CAR T Cell Therapy for Solid Tumors: Bright Future or Dark Reality?

Abstract

Chimeric antigen receptor (CAR) T cell therapy has garnered significant excitement due to its success for hematological malignancies in clinical studies leading to the US Food and Drug Administration (FDA) approval of three CD19-targeted CAR T cell products. In contrast, the clinical experience with CAR T cell therapy for solid tumors and brain tumors has been less encouraging, with only a few patients achieving complete responses. Clinical and preclinical studies have identified multiple “roadblocks,” including (1) a limited array of targetable antigens and heterogeneous antigen expression, (2) limited T cell fitness and survival before reaching tumor sites, (3) an inability of T cells to efficiently traffic to tumor sites and penetrate physical barriers, and (4) an immunosuppressive tumor microenvironment. Herein, we review these challenges and discuss strategies that investigators have taken to improve the effector function of CAR T cells for the adoptive immunotherapy of solid tumors.

Graphical Abstract

Currently, CAR T cell therapy has limited antitumor activity for solid tumors. In this review, Wagner and colleagues review the clinical experience with CAR T cells for solid tumors, including brain tumors, appraise CAR T cell therapy roadblocks, and discuss strategies that investigators have taken to improve their effector function.

Main Text

Adoptive cell therapies utilizing T cells expressing chimeric antigen receptors (CARs) have been propelled to the forefront of experimental cell therapies due to their clinical success for hematological malignancies targeting a range of antigens, including CD19, CD22, CD30, kappa, and B cell maturation antigen (BCMA).1, 2, 3, 4, 5, 6, 7, 8, 9, 10 The landscape of CAR T cell therapies for hematological malignancies, including successes and challenges, has been the subject of multiple recent reviews.11, 12, 13 We therefore do not discuss these in detail, except for highlighting the “lessons learned” as they relate to targeting solid tumors, including brain tumors. First, CAR T cells can eradicate chemorefractory cancer cells regardless of their underlying oncogenic driver mutations; second, lymphodepleting chemotherapy is at present a sine qua non to enable expansion and persistence of infused CAR T cells; third, inclusion of at least one co-stimulatory signaling domain in the CAR is critical for its success; fourth: antigen loss variants have emerged as one mechanism of therapeutic failure, even for antigens that are highly and homogeneously expressed such as CD19; and fifth, CAR T cells can induce significant side effects, including cytokine release syndrome (CRS) and neurotoxicity.11, 12, 13, 14, 15

In contrast to CAR T cell therapy for hematological malignancies, CAR T cell therapies for solid tumors, including brain tumors, have shown limited antitumor activity in early phase clinical testing despite targeting a variety of target antigens and tumor types.16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26 This therapeutic failure is most likely multifactorial and includes (1) CAR design and CAR T cell generation, (2) a limited array of targetable antigens and heterogeneous antigen expression (aka “antigen dilemma”), (3) limited T cell fitness, (4) inefficient homing to and penetration of solid tumors, and (5) the hostile tumor microenvironment (TME) (Figure 1).27, 28, 29, 30 Herein, we review CAR design, the current clinical experience with autologous CAR T cells for solid tumors, including brain tumors, and genetic approaches to overcome the aforementioned roadblocks. Since allogeneic CAR T cells are at present mainly explored for hematological malignancies in early phase clinical studies, we refer the interested reader to recently published reviews.^31,32 We also do not review CRS and neurotoxicity since several recent reviews have covered this topic.^14,33

Figure 1.

CAR T Cell Immunotherapy “Roadblocks” for Solid Tumors

Top left: antigen dilemma. Heterogeneous antigen expression results in immune escape variants, and antigen expression in normal tissues can lead to on target/off cancer toxicity. Top right: CAR T cell fitness. Expansion, contractions, and persistence of CAR T cells is often limited after infusion. Bottom left: homing/penetration. CAR T cells have a limited ability to traffic to and penetrate solid tumors. Bottom right: microenvironment. The solid tumor microenvironment is hostile, limiting CAR T cell effector function.

Design of CARs

CARs have a modular design that consists of an antigen-binding domain, most commonly a single-chain variable fragment (scFv) derived from a monoclonal antibody (mAb), a hinge and transmembrane domain, and an intracellular signaling domain.^27,34, 35, 36, 37, 38 In addition to scFvs, natural ligands of receptors such as cytokines or peptides that bind cell surface molecules are also being explored as antigen-binding domains.^22,39,40 While most CARs recognize epitopes of cell surface proteins in a major histocompatibility complex (MHC)-independent manner, scFvs have also been incorporated into CARs that recognize a peptide in the context of a human leukocyte antigen (HLA) molecule.41, 42, 43 While this approach allows CAR T cells to recognize intracellular molecules, it renders target antigen recognition dependent on a particular HLA type, restricting its application to a subset of patients. In addition, CAR T cells become sensitive to decreased HLA expression and defects in the antigen processing pathway, with both pathways used by tumor cells to actively evade immune responses.⁴⁴

First-generation CARs contained only a T cell activation domain, and the most commonly utilized domain is derived from CD3ζ.^37,38 Since optimal T cell activation also relies on co-stimulation, signaling domains from co-stimulatory molecules were included into CARs. Initial studies focused on domains derived from the canonical co-stimulatory molecules CD28 and 41BB.^37,38 Since then, signaling domains from a broad range of molecules have been explored, including OX40, CD27, and ICOS.^37,45, 46, 47, 48, 49 CD28 and 41BB co-stimulation in the context of CAR T cells has been extensively studied, including detailed phosphoproteomic and single-cell RNA sequencing (RNA-seq) analyses.50, 51, 52 They activate different pathways within T cells, with CD28 signaling promoting glycolytic metabolism and an effector memory phenotype, in contrast to 41BB signaling, which induces oxidative metabolism and a central memory phenotype.⁵³ Depending on the number of co-stimulatory molecules included in the CAR signaling domain, CARs are either designated second (one domain) or third (two domains) generation. In preclinical studies, the benefit of including two co-stimulatory molecules in CARs is model-dependent.^54,55 Results of one clinical study, comparing CD28-CAR versus CD28.41BB-CAR T cells targeting CD19, suggest that third-generation CARs endow T cells with a greater ability to expand after infusion in humans.⁵⁶

Optimizing CAR function remains a challenge since there is an intricate interplay between the functional (antigen recognition and signaling domains) and nonfunctional components (hinge and transmembrane domain) of CARs.^57,58 In addition, the location of the epitope within the targeted cell surface molecule determines CAR T cell activity.⁵⁹ For example, epitopes that are proximal to the plasma membrane induce greater CAR T cell activation than do distal epitopes.^57,59,60 Studies have also highlighted that too much CAR T cell activation is detrimental to T cell function. For example, CARs with two of three mutated immunoreceptor tyrosine-based activation motifs (ITAMs) in the CD3ζ chain endow CAR T cells with improved effector function.⁶¹ In addition, excessive CD28 co-stimulation has detrimental effects,⁶² and mutations in the YMXM motif of CD28 that reduce its signaling activity improve T cell function.⁶³ In addition to CAR signaling after activation, baseline (aka tonic) signaling determines CAR T cell activity.^64,65 Recently, investigators have also focused on studying the immunological synapse that is formed between CAR T cells and target cells, and they have correlated synapse formation with CAR T cell effectiveness.^66,67

Lastly, several studies have highlighted limitations of CARs with standard co-stimulatory molecules, which most likely cannot be overcome by further refining its structure.^27,37,38 Foremost, CARs only provide signal 1 (activation) and signal 2 (co-stimulation) of canonical T cell activation and not signal 3 (inflammatory cytokines). While CAR-activated T cells initially produce cytokines, they do not secrete sufficient amounts of cytokines after recursive exposure to antigen-positive target cells, leading to rapid erosion of their effector function.68, 69, 70 This occurs even in the absence of an immunosuppressive TME, highlighting that additional modification of CAR T cells or combinatorial therapies are needed to sustain and/or enhance their effector function. These are discussed in detail in the section Improving CAR T Cell Fitness.

