Left out in the cold: Moving beyond hormonal therapy for the treatment of immunologically cold prostate cancer with CAR T cell immunotherapies

Prostate cancer is primarily hormone-dependent, and medical treatments have focused on inhibiting androgen biosynthesis or signaling through various approaches. Despite significant advances with the introduction of androgen receptor signalling inhibitors (ARSIs), patients continue to progress to castration-resistant prostate cancer (CRPC), highlighting the need for targeted therapies that extend beyond hormonal blockade. Chimeric Antigen Receptor (CAR) T cells and other engineered immune cells represent a new generation of adoptive cellular therapies. While these therapies have significantly enhanced outcomes for patients with hematological malignancies, ongoing research is exploring the broader use of CAR T therapy in solid tumors, including advanced prostate cancer. In general, CAR T cell therapies are less effective against solid cancers with the immunosuppressive tumor microenvironment hindering T cell infiltration, activation and cytotoxicity following antigen recognition. In addition, inherent tumor heterogeneity exists in patients with advanced prostate cancer that may prevent durable therapeutic responses using single-target agents. These barriers must be overcome to inform clinical trial design and improve treatment efficacy. In this review, we discuss the innovative and rationally designed strategies under investigation to improve the clinical translation of cellular immunotherapy in prostate cancer and maximise therapeutic outcomes for these patients.


Introduction
Prostate cancer is the most prevalent cancer among men in many countries, and its diagnosis rates are expected to double by 2040 [1].Despite advancements in detecting and treating localized disease, managing metastatic prostate cancer remains a significant clinical challenge.Hormone therapy, or androgen deprivation therapy (ADT), has been a cornerstone treatment of metastatic prostate cancer for over 70 years.Inevitably, patients develop lethal castrate-resistant prostate cancer (CRPC) and new treatments for end-stage disease are urgently needed.Advances in hormone therapy have included the approval of androgen receptor signalling inhibitors, such as enzalutamide, abiraterone, and apalutamide, that are often used in combination with ADT for metastatic hormone sensitive and CRPC.Additionally, highly effective non-hormonal therapies have recently been approved, including 177 Lu-PSMA radioligand therapy for patients with PSMA-avid metastatic disease [2], and PARP inhibitors for patients who have specific genetic mutations, such as BRCA1, BRCA2, or ATM mutations [3].These treatments offer new options for patients who have exhausted existing therapies and support the notion that patients with CRPC may benefit from the development of targeted therapies beyond hormonal approaches.Additionally, it is crucial to consider novel therapies for patients with neuroendocrine prostate cancer (NEPC), as hormonal therapies are not effective for this prostate cancer subtype.
Recently, personalized medicine in the form of CAR T cell therapy has emerged as a promising approach in cancer treatment.CAR T cell therapy involves the genetic modification of T lymphocytes obtained from human donors to express a synthetic Chimeric Antigen Receptor (CAR) that recognises specific proteins or antigens present on the surface of tumors cells.CAR T therapy has shown remarkable efficacy in hematological malignancies and six CAR T cell products have been approved by the Food and Drug Administration (FDA) for blood cancer treatment to date in patients with acute lymphoblastic leukaemia, B cell lymphoma and multiple myeloma [4].With these clinical successes, CAR T cell therapy has become one of the most promising treatment options in cancer immunotherapy.Nonetheless, significant obstacles hinder the translation of these therapies into other cancer types, particularly solid tumors including prostate cancer [5].
Multiple factors limit CAR T cell efficacy in solid tumors, including heterogeneous antigen expression, a lack of truly specific tumorassociated antigens resulting in on-target, off-tumor toxicities and the immunosuppressive tumor microenvironment (TME), which impedes the infiltration and long-term effector function of CAR T cells in the tumor (Fig. 1) [6].In prostate cancer, tumors are described as "immunologically cold," referring to the limited ability of immune cells to infiltrate the TME and maintain the active effector functions necessary to eliminate tumor cells [7].Both physical and molecular barriers within the prostate TME contribute to the suppression of endogenous immunity and impact the success of adoptive immunotherapies.In a bid to overcome these anti-tumorigenic barriers, combination therapies are being trialled to pre-condition the TME and enhance CAR T cell efficacy [8].In addition, next generation CAR T cells are being modified via gene-editing to improve persistence, anti-tumor function and to deliver payloads that can potentiate anti-tumor immunity within the TME [9].Engineering of NK cells, as well as neutrophils and macrophages, are emerging as alternative systems for adoptive immunotherapy and may overcome some limitations associated with CAR T cells, such as the potential availability of allogeneic "off-the-shelf'' products.
Together, these approaches have led to a rise in the number of adoptive T cell immunotherapy preclinical studies in an effort to strengthen clinical translation and efficacy.This review provides a discussion of the most recent strategies to enhance cellular immunotherapy and improve anti-tumor responses in patients, with a focus on advanced and lethal prostate cancers (Fig. 1).

