Engineered T Cells: The Promise and Challenges of Cancer Immunotherapy

I. Preface

The immune system evolved to distinguish non-self from self to protect the organism. As cancer is derived from our own cells, defending ourselves from dysregulated cell growth presents a unique challenge. This is compounded by mechanisms of immune evasion and suppression that cancers themselves have developed. Natural cancer-specific immune responses occur ¹, and most often demonstrate impaired function. The modern genetic toolbox allows for creation of an immune system with enhanced anti-cancer function. Recent advancements have yielded stunning results in patients with relapsed/refractory hematologic malignancies, electrifying the field. Engineered T cells, so-called “living drugs” represent a new paradigm in anti-cancer therapy.

II. Introduction

Since Medawar and colleagues performed their seminal work², it has long been recognized that adoptively transferred T cells have potential to target and destroy cancer cells. In many cases, however, transferred polyclonal T cells lacked sufficient specificity or numbers sufficient to control tumor. T cells genetically engineered to express novel receptors have enhanced tumor specificity. In addition, advances in ex vivo expansion allow for production of clinically relevant doses of these therapeutic cells. Engineered T cells have produced unprecedented results in the clinic.

The earliest engineered T cell trials relied on expression of cloned T cell receptors (TCR) with targeted affinity. A TCR may recognize either intracellular or extracellular antigen in the context of MHC. When designing a TCR to target tumor, having the option to target intracellular tumor antigen may be advantageous. On the other hand, many tumors downregulate MHC expression, potentially masking their presence from a TCR engineered T cell. More recently, artificial receptors such as chimeric antigen receptors (CAR), combining B cell receptor derived and T cell receptor domains, have been employed to enhance T cell specificity (Figure 1). A CAR is commonly composed of (1) a specificity-conferring extracellular antibody single chain variable fragment (scFv), (2) a CD3z domain and (3) one or more intracellular costimulatory domains. CAR design has evolved over years to enhance efficacy and safety in particular immunologic settings (Figure 2). Unlike TCRs, CARs allow highly specific targeting of antigen in an MHC-independent fashion. Until recently, however, CAR T cell targets were limited to extracellular tumor antigens.

Figure 1.

Comparing basic structure of engineered T cell receptors and chimeric antigen receptors. Endogenous T cell receptors include paired alpha and beta chains associated with delta, epsilon, gamma, and signaling zeta chains. Most transgenic engineered T cell receptors also rely on recruitment of endogenous downstream signaling molecules such as LAT and ZAP70 to transduce the activation signal. Both endogenous and transgenic T cell receptors see intracellularly processed antigens that must be presented in the context of the Major Histocompatibility Complex and require costimulatory signals (not shown) for complete T cell activation. Chimeric antigen receptors, on the other hand, lack alpha and beta chains. The extracellular portion of a chimeric antigen receptor consists of single chain variable fragments derived from antibody heavy and light chain variable domains. Typically these are then fused to a transmembrane domain, an intracellular costimulatory domain and an intracellular zeta chain domain. Again, chimeric antigen receptors must recruit endogenous downstream signaling molecules to transduce activating signal, but costimulation is provided in cis and in response to the same activating signal. Chimeric antigen receptors see surface antigens independent of the MHC and are therefore not tissue type restricted.

Figure 2. CAR Design and Evolution.

CARs target surface antigens in an MHC-independent fashion and consist of an extracellular binding domain, hinge domain, transmembrane domain, and intracellular signaling domains. The first clinical trials tested CARs that had a binding domain composed of native CD4 that bound to gp120 on HIV-infected cells^183,184, with a single signaling domain composed of the CD3ζ chain^185–187. CAR’s with an extracellular domain composed of antibody single chain fragment variable portions were first reported by Kuwana¹⁸⁸ and later Eshhar and colleagues^189,190. Second generation CAR’s incorporating CD28 as a costimulatory domain were first developed by Roberts (US Patent 5,686,281) and reported by Finney¹⁹¹, and those incorporating 4-1BB as a costimulatory domain by Finney^192,193 Imai¹⁹⁴, and then others^195,196. CAR’s incorporating 3 or 4 signaling domains, so called “third and fourth generation”, have also been developed and are beginning clinical trials^71,197,198.

Adoptive transfer of T cells expressing engineered receptors has shown enormous promise in humans. CD19-directed CAR T cells (CART19) has generated complete and durable remissions in patients with refractory and relapsed B cell malignancies^3–6 NY-ESO-1–specific TCR–engineered T cells have generated clinical responses in patients with advanced multiple myeloma and synovial cell sarcoma^7,8. With the proof of concept established, engineered T cells have matured as a therapeutic option to treat malignancies. Building on this foundation, the field is broadening indications for current therapies, exploring, new targets, and employing the new techniques to create even safer and more effective therapies. We describe here some of the most recent and promising advances in engineered T cell therapy with a particular emphasis on what the next generation of T cell therapy will likely entail

III. Clinical trials with engineered T cells directed against B cell malignancies

B cell malignancies are the most common tumor type to be targeted by engineered T cells. There are a number of reasons for this. B cell malignancies are relatively common and express several conserved cell surface markers. Acquired B cell aplasia is a treatable condition with mild to moderate long term consequences. B cell tumors are often easily accessible by circulating immune cells, giving engineered T cell early and ample access to target cells. Finally, the use of engineered T cells to treat B cell tumors, specifically B cell acute lymphoblastic leukemia (B-ALL) has shown the greatest promise in the field to date.

The extracellular glycoprotein, CD19 is the most common B cell target for engineered T cell therapies (Table 1a). CD19 is an expressed on both benign and most malignant B cells with extremely limited non-B cell expression⁹. Clinical response to CD19 targeted T cell therapy, particularly in patients with B ALL, has been unprecedented. Several groups have reported response rates to CD19 targeted CAR T cells in over 80% of patients with relapsed and refractory B cell ALL^3–6. Several clinical trials have confirmed CD19 directed CAR T cells are effective for refractory non-Hodgkin lymphoma^10,11. Others have targeted rare CD19 positive plasma cell myeloma stem cells, demonstrating disease eradication at 12 months post transfer of CD19 targeted CAR T cells¹². Further, engineered T cells have been shown to persist for more than a decade after transfer¹³, suggesting that adoptively transferred T cells may be truly a “living drug”.

