Adoptive Cell Therapy in Breast Cancer: A Current Perspective of Next-Generation Medicine

Immunotherapy has become a cornerstone in the treatment of cancer and changed the way clinicians and researchers approach tumor vulnerabilities. Durable responses are commonly observed with immune checkpoint inhibitors in highly immunogenic tumors, while the infusion of T cells genetically engineered to express chimeric antigen receptors (CARs) has shown impressive efficacy in certain types of blood cancer. Nevertheless, harnessing our own immunity has not proved successful for most breast cancer patients. In the era of genomic medicine, cellular immunotherapies may provide a more personalized and dynamic tool against tumors displaying heterogeneous mutational landscapes and antigenic pools. This approach encompasses multiple strategies including the adoptive transfer of tumor-infiltrating lymphocytes, dendritic cells, natural killer cells, and engineered immune components such as CAR constructs and engineered T cell receptors. Although far from permeating the clinical setting, technical advances have been overwhelming in recent years, with continuous improvement in traditional challenges such as toxicity, adoptive cell persistence, and intratumoral trafficking. Also, there is an avid search for neoantigens that can be targeted by these strategies, either alone or in combination. In this work, we aim to provide a clinically-oriented overview of preclinical and clinical data regarding the use of cellular immunotherapies in breast cancer.


INTRODUCTION
Breast cancer (BC) is a leading cause of death worldwide and remains mostly incurable in advanced stages (1). Tumor initiation and progression is continuously controlled by innate and adaptive immune cells, which falter as cancer cells undergo mesenchymal dedifferentiation and/or evolve different mechanisms of tumor escape (2). In general, BC is not regarded as an inflamed tumor, triple negative BC (TNBC) and HER2 + tumors being more immunogenic than the most common luminal A-like subtype (3). Immunotherapeutic strategies against BC have traditionally been based on "passive immunotherapy" such as the HER2 blocking antibody trastuzumab. Encouraged by the success of immune checkpoint inhibitors (ICIs) in melanoma and lung cancer, numerous trials have tested the use of this "active immunotherapy" in BC with overall disappointing results (4). In the metastatic setting, the most significant achievement was observed in the IMpassion130 phase III trial, which demonstrated an increase in progression-free survival in TNBC patients receiving atezolizumab plus nab-paclitaxel compared to nab-paclitaxel alone (7.2 vs 5.5 months) (5). This humble benefit did not lead to a better overall survival and was not recapitulated when using paclitaxel as concomitant chemotherapy nor consistently associated to any predictive biomarker other than PD-L1 (6). Findings seem to be more clinically meaningful in the neoadjuvant setting, in which an increased pathological complete response rate has been reported in patients receiving atezolizumab (58 vs 41% for total population, 69 vs 49% in PD-L1 positive tumors) (7). This body of evidence underscores the need of a better understanding of the tumorimmune interaction, escape mechanisms, and the role of the microenvironment when a high tumor burden exists. Globally, the use of ICIs in BC would at best provide a nonspecific approach, guided by poorly understood biomarkers, to harnessing a debilitated immune system against a cold tumor. Instead, the development of omic-scale repositories and high-throughput technologies enable us to decode the genomic traits of each unique tumor and calls for the design of more specific and flexible immunotherapies, capable of targeting oncogenic addictions and overcoming temporal and spatial mutational heterogeneities. Thus, the aim of our work is to bridge the complex body of evidence on the different types of adoptive cell therapy (ACT) and the clinicians who everyday care for BC patients.

Tumor-Infiltrating Lymphocytes (TIL) Therapy
The adoptive transfer of lymphocytes to treat BC has been attempted in numerous occasions. Allogeneic stem cell transplants in addition to high-dose chemotherapy achieved successful longterm outcomes but arouse significant safety concerns, whereas ACT with autologous circulating lymphocytes conditioned in vitro was better tolerated but showed less efficacy (8)(9)(10)(11). Tumor-infiltrating lymphocytes (TILs) include a subset of naturally occurring T cells capable of targeting neoantigens encoded by genes harboring nonsynonymus somatic mutations (12,13). BC, particularly HER2 + and luminal-like tumors, have been traditionally considered as poorly immunogenic, with low numbers of TILs and a limited burden of neoantigens (3,14,15). However, a robust correlation exists between increased stromal TILs and a better prognosis in TNBC (16)(17)(18)(19).