Tumor-associated antigens
CAR T cells are directed towards tumor associated antigens (TAAs) expressed on the surface of cancer cells that can promote T cell activation and effector functions.Ideally, TAAs have high, uniform expression on cancerous tissue to prevent treatment resistance due to heterogeneous antigen expression and have minimal or no expression on normal tissue to limit on-target off-tumor toxicities.Notably, CAR T cells are designed to target TAAs and elicit an immunological response independent of antigen presenting cells (APCs) or epitope expression by major histocompatibility complexes (MHC).As CAR T cells recognise antigens in an MHC-independent manner, they are inherently resistant to immune escape due to MHC downregulation and can target cell surface antigens that are not normally recognized by T cell receptors, including glycolipids and glycosylated proteins [10].
Since the first clinical target antigen, CD19, was validated for CAR T therapy in patients with B cell malignancies, numerous cancer cellspecific cell surface markers have been investigated as potential CAR T cell and immunotherapeutic targets for different tumor types [11].In  prostate cancer, the most common TAA is prostate specific membrane antigen (PSMA).The development of small-molecule PSMA ligands has revolutionised prostate cancer diagnosis and clinical management with PSMA PET imaging and theranostics [2,12].PSMA-directed immunotherapies are now in clinical development, including CAR T cells (Table 1) [13].Unfortunately, targeting PSMA is not suitable for all patients, with heterogenous or no expression in up to 40 % of patients with castration-resistant prostate cancer (CRPC) [14][15][16], including predominantly PSMA-negative neuroendocrine prostate cancer (NEPC) [14,16], and thus the investigation of additional targets is warranted.
The use of PSMA-targeted imaging and treatment approaches has highlighted the potential of TAA-directed therapies for prostate cancer, and several antigens are in clinical or preclinical development for CAR T cell therapy, including PSCA, STEAP1, STEAP2, Lewis Y and Mucin-1 (Table 1) [13].A key consideration for the identification of potential immune targets is their expression on different subtypes of prostate cancer, particularly for patients with PSMA-negative or NEPC that have limited therapeutic options.Recently, Lewis Y, an oncofetal glycolipid antigen overexpressed in over 50 % of epithelial tumors [17], was shown to be upregulated on prostate tumors with diverse pathological and genomic features [18].This included both androgen receptor-positive and androgen receptor-negative neuroendocrine prostate tumors [18], demonstrating its potential as a target for a wider range of prostate cancer subtypes.Other cell surface molecules highly expressed on neuroendocrine tumors include delta-like ligand 3 (DLL3) and carcinoembryonic antigen-related cell adhesion molecule 5 (CEA-CAM5) [19][20][21].Targeting these molecules with adoptive cellular immunotherapy could expand treatment options for patients with neuroendocrine phenotypes.
Despite the plethora of emerging targets, preliminary clinical trials for CAR T cells as a monotherapy for prostate cancer have thus far demonstrated suboptimal efficacy [13].Many clinical trial results are still pending, and in some cases larger clinical studies are warranted to further investigate toxicities and treatment efficacy.However, it is generally considered that CAR T cells will not produce sufficient efficacy alone, and that novel strategies will be required to increase efficacy for prostate cancer, as well as other solid tumors.

Engineering strategies to improve CAR T cell efficacy in solid tumors
CAR T cell technology is constantly evolving, and to date multiple generations of CARs have been developed.The CAR construct has three main components: the extracellular domain, composed of a single-chain fragment variable (scFV) region that determines binding specificity; the transmembrane domain, to anchor the receptor within the cell membrane and mediate complementary target binding interactions; and the intracellular zone [22].The intracellular zone comprises a CD3ζ activation domain and one or more co-stimulatory domains, such as CD28 and 4-1BB, and harbours the immune receptor tyrosine-based activation motif (ITAM) for downstream signal transduction [13].Collectively, the CAR construct enables binding to cell surface antigens independent of MHC molecules and transmits primary and/or co-stimulatory signals for T cell activation.
First generation CARs contained one CD3ζ signalling domain and showed limited clinical efficacy, with poor T cell persistence and expansion [22].Second and third generation CARs incorporated one or multiple costimulatory domains respectively, most commonly CD28 or 4-1BB, to improve T cell activation and effector function [22], with the choice of the costimulatory domain aiming to improve tumor antigen sensitivity and CAR T cell survival and function [23].Next generation CARs are more sophisticated, arming CAR T cells with additional features to improve activation and persistence, prevent T cell exhaustion, facilitate trafficking into the tumor and/or limit cytokine toxicity [24].Next generation CAR T cells can incorporate cytokine inducers into the intracellular cassette, referred to as TRUCKs (T cell redirected universal cytokine-mediate killing), to deliver one or more cytokines or chemokines into the tumor to promote T cell trafficking, improve CAR T cell function, prevent T cell exhaustion and/or induce a shift towards a pro-inflammatory state within the TME.Next generation CAR T cells have also extended beyond cytokines, now arming CAR T cells with non-cytokine proteins, such as monoclonal antibodies, enzymes or transcription factors, to improve anti-tumor efficacy.
A single centre, phase I clinical trial was conducted for CART-P-SMA-TGFβRDN cells with 13 subjects (NCT03089203) [27].Dose-dependent toxicities were observed, with 5/13 patients developing cytokine release syndrome.Within two weeks post CAR T cell infusion, infiltration of these lymphocytes was observed in the tumors; however, this was not consistent across all subjects.Whilst early evidence of efficacy was seen in four patients with a PSA decline of >30 %, this was not durable and larger prospective studies are required to investigate safety and efficacy.
Combating immune suppression may also be achieved with knockdown or knockout of the PD-1 gene in CAR T cells, preventing CAR T cell inhibition through the PD-1/PDL-1 immune checkpoint axis.Zhou et al. silenced PD-1 using short-hair pin RNA technology in third generation CD19 and PSCA-specific CAR T cells [28].Silencing of PD-1 increased the ratio of CD8 + T cells and displayed increased killing and cytokine release [28].Using a PC3 cell xenograft model, the PSCA-ΔPD-1 CAR T cells decreased tumor growth and increased survival compared to control CAR T cells [28].Consistent with this, blocking PD-1/PDL-1 signalling through CRISPR/cas9 gene-editing or using a dominant-negative form of PD-1 has improved anti-tumor effects in preclinical models of triple negative breast cancer, glioblastoma, gastric cancer, non-small cell lung cancer (NSCLC) and hepatoma carcinoma, with improved CAR T cell expansion and survival [29][30][31][32][33].