CAR T Cells for Solid Tumors and Brain Tumors—Clinical Experience

CAR T cells have been evaluated in early phase clinical studies for more than a decade, targeting a broad range of tumor antigens, including B7-H3, CAIX, CEACAM5, CD133, CD171, epidermal growth factor receptor (EGFR), EGFRvIII, FRα, GD2, GPC3, HER2, interleukin (IL)-13Rα2, mesothelin, MUC1, prostate-specific membrane antigen (PSMA), ROR1, and vascular endothelial growth factor (VEGF)-R2.^19,20,22, 23, 24, 25, 26^,71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82 Table 1 lists selected published studies, and Table 2 lists ongoing clinical studies. Early studies focused on first-generation CARs specific for CAIX, CD171, FRα, GD2, or IL-13Rα2.^16,20,22,83 Two studies utilized polyclonal, activated CAR T cells (CAIX, FRα);^20,84 two studies focused on CAR T cell clones (CD171, IL-13Rα2);^75,83,85 and one study compared GD2-CAR T cells with GD2-CAR/Epstein-Barr virus (EBV)-specific T cells.^81,82 First-generation CAR T cells did not expand after infusion and had limited antitumor activity, except for the study with GD2-CAR/EBV-specific T cells, in which 3 of 11 patients had a complete response (CR). Despite having no antitumor activity, CAIX-CAR T cells induced “on target/off cancer” toxicity in the form of cholangitis.^19,20 In hindsight, limited antitumor activity was not surprising since CARs without co-stimulatory endodomains were grafted onto T cells, and patients did not receive lymphodepleting chemotherapy prior to the adoptive transfer of CAR T cells—two prerequisites of successful CAR T cell therapies for hematological malignancies.1, 2, 3, 4, 5, 6, 7, 8, 9, 10

Table 1.

Selected Published CAR T Cell Clinical Studies for Solid Tumors and Brain Tumors

Antigen	Diagnosis	Signaling Domain	Vector	T Cell Product	IL-2 after T Cells	Comment	References
Intravenous Administration, No Lymphodepleting Chemotherapy
FRα	OVCA	ζ	RV	ATCsa	Y	14/14 NR	21
CAIX	RCC	ζ	RV	ATCs	N	12/12 NR; on target/off cancer toxicity—bile ducts	19,20
CD171	NB	ζ	Plasmid	T cell clone	N	1/6 PR	83
EGFRvIII	HGG	4-1BB	LV	ATCs	N	1/10 SD	17
GD2	NB	ζ	RV	ATCs, VSTs	N	3/11 CR	81,82
HER2	sarcoma	CD28ζ	RV	ATCs	N	4/17 SD	23
HER2	HGG	CD28ζ	RV	VSTs	N	3/17 SD	26
Mesothelin	PCA	41BBζ	mRNA	ATCs	N	2/6 SD	80
Intravenous Administration after Lymphodepleting Chemotherapy
CD133	HCC, PCA, CRC	41BBζ	LV	ATCs	N	3/23 PR, 14/23 SD	24
CEACAM5	CRC, PDCA, stomach, esophagus	ζ	RV	ATCs	Y	7/14 SD; on target/off cancer toxicity—lung	25
EGFR	PCA	41BBζ	LV	ATCs	N	4/16 PR, 8/16 SD, 2/16 NE	86
EGFRvIII	HGG	CD28.41BBζ	RV	ATCs	Y	17/18 NR, 1/18 NE due to TRMb	16
GD2	NB	CD28.OX40ζ	RVc	ATCs	N	5/11 SD	87
GPC3	HCC	CD28ζ	LV	ATCs	N	2/13 PR, 1/13 SD, 4/13 NE	88
HER2	CRC	CD28.41BBζ	LV	ATCs	Y	1/1 TRMb; on target/off cancer toxicity—lung	18
HER2	sarcoma	CD28ζ	RV	ATCs	N	1/10 CR; 3/10 SD	89,90
Mesothelin	MPM, PCA, OVCA	41BBζ	LV	ATCs	N	11/15d SD	91
PSMA	prostate	ζ	RV	ATCs	Y	2/5 PR	71
ROR1	TNBC, NSCLC	41BBζ	LV	ATCs	N	4/6 mixed response; 1/6 SD; 1/6 not reported	72
VEGF-R2	metastatic CA	–	RV	ATCs	Y	1/23 PR	e
Intravenous Administration after Allotransplant
GD2	NB	ζ	RV	VSTs	N	3/3 PD	f
Locoregional Administration
B7-H3	meningioma	41BBζ	LV	ATCs	N	1/1 evidence of ALV	73
IL-13Rα2	HGG	ζ	plasmid	T cell clone	N	1/3 tumor necrosis	74
IL-13Rα2	HGG	ζ	plasmid	T cell clonea	N	imaging study; outcome not reported	75
IL-13Rα2	HGG	41BBζ	LV	ATCs	N	1/1 CR	22
Mesothelin	MPM, lung CA, breast CA	CD28ζ	RVc	ATCs	N	2/21g CR, 5/21 PR, 4/21 SD	76
MUC1	SVA	41BBζ or CD28ζ	LVh	ATCs	N	1/1 tumor necrosis	92

ALV, antigen loss variant; ATC, activated T cell; CA, cancer; CR, complete response; CRC, colorectal cancer; HCC, hepatocellular carcinoma; HGG, high-grade glioma; LV, lentivirus; MPM, malignant pleural mesothelioma; N, no; NB, neuroblastoma; NE, not evaluable; NR, no response; OVCA, ovarian cancer; PCA, pancreatic adenocarcinoma; PR, partial response; RCC, renal cell carcinoma; RV, retrovirus; SD, stable disease; SVA, seminal vesicle adenocarcinoma; VST, virus-specific T cell; Y, yes.

^aSeveral patients received allogeneic cell products.

^bPulmonary complications.

^cCAR vector also encoded inducible caspase-9.

^d6 of 15 patients received lymphodepleting chemotherapy.

^eClinicalTrials.gov: NCT01218867.

^fClinicalTrials.gov: NCT01460901.