Table 1.

Examples of CART cell clinical trials (clinicaltrials.gov, citations of clinical results where available)

a. CD19 or CD20-directed trials
Target	Indication	Reference
CD19 or CD20	Leukemia	NCT01044069
		NCT01860937
		NCT02146924
		NCT02228096
		NCT02435849
		NCT02028455
		NCT02614066
		NCT02625480
		NCT01747486
		NCT02030847
		NCT02535364
		NCT01683279
		NCT01475058
	Leukemia or lymphoma	NCT02443831
		NCT02529813
		NCT02546739
		NCT01087294
		NCT01430390
		NCT01593696
		NCT01626495
		NCT01853631
		NCT02050347
		NCT02456350
		NCT01865617
		NCT02081937
		NCT02132624
		NCT02349698
		NCT02537977
		NCT01864889
		NCT01029366
	Lymphoma	NCT02431988
		NCT02631044
		NCT02445248
		NCT02277522
		NCT02624258
		NCT00924326
		NCT01493453
		NCT01840566
		NCT02134262
		NCT02247609
		NCT02348216
		NCT02030834
		NCT01318317
		19
	Multiple myeloma	NCT02135406

b. Additional targets for hematologic CART cell trials
Target	Indication	Reference
CD22	B cell malignancy	NCT02588456
		NCT02315612
CD23	B cell malignancy	173
Kappa light chain	B cell malignancy	NCT00881920
CD5	T cell malignancy	174
CD30	Lymphoma	NCT02259556
		NCT02274584
CD70	Lymphoma	175
CD38	Multiple myeloma	176
CD138	Multiple myeloma	NCT01886976
BCMA	Multiple myeloma	NCT02546167
		NCT02215967
CD33	Myeloid malignancies	NCT01864902
		36
		177
CD123	Myeloid malignancies	NCT02623582
		NCT02159495
CD44v6	Various hematologic malignancies	34
CS1		NCT02203825
ROR1		NCT02194374

c. Solid tumor CART cell trials
Target	Indication	Reference
EGFR	EGFR-positive solid tumors	NCT02331693
		NCT01869166
EGFRvIII	Glioblastoma	NCT01454596,
		NCT02209376
		63
GD2	Neuroblastoma, Ewing’s sarcoma, osteosarcoma	NCT01822652,
		NCT01822652,
		NCT02107963
IL13Rα2	Glioma	NCT02208362
HER2	HER2+ solid tumors	94,97
Mesothelin	Mesothelioma, Pancreatic Cancer, Ovarian Cancer	NCT02159716,
		NCT02414269
		NCT01897415
		NCT02580747
		NCT02465983
		128
PSMA	Prostate cancer	NCT01140373
FAP	Malignant Pleural Mesothelioma	NCT01722149
GPC3	Hepatocellular Carcinoma	NCT02395250
cMet	Breast cancer	NCT01837602
Muc16	Ovarian cancer	178
CEA	Lung, Colon, Gastric, Breast, Pancreatic Cancer	NCT02349724
Lewis-Y	Solid tumors, myeloid malignancies	NCT01716364
Folate receptor β	Ovarian cancer	179
Muc1	Hepatocellular Carcinoma, NSCLC, Pancreatic Carcinoma, Triple-Negative Invasive Breast Carcinoma	NCT02617134
		NCT02587689

While frequently expressed, CD19 may be downregulated ¹⁴ or mutated¹⁵ in tumor cells, rendering these cells resistant to CD19 directed therapy. Relapse rates in ALL reported at the 2015 American Society for Hematology meeting ranged from 18–36%, with the majority of these (66–100%) due to CD19 negative relapses. Alternative markers, such as CD20 and CD22 are also frequently expressed in non-Hodgkin lymphoma¹⁶ and B-ALL¹⁷. Tolerability of anti-CD20 monoclonal antibodies (Rituximab) supports safe use of an anti-CD20 T cells. While shown to be safe, autologous CD20-targeted CAR T cells failed to persist in vivo in early trials¹⁸. Inclusion of dual costimulatory domains (CD28 and 4-1BB) enhanced CART20 persistence in patients with indolent B cell and mantle cell lymphoma¹⁹. CART20 cells could be detected up to one year post transfer and two of the three patients treated had progression free survival at 24 month follow up. Preclinical data have demonstrated CD22-directed CART cell anti-tumor capacity¹⁷ similar to that of CART20. Multiple phase I clinical trials using CART22 products are underway (Table 1b).

During B cell development, a given cell will express either kappa or lambda light chains. In humans, the ratio of kappa to lambda positive cells ranges from 4:1 to 0.5:1. When the ratio exceeds these limits, it is likely that a clonal, light chain restricted population has expanded. Light chain targeting by CAR T cells is a particularly attractive approach because, unlike CD19, light chain targeted CART cells have the potential to leave 20–80% of B cells and plasma cells untouched. In addition, kappa light chain deficiency does not appear to be associated with an increased risk of infection²⁰. Kappa targeted CAR T cells have been shown to generate specific cytotoxicity in response to kappa positive tumor cell lines²¹. These cells are now in use as part of a phase I clinical trial to investigate safety and efficacy in humans (Table 1b).