Dual targeting and adaptor-mediated CAR T cells
A considerable hurdle for cellular based immunotherapy in solid tumors is intra-tumoral or intra-patient antigen heterogeneity, with low or heterogeneous antigen expression within or across metastatic sites resulting in immune escape and therapy resistance.Tumor heterogeneity arises during prostate cancer progression as cancer cells evolve under selective treatment pressure; however, tumor heterogeneity is also detected early in prostate disease prior to the onset of treatment [34].Thus, engineering strategies that target two or more antigens concurrently may be used to strategically target tumor heterogeneity and more effectively eliminate solid tumors, as well as overcome disease relapse due to target antigen downregulation or phenotypic cancer cell plasticity that occurs during advanced disease [34].
Multiple antigens can be targeted by engineering CAR T cells that have dual CAR constructs or by using tandem CARs, which contain dual specificality for two different antigens within a single construct [35].Several clinical trials are ongoing for dual antigen targeting CAR T cells for haematological malignancies, including targeting CD19/CD22 or CD19/BCMA for B cell malignancies and multiple myeloma, with preliminary results showing promising safety profiles and efficacious responses [36][37][38].Tandem CAR T cells are also being explored preclinically for solid tumors, including HER2/IL13Ra2 and EGFRvIII/IL-13Rα2 in glioblastoma and ErbB2/MUC1 in breast cancer, with improved T cell activation, antitumor activity and survival [39][40][41].In addition to overcoming tumor heterogeneity, dual targeting antigens may also be a strategy to limit on-target, off-tumor toxicities.In the context of prostate cancer, dual targeting CAR T cells have been generated towards PSCA and PSMA where co-binding of both antigens was required for optimal CAR T cell activation and tumor cell apoptosis [42].This approach may limit T cell-mediated damage of healthy tissue expressing a single target antigen.
An alternative approach for overcoming both antigen heterogeneity and off-tumor toxicities are switchable or adaptor-mediated CAR T cells.Adaptor-mediated CAR T cells are engineered with an inert adaptorspecific recognition domain instead of an antigen-specific scFv extracellular domain [43].Activation of the CAR T cell requires infusion of a secondary bifunctional adaptor molecule that links the target antigen on the cancer cell with the CAR T cell, effectively switching the CAR T cell 'on' [43].In the absence of the adaptor molecule, tumor antigen recognition and T cell activation are uncoupled, effectively rendering the CAR T cell inert and switched 'off' [43].Adaptor-mediated CAR T cells hold several promising advantages over conventional CAR T cell therapies.Firstly, adaptor-mediated CAR T cells can be directed towards multiple antigens to overcome heterogenous antigen expression, or target antigen escape by infusing different target adaptor molecules either simultaneously or concurrently.Secondly, adaptor-mediated CAR T cells provide a level of control and flexibility as treatment can be rapidly halted by stopping the infusion of the adaptor molecule and effectively turning the CAR T cells 'off' in the event of damage to healthy tissue or other potentially lethal toxicities such as cytokine release syndrome and neurotoxicity [43].
There are multiple adaptor CAR T platforms in development, including the UniCAR and RevCAR platforms that have directed CAR T cells towards both PSCA and PSMA, as well as other target antigens [44][45][46].Using PSCA-or PSMA-expressing PC3 cells, Feldmanm et al. demonstrated activation of UniCAR T cells in a target-specific manner with similar killing efficiency to conventional CAR T cells.UniCARs could target both PSCA-and PSMA-positive PC3 cells simultaneously in the presence of both PSMA-and PSCA-adaptor molecules, termed target modules in this platform, demonstrating the potential to target multiple antigens either concurrently or sequentially [45].A phase 1 clinical trial is currently underway for UniCAR T cells in combination with PSMA adaptor molecules for PSMA-positive solid tumors in Germany (NCT04633148).Currently, limited clinical trial data exists for any of the adaptor CAR T platforms and clinical outcomes are required to determine their efficacy and safety profiles compared to conventional CAR T cells for different tumor types [43].

