Dendritic Cell–Based Cancer Vaccines

Dendritic cells (DC) are well known as the optimal APC for the priming of T cell responses. They are rare cells in the circulation and in tissues, and were very challenging to study for many years (1). When methods for ex vivo culture were developed, DC became much more actively investigated. More recently, with technologies that allow interrogation of single cells, insights into DC subsets and their biology have been made possible.

Before DC vaccines, cellular cancer vaccines were often based on genetic engineering of tumor cells, both autologous and allogeneic cells, and cell lines. Tumor cells engineered with cytokines like GM-CSF, allogeneic HLA molecules, or other xenoantigens have been a focus of preclinical and Phase I clinical trials, but their mechanisms of immunogenicity are thought to require Ag uptake and presentation by endogenous DC. Because there is the ability to culture DC in sufficient numbers, DC vaccines for cancer have been tested in Phase I, II, and III clinical trials. In this review, we present the biology of DC and the successes and failures to date with their use as vaccines against cancer.

Ralph Steinman first described and identified DC as a distinct cell type different from macrophages, due to their unique stellate shape and high expression of MHC in 1973 (1, 2). Since then, the field has greatly progressed and DC are often described as professional APC because of several key features (Fig. 1) (2–4). DC are mostly localized in tissues, acting as sentinels until Ag encounter. The specialized characteristics of DC allow for efficient Ag capture, internalization, and processing into peptides that are then presented in the context of MHC class I (MHC I) and II molecules. These complexes are subsequently able to be recognized by the TCR of CD8⁺ and CD4⁺ T cells (5, 6). DC that have captured Ags then migrate to lymphoid organs such as the spleen and lymph nodes to encounter and activate Ag-specific T cells through the TCR (signal 1) (7, 8). DC also provide costimulatory signals to T cells through the B7 family of molecules, (signal 2), transducing signals that result in expansion and clonal selection (4, 9, 10). Furthermore, DC can regulate and control the type and quality of T cell response elicited, via production of cytokines such as IL-12 p70 for Th1, IL-4 for Th2, or IL-17 for induction of a Th17 response (signal 3) (11–13).

Prior to Ag encounter, DC are immature. This is characterized by high expression of intracellular MHC II in late endosome-lysosomal compartments, low expression of costimulatory molecules, and low expression of chemokine receptors. In contrast, immature DC are biologically equipped for Ag capture and uptake through receptor-mediated endocytosis, pinocytosis, and phagocytosis (14–17). After Ag uptake and capture, Ag-loaded DC upregulate chemokine receptors like CCR7 to migrate to the draining lymph nodes (7, 18), allowing for occurrence of DC-T cell interaction critical for the initiation of T cell responses (19).

Conversion of DC from immature to mature DC is important for initiation of Ag-specific T cell responses. Effective induction of T cell response by DC can be functionally demonstrated in vitro through allogeneic MLR experiments. In addition, DC require very small amounts of Ag to stimulate T cell proliferation and are also shown to be superior stimulators of T cells, such that 100-fold more macrophages and B cells are needed to induce a proliferative MLR response (20, 21). During maturation, DC undergo physiologic changes resulting in increased expression of surface MHC I and MHC II molecules, increased expression of costimulatory molecules, expression of chemokine receptors, and secretion of cytokines to regulate the type of T cell response elicited (22, 23). DC maturation also results in lowering of the pH of endocytic vacuoles to activate proteolysis, and transport of peptide-MHC molecules to the cell surface while decreasing the capacity for Ag capture (6).

DC maturation can also be triggered by environmental stimuli because DC express a variety of receptors for sensing of viral and microbial pathogens as well as damaged, stressed, apoptotic, and necrotic cells, including autologous cells (4, 24, 25). DC can detect microbial and viral pathogens via pattern recognition receptors such as TLR, c-type lectin receptors, NOD-like receptors, and DNA/RNA receptors RIG-I and MDA5 (26). They can detect immune complexes via activating and inhibitory Fc receptors. DC respond to immunogenic and tolerogenic signals from other immune cells via receptors for TNF, IFN, IL-10, TSLP, and CD40L. Sensing of damaged cells and tissues by DC is possible due to release of molecules normally found intracellularly, such as ATP, heat shock proteins, and HMGB proteins. The recognition mechanisms that DC use to recognize endogenous and nonendogenous signals are also another strategy currently used to generate more effective DC vaccines against cancer, by taking advantage of highly immunogenic signals resulting from cell death (27).