Engineered T cells designed to target B cell malignancies serve as proof-of-concept that ex vivo modified T cells can eradicate tumor in humans. Highly effective, these engineered T cells have shown the ability to serially kill malignant B cells, suggesting that transfer of very few cells may be sufficient to achieve remission^22,23. Observations in treating B cell malignancies with engineered T cells have been both instructive and challenging. When the raw material for a drug is derived from a patient’s own cells, variability is unavoidable. A strategy to reduce variability may include enriching for central memory T cells, or to set the ratio of CD4 to CD8 T cells in the engineered product as 1:1 ^24,25. Rapid tumor clearance and associated immune activation indicates a need for careful management in patients with high tumor burden and developing approaches with control of in vivo function. Efficacy in treating different lymphoma histologies and the different response rates in CLL compared to ALL suggests that specific disease factors may need to be considered to enhance efficacy. Ultimately, the successful eradication of B cell malignancies by engineered T cells has provided the foundation upon which the field of adoptive T cell therapy is expanding.

IV. Moving beyond B cells

A. Novel T cell target selection for non-B cell haematological malignancies

Several T cell therapy targets in non-B cell malignancies are under investigation (Table 1b). Upon terminal differentiation, plasma cells downregulate many common engineered T cell targets such as CD19, CD20, CD22 and surface light chains. Therefore, to effectively target malignant plasma cells in conditions such as plasma cell myeloma, new targets must be considered. One such target, B cell maturation antigen (BCMA) is analogous to CD19, in that it is expressed in most cases of plasma cell myeloma and is not expressed on non-plasma cells^26,27. Unlike CD19, however, BCMA signaling can induce plasma cell proliferation and survival ^28–31. Therefore, plasma cell myeloma downregulation of BCMA to escape engineered T cell detection could limit tumor progression. BCMA-CART cells eradicate human multiple myeloma cell lines in xenograft models ³². Two phase I trials are currently investigating the feasibility, safety and efficacy of BCMA-CART cells against multiple myeloma (Table 1b). Cancer testis antigens, such as NY-ESO-1, are also upregulated on plasma cell myeloma cells and can be highly immunogenic³³. T cells engineered to express an affinity-enhanced, NY-ESO-1-specific TCR have been used to treat patients with advanced plasma cell myeloma. Clinical responses were observed, suggesting great promise in an otherwise incurable disease ⁷.

Treatment of myeloid malignancies has not changed substantially over the past decades; however, engineered T cell therapy may change this. Myeloid surface markers upregulated on malignant cells (eg. CD33, CD123, and CD44v6) are under investigation as T cell therapy targets ^34–36. Importantly, CD33 and CD123 are expressed on normal hematopoietic stem cells. Therefore, targeting these markers risks ablation of the hematopoietic stem cell compartment- an intolerable on-target, off-tumor effect. While preclinical animal studies are equivocal on the question of in vivo myeloablation ^35,37–39, some have proposed combining anti-myeloid T cell therapy with bone marrow transplant as salvage³⁵. A phase I clinical trial is investigating the use of CD123 targeted CAR T cells in treating myeloid malignancies (Table 1b).

Interestingly, some potential hematologic targets are not unique to hematologic malignancy. For example, receptor tyrosine kinase-like orphan receptor (ROR1) is a transmembrane glycoprotein expressed on embryonal tissue and aberrantly on many adult malignant tissues. Aberrant cell surface expression of ROR1 has been described in CLL, mantle cell lymphoma, B-ALL, and numerous types of solid tumors ^40–43. ROR1 expression appears to enhance cell survival and prevention of apoptosis, suggesting that tumor downregulation of ROR1 may confer a proliferative disadvantage^44,45. ROR1-targeted T cells generate cytotoxicity against human ROR1 positive B cell malignancies and sarcoma in preclinical studies ^43,46,47. Importantly, despite low level ROR1 expression in non-tumor tissue, transfer of ROR1-CART cells into nonhuman primates did not cause overt toxicity ⁴⁸. Autologous ROR1 directed CART cells are currently being investigated for safety and feasibility in a phase I trial to treat patients with CLL (Table 1b) ⁴⁹.

B. In search of specific solid tumor engineered T cell targets

Monoclonal antibodies directed against solid tumor antigens have shown promise in early clinical trials, though limited tissue penetration has restricted clinical responses ⁵⁰. Endogenous tumor infiltrating lymphocytes (TILs) have long been known to generate anti-tumor response and confer positive prognosis, however tumor immunosuppression prevents tumor clearance ^51–53. Given the ability of modified T cells to actively traffic to nearly every site in the body ^54,55 and to overcome tumor evasion ⁵⁶, engineered T cells possess unique potential to eliminate solid tumors. Selecting appropriate solid tumor targets, however, can be challenging. Most potential solid tumor targets are non-specific, being expressed on healthy tissue as well. At the same time, off-tumor effects may be less tolerable than the B cell aplasia associated with hematologic CART cell therapies. Different levels of surface marker expression may allow engineered T cells to preferentially target malignant cells ^47,57, however, low level expression on healthy tissue inherently increases the risk of on-target, off-tumor adverse effects. Those solid tumor targets that are highly specific for tumor tissue are rarely expressed throughout the tumor. T cell therapy directed against a tumor target that is not present on all tumor cells runs the risk of selecting for target-negative tumor outgrowth. To date, most solid tumor targets of engineered T cell therapy rely on overexpression in tumor tissue and are relatively non-specific (eg. GD2, IL13Ra, mesothelin, HER2). Nonetheless, a wide variety of potential solid tumor targets are under consideration (Table 1c, Table 2).

Table 2.