Adoptive transfer of autologous TILs was first described as a treatment modality by Rosenberg and colleagues in 1987 ( Figure  1A) (20). Substantial objective responses have been observed in patients with tumors with high mutation rates such as melanoma, lung or bladder cancer (20)(21)(22). However, with few exceptions, the infusion of unselected heterogenous TILs appears mostly ineffective in epithelial malignancies (23)(24)(25)(26)(27). In order to boost tumor recognition and killing efficacy, TIL therapy has been refined by selecting TILs reactive for tumor antigens (TAs) identified by whole-exome sequencing and RNA sequencing. Zacharakis et al. recently described the case of a 49 year-old woman with ER + /HER2metastatic BC refractory to multiple lines of chemotherapy, who exhibited a complete durable regression after ACT with TA-specific TILs in conjunction with IL-2 and an anti-PD1 agent (28). In this particular case, the genomic analysis of a right breast subcutaneous lesion revealed the presence of 62 nonsynonymous somatic mutations, of which the mutant versions of 4 proteins rendered the highest TIL reactivity. Further, a relevant impact of the concomitant anti-PD1 therapy was unlikely since no expression of PD-L1 was detected in tumor biopsies. A similar approach was used for a pulmonary metastasis of a TNBC patient, where an immunogenic mutation was found among 72 nonsynonymous mutations (29). However, outcome data from this tailored TIL therapy was not reported. Four clinical initiatives have been registered to date in ClinicalTrials.gov and are briefly displayed in Table 1. Notably, only two of them incorporate preconditioning with non-myeloablative chemotherapy regimens, and one of them will address the role of an anti-PD1 agent as concurrent medication. In sum, the transfer of selected autologous TILs primed against multiple MHC-restricted TAs may provide a safe and personalized option for patients with advanced BC.

Engineered T Cell Receptor (TCR) and Chimeric Antigen Receptor (CAR) Therapy
Gene transfer-based strategies have been developed to overcome the main challenges of TIL therapy, including the low yield of TIL expansion, the low affinity of human TCRs for TAs, and the immune tolerance elicited by the downregulation of MHC molecules and TAs (30). Both TCR and chimeric antigen receptor (CAR) gene transfer endow polyclonal T cells with reactivities that are not naturally present against TAs of choice and thus provide an adaptable and highly subtle tool for personalized medicine ( Figures 1A, B) (31).
The majority of engineered abTCRs recognize epitopes presented by MHC molecules, thereby narrowing down the group of potential targets to those which are MHC-restricted, and exhibit an increased specificity recognition and affinity for tumor cells ( Figure 1A) (32,33). Mounting clinical evidence on several tumor types along with preclinical data on BC underscores the rationale for TCR use in BC patients (34)(35)(36)(37). Of note, in both hormone-dependent and independent BC cell lines and in xenograft mice, Li et al. reported a notable enhancement of anti-tumor cytotoxicity by CD8 + T cells transduced with an MHC-A2-restricted placenta-specific 1 (PLAC1)-TCR molecule (38). However, to the best of our knowledge, evidence on humans is still lacking, with many ongoing clinical trials testing intravenous infusions of TCR-engineered T cells against TAs such as HER2, NYESO-1, and MAGE-A3 (Table 1). Interestingly, some of them will assess the value of adding anti-PD1 therapy to enhance immune reconstitution after lymphodepleting chemotherapy and cytotoxicity.
