The common technical issue in any DC preparation is the low prevalence of DC in the peripheral blood, ranging 0.1–1% of peripheral blood mononuclear cells (PBMC) (15). Thus, early DC preparations such as Sipuleucel-T use a density gradient to prepare an APC enriched preparation (Figure 1). “Second generation” DC vaccines use strategies that differentiate monocytes into dendritic like cells called MoDC (Figure 1), creating a more readily available source of APC as monocytes make up ~10% of PBMC compared to 1% for DC. MoDC are prepared by separating CD14⁺ cells from PBMC either by their ability to adhere to plastic overnight culture or by anti-CD14 microbeads and magnetic separation. CD14⁺ cells are then cultured with cytokines, typically GM-CSF and IL-4 for 4–5 days, after which they display an immature DC like phenotype (16). When cultured with tumor antigen in the form of peptide or protein these cells cross-present and induce T cell proliferation (16). In melanoma patients, whilst only a small proportion i.e., 4% of I.D., injected DC, migrate to local lymph nodes but those that do activate CD8⁺ T cells in a melanoma model, thus overcoming microenvironment of melanoma (17). There have been several clinic trials in PCa with MoDC (Table 1). They vary in their mode of antigen delivery (protein, peptide, apoptotic tumor cells, cell lysate from tumor cell lines or mRNA) (Table 1, Figure 1), whether the MoDC are immature or mature and if matured what activation agent was used (Table 1, Figure 1). All these nuances have a profound impact on efficacy and applicability thus it is worth exploring these differences in more detail.

Firstly, early trials used immature MoDC and as one would expect, immature MoDC have reduced expression of activation markers, reduced ability to stimulate T cells (54, 55) and reduced ability to migrate (55). A meta-analysis that extracted individual patient data from 10 clinical trials of DC vaccines in PCa confirmed that immature MoDC preparations had less clinical benefit than mature MoDC (56). In melanoma patients immature and mature MoDC preps were compared head-to-head, again immature MoDC were less effective (57).

Different maturation agents have been used (Table 1, Figure 1) and at least in vitro they activate different gene expression profiles in the MoDC which in turn causes differing T cell responses (58). Broadly, maturation agents haven been chosen that are GMP grade, induce activation markers and produce MoDC that stimulate T cells toward a type 1 helper T cell response. Human cytokine cocktail, consisting of TNF-α, IL-1β, IL-6, PGE2, has been most frequently used in MoDC trials (Table 1). This mix produces mature MoDC with a superior ability to stimulate T cells than immature MoDC (59) and improved migratory capacity to mobilize DC to lymph nodes where they can prime T cells (60). However, there is data that these MoDC preferentially recruit T-regs, thus, potentially dampening any immune response initiated (61, 62).

Polyinosinic-polycytidylic acid [poly(I:C)] is a synthetic analog of dsRNA and is a clinical grade TLR3 agonist that matures DC (63). These DC, unlike those produced by the cytokine cocktail, produced high levels IL-12 (64) which directs a Th1 type T cell response. In vitro experiments suggest better antigen specific T cell proliferation and less T reg development (60, 62). In clinical trials Poly I:C matured MoDC vaccines are reportedly well-tolerated producing immunological and clinical responses (50). However, there are no clinical trials that compare Poly (I:C) matured MoDC with cytokine cocktail matured MoDC directly.

A third combination of CD40L with IFN-γ has shown promise, similarly increasing IL-12 cytokine production (65). Whilst this combination has not been used in PCa, CD40L has been used in other cancer vaccines. In resected metastatic colorectal cancer, a small, randomized phase I DC vaccine trial randomized tumor lysate pulsed MoDC cultured with or without recombinant CD40L. CD40L induced CD86 and CD83 expression on DC but in this small study of only 26 patients, CD40L did not improve anti-tumor specific T cell proliferation, IFNγ ELISPOT response, DTH response or relapse free survival (66). Similarly, in melanoma patients where CD40L was compared to cytokine cocktail, no difference was found in immunological response (67).

There are numerous reasons why we do not see clinical effect of different DC maturation strategies despite promising preclinical data. One is that small gains in maturation state in vitro maybe overpowered by the immune environment in vivo. One strategy that aims to control this is the use of viral vectors to genetically modify DC. In PCa Sonpavde et al. (37) showed feasibility, safety, and the development of a peripheral immune response when DC were transfected with inducible human CD40 (37). In this trial an adenovirus vector was used to transfect DC with human CD40 that had its cytoplasmic domain fused to ligand-binding domains and a membrane-targeting sequence to allow CD40 to be regulated by lipid-permeable dimerizing drugs, in this case rimiducid (68). This allows control over the timing of CD40 expression. DC vaccine was given and 24 h after injection, when DC have migrated to the lymph node and are in close contact with T cells, rimiducid is given to activate CD40. In this phase 1 study, 86% of patients had stable disease, with just 10% with a partial response (37). In PCa PSA kinetics reported as PSA doubling time (PSADT) are an indicator of prognosis with a shorter PSADT indicating a worse prognosis (69). In this study, 53% of patients had an increase in their PSADT, a surrogate marker for improved clinical outcome. This proof of concept shows that we can co-ordinate both timing and activation state of DCs to improve clinical outcomes.