Another important feature exhibited by certain types of DC is the ability to cross-present intracellular Ags. As a result, DC can take up exogenous Ag, which can result in the DC presenting an antigenic peptide-MHC I complex that can elicit a CD8⁺ cytolytic response against immune complexes, nonreplicating microbes, and dying cells (28, 29). Although the mechanisms involved are still incompletely elucidated, it is clear that during cross-presentation, protein Ags that are endocytosed and most efficiently move toward MHC II presentation are also able to reach the cytoplasm where proteasome processing occurs prior to peptide-MHC I presentation (29).

Blood- and skin-derived DC are the focus of the majority of studies on human DC due to tissue accessibility. The different DC subsets are defined and classified based on characteristics such as morphology, tissue localization, function, developmental origin, and expression of surface markers (Table I). All DC progenitors originate from CD34⁺ hematopoietic progenitors localized in the bone marrow (30) and they proliferate in response to Flt3L (31). In addition, monocytes grown in culture with GM-CSF and IL-4 also result in differentiation into monocyte-derived DC (32), which allows for generation of large-scale DC in culture. Recently, Lee et al. (33) have shown that the specification of the DC lineage is established as early as the hematopoietic stem cell stage through lineage bias exerted by a combinatorial dose of expression of transcription factors IRF8 and PU.1, and that Flt3L reinforces IRF8 expression throughout DC lineage differentiation.

Blood DC are classified into three different subtypes (Table I). Plasmacytoid DC are characterized by expression of IL-3 receptor α-chain (CD123), BDCA-2 (CD303), and BDCA-4 (neuropilin) molecules (34). Plasmacytoid DC are well equipped to rapidly produce type I IFNs in response to viral infections due to high expression of TLR7 and TLR9 (35). They are also capable of immediate activation of memory cytotoxic CD8 T cells as well as expansion of viral Ag-specific CD8 effector T cells (36–38).

Conventional DC, also known as myeloid DC, are further subdivided into two types based on differential expression of BDCA-1 (CD1c⁺) and BDCA-3 (CD141⁺) (34). CD1c⁺ DC express various lectins, TLR 1–8 (39), and can produce high levels of IL-12p40 and p70 (40). CD1c⁺ DC are also believed to play a role in antimycobacterial immunity because CD1c is a nonclassical MHC molecule that presents mycobacterial isoprenoid glycolipids to T cells (41). Furthermore, CD1c⁺ DC can phagocytose Mycobacterium tuberculosis and produce IL-6 and TNF (42).

The second type of myeloid DC subset is CD141⁺ DC, best described as cross-presenting DC, similar to murine CD8⁺ DC, because of their ability to capture Ag ultimately presented in the context of MHC I (43). They are particularly suited for cross-presentation of cellular Ags, immune complexes, and Ag targeted to late endosomes due to expression of CLEC9A and TLR3 (44–46). CD141⁺ DC also express the receptor XCR1, which binds to chemokine lymphotactin/XCL1 produced by NK cells and activated CD8 T cells indicating role of CD141⁺ DC in antiviral and antitumor immunity (47, 48). More recently, the importance of tumor-resident DC, which express transcription factor BATF3 and can also express CD141, has been described as an important source of T cell–recruiting chemokines CXCL9 and CXCL10 (49–51). These findings have important implications for cancer immunotherapy by leveraging recruitment and/or activation of BATF3⁺ DC in the tumor microenvironment to promote T cell infiltration, especially in a T cell–poor tumor microenvironment, also known as a cold tumor (50).

