, the standard preconditioning protocol needs some optimization for each cancer entity.

Can the immune network be manipulated in order to induce an acute inflammatory response and to recruit innate immune cells?

CAR T cells can impact the cytokine network in a specific fashion by releasing inducible cytokines through so-called TRUCKs (e.g., with the inducible release of IL-12). CAR T cells with inducible IL-12 induce a second wave of immune cell responses to eliminate those cancer cells that are not recognized by the CAR T cells. ^73,86 Such cytokine-mediated conversion of a protective Th2 response within the solid tumor to a more acute inflammatory response will make a CAR T cell attack more efficacious. A more recent report indicates that activation of the STING pathway initiates immune responses to eliminate tumor cells that are not recognized by the adoptively transferred lymphocytes. ⁴²