^g18 of 21 patients received lymphodepleting chemotherapy and 13 of 21 patients received pembrolizumab.

^h41BBζ CAR vector also encoded IL-12.

Table 2.

Selected Actively Recruiting CAR T Cell Clinical Studies for Solid Tumors and Brain Tumors

Antigen	Diagnosis	Signaling Domain	Additional Engineering	Vector	T Cell Product	Other Therapy	NCT No.: ClinicalTrials.gov
Intravenous Administration, No Lymphodepleting Chemotherapy Listed onClinicalTrials.govSite
GPC3	HCC	–	TGF-β-CAR, IL-7-CCL19	–	ATCs	N	NCT03198546
	HCC	–	N	–	ATCs	N	NCT04121273
PSMA	solid tumors	–	N	–	ATCs	N	NCT04429451
Intravenous Administration after Lymphodepleting Chemotherapy
B7-H3	solid tumors	–	iC9	LV	ATCs	N	NCT04432649
CD171	NB	41BBζ or CD28.41BBζ	N	LV	ATCs	N	NCT02311621
Claudin 18.2	solid tumors	–	N	–	ATCs	N	NCT03874897
EGFR806	solid tumors	41BBζ	CD19-CAR	LV	ATCs	N	NCT03618381
IL-13Rα2	Melanoma	41BBζ	N	LV	ATCs	IL-2	NCT04119024
GD2	NB	–	iC9, IL-15	RV	ATCs	N	NCT03721068
	solid tumors	–	iC9	–	ATCs	N	NCT03373097
	solid tumors	–	C7R	RV	ATCs	N	NCT03635632
	DIPG, HGG	–	C7R	RV	ATCs	N	NCT04099797
	NB	–	IL-15	RV	iNKT	N	NCT03294954
	NB	–	N	–	ATCs	N	NCT02761915
	DIPG	41BBζ	iC9	RV	ATCs	N	NCT04196413
GPC3	solid tumors	41BBζ	N	RV	ATCs	N	NCT02932956
	HCC	41BBζ	N	RV	ATCs	N	NCT02905188
	HCC	–	N	–	ATCs	N	NCT03884751
Mesothelin	solid tumors	–	PD-1 KO	–	ATCs	N	NCT03747965
	solid tumors	–	N	LV	ATCs	N	NCT03054298
MUC-1	esophageal CA	–	PD-1 KO	–	ATCs	IL-2	NCT03706326
	NSCLC	–	PD-1 KO	–	ATCs	IL-2	NCT03525782
	breast Cancer	41BBζ	N	LV	ATCs	N	NCT04020575
PSCA	prostate CA	41BBζ	N	LV	ATCs	N	NCT03873805
	solid tumors	ζ	iMC	RV	ATCs	N	NCT02744287
PSMA	solid tumors	–	iC9	gene editing	ATCs	N	NCT04249947
	prostate CA	–	DNR TGFβ	LV	ATCs	N	NCT03089203
	prostate CA	–	DNR TGF-β	LV	ATCs	N	NCT04227275
Intravenous and Locoregional Administration After Lymphodepleting Chemotherapy
MUC16ecto	solid tumors	41BBζ	IL-12	LV	ATCs	N	NCT02498912
Locoregional Administration, No Lymphodepleting Chemotherapy
B7-H3	CNS tumors	–	N	LV	ATCs	N	NCT04185038
	CNS tumors	–	N	RV	ATCs	TMZ	NCT04385173
	CNS tumors	–	N	RV	ATCs	TMZ	NCT04077866
EGFR806	CNS tumors	41BBζ	N	LV	ATCs	N	NCT03618381
EGFR family	SCCHN	CD28ζ	IL-4/IL-2 CCR	RV	ATCs	N	NCT01818323
HER2	CNS metastases	41BBζ	N	LV	ATCs	N	NCT03696030
	CNS tumors	CD28ζ	N	RV	ATCs	N	NCT02442297
	CNS tumors	41BBζ	N	LV	ATCs	N	NCT03500991
IL-13Rα2	HGG	41BBζ	N	LV	ATCs	nivo/ipi	NCT04003649
	HGG	41BBζ	N	LV	ATCs	N	NCT02208362
Locoregional Administration after Lymphodepleting Chemotherapy
FRα	OVCA	41BBζ	N	LV	ATCs	N	NCT03585764
Mesothelin	pleural disease	CD28ζ	iC9	–	ATCs	Pembro	NCT02414269

ATC, activated T cell; CA, cancer; CCR, chimeric cytokine receptor; C7R, constitutive active IL-7 receptor α; DNR, dominant negative receptor; HCC, hepatocellular carcinoma; HGG, high-grade glioma; iC9, inducible caspase-9; iMC, inducible Myd88.CD40; ipi, ipilimumab; KO, knockout; LV, lentivirus; N, no; NB, neuroblastoma; nivo, nivolumab; OVCA, ovarian cancer; pembro, pembrolizumab; RV, retrovirus; SCCHN, squamous cell carcinoma of the head and neck; Y, yes.

In subsequent studies, CARs were evaluated with co-stimulatory endodomains (CD28, 41BB, CD28/41BB, CD28/OX40) targeting B7-H3, CEACAM5, CD133, CD171, EGFR, EGFRvIII, FRα, GD2, GPC3, HER2, IL-13Rα2, mesothelin, MUC1, PSMA, ROR1, and VEGF-R2.^72,86, 87, 88, 89^,91,92 Similar to first-generation CAR T cells, adoptive transfer of second- or third-generation CAR T cells or CAR virus-specific T cells did not result in significant T cell expansion in two studies, and variable expansion in one study.^23,26,87 Thus, inclusion of a co-stimulatory domain into the CAR by itself is not sufficient to consistently induce systemic T cell expansion. In contrast, lymphodepleting chemotherapy prior to CAR T cell infusion resulted in a significant expansion of CAR T cells irrespective of the utilized co-stimulatory domain and targeted antigen.^9,93,94 Despite significant in vivo expansion, antitumor activity remained limited (Table 1). However, single case reports of patients with recurrent refractory solid tumors or brain tumors, who achieved CRs after CAR T cell therapy, have been reported (Table 1).^22,90 Clinical studies have also highlighted safety concerns. Two patients died of acute respiratory failure after having received lymphodepleting chemotherapy, a high dose (1 to 6 × 10¹⁰) of third-generation CAR T cells, and IL-2.^16,18 CAR T cells either recognized an antigen that is expressed (HER2) or not expressed (EGFRvIII) on lung endothelial cells, suggesting that on target/off cancer toxicity as well as unspecific CAR T cell activation can result in fatal respiratory failure. Adoptive transfer of high cell doses (1 × 10⁹ to 5 × 10¹⁰) of CEACAM5-CAR T cells in conjunction with IL-2 also resulted in pulmonary toxicities (mild pulmonary edema), which was attributed to CEACAM5 expression on lung epithelium.²⁵

Tumor tissue after CAR T cell infusion has only been studied systematically in one study in which patients with high-grade glioma (HGG) received EGFRvIII-CAR T cells.¹⁷ Collectively, the analysis showed that CAR T cells were able to migrate to and proliferate at tumor sites. CAR T cells killed tumor cells as evidenced by the selection of target antigen-negative tumors, and, strikingly, induced a reactive immunosuppressive environment as judged by expression of IDO1 and PD-L1, and an influx of regulatory T cells (Tregs).

