Engineered cellular immunotherapies in cancer and beyond

Impact of tolerance and rejection mechanisms on cellular therapies

The holy grail of cell therapy is achieving both long-term engraftment and durable remission. However, the initial success of allogeneic cell therapies has been tempered by the short persistence of engineered T cells⁷. Development of efficacious allogeneic engineered cell therapies will require elucidation and manipulation of cellular rejection circuits to permit long-term engraftment. Here, we consider lessons from the field of immunology that are related to tolerance in the healthy state and mechanisms of engineered cell rejection. See Box 1 for considerations related to HSC transplantation (HSCT), the first successful allogeneic cell therapy to enter the clinic.

Understanding the dichotomy of solid tumor and blood cancer

CAR T-cell therapies have achieved durable responses in hematological malignancies, as the overall response rate across many trials in patients with refractory leukemia, lymphoma and myeloma is 50–90% (ref. ²²). However, in the context of solid tumors, CAR T-cell therapies are still in the early stages of development, and their efficacy is limited. This dichotomy is mediated in part by characteristics of the bone marrow niche of hematological malignancies and the additional barriers posed by the solid tumor microenvironment (TME), such as diminished infiltration and function of CAR T cells, and more complex mechanisms of tumor escape²³.

The success of CAR T-cells in patients with leukemia, lymphoma and multiple myeloma would have been predicted in part by previous favorable results with allogeneic HSCT; in contrast, outcomes with allogeneic HSCT in patients with solid tumors have generally been disappointing²⁴. In blood cancers, the availability of lineage-restricted antigens (for example, CD19 and B-cell maturation antigen (BCMA)) provides targets with acceptable toxicity, given expected on-target but off-tumor toxicity on healthy cells. The exception to this is acute myeloid leukemia (AML), in which acceptable surface targets that spare healthy HSCs are lacking; yet by analogy to allogeneic HSCT, a potent antileukemic effect would be predicted, given the strong graft-versus-leukemia effect (Box 1) mediated by T cells and NK cells in this context. The use of T-cell receptor (TCR)-engineered T cells, which target intracellular tumor antigens presented on MHC-I molecules, appears highly promising in AML²⁵.

Several scientific questions remain open and will need to be addressed so that there are uniform responses to engineered cell therapies for all forms of hematopoietic cancer. The bone marrow niche was initially thought to maintain HSCs and has since been demonstrated to be the principal site of residence for human plasma cells and memory T cells²⁶. Recent studies indicate that the bone marrow environment is composed of multiple micro-niches with distinct metabolic features and stromal cell populations²⁷. Leukemia and myeloma cells survive in distinct niches, but engineered T cells may have differential trafficking and persistence in these niches—perhaps contributing to the superior durability of responses to CD19-directed versus BCMA-directed CAR T-cell therapy. Cancer-specific issues remain a challenge, such as the observation that AML is more immunosuppressive than acute lymphocytic leukemia (ALL)²⁸. This is likely related to the close relationship between AML stem cells and HSCs²⁹, and the fact that several mechanisms may have evolved to protect HSCs from immune attack.

One observation now apparent from CD19 CAR T cells is that they induce durable responses in patients with many forms of B-cell malignancies²². The persistence of CD19 CAR T cells is dependent on the CAR endodomains and can be reliably measured by assessing the duration of B-cell aplasia after infusion³⁰. In our initial studies in patients treated with the CD19-directed CAR T-cell therapy tisagenlecleucel, we demonstrated ongoing B-cell aplasia and persistence of functional CAR T cells for a decade³¹. Emerging data suggest that long-term persistence of CD19 CAR T cells is unusual when compared with other cell therapies with CAR T and TCR T cells that target other cell lineages. This is likely owing to the ongoing production by HSCs of pre-B cells in the marrow that express CD19, thus presenting a target for CAR T cells, even after sterile elimination of tumor target cells has been achieved. A lesson from this observation is that various boosting strategies may be able to enhance persistence of engineered cells in the treatment of non-B-cell blood cancers and solid tumors^32,33.

Solid tumors, including some forms of lymphoma, create a robustly immunosuppressive microenvironment composed of numerous cell types and extracellular matrix (ECM)^34,35. Infusions of TILs, while commonly effective in metastatic melanoma³⁶, are rarely effective in other solid cancers such as adenocarcinoma. The increased interstitial pressure, creating compressive forces and high tensile strength created by the ECM combine to thwart T-cell infiltration. Thus, new approaches to control solid tumor biomechanical forces (such as heparanse³⁷) and improve T-cell trafficking are required.