Another variable amongst the different DC vaccination strategy is the type and form of antigen loaded onto DC.

MoDC vaccines have been administered in multiple different routes including intravenous (I.V.), intranodal (I.N.), intralymphatic (I.L), intradermal (I.D.), and subcutaneous (S.C). In a meta-analysis that pooled individual data from 84 patients, routes that allow migration to local lymph nodes i.e., I.D./I.L./I.N./S.C lead to better clinical response compared to the I.V. route (OR 3.2, 95% CI 1.1–9.0) (56). Fong et al. (41) showed similar findings using density enriched DC with a better cytokine profile seen with I.D. and I.L. route compared to I.V. with a trend to more transfusion reactions in the I.V. group.

Despite the variability in preparation as a whole these trials (Table 1) show that MoDC vaccinations in PCa are safe, produce a cellular immune response and a clinical response with a fall in PSA seen in up 27% (9/33) (18, 50, 83). However, it is important to note that an immunological response does not necessarily correlate with outcome (45, 46, 50) and often peripheral immune responses when detected are not sustained (46). Thus, the outcomes measured may not be clinically significant. Surrogate endpoints of reduction in PSA and difference in PFS may also not correlated with OS, as seen with Sipuleucel-T (5). Thus, despite a multitude of early trials we really need a Phase III trial of MoDC that looks at OS to determine clinical significance. The results of NCT02111577, a double blinded Phase III trial of MoDC loaded with apoptotic LNCaP cells added to standard chemotherapy for men with mCRPCa which has completed recruitment with 1,182 patients, should provide us with some clearer answers.

However, even without the results of this trial there are a number of reasons why MoDC preparations may not be the optimal approach. Monocytes are known to be dysfunctional in advanced cancer including in PCa. Most preclinical information on MoDC has been collected using healthy donor PBMC. However, when we compare MoDC prepared from healthy donors to those from patients with advanced cancer, patient MoDC are less efficient at phagocytosis, produce less IL-12 and express lower levels of the activation marker CD80 (84). In study of 24 patients with localized PCa, MoDC failed to upregulate CD80, CD83, and CCR7 after maturation with human cytokine cocktail, although for most patients, but not all, this was restored after surgery (85). In contrast two studies of only five patients each did show that MoDC from PCa patients were as good as healthy donors (32, 86).

The biggest issue with MoDC is that even from healthy donor PBMC, they do not perform as well as blood-derived DC; they do not stimulate T cells as well (54, 87), migrate as well (88) or have as much clinical efficacy (56). Thus, despite ease of production, MoDC vaccinations are unlikely to improve on the effectiveness of Sipuleucel-T.

Advances in efficiency of isolation protocols allow the use of blood DC as an alternative to MoDC or DC enriched preparations (Figure 1). Prue et al. (43) in a phase I trial showed that it was feasible to isolate CD1c⁺ DC from CRPCa patients via magnetic separation and vaccination, was well-tolerated with fever and pain the most common toxicity (43). More recently, in a phase II RCT Westdorp compared the efficacy of matured myeloid (m)DC vs. plasmacytoid (p)DC vs. combination of mDC and pDC (44). Again, this showed that blood derived DC were safe and induced an immune response, with a trend to a better response with mDC alone. These technical advances in isolating DC as a pure population and, as demonstrated by Westdorp et al. (44) isolating specific DC subsets and utilize the underlying specialization of human DC to take up antigen allows us to direct the immune response in a particular direction.

One way to target DC is to couple antigen to antibodies that bind endocytic cell surface molecules unique to DC. Preclinical data in mouse models show that delivering antigen in this way increases the efficiency of antigen presentation. Coupling OVA to the rat anti-mouse DEC-205 antibody (clone: NLDC-145) lead to a >100-fold increase in efficiency of DC antigen presented to mouse CD4⁺ and CD8⁺ T cells (99). Thus, targeting antigen directly to DC with antibody increases antigen presentation and in both in vivo mouse models and in vitro human models this leads to improved T cell response (100–102). However, in the absence of a maturation signal to the DC or indeed as a consequence of the function of the molecule targeted, this T cell response did not persist and in fact peripheral tolerance was induced (99). In contrast, in the presence of adjuvant such as anti-CD40 a strong memory response is formed after injection with OVA conjugated DEC-205, with CTL responses detectable up to 90 days after a single immunization (100). This need for a second “danger” signal to direct the immune system to form an inflammatory rather than tolerogenic response to the targeted antigen is not unique to DEC-205 antibodies but common to many surface antibody targets studied to date (99, 100, 103, 104). However, the selection of adjuvant in a clinical setting will need careful consideration to minimize side effects.