Examples of engineered TCR clinical trials (clinicaltrials.gov)

Target	Indication	Reference
MAGEA3	Various solid tumors	NCT02153905
		NCT02111850, 180
MAGEA4	Various solid tumors	NCT02096614
NY-ESO1 +/−LAGE-1 +/− MAGE3/6	Various solid tumors	NCT02366546
		NCT02457650
		NCT02070406
		NCT00670748, 8,181
	Various malignancies	NCT01697527
	Melanoma	NCT01350401
	Metastatic non-melanoma	NCT01967823
	Mesothelioma, NSCLC	NCT02408016
	Multiple myeloma	NCT01892293
	Multiple myeloma	NCT01352286, 7
WT1	Myeloid malignancy	NCT01621724
		NCT02550535
		NCT01640301
MART1	Metastatic melanoma	NCT02654821
		NCT00910650, 182
HPV16-E6	HPV associated cancers	NCT02280811
Thyroglobulin	Metastatic thyroid cancer	NCT02390739
Melanoma antigen tyrosinase	Melanoma	NCT01586403
CEA	Various solid tumor	NCT01723306

Target selection for T cell treatment of glioblastoma multiforme illustrates the variety of approaches available. Epidermal growth factor variant III (EGFRvIII), is a mutant form of EGFR, resulting from a coding sequence deletion, which generates a novel extracellular epitope. Unlike many other solid tumor markers, expression of EGFRvIII appears to be entirely limited to malignant tissue and is found in approximately 30% of cases of GBM. On the other hand, interleukin 13 receptor alpha 2 subunit (IL13Ra2) is also expressed in many cases of GBM (44–100% depending on methodology) ^58,59. Despite being present in more cases, IL13Ra2 is expressed on non-neoplastic tissues at either reduced ^59,60 or comparable levels ^58,61. Engineered T cell therapy targeting either EGFRvIII or IL13Ra2 has shown promise. EGFRvIII-CAR T cells have been shown to control growth of EGFRvIII positive human glioblastoma in preclinical models ^54,62. Phase I and I/II trials are now being conducted to determine the safety and efficacy of EGFRvIII CAR T cells in treating malignant gliomas ⁶³ (Table 2). Despite unclear non-neoplastic expression of the target, intracranial administration of IL13Ra2 CART cells has been shown to be safe and well tolerated in patients with GBM ⁶⁴. IL13Ra2 CART cell treatment of IL13Ra2 positive brain tumors is under investigation in an active phase I clinical trial (Table 2).

Ganglioside GD2, a glycosphingolipid, is expressed on both a variety of malignant and benign tissues. GD2 is highly expressed on neuroectodermal tumors (eg. neuroblastoma, melanoma, glioma), sarcomas, brain cancer, and small cell lung cancer ^65–67. Low level expression of GD2 is also found on non-malignant neurons, skin, melanocytes and peripheral nerves ⁶⁸. Anti-GD2 monoclonal antibodies have shown efficacy in the setting of minimal residual disease suggesting that enhanced immune mediated tumor clearance may be effective in non-minimal residual disease settings ^69,70. Anti-GD2 monoclonal antibodies have significant adverse effects including neuropathic pain, potentially due to targeting of GD2 expressed on peripheral nerves. Anti-GD2 CAR T cells are capable of generating an anti-tumor response in preclinical models ^71,72 and in phase I clinical trials ^73,74. Patients with active GD2 positive neuroblastoma were treated with GD2 CART cells and some experienced durable remission regardless of disease status at the time of infusion. Importantly, despite low-level GD2 expression on benign tissue, GD2 CART cells were well tolerated with no dose limiting toxicities observed ^73,74. These studies were done with first generation CAR T cells, and whether toxicity will be acceptable with more potent CAR designs remains to be determined. A phase I clinical trials is investigating GD2 CART cells in patients with a variety of GD2 positive malignancies (Table 1c).

Mesothelin is a 40-kDa cell surface glycoprotein expressed on normal pleura, pericardium and peritoneum ^75,76 and overexpressed on a variety of solid tumors including pancreatic cancer, mesothelioma and subsets of lung esophageal ovarian and breast cancers ^77–83. The physiologic function of mesothelin is unknown, however, some evidence suggests that in malignancy, the molecule is involved in metastasis making this an attractive therapeutic target ⁸⁴. Intra- and extrathoracic human mesothelioma lesions are eradicated by mesothelin targeted CART cells in preclinical models ^85,86. These findings confirm that mesothelin specific CART cells can traffic to appropriate body compartments and home to tumor while retaining anti-tumor effector function. The ability to localize while retaining function is essential for solid tumor eradication, in particular when targeting tumors in immune privileged sites or within a suppressive tumor microenvironment. Preliminary data from human clinical trials have shown mesothelin specific CART cells to be well tolerated and potentially effective against ovarian cancer, mesothelioma and pancreatic cancer ^86–88. Importantly, despite broad, low level mesothelin expression on benign tissue, on-target off-tumor toxicities have not been observed to date. However, these studies were done with a CAR comprised of a murine scFV, resulting in limited persistence of the CART cells. Whether CARTs using a fully human scFV would have durable persistence and acceptable toxicity remains to be determined. Numerous phase I studies are being conducted to further demonstrate the safety and efficacy of mesothelin CART cells.

Human epidermal growth factor receptor 2 (HER2) is a transmembrane tyrosine kinase receptor expressed on normal human gastrointestinal, respiratory, urinary tract, skin, breast and placental tissue. HER2 is also overexpressed in a variety of breast, head and neck, and nervous system cancers. Millions of women with breast cancer and other tumor histotypes have benefited from anti-HER2 antibody ⁸⁹, however monoclonal antibody localization and penetration have limited clinical response ^90,91. Further, HER2 expression on some malignancies, (eg. HER2+ sarcomas) is below the monoclonal antibody-mediated immune activation threshold ⁹². HER2-targeted T cells may overcome these limitations by actively trafficking to tumor sanctuaries and triggering in response to low target density.

HER2 targeted T cell therapy, however, also serves as an example of the challenge posed by low level benign tissue target expression. Lethal pulmonary toxicity was observed in a patient with HER2 positive colon cancer who was treated with 10¹⁰ HER2-CART cells⁹³. It is believed that low level HER2 expression on pulmonary endothelium triggered this response. This type of reaction was not seen with HER2 monoclonal antibody therapy. This not only suggests that HER-CART cells are able to activate in response to lower levels of target, but also confirms that monoclonal antibody data are insufficient to predict safety to T cell therapy. Subsequent HER2-CART cell trials have proceeded cautiously by using ultra-low doses of cells. In addition, lymphodepletive preconditioning, which removes endogenous competitors for growth factors was avoided, slowing initial in vivo response. Of note, despite these potential limitations, an anti-tumor response was still detected in patients with HER2 positive sarcoma treated with HER2-CART cells ⁹⁴. These findings are even more striking when one considers the relatively low expression of HER2 in these cases of sarcoma. Preclinical studies demonstrate that HER2-CART cells have efficacy in clearing HER2 positive GBM and medulloblastoma ^95,96. Alternatively, HER2 CART cells may be manufactured from CMV-specific autologous T cells, yielding a product that will engage CMV + target cells by the TCR or HER2 + target cells by the CAR, of potential benefit when CMV is also expressed in the tumor microenvironment. Preliminary clinical trials results have demonstrated safety and modest clinical responses associated with these bispecific CART cells ⁹⁷.