CAR T STEM -like cells
In solid tumors, CAR T cells are predisposed towards terminal differentiation when exposed to chronic antigen stimulation and lack of costimulatory signals [47].While the design of the CAR construct is crucial, it has also been shown that the anti-tumor activity of engineered T cells is highly dependent upon their differentiation status [48,49].Increasing evidence suggests that phenotypically less-differentiated CAR T cells are more likely to induce persistent anti-tumorigenic responses [50][51][52] and can facilitate improved tumor control [53].
Recently, several research groups have trialled novel CAR T cell production protocols to enrich for pathways that promote self-renewal capacity and long-term survival, which can increase the generation of effector-like progeny to facilitate prolonged anti-tumour responses.This includes expanding CAR T cells ex vivo in the presence of cytokines, including IL-7 and IL-15, to increase T cell "stemness", metabolic fitness and proliferative potential [48,[54][55][56].These expansion protocols increase the population of T CM (T cell memory; CD44 hi CD62 hi CCR7 hi ) and T SCM (T stem cell-like memory; CD95 + CD62L + CD45RA + ) cells, which demonstrate greater anti-tumorigenic effects when compared to conventional CAR T cells [48].For a more targeted approach, Chan and colleagues recently identified key transcription factors upregulated by IL-15, revealing strong enrichment for a Foxo1 gene signature [57].Genetic engineering of CAR T cells to overexpress the FOXO1 transcription factor was found to promote a stem-like phenotype with greatly improved mitochondrial capacity, persistence and therapeutic efficacy in vivo [57].Therefore, modification of the CAR T cell product to prolong effector functions has strong potential to improve translational efficacy and further preclinical studies testing CAR T STEM -like cells in prostate cancer models are warranted.

Combination approaches to improve CAR T cell recruitment and function
A complementary approach to engineering strategies is to investigate the use of combination therapies to modulate the TME and enhance CAR T cell efficacy.The objective of combination therapies is two-fold: 1) to enhance CAR T cell recruitment by altering the chemokine profile, modulating the endothelium, and disrupting the tumor stroma; and 2) to enhance CAR T cell effector function and persistence by promoting a pro-inflammatory microenvironment and reducing suppressive immune cells.To accomplish this, a range of agents are under investigation for their effectiveness in treating prostate cancer and other solid tumors.These drugs include immune checkpoint blockade and chemotherapies, as well as agents that specifically target vascular endothelial growth factor (VEGF) and TGFβ.

Immune checkpoint blockade
Consistent with engineered approaches that overcome inhibitory immune checkpoint signalling, combination treatment with immune checkpoint blockade may prevent CAR T cell exhaustion and reinvigorate anti-tumor responses in solid tumors [22,58].Immune checkpoint pathways and immunosuppressive molecules are upregulated in multiple cell types within the TME following CAR T cell therapy, resulting in immune suppression and CAR T cell exhaustion [22,[58][59][60].Yamaguichi et al. used an in vitro co-culture model and an in vivo humanised mouse model to demonstrate that CAR T cells increase PD-L1 expression on tumor cells and macrophages, particularly immunosuppressive M2 macrophages, reducing anti-tumor activity of CAR T cells in both PSCA + prostate cancer and CD19 + lymphoma models [59].PD-L1 blockade depleted M2 macrophages and restored CAR T cell killing [59].Interestingly, this was independent of PD-L1/PD-1 signalling as the response was only observed in combination with anti-PD-L1 inhibitors atezolizumab or avelumab, but not the anti-PD-1 inhibitor nivolumab [59].In PDX models of prostate cancer, nivolumab had no added benefit when combined with CAR T cells directed towards Lewis Y in vivo [18].It is possible that targeting M2 macrophages through PD-L1 blockade may have improved Lewis Y-CAR T cell efficacy.Alternatively, responses to immune checkpoint blockade may be tumor-specific, and further studies in PDX models may be required to study heterogeneous responses across a diverse range of tumors.