Thus, CAR T cells at present rarely induce CRs in patients with solid tumors or brain tumors. In addition, several studies have highlighted their potential to induce on target/off cancer toxicity. In the following sections we review approaches that investigators have taken to improve their efficacy and safety, with special focus on (1) clinical grade CAR T cell production and T cell subsets for CAR T cell therapy, (2) expanding the array of targetable antigens and overcoming heterogeneous antigen expression, (3) CAR T cell fitness, (4) CAR T cell trafficking and tumor penetration, and (5) counteracting the immunosuppressive TME (Figure 2).

Figure 2.

Strategies to Overcome CAR T Cell Immunotherapy Roadblocks for Solid Tumors

Summary of strategies that are discussed in detail in this review.

Improving Clinical-Grade CAR T Cell Production

Preclinical studies have highlighted that the effector function of T cells progressively decreases as T cells differentiate from naive (T_N) to stem cell memory (T_SCM), central memory (T_CM), effector memory (T_EM), and terminally differentiated T cells that express the CD45RA effector (T_EMRA).⁹⁵ Transcription factors govern T cell differentiation. For example, TCF7 promotes self-renewal and memory formation, while TOX, BLIMP1, and TBET promote terminal T cell differentiation.96, 97, 98, 99, 100, 101, 102 More recent studies have also highlighted the critical role of epigenetic programs enforced by the de novo DNA methyltransferase 3A (DNMT3A) in determining T cell fate.103, 104, 105 In addition, a recent case report highlighted that disruption of TET2 in CD19-CAR T cells, an enzyme that regulates DNA demethylation, resulted in clonal T cell expansion.¹⁰⁶

Despite these insights in T cell biology, most initial clinical studies utilized polyclonal, CD3/CD28-activated T cells that are expanded in the γ-cytokine IL-2. These expansion protocols in general result in significant T cell differentiation paired with inferior effector function in preclinical models. Investigators have explored alternative γ-cytokines to preserve CAR T cell function.¹⁰⁷ For example, replacing IL-2 with a combination of IL-7 and IL-15 for the ex vivo expansion of CAR T cells increases the frequency of CD8⁺CD45RA⁺CCR7⁺ stem cell-like CAR T cells with superior antitumor functionality.¹⁰⁸ Based on these findings, many clinical-grade CAR T cell production protocols are currently using IL-7 and IL-15. In addition, IL-21 has been shown to prevent differentiation of genetically modified T cells paired with improved effector function in preclinical studies.^109,110 Inhibiting or activating pathways in T cells that promote or prevent T cell differentiation are alternative strategies that have been explored. These include activating Wnt/β-catenin signaling pathways or inhibiting mTOR or AKT signaling.^111,112

Notwithstanding these advances, there is at present a lack of generally accepted biomarkers that predict efficacy of the infused CAR T cell product. Several studies with CD19-CAR T cells are starting to shed light on this issue. In one study in patients with chronic lymphocytic leukemia (CLL), the presence of CD27⁺PD-1⁻CD8⁺ CD19-CAR T cells expressing high levels of the IL-6 receptor in the T cell product correlated with antitumor activity,¹¹³ and in another study in patients with acute lymphocytic leukemia (ALL), the presence of tumor necrosis factor (TNF)-α⁺CD8⁺ CD19-CAR T cells, respectively.¹¹⁴ One recent study, using RNA-seq analysis, demonstrated that CD19-CAR T cell clones that expand after infusion are derived from clusters of CAR T cells in the product that express higher levels of genes associated with T cell proliferation and cytolytic activity.¹¹⁵ While this study found no correlation with vector integration sites, another recent report demonstrated that vector integration site distribution in CD19-CAR T cells is linked to clinical outcome.¹¹⁶ Clearly, additional studies are needed to identify biomarkers not only for the CAR T cell product, but also for the targeted tumor.

Besides improving culture conditions for CAR T cell generation, investigators have also focused on the genetic modification step of T cells during CAR T cell production. Most clinical studies so far have utilized CAR T cells that were generated with retroviral or lentiviral vectors with an excellent safety record.117, 118, 119 However, clinical-grade production and release testing of recombinant viral vectors is expensive, and patients require additional monitoring. Therefore, alternative approaches have been developed, including the use of PiggyBac or Sleeping Beauty transposon systems.^120,121 More recently, combining gene-editing technologies (e.g., CRISPR-Cas9, TALEN, ZFN, meganucleases) with homology-directed DNA repair (HDR) has enabled the integration of transgenes into specific loci.122, 123, 124 This approach allows for control of transgene expression with endogenous promoters. For example, CD19-CARs have been knocked in to the T cell receptor (TCR) α constant (TRAC) locus using CRISPR-Cas9 gene editing to drive CAR expression from the endogenous TCR α chain promoter. This resulted in consistent CAR expression, reduced baseline signaling, and improved antitumor activity compared to CD19-CAR T cells generated by viral transduction.¹²⁵ The clinical experience with CAR T cells generated with non-viral methods is so far limited,⁷⁸ and most ongoing studies are focused on targeting hematological malignancies.

T Cell Subsets for CAR T Cell Therapy

Investigators have explored CAR T cell products with a defined CD8⁺/CD4⁺ T cell subset ratio and products that were derived from T_SCM, T_CM, or T helper (Th17) cells.^48,126, 127, 128, 129 These studies have highlighted that T cell subset-derived CAR T cell populations have improved effector function in comparison to polyclonal, CD3/CD28-activated CAR T cells. Importantly, the co-stimulatory domain of the CAR might have to be tailored for a specific T cell subset with one study demonstrating superior antitumor activity when CD4⁺ T cells, expressing CARs with an ICOS co-stimulatory endodomain, were combined with CD8⁺ T cells, expressing CARs with a 41BB co-stimulatory endodomain.¹³⁰

However, the translation of these approaches into the clinic might be limited due to the frequency of some of these T cell subset populations; for example, T_SCM cells are very low in the peripheral circulation.¹³¹ In addition, sequential cycles of chemotherapy have depleted T cell subsets, including T_N cells, in heavily pretreated cancer patients, who are the current CAR T cell study population.¹³²

Selecting virus-specific T cells for CAR T cell generation is an alternative approach to selecting T_CM and T_EM cells with unknown specificity. Examples include the aforementioned clinical study with GD2-CAR/EBV-specific T cells in which investigators demonstrated improved persistence in comparison to GD2-CAR T cells.⁸¹ In addition, a clinical study is ongoing, which evaluates the safety and efficacy of GD2-CAR/varicella zoster virus (VZV)-specific T cells (ClinicalTrials.gov: NCT01953900). Advantages of this approach are not only that CAR/virus-specific T cells might have improved effector function in comparison to their unselected CAR T cell counterparts, but these CAR/virus-specific T cells can be boosted after infusion through their endogenous, virus-specific αβ TCR either by vaccination (e.g., VZV vaccine) or reactivation of the latent virus (e.g., EBV). Lastly CAR-expressing γδ T cells¹³³ and invariant natural killer (iNK)T cells have been successfully explored in preclinical studies,¹³⁴ and a clinical study with iNK T cells expressing GD2-CARs and IL-15 is in progress (ClinicalTrials.gov: NCT03294954).

