Immunologically “hot” solid tumors (e.g. melanoma) (1) with a tumor microenvironment (TME) marked by infiltrating CD8+ T-cells (2, 3), high programmed death ligand 1 (PD-L1) expression (4), or a high tumor mutational burden have shown remarkable responses to immunotherapy including immune checkpoint inhibitors (ICIs) (5). Unfortunately, these benefits have not extended to “cold” tumors (e.g. prostate or pancreatic cancer) (1) where T-cells are either entirely absent (“immune desert”) or sequestered at the periphery (“immune-excluded”) (3, 6). Many pediatric/adolescent and young adult solid tumors are cold tumors (7, 8) and have failed to respond to ICIs (9).

Several approaches have attempted to harness cellular therapy to cure these tumors. Autologous hematopoietic stem cell transplant (HSCT) has enabled maximal chemotherapy dosing in susceptible tumors with varying levels of effectiveness in neuroblastoma (10), Ewing sarcoma (11), breast cancer (12), retinoblastoma (13), hepatoblastoma (14), and other diseases. Recently some groups have piloted allogeneic HSCT to treat solid tumors. Though durable responses are rare, evidence for graft-vs-tumor effect has been observed (15). Finally, as adoptive cellular therapy (ACT) has proven transformative for leukemia and lymphoma, the development of novel ACT for solid tumors has exploded ( Figure 1 ). In this review, we discuss ACT in solid tumors in clinical development, consider challenges plaguing the field, and highlight proposed strategies which will be tested in future clinical trials.

T-cells are critical in immune surveillance for cancer. The T-cell receptor (TCR) can recognize cancer-specific antigens processed by major histocompatibility complex (MHC) and presented on the cell surface. TCR engagement by MHC-presented non-self antigens leads to activation and T-cell mediated killing (16). T-cell cytotoxicity in solid tumors has been leveraged using both native T-cells and autologous T-cells genetically engineered to express a specific TCR. Future efforts in allogeneic “off-the-shelf” approaches are being actively studied.

NK-cells are innate immune cells and protect against infections and cancer (57, 58). Efforts to harness NK-cell biology for ACT in cancer treatment has gained considerable interest as an alternative to T-cell based immunotherapeutics. NK-cells possess qualities which may allow them to overcome the hostile TME (58–60). While T-cells recognize unique tumor antigens, NK-cell-mediated cytotoxicity depends on the sum of activating and inhibitory signals, including tumor cell lack of MHC class 1 expression or antibody-dependent cell-mediated cytotoxicity (61). Furthermore, NK-cells can produce inflammatory cytokines such as IFNγ and TNFα which can activate CD8+ TILs and enhance their cytotoxicity (62).

These properties allow NK-cells to be engineered or manipulated via different mechanisms from T-cell-centric immunotherapies. Examples include the administration of agonist cytokines or engineering NK-cells which constitutively secrete these cytokines (63). Others have proposed NK-cells which constitutively secrete chemotactic factors to recruit cytotoxic lymphocytes to the TME (64). Tri-specific NK-cell engagers (TriKEs) have been proposed to confer tumor-specificity to NK-cells and enhance NK-cell activation by engaging stimulatory receptors such as the IL-15 receptor (65, 66). Additionally, CAR NK-cells (CAR-NK) designed from stem cell progenitors represent another way to generate tumor-specific NK-cells. Attractively, CAR-NK may be less toxic and could be produced at lower cost than CART (67). Recent experience with CD19-CAR-NK in B-cell malignancies provides proof-of-concept that this strategy can be safely and effectively utilized and with potential for persistence (68). Barriers remain to production and monitoring of persistence of these cells, but additional alterations to the NK-cell product and manufacturing strategies have been proposed to mitigate these issues. Finally, NK-cells also express immune checkpoint molecules such as PD-1, and either combination with ICIs or intrinsic downregulation of these checkpoint molecules have been proposed as mechanisms to further enhance the efficacy of NK-cell-based approaches (69, 70).