One implication of the biomechanical properties of the solid TME is that therapeutic cell doses used in many studies may be inadequate. Recent modeling of human and rodent circulatory systems has revealed that the delivery rate of CAR T cells to solid tumors is 10,000-fold greater in mice than in humans³⁸, providing a potential explanation for the disappointing results of cell therapy trials in patients with solid tumors. This work challenges allometric-based approaches in preclinical studies that are traditionally used in dosing considerations for solid tumors³⁹. While engineered T-cell therapy dosing has always required additional individualized features in the transition from mice to humans (for example, consideration of tumor mass and receptor expression), these findings suggest that improved dosing for solid tumors follows a less linear path than current algorithms suggest.

Solid tumors also harbor a strong tolerogenic environment, leading to T-cell exhaustion and dysfunction^40-42. In the presence of continuous stimulation by tumor antigens, T cells develop a progressive state of hypofunctionality characterized by distinct epigenetic, metabolic and phenotypic signatures⁴³. To assess this dysfunction in CAR T cells directed at pancreatic cancer, our group has recently developed an in vitro model of continuous antigen stimulation, enabling the investigation of dynamic temporal changes in the induction of T cell dysfunction⁴⁴. CD8⁺ CAR T cells displayed altered chromatin dynamics and upregulated many genes typically associated with NK cells. The NK-like T-cell transition was associated with the upregulation of ID3 and SOX4, and knockout of the transcription factors encoded by these genes prevented or delayed the state of T-cell dysfunction. New strategies to overcome the toxic effects of the solid TME are imperative as the field seeks to improve the therapeutic efficacy engineered cells for solid tumors.

Toxicities from engineered immune cell therapies

The toxicities of engineered cellular immunotherapies are distinct from those of other classes of therapies, such as cytotoxic chemotherapies, and, surprisingly, they are also distinct from those of immune checkpoint inhibitors⁴⁵. This realization has required the development of classification systems that are dedicated to the grading of adverse events resulting from engineered cell therapies^46,47.

As with all therapeutics, toxicities with engineered cells can be classified as on-target or off-target. The longest running clinical experience in the CAR T-cell field has been with CD19-directed CAR T cells, for which on-target toxicities, including B-cell aplasia, cytokine-release syndrome (CRS) and immune-effector-cell-associated neurotoxicity syndrome (ICANS), have been described⁴⁶. ICANS was initially thought to result from a cascade of off-target effects due to systemic inflammation and cytokine release⁴⁸; however, some aspects of the syndrome may be related to the expression of CD19 in pericytes in the central nervous system⁴⁹, an unexpected observation because CD19 was initially described as a B-lineage-restricted protein. An important and somewhat surprising recent finding suggests that the host microbiome has substantial effects on the efficacy and toxicity of CAR T cells (Box 2).

CRS and ICANS toxicities are now regarded as a class effect of CAR T cells, as these syndromes have been observed in patients treated with CAR T cells targeting CD19, BCMA, and many other cell surface structures⁵⁰. Increasing clinical work has led to the emergence of specific toxicity profiles for the various targets. Although both CD19- and BCMA-targeted CAR T cells trigger ICANS, the clinical aspects differ; for example, patients with multiple myeloma treated with BCMA-specific CAR T cells may develop a Parkinsonian syndrome⁵¹, potentially related to expression of BCMA in the central nervous system⁵⁰.

Unlike some other therapies (such as chemotherapy), the toxicities of which are predominantly off-target, the toxicity of engineered cells is predominantly on-target. Therefore, CRS and ICANS induced by engineered cells may be more a feature of increasingly potent therapies than a class effect. Further, these toxicities are generally not observed in preclinical studies owing to a lack of sufficiently representative animal models. Emerging single-cell technologies can precisely map target expression levels and thereby predict on-target, off-tumor toxicities with CAR T cells⁵⁰. However, this remains a challenging technical issue with TCR T cells, which can cause severe and unexpected toxicity⁵². An emerging lesson for the field is that the expression of targets in healthy patients may not be the same as in patients with cancer. The inflammation triggered by initially on-target recognition by engineered cells can trigger the subsequent induction of off-tumor target expression that was not present at the time of cell infusion⁵⁰. Real-world data indicate that, with increasing clinical experience, the severity of CRS and ICANS has decreased, likely reflecting earlier intervention and possibly the treatment of patients with lower disease burden^47,53.

Genotoxicity is an important challenge with engineered cell therapies. As engineered receptors are primarily introduced by viral-vector integration, safety concerns arise from the potential for insertional mutagenesis and cellular transformation. Clinical experience indicates that these risks are cell-type-specific. For example, the risk of transformation is higher with HSC than with T cells⁵⁴. Our group has found that CAR T cells can safely persist for a decade in patients^31,55, and despite the many thousands of patients who have been treated with engineered T cells globally, there are, to our knowledge, only two reported instances of T-cell transformation⁵⁶. Both these cases of CAR T lymphomas occurred in the setting of allogeneic HSCT, with a CAR T product manufactured using the piggyBac non-viral gene transfer technology⁵⁷. In a pilot trial of multiplex engineering of T cells using CRISPR–Cas9 technology, we found that chromosomal translocations, rather than off-target edits, were the primary safety concern⁵⁸.