Two major DC subsets have been characterized in human skin: dermal interstitial DC (dermal DC) and epidermal Langerhans cells. Dermal DC subsets can be further classified into CD1a⁺ DC and CD14⁺ DC (52). Whereas not much is currently known about CD1a⁺ DC, several studies suggest that CD14⁺ DC are specialized in the development of humoral responses primarily by the differentiation of naive CD4 T cells into follicular helper T cells and by providing direct help to activated B cells (43, 53). In comparison with CD14⁺ DC, Langerhans cells are more efficient in cross-presentation of protein Ags and can prime naive CD8 T cells to differentiate into effector CTL (43, 53). Thus, the continued study of DC biology can improve our understanding of the diverse functionality of DC, leading to better strategies in designing more effective DC vaccines against cancer.

These findings and the paucity of studies on non-skin resident DC suggest that further studies are needed to fully understand different DC subsets and their functional specializations. The complexities of phenotyping myeloid cells have made characterizing the diverse cells in tumors quite challenging. Some limited overlap in cell surface markers between human and murine DC has added difficulties. However, more recent data have identified key surface and intracellular markers of human DC that now allow subsets to be identified. CD103⁺ DC in tumors and their role in allowing T cell infiltration is an example (see above) (49, 50).

Based on the potent Ag presentation and T cell activation activity of DC, and the development of straightforward in vitro culture conditions to promote DC differentiation from monocytes (54–56), DC were subsequently tested as cancer vaccines (57) (Fig. 2). Between 1995 and 2004, the first series of clinical trials testing monocyte-derived DC cultured in GM-CSF + IL-4 reached the clinic. They were loaded with tumor Ags in a variety of ways and used to promote tumor Ag-specific antitumor T cell immunity. Many were tested in melanoma patients based on the known immunogenicity of that tumor, as well as the characterization of shared tumor Ags like MART-1/Melan-A and gp100 (58–60). Other tumor types that served as early tests of DC vaccines were B cell lymphoma (61), and, later, acute myelogenous leukemia (62), myeloma (63), and hepatocellular cancer (HCC) (58). The one Food and Drug Administration–approved cancer vaccine, Provenge, is based on GM-CSF–activated APC targeting prostate cancer (64, 65), (although the extent to which this vaccine is based on DC and not other constituents is much discussed). These early trials showed that DC vaccines were both safe and immunogenic. Tumor-specific CD8⁺ T cell responses were detected in vaccinated patients as quickly as 7 d after a single DC injection (58). Most trials have been tested in late-stage cancer patients who have progressed after standard-of-care treatments, and who may not have optimally functioning immune systems due to tumor-induced immune suppression and previous treatments. Regardless, up to 10% of vaccinated patients can have durable tumor regressions. Ex vivo cultured monocytes are not the only clinically tested DC platform. CD34⁺ stem cells from mobilized donors cultured in GM-CSF, FLT3-L, and TNF have been tested in melanoma with positive immunologic response and clinical results (66). Plasmacytoid DC have also been tested successfully (67). More detailed knowledge of DC subsets with both unique and shared functional capabilities is a more recent development; hence, the majority of trials tested the DC most easily prepared for clinical testing: CD14⁺ peripheral blood monocyte-derived DC.

Because DC are highly sensitive environmental sensors, the reagents they are exposed to can strongly impact their function; hence, the culture conditions are critical. The starting cell populations, CD14⁺ monocytes, CD34⁺ stem cells, CD123⁺ plasmacytoid DC, and BDCA-1⁺ circulating myeloid DC (68–71), are all immunogenic, but distinct. The cytokines used for in vitro differentiation (GM-CSF plus IL-4 versus IL-15 for monocyte-derived DC) (72) can impact downstream T cell responses in vitro, although the distinctions in vivo can be challenging to determine above the noise of human variation. The area of DC maturation signals has been an active area of investigation for many years. In vitro (73) and in in vivo models, immature DC were found to be less effective activators of immunity, expressing lower levels of MHC I and II, and of costimulatory molecules, having reduced cytokine production, and being less able to traffic to lymph nodes for T cell activation (74, 75). Several studies of in vivo labeling of DC have shown that a small minority travel to lymph nodes, and maturation state at the time of injection may not be the most critical element (76). Therefore, in an effort to promote DC traffic to lymph nodes, potent T cell activation, and type 1–skewed immune responses, a variety of mixtures have been tested. They have been tested for the ability to upregulate surface CD83 (indicating maturation) and CCR7 (indicating the ability to traffic to lymph nodes) and have high IL-12p70 production (to skew T cells to type 1 immunity), and strong in vitro T cell activation (often using healthy donor blood). Because a potency assay for DC vaccines has yet to be identified which would predict in vivo activity, identifying an optimal maturation mixture remains a challenge. In addition, data obtained from healthy donor cells are not always replicated in patient cells. A novel DC delivery approach to direct DC to lymph nodes via lymphatic ports and intralymphatic injection has recently been shown to be feasible (77).