In order to bypass the limitations of MHC restriction of conventional abTCRs, intensive research has focused on the development of CARs and, more recently, on the gdT cell In general, TILs are enzymatically isolated, activated with high-dose IL-2, and eventually expanded for therapeutic use. More recently, they can also be screened for a high avidity for TAs. TCR transfer, usually using viral vectors on circulating T cells, endows T cells with TCRs with high affinity for TAs. Further, to help condition the body for the T cell transplant, patients often receive a non-myeloablative lymphodepleting chemotherapy regimen before IV infusion, which facilitates the access to growth-promoting cytokines and removes suppressor cells. The role of concomitant immunomodulatory therapies is yet to be elucidated. In both approaches, recognition of cognate TAs is MHC-restricted. (B) CAR engineering of circulating T cells has been progressively refined. First-generation CARs include only a CD3z chain as intracellular signaling domain; secondgeneration CARs add a single co-stimulatory domain, such as CD28, 4-1BB (CD137), CD27, or OX40; third-generation CARs add two or three co-stimulatory domains; fourth-generation CARs, also known as TRUCKs (T cells redirected for antigen-unrestricted cytokine-initiated killing) are further armored with potent antitumor cytokines and co-stimulatory ligands. CARs target a wide range of surface TAs in an MHC-independent manner, and multiple trials are currently testing the feasibility and efficacy of different administration routes. (C) DCs can be generated from PBMNC and HSPCs and become mature after being pulsed using a growing set of TA loading mechanisms. In trials, DCs are infused IV but also as IT or IN vaccines.  compartment. gdT cells exhibit potent anti-tumor responses by bridging innate and adaptive immunities, since they incorporate both gdTCRs and killer cell immunoglobulin-like receptors (KIRs) (39,40). Also, gdT cell ligand recognition requires the expression of accessory costimulatory molecules, which may prevent harmful selfreactivity. Infiltration by gdT cells has been associated with improved outcomes in a small cohort of TNBC patients (41). Consistently, ACT of gdT cells together with trastuzumab improved control of tumor growth as compared to trastuzumab alone in a mouse model of HER2 + BC (42). However, the function of the gdT cells may be extremely pleiotropic. In this regard, Peng et al. described a BC-infiltrating gdT cell subset with strong immunosuppressive effects on T cells and DCs regulated via the Toll-like receptor 8, thus suggesting that its depletion or reversal could enhance anti-tumor responses (43). ACT with unmodified or engineered gdT cells emerges as an appealing prospect for BC immunotherapy, but further functional characterization and data on clinical interventions are still required (44). On the other hand, CAR-T cells are T cells engineered to express an artificial receptor with a modular design consisting of an extracellular ligand-binding domain, usually a single-chain antibody, a hinge, a transmembrane domain, and a cytoplasmic signaling domain, with increasing complexity and functionality across the four generations of CAR constructs ( Figure 1B) (45)(46)(47). Compared to TILs, CAR-T cells are not as affected by the hurdles of isolation, expansion, and persistence limitation of natural tumor-specific T cells. Moreover, CAR recognition occurs in an MHC-independent manner, which helps overcome MHC downregulation as a mechanism of tumor escape, and can also recognize carbohydrate and glycolipid antigens (46). Yet, cognate antigens are consequently restricted to surface molecules. Numerous preclinical studies in vitro and in vivo have evaluated the use of CAR-T cells armed to specifically target TAs in BC, with HER2-CAR constructs attracting the most attention and achieving robust tumor regressions (48)(49)(50)(51)(52)(53)(54)(55)(56)(57)(58)(59)(60). To our knowledge, only one phase I trial has been published testing a HER2-CAR in BC patients. In the study by Lum et al., 23 metastatic BC patients independent of their HER status received 8 infusions of anti-CD3/HER2 bispecific antibody-armed T cells. In the evaluable patients at 14.5 weeks, 13 patients experienced clinical benefit, including 2 objective responses (61). Notwithstanding, serious adverse events have been reported following the use of HER2-CARs. The first evidence on the clinical use of HER2-CAR-T cells was a case report of a patient with metastatic HER2 + colon cancer in whom the administration of a 3 rd generation HER2-CAR was followed by multiple cardiac arrests, respiratory distress, and multiorgan damage (62). This harm was attributed to an inflammatory cytokine release elicited by the immune-mediated recognition of HER2 in normal epithelial tissues, which is referred to as "on-target, off-tumor" toxicity.