The emerging clinical efficacy of engineered HSCs and engineered induced pluripotent stem cells (iPSCs; generated by reprogramming of adult cells) highlights an urgent need to further characterize and enhance the safety of engineered stem cells. In preclinical models, tumor formation from engineered iPSCs is increased when cells are derived from patients with some baseline genetic abnormalities⁵⁹. The convergence of multiplex human genome engineering with cell therapies illustrates the increasing need to refine genome-editing technologies and to develop assays for monitoring the safety of infused cellular products.

In addition to cell-type specificity, another issue facing the field is the broad safety of allogeneic cell therapies⁶⁰, which have substantially higher safety hurdles than do autologous cell therapies. Autologous T cells have specialized cell-intrinsic and cell-extrinsic safety features, such as control of TCR clonal abundance and resistance to oncogene transformation, which may provide an additional measure of safety lacking in the T cells, HSCs and iPSCs used for allogeneic therapies^61,62. It is reassuring that, at present, the collective safety of engineered cell therapies compares favorably to cytotoxic agents that have been the standard of care for the previous 50 years^53,63.

Combination cancer therapy

The feasibility of engineered cells of diverse types and diverse sources (autologous versus allogeneic) creates the permutational possibility of many combination strategies. As well as combining different cellular therapies, it is also possible to combine cellular therapies with other drugs or recombinant proteins. There is often confusion between the terms ‘combination therapy’ and ‘combination product’. Combination therapy refers to the common practice in medicine of the administration of, for example, several chemotherapy reagents to a patient in combination for treatment of a particular cancer. In contrast, a combination product is defined by health authorities as a product that involves a drug and/or a biologic and/or a medical device. Thus, combining an infusion composed of dendritic cells with one composed of T cells would be a combination therapy but not a combination product, because both components are considered biologics by the FDA. In contrast, engineered T cells injected into the liver with a specific catheter for treatment of cancer would be a combination product. For information on the FDA guidance and policy, see ref. ⁵. Below, we discuss potential combination strategies, some of which would be considered combination products from a regulatory perspective.

Combination therapies can alleviate the unique hurdles of engineered cellular therapies to enhance antitumor effects (Fig. 4). Indeed, preclinical studies of combination therapy incorporating oncolytic virotherapy^64-66, immune checkpoint blockade⁶⁷, bispecific antibodies⁶⁸ or small molecules with cellular therapy have demonstrated enhanced antitumor activity. Oncolytic adenoviruses (OAds) have engineered modifications that allow them to selectively enter tumor cells and replicate, resulting in tumor cell death through oncolysis (thereby reducing the modalities of tumor escape); they can also be modified to express therapeutic transgenes in the TME^69,70 and have demonstrated safety in clinical trials^71,72. Oncolytic virotherapy and CAR and TCR T-cell therapy are synergistic modalities for solid tumor treatment. Combinatorial studies have exploited the transgene delivery potential of OAds to drive tumor-antigen-specific expression (that is, CD19)⁷³. Preclinical studies published by our group have demonstrated that the new combination of OAds expressing tumor necrosis factor (TNF)-α and interleukin (IL)-2 with mesothelin-targeting CAR T cells was able to modulate the immunosuppressive TME and induce CAR-dependent and CAR-independent host immunity in pancreatic cancer⁷⁴.

Immune checkpoint blockade is another approach to overcome tumor-associated immune suppression⁷⁵ and revitalize T cells⁷⁶. In clinical trials, the combination of pembrolizumab with mesothelin-targeting CAR T cells further enhances the persistence and function of the latter in patients with malignant pleural diseases⁶⁷. Effective checkpoint blockade and expression of checkpoint molecules are key to the success of such combinatorial therapies; for instance, one study demonstrated that two CARs, one targeting EGFRvIII and the other targeting IL-13Ra2, preferred different checkpoint blockades within the same tumor model⁷⁷.

An alternative approach to checkpoint blockade is to redirect non-specific bystander T cells against tumors by engineering CAR T cells to produce bispecific antibodies, as shown in a preclinical study for the treatment of glioblastoma⁶⁸. Research by our team has also explored the combination of folate receptor alpha-targeting CAR T cells with OAds expressing a localized bispecific antibody targeting EGFR, which led to enhanced T cell activation and antitumor effects⁷⁸.