Chemotherapeutic agents
Chemotherapeutic agents, particularly cyclophosphamide and fludarabine, are often given as lympho-depleting agents prior to CAR T cell therapy to enhance CAR T cell engraftment [6].Clinical trials have demonstrated improved CAR T cell expansion and clinical responses using this approach for haematological malignancies [61,62], and studies are underway to determine whether these benefits are also seen in solid tumors.Using PSCA-directed CAR T cells, Murad et al. demonstrated that preconditioning with cyclophosphamide was required for treatment efficacy in a PSCA knock-in mouse model with an intact immune system [63].Cyclophosphamide increased tumor-specific CAR T cell trafficking and expansion, and resulted in widespread transcriptomic changes in the TME, including pro-inflammatory myeloid and T cell signatures [63].Therefore, in addition to lymphodepletion, cyclophosphamide may also induce a pro-inflammatory state within the TME to aid anti-tumor efficacy.A single centre phase I clinical trial for PSMA-CAR T cells in patients with metastatic castrate-resistant prostate cancer demonstrated improved T cell proliferation and function following preconditioning with cyclophosphamide/fludarabine compared to without preconditioning [27].Based on this outcome, a second phase I multicentre dose escalation study is ongoing with lymphodepleting preconditioning used for all cohorts.
Beyond conditioning regimes, neoadjuvant or adjuvant chemotherapy treatment may break down barriers for CAR T cell therapy in solid tumors by remodelling the immunosuppressive TME.Using PDX models of prostate cancer, our laboratory tested second generation CAR T cells directed towards the oncofetal Lewis Y glycolipid antigen, which have demonstrated efficacy in acute myeloid leukemia [64].Whilst docetaxel and nivolumab had no effect on CAR T cell recruitment and function, a single dose of carboplatin one week prior to CAR T cell infusion substantially reduced tumor burden, with a sustained response to treatment [18].Carboplatin induced a cascade of pro-inflammatory changes within the TME, including polarisation of macrophages to a pro-inflammatory M1 phenotype, an altered cancer-associated fibroblast phenotype with enhanced extracellular matrix degradation, increased chemokine expression and altered endothelial cell phenotype to promote T cell trafficking [18].Collectively, this shift in the TME facilitated increased and sustained T cell recruitment into the tumors to promote anti-tumor CAR T cell activity.This response was dependent on changes induced within the TME as, in a PDX model that was less sensitive to carboplatin, there was a reduced response to combination treatment [18].Additional studies have shown that carboplatin, along with other platinum compounds, downregulate PD-L1/L2 expression on tumor and immune cells and increase chemokine expression to promote T cell recruitment in melanoma, colorectal, breast and lung cancer models [65][66][67], collectively providing preclinical evidence for the use of carboplatin as a modulating agent for CAR T cell immunotherapy.
While we found no benefit of the taxane chemotherapy docetaxel on CAR T cell efficacy in PDX tumors of prostate cancer [18], docetaxel has previously been shown to decrease T regulatory cells (Tregs) in the peripheral blood of patients with NSCLC [68], as well as polarise macrophages to an M1-phenotype, decrease myeloid-derived suppressor cells, remodel the tumor stroma, decrease exhaustion markers on CAR T cells and enhance the anti-tumor activity of CAR T cells in murine mammary tumors and a C4-2 cell model of prostate cancer [69][70][71].It is likely that the response to agents will differ between tumors, with TME modifications at the transcriptional level likely playing a key role in response.
In addition to carboplatin and docetaxel, several chemotherapeutic agents have been shown to modulate the TME, including alkylating agents, anti-metabolites and other taxane and platinum-based agents [6].Paclitaxel decreased immune suppression by depleting and inhibiting the function of Tregs and myeloid-derived suppressor cells, promoting an M1 macrophage phenotype and disrupting the tumor stroma in preclinical models of renal cell carcinoma, melanoma and breast cancer [72][73][74][75][76].These changes within the TME may collectively improve CAR T cell efficacy, and a phase I trial testing Claudin18.2-CART cells for gastric cancer demonstrated a high response in the majority of patients (21 out of 28) following treatment with a low-dose of nab-paclitaxel, with acceptable tolerability [77].Oxaliplatin in combination with cyclophosphamide also has immunomodulatory effects on the TME in KRAS,p53 models of lung adenocarcinoma, increasing CD8 + T cell infiltration, enhancing chemokine expression by tumor infiltrating macrophages and improving tumor control by CAR T cells [78,79].Oxaliplatin and cyclophosphamide treatment also increased PD-L1/PD-1 expression, resulting in both increased sensitivity to immune checkpoint blockade and synergistically improving the response to CAR T cell therapy [78,79].Consistent with these studies in lung cancer, oxaliplatin increased sensitivity to anti-PD-1 treatment in mouse and human prostate cancer cell lines [80], suggesting that oxaliplatin may also act as an immune modulatory agent in prostate tumors.
Collectively, this data suggests that chemotherapeutic agents have a range of positive impacts on the TME, and studies are now required to systematically evaluate the most appropriate treatment strategies across different tumor types.Considering there is significant tumor heterogeneity, it is likely that multimodal approaches will be necessary to target different tumor subtypes and overcome the immunosuppressive TME.Understanding how a tumor's unique TME composition impacts response and identifying pro-inflammatory gene signatures that incorporate multiple components of the TME may be beneficial to determining the best candidates that respond to chemo-CAR T cell therapy.In addition to identifying the most appropriate agent/s, treatment dose and schedule will likely impact on treatment outcomes.A fine balance might exist between modulating the TME and avoiding cytotoxic effects of chemotherapy on CAR T cell viability, and engineering approaches to generate chemo-resistant CAR T cells may be needed to facilitate this treatment strategy.

Other TME modulating combination approaches
In addition to immune checkpoint blockade and chemotherapies, a suite of alternative approaches to enhance CAR T cell function are under investigation [22,81].As well as engineering TGFβ signalling in CAR T cell design, TGFβ inhibitors may prevent CAR T cell exhaustion and promote CAR T cell cytotoxicity and cytokine production [82,83].Inhibiting other cytokines, including granulocyte-macrophage colony-stimulating factor, may also be efficacious [81].Normalising the vasculature of solid tumors may also promote CAR T cell infiltration and intratumoral distribution.Vasculature normalisation by inhibition of vascular endothelial growth factor (VEGF) or p21-activated kinase 4 (PAK4), a regulator of genetic reprogramming and aberrant vascularisation in glioblastoma endothelial cells, has been shown to enhance the delivery of EGFRvIII-CAR T cells and delay tumor growth in preclinical models of glioblastoma [84,85].Alternative strategies include combining CAR T cells with oncolytic viruses, bispecific tumor-targeted T cell engagers (BiTEs), vaccines or nanoparticles [22].Combination therapies with CAR T cells is an expanding research field and holds promise for translating the success of CAR T cell therapy to the treatment of solid tumors.