CAR T Cell Therapy Targets—The Antigen Dilemma

An “ideal” CAR target should be highly and homogeneously expressed throughout the tumor, across multiple patients, and have minimal to no expression in vital normal tissues. Unfortunately, these characteristics do not apply to most CAR targets that are currently being explored for solid tumors and brain tumors (Table 2). In general, current targets are differentially expressed antigens, which may be present at low levels in normal tissues. The only exceptions are EGFRvIII, a cancer-specific splice variant expressed in HGG, and EGFR806, a conformational EGFR epitope that is also present in HGG.^135,136 While neoantigens are a potential solution to this “antigen dilemma,” most of these are patient specific, raising concerns about feasibility. In addition, most neoantigens described so far are derived from intracellular proteins, which are difficult to target with standard CAR T cells.¹³⁷ Investigators have therefore embarked on discovering novel CAR targets using genomic and proteomic approaches.138, 139, 140 In addition, there is an intense focus on engineering CARs and CAR T cells to increase their specificity for antigens expressed on tumor cells. These include (1) modulating the affinity of the scFv, (2) targeting multiple antigens, and (3) restricting CAR activity to tumor sites. In addition, investigators have developed safety switches that can be activated in the event of adverse events secondary to CAR T cells, including on target/off cancer toxicity.

Modulating scFv Affinity

This approach is only applicable to antigens that are expressed at very high levels such as gene-amplified HER2 in HER2-positive breast cancer. For example, investigators have shown that reducing the scFv affinity of HER2-CARs or EGFR-CARs decreases the risk of on target/off cancer toxicity.¹⁴¹ In addition, in murine models, the incidence of neurotoxicity of GD2-CAR T cells varied according to the affinity of the expressed CAR.¹⁴² In addition to modulating scFv affinity, recent studies have also highlighted that the choice of hinge, transmembrane, and/or costimulatory domains of CARs can influence antigen density requirements for CAR T cell activation.^58,143,144

Targeting Multiple Antigens

Heterogeneity is a hallmark of cancer and a major roadblock for all targeted cancer therapies, including immunotherapy.¹⁴⁵ Heterogeneous antigen expression is one consequence of tumor heterogeneity leading to the outgrowth of tumor cell subpopulations when a single antigen is targeted. Investigators have therefore focused on targeting cancer stem cells (CSCs) or tumor-initiating cells (TICs),146, 147, 148 or multiple antigens to prevent immune escape. In addition, multi-specific CAR T cells have the potential to reduce the risk of on target off/cancer toxicity if they target an array of antigens present in a particular tumor and not in normal tissues. Several approaches are actively being pursued, including expressing CARs with two scFvs. Investigators have either assembled two scFvs in a linear (aka tandem CAR) fashion or by alternating heavy and light chains of the two scFvs (aka loop CAR), similar to the dual-affinity retargeting (DART) bispecific antibody platform.149, 150, 151 Investigators have also expressed two or three different CARs in T cells.¹⁵² The clinical experience with T cells expressing bispecific CARs or two CARs is currently limited to CD19/CD22- and CD19/CD20-targeted approaches for leukemia and/or lymphoma. An alternate approach is to design CARs in which the antigen-binding domain is derived from a ligand that binds multiple antigens, and the most relevant example for solid tumors is the T1E peptide that binds to the EGFR family of receptors.⁴⁰ So-called “universal CARs” have also been developed to target multiple antigens that contain an ectodomain that can be paired with multiple antigen recognition domains. Examples include (1) avidin-CARs/biotin-labeled scFvs, (2) CD16-CAR/mAbs, (3) anti-fluorescein isothiocyanate (FITC)-CARs/FITC-labeled scFvs, (4) coiled-coil CARs (SUPRA CARs), (5) anti-PNE-CARs/PNE-scFvs, and (6) NKG2D-CARs/ULBP2-mAbs.153, 154, 155, 156, 157, 158, 159 Lastly, bispecific T cell engagers (BiTEs) have been expressed in T cells,^160,161 and more recently adapted to CAR T cells, opening up the opportunity to target multiple antigens and redirecting bystander T cells to tumor cells.¹⁶²

In addition to designing CAR T cells to target multiple antigens, strategies are being pursued to engineer T cells to activate bystander T cells to recognize tumor cells (aka induce antigen/epitope spreading). These include the transgenic expression of cytokines (e.g., IL-18 or IL-36γ) or CD40L.163, 164, 165 Likewise, one recent study has demonstrated that engineering T cells to secrete the FLT3 ligand promotes epitope spreading.¹⁶⁶

Restricting CAR Activity to Tumor Sites

CAR activity can be restricted to tumor sites by several measures. First, CAR expression can be induced once T cells have migrated to tumor sites with small molecules, such as a doxycycline-inducible expression system.¹⁶⁷ Second, CAR T cells can take clues from the TME to express a CAR. These include hypoxia-inducible expression systems,¹⁶⁸ as well as synthetic signaling circuits that take advantage of Boolean logic.169, 170, 171 One of the best examples of Boolean logic-gated CARs is the synthetic notch (synNotch) system,169, 170, 171 in which an antigen-specific synNotch receptor induces the expression of a second CAR upon activation. For example, EpCAM- or B7H3-specific synNotch receptors have been used to restrict ROR1-CAR expression to tumor sites, reducing on target/off cancer toxicity in preclinical models.¹⁷² Other strategies include engineering T cells to express two CARs with different specificity in which one CAR provides signal 1 and the other signal 2 to limit full CAR T cell activation to tumor sites at which both antigens are present.^173,174 With this approach full CAR T cell activation has been restricted to (1) PSMA- and PSCA-positive or (2) MUC1- and HER2-positive tumor cells in preclinical models.^173,174 Expressing inhibitory CARs that recognizes an antigen expressed on normal tissues in conjunction with a tumor-specific CAR is another approach that has been evaluated using CD19 as a tumor-specific antigen and PSMA as a model antigen expressed on non-targeted cancer cells.¹⁷⁵ Lastly, inactive CARs have been designed (e.g., EGFR)¹⁷⁶ that require proteolytic processing into an active form by matrix metalloproteinases (MMPs) within the TME.