A main thrust in the search for new cell targets lies in discovery of methods to target “neo-antigens” with TCRs that are particular to each mutated tumor. T cell epitopes associated with impaired peptide processing (TEIPP) antigens are unique T cell epitopes resulting from impaired peptide processing. TEIPP are significant because they are derived from broadly expressed self Ag, and similar to other antigens such as viral antigens, are not restricted by central tolerance⁹⁸. TEIPP don’t require the cellular transporter associated with antigen processing (TAP). Accordingly, tumors that have defects in TAP (such as 30 to 50% of ovarian cancer) have relatively more of these peptides at the cell surface because there is less competition from endogenous natural peptide epitopes.

C. Conceptual evolution in redirected T cell targeting in solid tumours

Clinical feedback has allowed re-evaluation of some basic tenets of CART cell targeting. Whereas prior approaches emphasized efficacy, minimization of off-tumor effects is now the primary driver of target selection when potent CARTs are used. CART cells are able to respond to minimal target expression, making target specificity particularly important. Off-tumor effects can be lethal and currently limit clinical applications, particularly with regard to solid tumor therapy. While intracellular tumor markers have been classically excluded as potential targets, recent work forces their reconsideration. CARs, by definition, are designed with affinity to an extracellular ligand. However, human antibodies with affinity for an epitope of Wilms tumor antigen 1 presented by HLA-A2 have been developed ^99,100. Further modifications of these antibodies have enhanced antibody-dependent, cell mediated cytotoxicity ¹⁰¹. Thus, where TCRs had advantages of recognizing intracellular antigens presented by Major Histocompatibility Complex to T cells, antibodies that can be incorporated into CAR constructs have now been generated. It is likely that antibodies to additional intracellular antigens presented by MHC will be generated in the future. Inclusion of these intracellular markers as potential targets could improve therapeutic specificity and therefore safety, however the potential for off-target recognition of this class of CARTs remains to be tested.

As the repertoire of potential targets expands, better understanding of cancer biology may allow more precise targeting. Cancer stem cell (CSC) populations have now been characterized in many cancers. Subpopulations of tumor cells with “stem-like” properties have been identified in ovarian cancer ¹⁰², glioblastoma multiforme ^103–105, multiple myeloma⁷, and acute myeloid leukemia ^106,107 among others. It follows that elimination of the CSC subpopulation is crucial to achieve durable remission. Therefore, precise targeting of these subpopulations may be critical to prevent relapses. Several varieties of CART cells have been shown to eliminate CSC subpopulations along with other tumor cells ^{12,96,108–110}. Future strategies to target CSC subpopulations may maximize clinical effect while minimizing off-tumor effects.

Finally, new findings force us to rethink what it means for a T cell therapy to be “specific” for a target. Two step approaches are being employed, wherein T cells are engineered to express a receptor with affinity for a non-specific molecule and this molecule is then fused to a specific and targetable agent. Preclinical models based on CART cells with affinity for either a bispecific small molecule ¹¹¹ or the Fc gamma receptor ¹¹² have shown promise. The advantages of this approach are that the targetable agent may control response and allow for simultaneous multivalent targeting by a single population of engineered T cells. Alternatively, others have generated T cells specific for tumor antigen, that upon binding, produce cytokines that are intended to recruit endogenous immune cells and mediate tumor clearance. T cells redirected for universal cytokine killing (TRUCKs) have been engineered to express inducible or constitutive IL12, which induces innate immune anti-tumor response and alters tumor immunosuppression ^113,114 (Figure 4a). TRUCKs have the ability to enhance tumor penetration. Finally, despite a great deal of effort to define engineered T cell specificity ex vivo, the specificity of these cells may evolve upon in vivo stimulation by tumor. After EGFRvIII positive tumor clearance by EGFRvIII-CART cells, mice have been shown to be resistant to subsequent EGFRvIII negative tumor challenge. This demonstrates that engineered T cells have the ability to generate immunity to non-target tumor antigens after in vivo anti-tumor response ^62,63. Together, these findings serve as a reminder that an engineered T cell anti-tumor response is a dynamic process that relies on both cell design and host factors.

Figure 4. New CAR Models and Concepts [Au; please also expand the examples here to non-CAR T cells.].