Natural killer cells and neutrophils
Different immune cells are under investigation as alternative options to T cells for adoptive cell therapy, including natural killer (NK) cells, macrophages and neutrophils, with several ongoing clinical trials for haematological and solid cancers.NK cells are innate lymphoid cells that can recognize antigens in an MHC-independent manner and kill foreign cells through several mechanisms, including Fas/FasL-mediated apoptosis, granzyme and perforin release, and cytokine secretion to recruit innate and adaptive immune cells [86,87].
CAR NK cells have the potential to overcome several of the limitations associated with CAR T cell therapy.CAR T cells require autologous T cells to prevent alloreactivity and graft-versus-host-disease (GvHD).In addition to manufacturing costs, harvesting sufficient autologous T cells from patients is not always achievable, particularly following multiple lines of cancer therapy.There is the potential for patients to progress and/or require bridging therapies whilst T cells are being cultured ex vivo [88].In contrast, NK cells have reduced risk for alloreactivity as they are activated in an MHC-independent manner, and therefore allogenic CAR NK cells can be derived from healthy donors, NK cell lines, cord blood or induced pluripotent stem cells, making 'off-the-shelf' large scale production feasible [88][89][90][91][92]. CAR NK cells derived from the NK-92 cell line have shown preclinical anti-tumor efficacy in several cancers, including prostate cancer, and are currently being investigated in several clinical trials [88,90,93,94].The immune-related toxicities of CAR T cells, including cytokine release syndrome, may also be avoided with CAR NK cells due to a different cytokine profile upon activation [89,95].A published phase I trial with CD19-directed CAR NK cells for lymphoid tumors demonstrated no increase in inflammatory cytokines following treatment in 11 patients, and none of the patients developed cytokine release syndrome, neurotoxicity or GvHD (NCT03056339) [89].NK cells also have a shorter life span than T cells [96].Whilst a drawback of this is the likelihood of increased infusions of CAR NK cells, a benefit is the ability to cease treatment if immune-related or off-target toxicities occur [88].
Currently, one phase I clinical trial is evaluating the use of anti-PSMA CAR NK cells for the treatment of CRPC (NCT03692663).Anti-PSMA CAR NK cells derived from the NK-92 cell line and using a CAR construct with CD28 and CD3ζ intracellular domains showed promise in preclinical studies, with on-target specificity demonstrated in vitro and a reduction in the growth of subcutaneous and orthotopic PSMA-expressing PC3 tumors in vivo [90].Wu et al. screened different ITAM constructs to identify the best candidate for PSMA-directed CAR NK cells, and the resulting pPSMA-CAR-NK cells decreased growth of C4-2 cells in vivo, with no increase in IL-6, which has been associated with cytokine release syndrome in CAR T cell therapy [97].Combination therapy with immune checkpoint blockade may further potentiate the effect of CAR NK cells as PD-L1 expression is upregulated on CAR NK cells via CAR-triggered PI3K/AKT/mTOR pathway [93].Blocking the PD-L1/PD-1 axis using the anti-PD-L1 antibody atezolizumab improved the efficacy CAR NK-92 cells in reducing the growth of subcutaneous C4-2 tumors in vivo [93], and thus could be explored in future clinical trials.Beyond PSMA, targeting other prostate cancer tumor-associated antigens is also a possibility for CAR NK cell therapy, with CD24-directed NK CAR cells showing efficacy in a preclinical study for urological tumor cell lines, including PC-3 and DU-145 prostate cancer cell lines [98].
Neutrophils may be another cell vehicle for adoptive cellular immunotherapy.Similar to NK cells, the generation of CAR neutrophils may provide an 'off-the-shelf' option with reduced risk of alloreactivity and can be generated from specific progenitors or pluripotent stem cells.Harris et al. engineered anti-PSMA CAR-neutrophils from human pluripotent stem cells with an CD3ζ intracellular domain [99].The CAR-neutrophils demonstrated on target cytotoxicity against the LNCaP prostate cancer cell line in vitro; however, further validation in vivo is required [99].
While CAR NK cells and neutrophils may hold promise for the treatment of solid tumors, and have potential advantages over CAR T cell therapy, considerable hurdles remain.Many of the challenges for CAR T cell therapy are applicable to CAR NK cells therapy, particularly recruitment into the tumor and survival within the hostile tumor microenvironment.Indeed, consistent with results from CAR T cell therapy, monotherapy with CAR NK cells in the above preclinical studies did not result in tumor regression in vivo [90,93,98].Arming CAR immune cells, for example with IL-15 [91], as well as combination approaches to 'warm' the TME, are likely required to optimise treatment for solid tumors, including prostate cancer.