Safety Switches

Neither of the aforementioned strategies may end up being sufficient to prevent on target/off cancer toxicities. Investigators have therefore developed inducible safety switches to ablate CAR T cells when the need arises. These safety switches can be conceptually divided into four approaches. The first approach relies on activating a prodrug that is toxic to T cells. The best studied example is the antiviral ganciclovir, which readily kills T cells that are genetically modified with the herpes simplex virus thymidine kinase (HSV-tk).¹⁷⁷ The second approach takes advantage of dimerizer drugs to activate a molecule that is transgenically expressed in CAR T cells. Several dimerizer systems have been developed;¹⁷⁸ of these, the inducible caspase-9 (iC9) system has been evaluated in humans with encouraging results.^179,180 Other promising strategies to increase the safety profile of CAR T cells include splitting CARs into two domains, which need to be linked with a small molecule to be active,^181,182 or taking advantage of the small molecule-assisted shut-off (SMAsh) technology.¹⁸³ The third approach relies on expressing a molecule on the cell surface of T cells that can be targeted with a clinically approved mAb (e.g., CD20 and truncated EGFR).^184,185 At present the clinical experience with these approaches is limited. Finally, the last approach relies on exploiting intrinsic vulnerabilities of T cells with a prime example being the tyrosine kinase inhibitor (TKI) dasatinib, which inhibits TCR and CAR signaling through inhibition of Lck.¹⁸⁶

Improving CAR T Cell Fitness

As mentioned in the Design of CARs section, even outside the immunosuppressive TME, CAR T cells have a limited ability to produce cytokines, expand, and kill when recursively exposed to tumor cells. T cell exhaustion is a widely used term to describe this T cell response to chronic antigen exposure, and there is significant debate in the field how to best define T cell exhaustion.¹⁸⁷ While traditionally cell surface markers, such as PD-1, LAG-3, TIM3, and TIGIT, have been used to define T cell exhaustion, more recent studies have highlighted that expression of transcription factors such as TOX or epigenetic programs might be more suitable to reliably define T cell exhaustion.100, 101, 102, 103^,105 To improve the fitness of CAR T cells, investigators have focused on (1) providing additional signals to promote T cell activation/co-stimulation, (2) providing signal 3 by transgenic expression of cytokines or constitutively active cytokine receptors, (3) silencing or deleting molecules that restrict T cell activation, and (4) transgenic expression of transcription factors.

Provision of Additional Signals to Promote T Cell Activation/Co-stimulation

Standard CARs only provide co-stimulation, and not directly signal 3, a necessary signal for T cell expansion. To overcome this limitation, investigators have incorporated the truncated cytoplasmic domain of IL-2Rβ and a STAT3-binding YXXQ motif into CD28.ζ-CARs targeting CD19.¹⁸⁸ Other groups have shown that T cells expressing CD28.ζ-CARs that incorporate the Toll/IL-1 receptor domain of Toll-like receptor (TLR)2 improve CAR T cell effector function.¹⁸⁹ In both approaches, signals are provided simultaneously, which is in contrast to physiological T cell activation in which signals are provided in a temporospatial fashion. T cells express 41BB upon activation, and studies have demonstrated that expression of the 41BB ligand (41BBL) on the cell surface of CD28.ζ-CAR T cells endows CAR T cells with superior effector function in comparison to incorporating the 41BB signaling domain directly into the CAR.^55,190 Other ligands of the TNF superfamily have also been expressed on the cell surface of CAR T cells, including CD40L.¹⁶⁵ Lastly, investigators are actively exploring the activation of TLR pathways through inducible co-stimulatory molecules containing signaling domains of MyD88,^69,191 the central signaling molecule of TLRs, IL-1β, and IL-18.¹⁹²

Transgenic Expression of Cytokines or Constitutively Active Cytokine Receptors

Cytokines or Cytokine Receptors That Signal through the JAK/STAT Pathway

Common γ-cytokines (IL-2, IL-7, IL-15, IL-21) and IL--12 and IL23 activate the JAK/STAT signaling pathways in T cells. While common γ-cytokines signal through JAK1/JAK2, and predominantly STAT5, IL-12 and IL-23 signal through JAK2, TYK2, and STAT3 and/or STAT4.193, 194, 195 Transgenic expression of these cytokines, demonstrated in numerous preclinical studies, improves the ability of CAR T cells to expand and/or persist, resulting in improved antitumor activity.196, 197, 198 Investigators have not only explored secretory but also membrane-bound versions of cytokines, which might be advantageous since they have the potential to improve cytokine activity (e.g., IL-15) and also restrict most of the cytokine’s action to the genetically modified cell.^199,200 To mitigate systemic side effects of secreted cytokines, investigators have engineered T cells in which cytokine expression is controlled by the nuclear factor of activated T cells (NFAT) promoter. However, while systemic side effects of IL-12 could be mitigated in preclinical studies, one clinical study with NFAT-IL-12-modified tumor-infiltrating lymphocytes (TILs) indicated that the NFAT promoter is insufficient to restrict gene expression to activated T cells.²⁰¹ Another approach to reduce cytokine secretion of gene-modified CAR T cells is to use a construct in which the cytokine gene is placed 3′ prime of an internal ribosomal entry site (IRES).²⁰² A clinical study with MUC16^ecto-CAR/IRES-IL-12 T cells is in progress (ClinicalTrials.gov: NCT02498912). GD2-CAR T cells (ClinicalTrials.gov: NCT03721068) or GD2-CAR iNK T cells (ClinicalTrials.gov: NCT03294954) expressing IL-15 are also currently being evaluated in patients with neuroblastoma. In one of these studies (ClinicalTrials.gov: NCT03721068), CAR/IL-15 T cells are also modified with the iC9 safety switch to allow for the control of potential side effects. Alternative approaches to activate the JAK/STAT pathways are also being pursued, including expression of constitutively active cytokine receptors (e.g., constitutively active IL-7 receptor α [C7R]).⁶⁸

Cytokines That Signal through MyD88

IL-1β and IL-18 signal through MyD88, and at present only transgenic expression of IL-18 has been explored by several groups of investigators. Consistently, IL-18 improved the effector function of CAR T cells not only in xenograft models but also in syngeneic murine models.^163,203,204 However, in some models, administration of CAR/IL-18 T cells was associated with significant toxicities, which could be mitigated in one study by expressing IL-18 under the transcriptional control of the NFAT promoter.²⁰⁴

Silencing or Deletion of Molecules That Restrict T Cell Activation

T cells have developed an intricate system to limit their activation, and investigators have conducted unbiased screens to discover key molecules within this tightly regulated system. While the first screens relied on short hairpin RNA (shRNA) approaches identifying the phosphatase pp2r2d as a negative regulator of αβ TCR T cell activation in tumor models,²⁰⁵ more recent studies have taken advantage of CRISPR-Cas9 gene-editing technology.²⁰⁶ One recently published in vitro screen identified TCEB2, SOCS1, CBLB, and RASA2 as negative regulators of αβ TCR activation.²⁰⁶ An in vivo screen also has highlighted REGNASE1 as a key negative regulator, and REGNASE1 knockout CD19-CAR T cells had improved antitumor activity in a syngeneic leukemia model.²⁰⁷