A) T cells redirected for universal cytokine killing (TRUCKs) co-express a CAR and an anti-tumor cytokine. Cytokine expression may be constitutive or induced by T cell activation (eg. IL-12). Targeted by CAR specificity, localized production of pro-inflammatory cytokines recruit endogenous immune cells to tumor sites. B) Universal CAR T cells are engineered to no longer express endogenous TCR and/or HLA molecules preventing GVHD or rejection respectively in the allogeneic setting. C) Self driving CARs co-express a CAR and a chemokine receptor, which binds a tumor ligand (eg. CCR2b-CCL2), thereby enhancing tumor homing. D) CAR T cells engineered to be resistant to immunosuppression (Armored CARs) may be genetically modified to no longer express a variety of checkpoint molecules (eg. CTLA4, PD1), with a checkpoint switch receptor, or may be administered with a monoclonal antibody checkpoint blockade. E) A self-destructing CAR may be designed by using RNA delivered by electroporation to encode the CAR^86,128. Alternatively, inducible apoptosis of T cell as shown in the right hand section of panel G may be achieved based on ganciclovir binding to thymidine kinase in gene modified lymphocytes²⁰³ or the more recently described system of activation of human caspase 9 by a small molecule dimerizer^16,204. F) A Conditional CAR T cell is by default in the “off” position, until the addition of a small molecule to complete the circuit turning the CAR to the “on” position^111,154. Alternatively, a receptor may be delivered to a T cell that serves as an adaptor to subsequently administered secondary antibodies directed at target antigen¹¹². G) Marked CAR T cells express a CAR plus a tumor epitope to which an existing monoclonal antibody agent binds. In the setting of intolerable adverse effects, administration of the monoclonal antibody clears the CAR T cells and alleviates symptoms with no additional off-tumor effects. H) A tandem CAR (TanCAR) T cell expresses a single CAR consisting of two linked scFvs that have different affinities fused to intracellular costimulatory domain(s) and a CD3ζ domain. TanCAR T cell activation is achieved only when target cells co-express both targets. I) A dual CAR T cell expresses two separate CARs with different ligand binding targets; one CAR includes only the CD3ζ domain and the other CAR includes only the costimulatory domain(s). Dual CAR T cell activation requires co-expression of targets on tumor. J) A safety CAR (sCAR) consists of an extracellular scFv fused to an intracellular inhibitory domain (eg. CTLA4 or PD1). iCAR T cells co-expressing a standard CAR become activated only when encountering targets cells that possess the standard CAR target but lack the iCAR target.

V. Building Smarter Redirected T Cells

A. Novel gene transfer and editing

Current gene modification techniques used to produce engineered T cells must balance efficiency, safety and cost. Due to robust efficiency, viral vector-based protocols are the most frequently employed methods of T cell transduction ¹¹⁵ (Figure 2). Both retroviral and lentiviral vectors are able to deliver moderate sized payloads, which integrate into host genomes and consistently express the construct. Lentiviral vectors are preferred to retroviral vectors as they may integrate in non-dividing human primary cells and confer a decreased risk of insertional oncogenesis, at least as observed in hematopoietic stem cells^116–120. However, to date, no lentiviral transduced engineered T cell products have been reported to demonstrate insertional mutagenesis despite hundreds of treated patients.

DNA transposons have been used to efficiently insert gene cassettes in the host genomic DNA ^115,121,122. Transposon-based systems, such as the Sleeping Beauty (SB) transposon system have been developed to successfully produce CART cells of suitable quality for clinical investigations ^123,124. Safety and efficacy of transposon-engineered CART19 cells are currently under investigation ¹²⁵ (NCT00968760). Alternative approaches such as, the piggyBac (Systems Biosciences Inc.) transposon system have also been used to generate several types of CART cells (eg. CART19 cells ¹²⁶, EBV-specific HER2 CART cells ¹²⁷). With viral and non-viral methods of integration, a theoretical risk of insertional oncogenesis remains.

Along with advances in electroporation techniques, efficiency of non-integrating, non-viral methods of gene modification are showing promise as an alternative or complement to viral vector based methods. Electroporation also allows provision of non-integrating constructs, such as mRNA, which eliminates risk of insertional oncogenesis. For these reasons electroporation of engineered T cells is an emerging strategy for gene modification and interrogation of new tumor targets. Though relying on the same principle of electrical disruption of membranes, once electroporated mRNA has entered the cell, it does not need genomic integration for construct expression. Whereas integrated constructs have been observed for more than a decade post transfer ¹³, electroporated mRNA rapidly degrades and is associated with transient expression ^86,128. In clinical application, transient expression of a construct may require repeated doses to achieve adequate effector function ^86,129. In humans, RNA modified mesothelin-CART cells have been shown to be safe, however repeated doses may be problematic if the engineered cells are themselves immunogenic. While preliminary evidence suggests that these cells are effective at targeting mesothelin positive tumors ⁸⁷ one case of anaphylaxis has been described in the setting of infusions that were separated by 49 days ¹³⁰. Numerous active clinical trials are using mRNA-modified CART cells to target malignancy (Tables 1 and 2).

Gene editing is one of the most exciting recent developments in the modernization of redirected T cell manufacturing. The overarching term “gene editing” refers to a variety of techniques that confer particular advantages or disadvantages depending on application. What they share, however, is the ability to efficiently knock-out and/or knock-in genetic elements. Both protein-based (zinc finger nucleases, transcription activator-like effector nucleases) and RNA-based (clustered regularly interspaced short palindromic repeats (CRISPR)/Caspase 9) techniques are effective at specific gene disruption or insertion. To produce superior engineered T cells, gene editing may be used to knockout inhibitory receptors rendering the cells resistant to tumor immunosuppression and/or knock-in an array of function-enhancing molecules.

Efficient gene editing of primary human T cells has been demonstrated ^131–134. Safety of gene-edited T cells in humans was demonstrated with the adoptive transfer of CCR5 zinc-finger mediated knockout, autologous T cells in 2014 ¹³⁵. The manufacture of gene edited human CART cells has been shown to be feasible with the production of TCR TALEN or CRISPR/Cas9 mediated knockout, CAR T cells^136,137 (Figure 4b). A preliminary description of the first use of gene-edited CART cells in humans was recently reported in an infant with CD19+ ALL ¹³⁶. Autologous CART19 cells were unable to be produced. The patient was heavily preconditioned with chemotherapy to delay CAR T cell rejection by the patient and thereby enhance CAR T cell persistence. CARTs were produced from an unrelated donor by deleting the endogenous TCR to prevent GVHD. In addition, CD52 was deleted from the CART cells, permitting in vivo deletion of patient lymphocytes while sparing the infused CD52-negative CARTs. The administration of donor-derived, gene-edited T cells has the potential to revolutionize the current manufacturing paradigm; a single donor could provide starting material to manufacture products for numerous recipients. While promising, this exciting step forward will require further investigation on more patients to demonstrate the role of allogeneic CART cells in tumor control. Safe and effective use of allogeneic CART cells may require additional editing of endogenous molecules such as HLA¹³⁷.