Another critical area of DC vaccine design and translation is Ag loading. Many sources of Ag have been tested ex vivo: MHC I–restricted peptides (78), synthetic long peptides (79), full length proteins, tumor heat shock proteins, containing peptides (80), autologous tumor cells (lysates and cells), as well as killed allogeneic tumor cells (81). In addition, transfection with DNA and RNA or transduction with viral vectors (82, 83) has been tested. All of these approaches have been successful in that they provide Ag, and are immunogenic in vitro, particularly when combined with maturation signals. Identification of the superior approach remains unknown. Short (8–11 aa) peptides are limited by the ability to only activate CD8⁺ T cells, which could be more limited in function [“helpless” (84)], and the heterogeneity of HLA types in patients might support an overlapping long peptide or full-length Ag approach, but addition of antigenically heterologous help in the form of keyhole limpet hemocyanin has been used by many groups. There are data suggesting tumor Ag-specific help is superior (85, 86); there are also data indicating the utility of keyhole limpet hemocyanin addition (87).

Two recent studies have directly compared whole tumor cell-based loading: DC-tumor hybrids (which can be accomplished by several methods) or DC-tumor cell mixtures, one in an in vitro study (88) showing electroporation-mediated fusion to be superior for T cell activation. A second report of a clinical trial in melanoma compared DC-tumor coculture, PEG-mediated tumor-DC fusion and tumor lysate-loading of DC (L. Geskin, submitted for publication). This trial indicated that a mixture of tumor cells plus DC may be inferior to the fusion and lysate loading (however, small numbers precluded a significant correlation). The optimal approach for Ag loading to produce the strongest CTL response has yet to be identified. In vivo DC targeting has also been investigated, particularly targeting Ags to the DEC-205 DC surface receptor (89). This area has been limited to date by specificity and efficiency of targeting (90).

There have been several high-profile DC-based vaccine reports in the last 2–3 y that highlight the state of the art. In one proof-of-principle study focused on neoantigens, three stage III resected melanoma patients received mature autologous DC pulsed with peptides derived from mutated neoantigens (91). Identification of the patient-specific neoantigen peptides (which were short, HLA-A2–binding MHC I–restricted epitopes) required substantial effort due to current limitations in epitope prediction software and in vitro testing. Notably, all three had previously received CTLA-4 blockade. In the blood of the vaccinated patients (after brief in vitro stimulation), peptide-specific T cell responses were identified. Importantly, the postvaccine blood samples showed a more diverse TCR repertoire, which suggests the promotion of determinant/epitope spreading (58). In a study of 12 glioblastoma patients that addressed DC vaccine injection, preconditioning the DC vaccine injection site with a potent recall Ag (tetanus toxoid) versus empty DC (without Ag) was tested (92). The subsequent DC vaccines were loaded with CMVpp65 RNA (expressed in a large percentage of glioblastoma tumors), and the DC injected in tetanus toxoid preconditioned sites resulted in superior lymph node trafficking of DC and superior patient survival. Although a very small study should not be overinterpreted, the companion murine model studies implicated the chemokine CCL3 as mediating superior DC migration and efficacy.

Another recent trial testing DC transfected with HSP70 mRNA was investigated in HCV⁺ HCC patients (93). The trial was a dose escalation, with 1 × 10⁷, 2 × 10⁷, and 3 × 10⁷ DC delivered per injection. What was remarkable was the objective clinical responses in this disease setting (two complete clinical responses, five disease stabilizations, and five progressive diseases of 12 patients). Although some tumor histologies like melanoma might be expected to show responses like this, in HCC it is unusual (94).