Modulating Transcription Factors

Transcription factor networks are critical for T cell plasticity, and several recent studies have highlighted their importance. For example, c-Myb (1) is a transcriptional activator of TCF7, a transcription factor critical for memory formation, and (2) represses ZEB2, a transcription factor that promotes T cell differentiation.²⁰⁸ In addition, TOX has emerged as a key transcription factor that enforces T cell exhaustion, and investigators have shown that TOX- and TOX2-deficient murine CAR T cells have improved effector function.100, 101, 102^,209 Other studies have highlighted that the nuclear receptor transcription factors NR4A1 (NUR77), NR4A2 (NURR1), and NR4A3 (NOR1) orchestrate transcription programs that promote T cell exhaustion and that deleting all three factors in CAR T cells is necessary for optimal function.²¹⁰ Lastly, overexpression of c-Jun (transcription factor AP-1) in CAR T cells increases their functional capacities and reduces terminal T cell differentiation, resulting in improved antitumor activity.²¹¹

CAR T Cell Trafficking and Tumor Penetration

In hematological malignancies, CAR T cells and their corresponding target malignancies share hematopoietic origins, and thus have a higher propensity to migrate to similar areas such bone marrow or lymph nodes. In contrast, many solid tumors do not attract T cells. These tumors are called cold tumors, and they can be further subdivided into tumors that do not (cold, ignored) or have (cold, excluded) a T cell infiltrate in their periphery.²¹² T cell migration into tumors is influenced not only by chemokines and adhesion molecules, but also the tumor vasculature and the presence of other immune cells within the tumor.212, 213, 214 Recent studies have highlighted the role of tissue-resident memory T cells for cancer immunotherapies, which express a unique pattern of adhesion molecules and residency markers.²¹⁵ Engineering approaches have so far mainly focused on taking advantage of chemokines that are secreted by tumor cells for which T cells do not express the corresponding chemokine receptor (e.g., CXCL8 by melanoma or CCL2 by neuroblastoma).^216,217 In preclinical models, transgenic expression of chemokine receptors on tumor-specific T cells, including CAR T cells, has been shown to enhance their ability to home to tumor sites, resulting in improved antitumor activity,^216,217 and this approach is being evaluated in a clinical study for the adoptive immunotherapy of TILs in melanoma (ClinicalTrials.gov: NCT01740557). Alternatively, combining the infusion of CAR T cells with the local delivery of an oncolytic adenovirus encoding RANTES and IL-15 has improved homing to and persistence of CAR T cells at tumor sites in preclinical models.²¹⁸

The tumor stroma itself can be extremely dense with various cell types protecting the tumor. Targeting the extracellular matrix (ECM) and degrading these physically protective proteins has also shown promise preclinically. Heparin sulfate proteoglycans (HSPGs) are expressed in the ECM as well as neuronal tissue during development.^219,220 Many cancer cells, especially neuroblastoma, express high levels of HSPGs. Expression of heparinase on CAR T cells to degrade HSPGs has demonstrated improved tumor infiltration and antitumor activity in vivo.²²¹ However, the addition of ECM-degrading enzymes to standard chemotherapies has provided mixed results, with some clinical trials showing decreased survival in patients with the combination therapy.²²²

Counteracting the Immunosuppressive TME

Solid tumor cells are intermixed with suppressive cell populations such as tumor-associated macrophages (TAMs), myeloid-derived suppressor cells (MDSCs), Tregs, and cancer-associated fibroblasts (CAFs) (Figure 3).223, 224, 225 Tumor cells and/or cells of the TME express a broad range of immune checkpoints, including PD-L1 and ligands for LAG-3, TIM-3, and TIGIT. MDSCs, Tregs, and/or CAFs also secrete immunosuppressive cytokines such as transforming growth factor (TGF)-β and IL-10.^226,227 In addition, amino acids that are critical for T cell function such as arginine and glutamine are often depleted within the TME, contributing to limited T cell fitness (Figure 3).^228,229 Some of the approaches that counteract the immunosuppressive microenvironment rely on strategies that were discussed in the section Improving CAR T Cell Fitness. For example, transgenic expression of IL-15 in cytotoxic T cells counteracts Treg-mediated suppression.²³⁰ Herein, we review additional strategies, including (1) directly counteracting immunosuppressive factors, (2) hijacking cytokines or growth factors to promote T cell effector function, (3) improving T cell metabolism in the TME, and (4) targeting non-malignant cells of the tumor stroma.

Figure 3.

Immunosuppressive Solid Tumor Microenvironment

Scheme of the solid tumor microenvironment that consists of (1) tumor vasculature; (2) stromal cells, including cancer-associated fibroblasts; (3) immunosuppressive cells, including myeloid cells, macrophages, and regulatory T cells (Tregs); (4) suppressive metabolites; (5) inhibitory ligands, including PD-1; and (6) suppressive cytokines. See text for additional details.

Directly Counteracting Immunosuppressive Factors

Several preclinical studies have shown that combining PD-1/PD-L1 blockade with CAR T cell therapy improves antitumor activity.²³¹ In an effort to reduce side effects from systemic checkpoint blockade, CAR T cells were genetically modified to express a PD-1/CD28 switch or a truncated PD-1 receptor that functions as a dominant negative receptor (DNR).^70,232 In addition, PD-1 was deleted in CAR T cells by CRISPR-Cas9 gene editing.233, 234, 235 All three approaches resulted in improved effector function of CAR T cells in preclinical models. Thus far the clinical experience with these approaches is limited, and only PD-1 knockout αβTCR T cells have been evaluated clinically.²³⁶ Other approaches include silencing CTLA-4 or FAS expression on the cell surface of tumor-specific or CAR T cells.^237,238 More recently, investigators demonstrated that expression of a DNR FAS receptor also improved the effector function of adoptively transferred tumor-specific T cells.²³⁹

Hijacking Cytokines or Growth Factors to Promote T Cell Effector Function

The TME is rich with cytokines that either inhibit T cell function or for which T cells do not express the corresponding receptors. The inhibitory effects of these cytokines can be negated by expressing DNRs or cytokine switch receptors (CSRs) that convert an inhibitory signal into a growth-promoting signal. Examples of cytokine DNRs include DNR-TGF-β receptors.²⁴⁰ DNR-TGF-β-expressing EBV-specific T cells for EBV⁺ lymphoma have been evaluated in early phase clinical studies, demonstrating safety and suggesting improved efficacy in comparison to their unmodified counterparts.²⁴¹ CSRs have been developed to take advantage of IL-4 produced within the TME, including IL-4/IL-2, IL-4/IL-7, and IL-4/IL-21 CSRs.242, 243, 244 In addition, a TGF-β CSR, consisting of a TGF-β-specific scFv and a CD28z signaling domain, has shown promise in preclinical CAR T cell therapy models.²⁴⁵ Of these, the IL-4/IL-2 CSR is currently being evaluated in one early phase clinical study in the context of TE1-CAR T cells (ClinicalTrials.gov: NCT01818323). Lastly, colony-stimulating factor-1 (CSF-1) is an example of a cytokine that is present in the TME and for which T cells do not express the cognate receptor. One preclinical study has shown that expression of CSF-1R in CAR T cells improves their effector function in a CSF-1-dependent manner.²⁴⁶