Macrophages
CAR macrophages provide an alternative approach to improve immune cell trafficking into the tumor and overcome the immunosuppressive TME.Whilst lymphocyte infiltration is typically low in immunologically cold tumors, macrophages are still able to penetrate these tumors and tumor-associated M2 macrophages are one of the main immunosuppressive cell types of the TME [100][101][102].Tumor-associated macrophages have been observed in both localised and metastatic prostate tumors, and may promote proliferation and invasion of cancer cells, with high levels of M2 macrophages associated with decreased time to progression [103][104][105].The ability of macrophages to infiltrate and persist within the TME, as well as their ability to change from an immunosuppressive M2 to a pro-inflammatory M1 phenotype, may be of advantage for cellular immunotherapy.
The use of CAR macrophages is mainly in the preclinical setting, and there is only one registered clinical trial for HER-positive solid tumors (NCT04660929) [88].Whilst no known preclinical studies have been published for CAR macrophages for prostate cancer, anti-tumor efficacy in other solid tumors, including breast and ovarian cancer has been reported, due to CAR-mediated phagocytosis of target cells as well as stimulation of anti-tumor T cell responses and the induction of a pro-inflammatory environment [106][107][108][109][110]. In addition to targeting tumor cells directly, CAR macrophages have also been generated to target the extracellular matrix to decrease collagen content of tumors as well as CCR7-expressing immunosuppressive cells in breast tumors: both studies demonstrating increased T cell infiltration into the tumors and decreased tumor growth [108,111].Therefore, CAR macrophages may play an important role in cellular immunotherapy, both by directly targeting the tumor cells as well as the surrounding microenvironment and future studies are warranted to assess their potential for the treatment of prostate cancer.

TCR-based immunotherapy to complement CAR T cell therapy
A complementary approach to CAR T cell therapy is T cell receptor (TCR) engineered T cells (TCR-T cells).CAR T cells are restricted to recognizing tumor-associated antigens overexpressed on the cell surface of cancer cells, with low expression on normal tissue [112].However, these extracellular targets constitute less than 15 % of a cell's total protein population, greatly limiting the number of antigens available for targeting [113].In contrast, TCR-T cells recognize epitopes in the context of MHC molecules, and therefore can be directed towards either cell surface or intracellular antigens, opening the possibility for targeting tumor-specific antigens or neoantigens [112].Neoantigens are found exclusively on cancer cells as the result of aberrant protein production due to genetic alterations, dysregulated RNA splicing or post-translational protein modification [114].The majority of neoantigens are 'private' or patient-specific due to the vast heterogeneity between the genome and transcriptome, and the resulting immunopeptidome, of tumors [115].Personalised vaccines targeting private neoantigens have shown efficacy in preclinical studies for some tumor types, such as melanoma, and are currently being investigated in clinical trials [114].Autologous TCR-T cells directed towards private neoantigens are also being investigated in clinical trials for solid tumors, including ovarian cancer, lung cancer, colorectal cancer and pancreatic cancer [114].However, both logistical and economical barriers may reduce the widespread use of targeting private neoantigens.In contrast, genetic mutations shared by a subset of patients that are HLA-matched, referred to as 'public' neoantigens, offer the potential for broader applications and off-the-shelf therapies.Promising examples of public neoantigens include fusion genes CBFB-MYH11 in fusion-driven acute myeloid leukemia [116], and mutated KRAS variants G12V and G12D in numerous cancers including pancreatic and colorectal cancer [114,117,118].
Immunologically cold tumors with a low tumor mutational burden, including prostate cancer, have a smaller pool of potentially targetable neoantigens from somatic mutations compared to immunologically hot tumors with high mutational burdens.However, certain subclasses of prostate cancers are more genetically unstable with a higher tumor mutational burden, including tumors with DNA mismatch repair defects and biallelic inactivation of CDK12 [119][120][121][122].In addition, fusion peptides as the result of gene fusions, which are more common in prostate cancer, may also stimulate cytotoxic T cell responses, potentially with a higher level of immunogenicity compared to peptides derived from somatic mutations [116,123,124].RNA sequencing and computational analyses of 85 prostate adenocarcinoma cases from The Cancer Genome Atlas identified gene fusions in 87 % of tumors, out of which 41 % expressed a chimeric amino acid sequence (CASQ) that spanned the breakpoint of the two fused genes [125].The majority of CASQs (87 %) were predicted to generate HLA-restricted epitopes, including a CASQ encoded from a TMPRSS2 and ERG fusion that was recurrent in four cases [125].Therefore, targeting fusion proteins may provide an alternative approach for immunologically 'cold' tumors, including prostate cancer.
Despite holding promise for targeted therapies, the identification of neoantigens for cancer therapy is complex, with only a proportion of neoantigens having immunogenic potential and it may not always elicit a stimulatory response but instead an anti-cancer inhibitory response [126][127][128].Furthermore, whilst targeting tumor-specific antigens decreases the potential of off-target toxicities from antigen binding on normal tissue, there is the potential of cross-reactivity to structurally similar peptides derived from self-antigens [115].Multiple screening tools and pipelines are currently being used, or are in development, to identify immunogenic tumor-specific neoantigens and predict T cell reactivity to self-peptides [115,126,129], which are required to unlock the full potential of TCR-T cell therapy.
As for CAR T cell therapy, TCR-T cell therapy also needs to overcome the many barriers to cellular immunotherapy in solid tumors, including the time and cost associated with adoptive T cell transfer, tumor antigen heterogeneity, the immunosuppressive TME and immune escape due to loss of target antigens, or in the case of TCR-Ts, immune evasion by downregulation or loss of surface HLA [130][131][132].Common approaches can be employed to overcome some of these limitations, including TME modulating agents.Non-cellular soluble TCR-based agents may also overcome some of the limitations of T cell approaches, including bispecific T-cell engager molecules (BiTEs), Immune mobilizing monoclonal TCRs Against Cancer (ImmTAC) and TCR-mimic monoclonal antibodies (TCRm) [115].Multiple TCR-based approaches are under preclinical and clinical investigation and may be able to address the current barriers to cellular immunotherapies for prostate cancer.