Overall, combination strategies involving engineered cell therapy for solid tumors show promising results; the interaction between the OAd and tumor immune cells and the selection of appropriate CAR signaling domains and of effective checkpoint blockade have all been highlighted as key determinants of antitumor efficacy.

New directions in cancer and beyond

Clinical considerations: bottlenecks and challenges

Autologous CAR T cells are personalized, living drugs that require a rethinking of the traditional manufacturing paradigm used for other biologics, such as recombinant proteins and vaccines¹¹¹. The current approach to manufacturing autologous CAR T-cell therapies relies on a centralized model, in which patient materials are cryopreserved, shipped to the manufacturing facility, genetically modified and expanded, cryopreserved a second time, and then shipped back to the clinical site^112,113. In addition to complex logistics, manufacturing of currently approved CAR T-cell products requires manual processing steps that must be conducted by highly trained personnel in good manufacturing practice (GMP)-certified facilities¹¹⁴. This cumbersome, centralized model presents a bottleneck for translation of preclinical candidates into phase 1 trials and will present challenges for commercialization of CAR T-cell and HSC therapies directed for indications with larger patient populations. It also presents challenges for translation in emerging markets that lack the required infrastructure for receipt and storage of cryopreserved materials.

Several approaches are being pursued to address the manufacturing challenges for autologous CAR T-cell therapy. Off-the-shelf CAR T-cell products—whether derived from allogeneic T cells or iPSCs—have received considerable attention owing to the elimination of the complications associated with a personalized therapy, and are currently being evaluated in numerous clinical trials^7,115. Although the use of non-patient starting material considerably reduces the cost per dose and simplifies logistics¹¹⁶, manufacturing is still highly complex owing to the need for genetic editing¹¹⁷, and durability is not yet on par with autologous therapies^7,118.

Another method to mitigating the manufacturing bottleneck is the use of in vivo cell engineering, in which a patient’s T cells are transduced or transfected directly within the patient. In vivo cell transduction eliminates the ex vivo manufacturing step, and if successful, would enormously simplify the delivery of CAR T cell therapy and allow for translation to a broader patient population. To date, there have been multiple demonstrations of in vivo CAR T-cell generation in mice using CD4− and CD8-targeted lentivirus^119,120, as well as an engineered adeno-associated virus¹²¹. Given the risks of off-target transduction, a potentially safer approach might be to transiently transfect T cells using T-cell-targeted lipid nanoparticles loaded with mRNA encoding the CAR. This approach was used to successfully treat cardiac fibrosis in mice using CD5-targeted lipid nanoparticles containing a mRNA encoding a FAP-directed CAR¹⁰³.

As with food products, cell therapies may be broadly classified as natural or genetically modified. The policy of the FDA is to regulate cell and tissue products that are not homologous and are more than minimally manipulated¹²². Broadly speaking, the homologous use of cells refers to regenerative medicine applications that repair or replace damaged cells and tissues. Cell products that are more than minimally manipulated have altered characteristics (such as being genetically engineered), another feature that triggers full regulation by health authorities. These distinctions have major implications for clinical medicine and the development of cellular therapeutics. The prospect of the cure for hemoglobinopathies with genetically engineered HSCs is on the horizon, offers hope to afflicted patients, and has major implications for the pharmaceutical industry and healthcare policies. The manufacturing of genetically modified autologous HSCs will require higher standards and more stringent regulation than for red blood cells derived from healthy donors.

Summary and perspectives

Transplant studies have laid the groundwork for our current understanding of immune tolerance and highlight the underlying complexities of cellular therapeutics today. The expansion of the cellular immunotherapeutic field to numerous diseases, cell sources, and cell types indubitably reflects the careful integration of engineering innovations with a deepened understanding of underlying mechanisms of action.

Lessons learned from cancer immunotherapies have set the stage for application of engineered cellular therapy to autoimmune diseases, genetic disorders, infectious diseases and regenerative medicine. However, we need to identify more selective targets, leveraging our understanding of peripheral tolerance. TCRs targeting shared mutations in driver oncogene pathways hold promise for treatment of solid tumors⁸⁸, as do CARs that bind to tumor-specific glycans, splice variants, peptide central targets presented by human leukocyte antigens^123,124. Advances in synthetic protein engineering and computational modeling also offer novel opportunities to increase CAR T-cell specificity in the TME^125,126.

Going forward, as the field implements allogeneic engineered cell therapies, we are likely to face obstacles in engraftment and persistence. However, as we break down their mechanisms, these hurdles are likely to be overcome by novel applications of synthetic biology and increased precision in genome editing. Science, medicine and the public need to embrace these developments to realize the full potential of engineered cell therapies.