Besides HER2, the single injection of accessible lesions with CAR-T cells targeting c-Met, a cell-surface protein tyrosine kinase aberrantly expressed in BC, in a group of 6 patients with metastatic BC comprised by two ER + tumors and 4 TN tumors, did not render measurable responses but elicited extensive tumor necrosis and loss of c-Met immunoreactivity at the injection site, and also translated into detectable levels of c-Met-CAR-T cell mRNA in peripheral blood (63). Similarly, Specht et al. recently communicated preliminary safety results of a phase I trial targeting ROR-1, a tyrosine kinase protein expressed in TNBC and associated with a worse prognosis (64,65). Interestingly, patients received a 2 nd generation ROR1-CAR engineered with a truncated EGFR molecule to permit the elimination of infused cells in case of toxicity (64). Only 6 patients had been enrolled with no adverse events observed, but further update is expected to support this innovative approach. A considerable number of trials are testing CAR constructs against multiple TAs in BC ( Table 1). We expect that these studies also convey relevant information about on-target, offtumor effects, and the benefits of the different administration routes, preconditioning or concomitant immunomodulatory therapies. In addition, it seems clear that a thorough genomicscale understanding of molecular vulnerabilities and antigenic shifts will be of paramount importance in the design of CARbased strategies.

DENDRITIC CELL (DC) THERAPY
Dendritic cells (DCs) are particularly well-suited for BC immunotherapy due to their ability to sensitize CD8 + T cells and also CD4 + T cells capable of generating memory T cells and contribute with additional cytotoxicity against tumors (66). DCs have been found infiltrating BC specimens in nearly half of the patients with either early or advanced disease, but are mostly relegated to the periphery, functionally compromised, and show a poor correlation with outcome (67)(68)(69)(70).
Autologous DCs may be fused with tumor cells or pulsed with tumor lysates or TAs to activate T cells against tumors ( Figure  1C) (71-74). Across these strategies, DCs may be either exposed to one particular neoantigen or to the entire repertoire of TAs, including those yet to be identified. In contrast to what was observed in TIL and TCR therapies, DCs can be obtained in large numbers from bone marrow precursors and monocyte-derived DCs from peripheral blood (75). The pioneering study by Brossart et al. evaluated the vaccination with autologous DCs pulsed with HER2 or MUC1-derived peptides in 7 BC patients. Although the clinical outcomes were disappointing, peptidespecific T cell responses could be detected even at 9 months after initiation of vaccinations, and T cell responses against epitopes not used for vaccination were identified as a result of cross priming (76). More encouraging objective responses were achieved by Avigan et al. in a phase I trial testing the vaccination with DCs fused with autologous tumor cells in 16 patients with metastatic BC (77). These included 2 patients attaining a partial response and 6 patients attaining a stable disease, although the anti-tumor effects were not maintained over time. In the neo/ adjuvant setting, vaccination with autologous HER2-pulsed DCs achieved a modest rate of pathological complete responses in HER2 + BC patients, which yet correlated poorly with immune surrogates in peripheral blood (78). This study, however, demonstrated that intralesional and intranodal routes of administration may not substantially differ in terms of antitumor efficacy, thus facilitating vaccination when tumor locations are challenging. Likewise, the trial conducted by Qi et al. in stage II-IIIA ER -/PR -BC patients reported a 3-year relapsefree survival of 71% versus 31%, with and without vaccination, respectively (79). Other promising approaches consist of adding cytokine adjuvants, such as IL-2, or targeting both the innate and adaptive immune systems by complementing DCs with cytokine-induced killer cells, although the response to these strategies has so far been humble or confused by the effect of concurrent chemotherapies (80)(81)(82). More than 20 trials are registered to date testing DC vaccinations in BC patients of all major pathological and most of them are designed to pulse DCs with TAs of choice ( Table 1). Although ACT with DCs has not yet materialized in a relevant clinical benefit, we believe that the role of DCs as stimulators of T-cell response and long-term memory, and their safety and ease of manufacture, may justify further development alone or in combination with other T cell therapies.

NATURAL KILLER (NK) CELL THERAPY
Different from the previous approaches, NK cells represent an attractive asset for cancer immunotherapy due to their innate ability to eliminate cancer cells in an MHC-independent and non-TA-restricted manner. The "loss of self" mediated by the downregulation of MHC molecules as a mechanism of tumor escape hinders the recognition of cancer cells by CD8 + T cells but unleashes the activity of NK cells, which are regulated by the interplay of activating and inhibitory receptors such as KIRs and natural killer group 2D (NKG2D) (83,84).