Preclinical models for cellular immunotherapy
Despite the need to improve CAR T cells for the treatment of prostate cancer, preclinical models for cellular immunotherapies are complicated and all models have inherent advantages and limitations.Most preclinical studies thus far have been conducted using a handful of cell lines, either in in vitro assays or established as xenografts in vivo.Whilst these have provided important insights into potential targets and the use of CAR T cells for the treatment of prostate cancer, they fail to reflect tumor heterogeneity nor the complex interaction between the TME, tumor cells and CAR T cells.A wide range of additional preclinical models are also needed to understand the effectiveness of and biological responses to immunotherapies.
One way to model clinical heterogeneity is via PDXs, which are established from human prostate tumors collected at surgery, biopsy or after death (Fig. 2).Using the Melbourne Urological Research Alliance (MURAL) collection of PDXs [133], Porter et al. demonstrated that in one PDX model there was a significant reduction in tumor burden following carboplatin and Lewis Y-directed CAR T cell combination treatment.However, a second PDX model showed a modest response, with the extent of the response dependent on changes within the tumor [18].This demonstrates the impact of tumor heterogeneity on treatment response, not only within the tumor cells but also the TME and showcases the need for diverse preclinical models to understand treatment responses across different tumors.
As PDXs are grown in immunocompromised host mice to prevent graft rejection, they lack the full complement of immune cells, principally the lymphoid compartment [134].Despite this, they can still provide important insights into the interactions between other components of the TME, including fibroblasts, endothelial cells, the extracellular matrix, and host myeloid cells.To complement PDX studies, models that incorporate the full repertoire of immune cells are warranted in order to study the dynamic cross talk between host immune cells, the tumor, and immunotherapeutic agents.An option is the use of humanised PDX models, which incorporate human immune cells and anti-cancer immune responses into immunodeficient mice by infusing human peripheral mononuclear blood cells or CD34 + cells (Fig. 2) [134].Whilst this provides a platform to study immunotherapy in the context of a more extensive human immune system, it does present its own challenges, including GvHD, incomplete engraftment of human cells and a lack of human cytokines and growth factors [134].
An additional approach are syngeneic mouse models, where tumors are established from mouse cancer cells with the same genetic background as the host mice (Fig. 2).These models provide crucial insight into the efficacy of CAR T cells in the context of an intact immune system; however, there are only a few syngeneic mouse models are available for prostate cancer research, and they lack tumor heterogeneity, both in the context of the tumor cells and potentially also the TME.
Bhatia et al. performed an elegant study using a syngeneic RM9 prostate cancer model to demonstrate anti-tumor efficacy and safety of STEAP-1 directed CAR T cells [135].The burden of RM9 prostate cancer cell line tumors, engineered to express human STEAP1 (RM9-hSTEAP1 cells), was reduced following infusion of murinized STEAP1-CAR T cells in both NSG mice, as well as a human STEAP1 knock-in mouse [135].In addition to showing anti-tumor efficacy, the humanized STEAP1 mouse model, which displayed STEAP1 expression in the prostate and adrenal gland, showed no gross toxicities nor tissue disruption [135].Bhatia et al. also demonstrated that treatment resistance, due to STEAP1 antigen escape, can be combated through combination treatment with IL-12 therapy [135].This study highlights the potential of using complementary experimental models to evaluate treatment efficacy, potential off-target toxicities, and biological responses to CAR T cell therapy.
Therefore, a variety of approaches are required to realistically determine the potential of adoptive cellular immunotherapy for the treatment of prostate cancer and ensure that the correct treatment strategies with the correct patient selection are being implemented in clinical trials.

Conclusion
Ultimately, responses to cellular immunotherapy are dependent on cell trafficking to the tumor, and persistent and durable function of the T cells within the TME.To achieve this, dual approaches using next generation CAR T cell technology and strategies to overcome the immunosuppressive TME are required to optimise responses to CAR T cell therapy for prostate cancer.Appropriate preclinical studies are warranted to investigate the efficacy of different treatment strategies, and complementary approaches incorporating different experimental approaches, including syngeneic mouse models and PDXs are required to assess the impact of the TME and tumor heterogeneity.To date, most of the work has been conducted in other tumor types, and more sophisticated study designs are required to assess the feasibility of novel CAR T cell approaches for the treatment of prostate cancer.Fig. 2. Preclinical research for cancer immunotherapy relies on several experimental models.Cancer immunotherapy cannot be studied using a single preclinical model or experimental approach as the response is likely influenced by multiple factors, including tumor heterogeneity, the complex tumor microenvironment and dynamic immune cell interactions.Several preclinical models are available, each with their own advantages and disadvantages, that are appropriate for different in vivo studies and should be used as complementary approaches.

Fig. 1 .
Fig. 1.Current challenges of CAR T cell immunotherapy for the treatment of prostate cancer and strategies to overcome them.
L.H.Porter et al.