Promoting tumor-specific immunity with a vaccine and then amplifying it with checkpoint blockade is an attractive hypothesis that is being tested in different ways. In a larger, 39 melanoma-patient DC vaccine and checkpoint CTLA-4 combination trial, autologous, standard monocyte-derived DC were loaded with several shared melanoma Ag mRNAs and functionally enhanced with adding costimulation mRNAs by electroporation (95). In this study, based on previous trials in the same electroporated DC platform, the vaccines given both i.v. and intradermally, were combined with simultaneous CTLA-4 blockade (with ipilimumab). There was an encouraging overall response rate with 8 complete clinical responses and 7 partial clinical responses from the 39 advanced melanoma patients, suggesting that this is a combination setting worthy of further testing. Given the previous study of vaccination after CTLA-4 blockade, the combination schedule will need to be investigated in future trials. Such testing is complicated by the Food and Drug Administration approvals of CTLA-4 and PD-1 blocking Abs; hence, many patients will now receive these therapies after initial diagnosis.

Another trial tested the use of primary circulating CD1c⁺ blood DC (70). These cells were cultured for only 16 h, and the 14 stage IV melanoma patients in this study received 3 × 10⁶ to 1 × 10⁷ DC. In total, 4 of the 14 patients showed strong antitumor T cell responses that correlated with favorable progression free survival (12–35 mo). This trial showed that harvesting these primary DC from blood is feasible, the cells are immunogenic and could be tested further.

There are many disease settings in addition to cancer where activation of Ag-specific immunity or skewing of spontaneous immunity toward a more therapeutic immune response is desirable. One of the first DC trials was in an infectious disease prevention setting, and examined influenza virus (Flu) M1 peptide-pulsed DC (96), showing strong Flu-specific CD8⁺ T cell responses as soon as 7 d after vaccination. Although a reasonably efficacious Flu vaccine exists, a very active infectious disease investigative clinical setting for vaccination is chronic HIV infection. DC vaccines for HIV have been developed (67) and tested in vitro (97, 98) and in vivo in several clinical trials (99, 100). These studies, demonstrating safety and immunogenicity of type-1 immunity promoting DC, are sufficiently promising to lead to a new series of next-generation clinical trials that are in preparation. The lessons learned in cancer and in chronic infection may help inform each other.

The following examples provide another useful counterpoint to stimulatory DC vaccines for cancer. Based on the studies on DC maturation and tolerance induction from immature DC, culture conditions were developed for induction of an immune-suppressive state in DC with the goal of specifically inducing suppression or prevention of immune responses against self-antigens (like those targeted in autoimmunity) or highly immunogenic allogeneic Ags (in settings of transplantation). Some of these strategies involve specifically downregulating key molecules of costimulation (testing antisense oligonucleotides against CD80, CD86, and CD40 (101). Others are based on culture conditions, including vitamin D3 (NF-κB inhibition) and IL-10 (tolerance promotion), designed to skew DC to a stable suppressive state (102, 103). These modulations of DC culture and resultant phenotype are stable enough to resist strong maturation signals from the TLR ligand LPS and can increase inhibitory checkpoint molecule expression (104) including PD-L1 on the surface of DC.

DC have been known as the environmental sensors and potent APC for decades. The over 200 clinical trials testing DC vaccines have shown that DC are safe vaccines, highly immunogenic, and periodically able to activate an antitumor immune response capable of inducing a durable, objective tumor regression and clinical response in a previously treated, late-stage cancer patient. This is why DC will continue to be tested as cancer vaccines in new stages of cancer (earlier stages), new tumor histologies (melanoma is not the only potentially responsive tumor), and new combinations (preceding simultaneously with checkpoint blockade or after checkpoint therapy). Research into DC remains critical. We do not yet know which DC subsets are optimal for clinical translation, nor do we yet know the best Ags, loading strategies, doses, routes of administration, or potency assays.

The authors have no financial conflicts of interest.