Despite the need for adjuvant, the safety and ease of delivery of in vivo DC targeting has been demonstrated in a phase I clinical trial of CDX-1401, a fully human anti-DEC-205 (CD205) mAb (3G9) genetically fused to the full-length NY-ESO-1 protein. The vaccine was used in combination with resiquimod (TLR7/8 agonists) and poly-(I:C) as adjuvants. It was well-tolerated in the 45 patients who entered the study and, induced a cellular immune response in 56% and humoral immune response in 79% of cases. Thirteen patients developed stable disease and 2 a partial response (105). This demonstrates that using antibody to target antigen to DC is safe and feasible and can induce an immune response in humans.

While safety has been demonstrated, reports on trials in ovarian cancer and acute myeloid leukemia (AML) are awaited. There remains the question of choice of molecule to target as targeting DEC-205 which naturally trends toward tolerance may be superceeded. Clec9a (CD370) is another endocytic surface marker with a much narrower expression profile. Whilst DEC-205 is highly expressed on cDC1 it is also expressed on monocytes, B lymphocytes and low levels on T cells and NK cells. In contrast, Clec-9A expression is limited to cDC1, which is the DC subset known for their ability to cross present antigen and elicit a CD8⁺ cytotoxic T cell response, ideal for a tumor vaccine. The other interesting ability of Clec9a is its ability to drive a memory immune response without adjuvant, however, in mice tumor models, adjuvant is still required (106). Whilst this looks like a promising target, it has been demonstrated that the cDC1 population, which is targeted by Clec9a is reduced in PCa (107) and is less responsive to activation with poly (I:C). This suggests underlying functional impairment and testing these treatments in a PCa tumor model is awaited.

Antibody-directed antigen uptake demonstrates that DC can be loaded “in vivo” (105), is safe and produces an immune response. However, antibodies are limited by the amount of antigen that they can deliver through coupling protein to antibody before the latter's ability to bind and be endocytosed is impaired. This has led to the development of co-delivery systems.

Co-delivery systems have two advantages, they allow the co-administration of adjuvant with antigen and can deliver multiple antigens. Some co-delivery systems are easy to adapt to different antigen make ups thus allowing personalized vaccine with “neoantigens' matched to each patient. There are two main vehicles studied: modified viral vectors and nanoparticles. Viral vectors include the filamentous bacteriophage antigen display system and modified adenovirus. The filamentous bacteriophage system is based on a non-pathogenic prokaryotic virus which can be engineered to express exogenous peptides as fusions to viral capsid proteins (108, 109). The bacteriophage is the adjuvant and in a mouse model it has been manipulated to express both mouse DEC-205 and OVA. In this system it produces an enhanced T cell response compared to injection with OVA: DEC-205 antibody conjugate (108). Similarly, a model where attenuated adenovirus was manipulated to express OVA and anti-mouse DEC-205 (110), produced a memory CD8 T cell response. Whilst this shows promise in pre-clinical models, translation to humans is yet to come.

Nanoparticles perhaps are closer to translation, in particular Poly(DL-lactide-co-glycolide (PLGA), a biodegradable slow-release polymer that is FDA approved to encapsulate drugs, can be adapted to encapsulate antigen and adjuvant (111). Due to their size nanoparticles readily taken up by DC (112) and in vitro studies show human DC take up peptide more efficiently if it is delivered inside a PLGA nanoparticle rather than soluble form (113). Nanoparticles not only direct peptide to the DC but also protect peptide from degradation, thus increasing the length of time to which DC are exposed to peptide. PLGA delivery of peptide induced T cells with a much greater CTL response than peptide loaded DC both in vitro (113) and in vivo (113, 114). Nanoparticle delivery has been tested in a mouse models of prostate cancer with the mouse prostate tumor antigen, six-transmembrane epithelial antigen of the prostate (mSTEAP). In this model, a single dose of mSTEAP on PLGA nanoparticles was compared to mSTEAP peptide plus adjuvant. The nanoparticle bound mSTEAP reduced both growth of TRAMP-C2 tumor cells in C57BL/6 mice and increased OS of the mice compared to peptide combined with adjuvant (114). Thus, in a PCa model, nanoparticles were more effective than a peptide vaccine. It is important to note though that comparison to DC vaccination strategies, antibody directed antigen uptake or other novel vaccination strategies remains to be assessed.

Nanoparticles have been used as a co-delivery system for antibody directed antigen uptake. Nanoparticles coated in anti-DEC-205, anti-CD40, and anti-CD11c antibodies to deliver antigen and adjuvant direct to DC all lead to increased CD8 and CD4 T cell proliferation and cytotoxicity in vitro and in vivo above non-targeted nanoparticles. CD40 targeted nanoparticles improved antigen specific T cell proliferation in the draining LN above other target receptors, and also cytotoxicity against target cells (115). In a mouse tumor model, CD40 nanoparticles containing OVA improved OS of B16-OVA inoculated mice compared to isotype control (116).

Whilst these emerging technology show promise in improving deliverability and efficacy of a DC based vaccine, they are yet to be translated into clinical trials in prostate cancer.

If we are to successfully translate in situ targeting of DC, clinical benefit will not occur without understanding what drives the immunosuppressive microenvironment of PCa.