B. Enhancing trafficking

Engineered T cell localization at target sites is crucial for clinical efficacy, particularly when targeting solid tumors ¹³⁸. Route of administration and effective trafficking to the tumor site both play significant roles in granting T cell access to target tissue. While T cells are capable of migrating to nearly all body compartments, including immune privileged sites ^54,88, accumulation of engineered T cells may be enhanced by local administration. In several preclinical solid tumor models local administration of CART cells demonstrated superior accumulation at tumor sites and control of tumor growth compared to systemic administration ^85,123,139. Notably, intrapleurally injected Meso-CART cells outperformed systemically administered cells in clearance of intrathoracic and extrathoracic mesothelioma lesions ⁸⁵. The superior extrathoracic tumor clearance suggests that early exposure of engineered T cells to target may enhance the overall ability of these cells to traffic to and clear the tumor. Further, engineered T cells can be modified to enhance trafficking (Figure 4c). Chemokine receptor-ligand interactions play an important role in mediating endogenous immune cell trafficking. In fact, efficacy of conventional chemotherapeutics is linked to upregulation of chemokine ligands on tumor that is mediated by these drugs ¹⁴⁰. CART cells may be engineered to express chemokine receptors to enhance trafficking into tissue and homing to tumor sites. Co-expression of the chemokine receptor CCR2b in CART cells targeting either GD2 or mesothelin has been shown to enhance tumor infiltration and anti-tumor effects in animal models ^141,142.

C. Avoiding tumor suppression and escape

Malignancy may be refractory to engineered T cell therapy by immune escape or tumor immunosuppression. A variety of CD19 mutations and alternative splicing have been associated with development of CART19 resistant ALL ¹⁵. In this setting multivalent targeting may prevent single agent resistance. The combination of CD123 targeted and CD19 targeted CAR T cells prevents the outgrowth of CD19 negative escape mutants in preclinical models ¹⁴³. The tumor microenvironment may also directly inhibit a potential immune response. By definition, tumor existence is dependent on inhibition of endogenous immune control. This is achieved through a variety of mechanisms including cell-cell signaling and release of soluble cytokines. Importantly, like the endogenous immune system, adoptively transferred T cells are also susceptible to tumor-mediated immunosuppression ¹⁴⁴. Further, chronic T cell activation induces upregulation of inhibitory ligands on the activated cells ¹⁴⁵. A variety of methods can be used to engineer T cells to be intrinsically resistant to tumor immunosuppression (Figure 4d). Expression of a dominant negative TGFβ receptor confers T cell resistance to this tumor-produced, suppressive cytokine ¹⁴⁶. Others have transduced tumor specific T cells with hybrid receptors comprised of an IL4 exodomain and an IL7 endodomain ¹⁴⁷. Tumor generated IL4, a suppressive cytokine, produces an activating signal in these cells. The addition of anti-PD1 monoclonal antibody has been shown to enhance function of CART cells in preclinical models ¹⁴⁸. This finding suggests that modifying T cells to be intrinsically resistant to checkpoint inhibition could enhance engineered T cell efficacy in humans. Many groups are now attempting to generate CART cells resistant to PD1-PDL1 and CTLA4-CD80/CD86 signaling ^137,149. Future T cell therapies will incorporate multiple forms of checkpoint blockade to further enhance efficacy.

D. Improving safety: Boolean logic gates

While treatment related mortality is far below that seen with conventional treatments for relapsed/refractory cancers, serious adverse events have been observed following infusions of engineered T cells. Excessive and rapid tumor clearance has been associated with serious and occasionally fatal cytokine release syndrome. On-target, off-tumor activation of engineered T cells by very low level of target on non-malignant tissue has been associated with dose limiting toxicities ¹⁵⁰ and death in some cases ^93,109. Finally, unexpected and fatal cross-reactivity seen with an engineered TCR T cells demonstrates current limitations of in vitro screening for cross-reactivity^151,152.

Molecular “switches” allow for greater control over engineered T cell in vivo performance and may improve safety. Cells may be engineered to express pro-death signals that can be induced with an exogenous element (off-switch, see Figure 4e). Examples of off-switches or “suicide genes” include Herpes simplex virus thymidine kinase (HSV-Tk) and inducible human caspase 9 (iCasp9). Provision of ganciclovir or FK506 binding protein (FK506BP) respectively induces selective cell death specific to those cells expressing the suicide gene. Deletion of CART cells in animal models has been achieved via both HSV-Tk/gancyclovir ¹⁵³ and iCasp9/FK506BP systems ¹⁵⁴. Alternatively, T cells may be engineered to conditionally activate only in the presence of an exogenous molecule, withdraw of which terminates signaling (on-switch) (Figure 4f). On-switches are currently under development, though this technology is less mature. On-switches may prove safer to off-switches as the default is to ablate signaling. In addition, removal of the exogenous activator molecule does not necessarily lead to cell death. One can envision repeated dosing of the activator molecule as tolerated by the patient. The feasibility of producing CART cells with small molecule dependent signaling have been established^155,156. In this system, the switch redirects activity of orthogonal receptor through the selective formation of immunological synapses in a temporally controlled manner. Further, this system is readily adaptable to different antigen targets. Another type of flexible receptor targeting system has recently been described by Lim and colleagues ¹⁵⁷. Based on synthetic Notch receptors, this system allows for conditional expression of a targeting receptor upon engagement with a tissue specific ligand.

Engineered T cells may be marked with unique cell surface molecules to which existing monoclonal antibodies bind (Figure 4g). If this epitope is also expressed on tumor cells, treatment with these monoclonal antibodies could eliminate CART cell mediated adverse effects while simultaneously treating the tumor. A fusion of CD34 and CD20 epitopes (RQR8) ^136,158 and a truncated form of human EGFR polypeptide ¹⁵⁹ have separately been expressed in CART cells. In the setting of intolerable adverse effects, these CART cells would be susceptible to elimination by rituximab (monoclonal anti-CD20) or cetixumab (monoclonal anti-EGFR) respectively. Given the availability of such a wide array of inducible and specific methods of CART cell deletion, it is likely that more clinical trials will include such constructs moving forward ^{16,154,158,159}.