Improving Metabolic Fitness of CAR T Cells in the TME

The TME is hypoxic and deprived of critical nutrients that are required for T cell proliferation. In addition, the TME (1) is enriched in metabolic end products that are immunosuppressive, for example, R-2-hydroxyglutarate in tumors with isocitrate dehydrogenase 1/2 mutations;247, 248, 249 (2) has increased concentrations of electrolytes, such as potassium, that are immunosuppressive;²⁵⁰ and (3) harbors reactive oxygen species (ROS) that impair T cell function.^251,252 Several strategies are being actively explored to improve the metabolic function of adoptively transferred T cells. These include ex vivo loading of tumor-specific T cells with arginine,²²⁸ or the genetic modification of CAR T cells with enzymes that are critical for arginine re-synthesis, including argininosuccinate synthase or ornithine transcarbamylase.²⁵³ Investigators have also focused on manipulating glutamine metabolism in the TME, and recent studies have highlighted that glutamine antagonists and transient inhibition of glutaminase increase T cell effector function.^254,255 To protect CAR T cells from ROS, investigators have genetically modified T cells to secrete catalase. This not only protected T cells from ROS damage but also protected unmodified bystander immune cells.²⁵⁶ Lastly, small molecules such a metformin are being explored to reduce tumor hypoxia.²⁵⁷

Targeting Non-malignant Cells of the Tumor Stroma

CAFs, MDSCs, TAMs, and tumor-associated endothelial cells are present in the tumor stroma, and while not malignant, they promote tumor growth/metastasis and immunosuppression and/or secrete components to the ECM that hinder immune cell penetration. CAFs have been targeted by several groups with fibroblast activation protein (FAP)-CAR T cells. While FAP-CAR T cells had potent antitumor activity in preclinical models, some safety concerns were raised due to FAP expression on stromal cells in healthy tissues; nevertheless, this approach has been translated into the clinic.258, 259, 260, 261 TAMs have been targeted with CD123- or folate receptor β (FRβ)-CAR T cells in preclinical models,^262,263 and CSF1R-CARs have been developed, opening up the possibility to target CSF1R-positive TAMs.²⁶⁴ MDSCs express high levels of NKG2D ligands, and in one preclinical model, combining NKG2D-CAR NK cells with GD2-CAR T cells resulted in improved antitumor activity for neuroblastoma.²⁶⁵ Lastly, endothelial cells of the tumor vasculature have been targeted with CARs specific for VEGF-R2, PSMA, or the EDB/EIIIB splice variant of fibronectin in preclinical models.266, 267, 268 Of these approaches, only VEGF-R2 CAR T cells have been successfully evaluated in one clinical study. While the results have not been published, they are available (ClinicalTrials.gov: NCT01218867), and out of 23 infused patients, only 1 patient achieved a partial response.

Combinatorial Therapies

While this review has focused on additional genetic modification to improve the effector function of CAR T cells, numerous studies are underway to combine CAR T cell therapy with other treatment modalities to improve outcomes by creating therapeutic synergies.^269,270 These include efforts, as already mentioned in this review, to combine CAR T cell therapy with cytokine administration, checkpoint blockade, oncolytic viruses, radiation, and/or vaccines. Examples for oncolytics/CAR T cell therapy combinations include the administration of oncolytic viruses encoding BiTEs, cytokines, and/or checkpoint inhibitors,271, 272, 273 or oncolytic viruses that encode CAR targets.^274,275 In addition, investigators have explored the use of T cells to directly deliver viruses into tumors.276, 277, 278 Checkpoint blockade/CAR T cell therapy combinations are actively being pursued,²⁷⁹ including the aforementioned clinical studies combining mesothelin-CAR T cells with pembrolizumab for the treatment of mesothelioma.⁷⁶ Furthermore, several studies have highlighted the benefit of combining radiotherapy with the administration of CAR T cells, and early clinical studies for patients with B cell lymphoma have produced encouraging results.²⁸⁰ Recent preclinical studies have also demonstrated the benefit of boosting CAR T cells outside the immunosuppressive TME with amphiphile CAR ligands or RNA-based vaccines.^281,282

Other approaches include combining CAR T cells with small molecules that target cancer cells that do not affect CAR T cells, with the best example being combining the BTK inhibitor ibrutinib with CD19-CAR T cell therapies for lymphoma.²⁸³ In addition, combining CAR T cells with chemotherapeutic agents that upregulate the expression of the targeted antigens is an attractive option to enhance the antitumor activity of CAR T cells that is actively being pursued.²⁸⁴ Finally, combinational treatment of HER2-CAR T cells with the smac-mimetic birinapant, which promotes apoptosis, significantly enhanced antitumor activity in a TNF-dependent manner.²⁸⁵ Thus, there is a myriad of combinatorial approaches, and selection of an optimal combination will most likely depend on the targeted tumor type and disease status. For example, brain tumors, which often present as local disease or with limited sites of recurrences, might be ideal tumors to combine local radiation or delivery of an oncolytic virus with CAR T cell therapy. In contrast, combining local delivery of agents with CAR T cell therapy for metastatic cancer should only be pursued if treating selected metastases induces an abscopal effect,²⁸⁶ leading to the regression of distant, untreated metastases. Feasibility and costs are also a major consideration; in this regard, repurposing FDA-approved drugs for combinatorial CAR T cell therapy should actively be explored as for other cancer therapeutics.^287,288

Conclusions

CAR T cells for solid tumors so far have limited activity in early phase clinical studies, raising concerns of whether CAR T cells can be developed into a curative therapeutic approach in the future. While currently ineffective, clinical and preclinical studies have identified roadblocks, including (1) an antigen dilemma, (2) T cell fitness and survival before reaching the tumor site, (3) inability of T cells to efficiently traffic to tumor sites and penetrate physical barriers, and (4) an immunosuppressive TME. In the last decade, investigators have developed elegant solutions and/or countermeasures in preclinical studies, engineering CAR T cells with significantly improved effector function and/or combining them with other treatment modalities. Thus, we strongly believe that the future of CAR T cells for solid tumors is bright, and they will not become a mere footnote in the history of cancer immunotherapy. A key challenge will be how to prioritize and streamline the clinical evaluation of the developed approaches. In addition, there is a continued need to advance our understanding in basic CAR T cell biology and how cancer cells and resident immune cells interact with adoptively transferred CAR T cells. Lastly, future CAR T cell design might benefit from considering the “ten principles for good design” that were put forward by the eminent industrial designer Dieter Rams:²⁸⁹ good design (1) is innovative, (2) makes a product useful, (3) is aesthetic, (4) makes a product understandable, (5) is unobtrusive, (6) is honest, (7) is long-lasting, (8) is thorough down to the last detail, (9) is environmentally friendly, and (10) involves as little design as possible.

Author Contributions

J.W., E.W., and S.G. wrote the manuscript. All authors (J.W., E.W., C.D., and S.G.) reviewed and edited the manuscript.

Conflicts of Interest

J.W., C.D., and S.G. have patent applications in the fields of T cell and/or gene therapy for cancer. S.G. has a research collaboration with TESSA Therapeutics, is a DSMB member of Immatics, and is on the Scientific Advisory Board of Tidal.