Activated NK cells can be manufactured in large numbers ex vivo from primary NK cells, hemopoietic stem cells, and clonal cell lines, of which the NK-92 is approved by the US FDA for use in clinical trials ( Figure 1D) (85)(86)(87). So far, adoptive transfer of autologous NK cells has been tested in a wide range of solid malignancies with poor clinical efficacy, which has been explained by the immunosuppressive state of the host and because the inhibitory receptors on autologous NK cells matched molecules exhibited on the tumor cell surface (87)(88)(89). Anecdotally, a report by Tian et al. described a partial response in a patient with progressing metastatic HER2 + BC who underwent treatment with trastuzumab-treated NK cells, which was consistent with an increased activation and expansion of NK cells mediated by trastuzumab in vitro (90). Allogeneic NK cells, however, have not proved to do much better in BC patients, with only one phase II trial published describing 4 patients with stabilized disease from a total of 6 patients evaluated at 4-6 weeks from infusion and after pre-conditioning with lymphodepleting chemotherapy and total body irradiation (91).
In order to enhance their cytotoxic properties, NK cells are also being modified with the addition of CARs against specific TAs. Compared to CAR-T cells, CAR-NK cells are theoretically less potent due to their lack of clonal expansion, relatively short lifespan, and less cytotoxic cytokines (87). Although CAR-T cells may mediate more incisive and long-term responses, the use of CAR-NK cells would minimize the risk of cytokine release syndrome and tumor-lysis syndrome, thereby increasing overall treatment safety (92). Importantly, CAR-NK therapy is expected to be much less expensive, considering that NK cells can be derived from multiple sources. Encouraging results have been reported in a phase I/IIa trial using cord blood-derived CAR-NK cells targeting CD19 in patients with relapsed or refractory non-Hodgkin's lymphoma and chronic lymphocytic leukemia, with up to 64% of patients achieving a complete response (93). In BC, tissue factor (TF) was recently described by Hu as a novel and common yet selective molecule on TNBC, whose targeting by TF-CAR NK cells resulted in an increased cytotoxicity against TNBC cell lines and was effective and safe for the treatment of TNBC in an orthotopic mouse model (94). Chen et al. recapitulated these findings when investigating the effect of EGFR-CAR NK cells in TBNC cell lines and in mice preinoculated with brain metastases (95). To the best of our knowledge, there is not published data on human trials on BC to date, although several initiatives can be found registered in the Clinical Trials.gov repository including multiple trials evaluating the intravenous infusion of ex vivo expanded, autologous NK cells and also the administration of NK cells incorporating HER2-and MUC1-CAR constructs ( Table 1).

CONCLUDING REMARKS
ACT offers a growing toolkit to overcome antigenic heterogeneity and the broad repertoire of immune escape mechanisms occurring in advanced BC. To fully capitalize these set of highly personalized treatments, we must address both approach-specific and cross-cutting challenges. ACT with autologous TILs may benefit from the standardization of TIL assessment in routine biopsies and the effective expansion of those TILs with the highest anti-tumor reactivity. Gene transferbased TCR therapies increase antigen specificity but still fail to target those not presented by the MHC, whereas CAR engineering may provide additional versatility but entails elevated costs and significant on-target, off-tumor toxicity. Additionally, although DC and NK cell therapies may have not achieved relevant tumor responses, their better safety profile and reduced costs make them suitable companions for multimodal strategies. The successful transition of the different ACTs to the clinic poses a number of common considerations. The discovery of TAs that can guide ACT against BC is critically linked to its success and relies on comprehensive strategies integrating genomic sequencing, in silico prediction, and immunogenicity evaluation. Methodological refinement is also required to improve our ability to isolate immune components and modify them ex vivo and in vivo, and to enhance cell persistence and intratumor trafficking. Finally, clinical trials testing ACTs will progressively need to be more adaptable, explore the reliability of predictive biomarkers, and generate quality data from small sample sizes. Both puzzling and fascinating, this is the path ahead to materialize ACT and transform the therapeutic landscape of BC patients.

AUTHOR CONTRIBUTIONS
JF-A, KG-H, and AO contributed to the conception and scope of the study. JF-A and AO wrote the first draft. JF-A and KG-H composed the figures/tables. All authors contributed to the article and approved the submitted version.