Myeloid cells readily infiltrate primary tumors and metastases. Harnessing this property for ACT shows promise in the treatment of solid malignancies (71). Myeloid cells are highly plastic and may acquire a wide spectrum of immune-stimulatory or immune-suppressive phenotypes in response to the local milieu. Tumor associated macrophages (TAMs) are polarized to an anti-tumor M1 phenotype in response to pro-inflammatory factors such as IFNγ, GM-CSF and lipopolysaccharide. M1 TAMs promote Th1 responses, phagocytosis of tumor cells, and antigen presentation. Tumor-associated cytokines such as IL-10, IL-4, IL-13 and TGF-β promote polarization towards an immunosuppressive M2 phenotype. M2 TAMs promote tumor progression through mechanisms including angiogenesis, extracellular matrix (ECM) remodeling and regulatory T-cell recruitment (72). This M1/M2 classification is an oversimplification, however induction of an M1-like, anti-tumor phenotype is important for the success of myeloid-based ACT. The first myeloid-based ACT utilized macrophages polarized to the M1 phenotype ex vivo with IFNγ. Clinical trials showed limited efficacy, but these therapies were generally well-tolerated (73–75).

Subsequent work has focused on engineering myeloid cells towards a more potent and durable anti-tumor phenotype. Anti-HER2 CAR-macrophages (CARM) reduced tumor growth and prolonged survival while reprogramming the immune-suppressive TME in xenograft models (76). A first-in-human trial evaluating CARM is now underway in HER2-overexpressing solid tumors (NCT04660929). Preclinical work has shown that myeloid cells can also be used to deliver cargo to the TME. Administration of myeloid cells genetically engineered to express IL-12, a potent anti-tumor cytokine, resulted in durable cures in a syngeneic model of embryonal rhabdomyosarcoma through activation of T-cell responses in the tumor and metastatic microenvironment (77).

Significant remaining challenges for optimization of solid tumor ACT are outlined in this section. Additionally, we will summarize proposed strategies to overcome these challenges ( Figure 2 ).

Selection of antigens such as GD2 (78) and CTAs (79), which are expressed on numerous solid tumors, leverages the possibility that a single ACT could be active across multiple histologies. However, few antigens are tumor-specific. Thus, identifying a target antigen which will allow tumor clearance without unacceptable normal tissue toxicity (on-target/off-tumor effect) is problematic. In addition to selecting the proper target antigen, low antigen density and antigen downregulation within heterogeneous solid TMEs have emerged as additional barriers to ACT (80, 81).

ACT trafficking is also challenging in solid tumors. Trafficking can be inhibited by physical barriers, loss of MHC class 1 expression, repellent cytokine gradients, expression of inhibitory ligands such as PD-L1, and abnormal tumor vasculature (82). CNS tumors are further shielded by the blood-brain barrier (83). If ACTs cannot traffic to the tumor and engage their target antigen, they fail to be activated and expand, leading to rapid loss of ACT.

Finally, the TME present in many solid tumors is hostile to the ACT. Tumors recruit immunosuppressive TAMs and myeloid-derived suppressor cells (MDSCs) (84) which express inhibitory molecules such as PD-L1 (84), secrete inactivating cytokines such as IL-10 (85), and promote a hypoxic TME (86) which can thwart ACT cytotoxicity. These tumor-sustaining programs promote rapid and irreversible ACT exhaustion, inhibit expansion, and result in failure of tumor clearance. Further ACT engineering or combination with agents to allow ACTs to overcome these challenges, will be necessary for ACT optimization in solid tumors.

Aberrant tumor vasculature and ECM deposition impede ACT trafficking. Regional ACT administration is one strategy to overcome this hurdle. A recent phase I trial demonstrated that intrapleural administration of mesothelin-CART combined with pembrolizumab was safe and feasible, and showed potential efficacy with 2 patients demonstrating metabolic CR on PET scan (87). Intraventricular CART administration for both primary brain tumors and CNS metastases is also under evaluation in early-phase clinical trials and in preclinical models (88, 89) (NCT04196413). Additionally, methods to disrupt the blood-brain barrier to allow trafficking of ACT to CNS tumors, such as focused ultrasound (90) or other mechanical or pharmacological methods (91) have been piloted.