Deletion of CART cells may limit adverse effects, but will also terminate the anti-tumor clinical effect. Off tumor toxicity can also be prevented by designing CART cells with enhanced specificity. To achieve this, CARs have been designed to only transmit activating signals in response to a particular combination of targets. For example, bispecific CARs have been generated such that the extracellular portion of the CAR contains two linked scFvs with different specificities (Figure 4h). T cells expressing these tandem CARs (TanCAR) are only activated in the presence of both targets; a target cell positive for a single antigen is insufficient to trigger activation and killing. TanCARs against HER2+CD19 and HER2+IL13Ra2 have been developed ^160,161. An alternative to this method is to combine one CAR that transmits only primary signal with a second CAR with distinct specificity that transmits only costimulation (Figure 4i). In this approach, a single T cell expressing a CD3ζ only-CAR against the first target and a costimulatory domain only-CAR against a second target will only become fully activated in the presence of both targets. Such dual CARs against mesothelin + alpha folate receptor and HER2 + MUC1 have been shown generate specific target cytotoxicity against dual expressing targets ^162,163. Lastly, extracellular scFv fused to inhibitory signaling domains are capable of specifically inhibiting CART cell activation (Figure 4j). These inhibitory signals allow protection of cells with a particular immunophenotype from CART cell killing ¹⁶⁴. Incorporation of all activation and inhibitory signals creates a complex computational algorithm for engineered T cell receptor targeting and decision-making. Importantly, this has allowed for reconsideration of targets previously thought to be undesirable due to off-tumor toxicities. Further, many of the same types of receptor algorithms shown in figure 4 may be applies to the next generations of engineered T cell receptors to improve targeting and control. For example, a self destruct or a conditional switch may be inserted, along with the TCR. A switch receptor, or armored TCR, may be created by inserting a decoy receptor that binds to PD-L1 on tumors, but provides an accessory signal to augment engineered TCR signaling^149,165. These new molecular systems embedded in a cellular drug will soon allow highly specific immunophenotypes to be targeted, off tumor effects to be minimized and safety to be enhanced in the clinic.

VI. The rapidly approaching future of cancer immunotherapy

The advent of kinase targeted drug therapies and checkpoint blockade antibodies has increased survival in some patients with cancer. In the previous decade, patients with myeloma had an average survival of 2 to 3 years, and with the advent of improved therapies it is now 7 to 8 years, and still increasing ^{166,167,148, 149}. Although CLL has remained incurable with standard treatments ^168,151, the advent of effective targeted therapies such as ibrutinib and idelalisib has significantly extended survival ^169,152. Checkpoint therapies are a new class of cancer drugs that are one of the major advances in cancer therapy in the past decade, with reproducible benefit observed in 20 to 30% of patients with a variety of previously incurable cancers^145,128. However, there are significant costs associated with recurrent administration and the majority of patients do not currently benefit from these therapies. Thus these therapies, which must be administered long term, present a significant economic burden for patients and the economy.

In contrast, adoptive therapy with engineered T cells has two characteristics that may complement the limitations of kinase targeted and checkpoint therapies. First, engineered T cells require only one treatment for durable benefit ¹⁷⁰. Secondly, nearly all patients (>90%) with acute lymphoid leukemia respond to CART cells ^4,6, a response rate not previously observed with other forms of cancer immunotherapies. While not yet tested clinically, pre-clinical models reveal a potent enhancement of anti-tumor efficacy with the combination of CART cells and checkpoint blockade ¹⁷¹. It is possible that the combination of these therapies could result in the long-term survival and eventual cure of a number of cancers after only a few treatments ¹⁷². Even today, with early generation manufacturing, the production and delivery of a one time treatment delivering durable benefit is disruptive to health care financing and reimbursement models. The expanded availability of redirected T cell therapeutics in cancers beyond hematologic malignancies is dependent on the development of automated cell engineering and potentially on the development of universal sources of allogeneic T cells.

Figure 3. Engineered T Cell Manufacturing.

Leukocytes are generally collected by leukapheresis (1) and lymphocytes can be enriched (2) by counterflow centrifugal elutriation¹⁹⁹ or subsets selected (not shown). The enriched lymphocytes are placed in to culture and (3) stimulated with bead-based artificial antigen presenting cells^200,201 and viral vector (4) added²⁰². The culture is expanded in a bioreactor for several days (5) and then the T cell bulk product (6) is washed and concentrated, samples removed for quality control release testing (7) and quality assurance review. The final formulation is cryopreserved (8), allowing facile shipment to distant infusion sites, where the final product bag (9) is thawed and infused. Manufacturing time is generally 5 to 10 days, and collection to infusion times can range from 2 to 4 weeks depending on patient clinical status and chemotherapy conditioning regimens.

Table 3.

Commercial Developers of Engineered T cells

Company	Engineered T Cell Technology in Development
Adaptimmune	TCRs
Autolus	CARs
Beijing Doing Biomedical	CARs
Bellicum	CARs, “suicide” switch
bluebird bio	CARs
CARsgen	CARs
CBMG	CARs
Celgene	CARs
Cellectis	Allogeneic CARs
Celyad	CARs
Formula Pharmaceuticals	CARs (in cytokine induced killer cells)
Juno Therapeutics	CARs, TCRs
Kite Pharma	CARs, TCRs
Medigene	TCRs
NantKwest	CARs (in NK cells)
Novartis	CARs
Opus Bio	CARs
PersonGen Biomedicine	CARs
Poseida Therapeutics	CARs
Takara Bio	CARs, TCR’s
Theravectys	CARS, regulated expression
Unum	CARs
Ziopharm/Intrexon	CARs, regulatable expression