Further genetic modification of ACTs to overcome and leverage features of the hostile TME is currently being explored. Many solid tumor types recruit TAMs by producing chemokines such as CXCL8 and CXCL2. Preclinical data suggest that chemokine secretion can be leveraged to enhance CART trafficking by engineering CART to express chemokine receptors. For example, CXCR2-modified GPC3-CART had improved trafficking in a hepatocellular carcinoma model (92), while a CXCR1/2-modified CD70-CART enhanced CART trafficking and efficacy in murine GBM, ovarian cancer and pancreatic cancer models (93). Many groups have also sought to generate a more “fit” ACT through enhanced cytokine secretion [thoroughly reviewed by Bell and Gottschalk (94)]. Additional modifications include creation of ACT which is resistant to exhaustion [e.g. DNA methyltransferase 3 alpha knock-out (95) or PD-1 deletion (96)] or tuning ACT to be effective despite low antigen density [e.g. c-Jun overexpression (97, 98)]. These modifications of ACTs are now entering clinical trials (e.g. TGF-βR knockout CART NCT04976218).

Tumor-associated vasculature is characterized by pericyte loss, resulting in leakiness and adhesion molecule down-regulation impairing T-cell migration into the tumor (99). VEGF inhibitors, which promote vascular normalization, may enhance CD8+ T-cell infiltration into tumors (100). Anti-VEGF agents have shown synergy with ICIs in select solid malignancies, resulting in FDA approval of these combinations in hepatocellular carcinoma and renal cell carcinoma (101). Preclinical studies suggest that antiangiogenics can also improve ACT trafficking (102, 103). Combining ACTs with antiangiogenics warrants further study in clinical trials.

ECM-remodeling agents may enhance the ability of ACTs to infiltrate tumors. In gastric cancer models, hyaluronic acid reduced mesothelin-CART infiltration, however these CART had superior efficacy when combined with infusion of a secreted form of the human hyaluronidase PH20 (104). CART engineered to express heparinase, which degrades heparan sulfate proteoglycans, showed superior anti-tumor activity and were associated with increased T-cell infiltration in preclinical models (105).

The solid TME contributes to T-cell exhaustion via multiple mechanisms, including repeated TCR stimulation and metabolic stress, thereby reducing the ACT efficacy. Engineering CART to reduce tonic signaling through incorporation of the 4-1BB costimulatory domain vs CD28 costimulatory domain showed reduction in CART exhaustion and enhanced persistence and efficacy in preclinical studies (106). Induction of transient rest periods in CART, such as by dasatinib utilization, has shown exhaustion reversal and improved efficacy (107). A dasatinib-containing culture platform is being used to manufacture GD2-CART in ongoing clinical trials (NCT04539366, NCT04196413). CART combination with ICIs is also under evaluation in clinical trials (108).

The ability of myeloid cells to orchestrate immune responses in the TME makes them an attractive therapeutic target. Low-dose chemotherapy has shown reduction of tumor MDSCs (109–111). MDSC differentiation with ATRA reduced their immune-suppressive function and enhanced efficacy of GD2-CART in preclinical models (112). In a pilot trial studying ipilimumab vs ipilimumab combined with ATRA, patients receiving ATRA had fewer circulating MDSCs (113). Inhibiting myeloid cell trafficking through CSF1R inhibition is another potential avenue to reduce myeloid cell immune-suppression in the TME. CSF1R-targeting agents are generally well-tolerated in the clinic, and the multi-TKI CSF1R inhibitor Pexidartinib is FDA-approved to treat tenosynovial giant-cell tumor (114, 115). Clinical trials studying CSF1R inhibitors with ICIs are underway (NCT02777710, NCT02829723, NCT03502330, NCT04848116, NCT02526017).

While ACT has yet to yield the transformative results in solid tumors that CART have shown for hematologic malignancies, evidence exists that some patients with solid tumors may respond to ACT. T-cells, NK-cells, and myeloid cells have each been engineered to target these tumors, and each have advantages and unique challenges. Further engineering ACTs to overcome tumor immune resistance mechanisms and better understanding how to combine with TME-modifying agents will be critical to expanding the number of patients with solid tumors who may derive therapeutic benefit.

JL, KW, and JG wrote the first version of the manuscript. NS provided critical feedback and additions. All authors contributed to the final version of the manuscript.

The content of this publication does not necessarily reflect the views of policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.