Active immunization against cancer with dendritic cells: The near future


DCs are antigen-presenting cells that regulate several components of the immune system. The mature or terminal stage of DC development induces specific T-cell immunity and resistance to experimental tumors in vivo. However, DC maturation is induced by inflammatory and microbial stimuli, so it is unlikely that mature DCs normally present antigens from tumor cells in cancer patients. Accordingly, clinical studies have begun in which DCs are generated ex vivo, charged with tumor antigens, exposed to maturation stimuli and reinfused to immunize patients. This approach has the potential to control responses to cancer antigens in a specific and nontoxic manner, in both vaccination and therapeutic settings.

DCs can mobilize several immune resistance mechanisms. These include CD8+ CTLs, CD4+ helper T cells, NK and NKT cells. Each of these lymphocytes recognizes targets through a distinct mechanism and has the capacity to kill tumor cells and release valuable protective cytokines like IFN-γ. CD4+ T cells also provide essential help for the expansion and maintenance of CD8+ cytolytic cells, while NK and NKT cells can eliminate targets that dampen presentation on MHC class I to escape CTL recognition. The stimulation and concerted action of these classes of lymphocytes can now be studied directly in patients, using ex vivo–derived DCs.

Several findings have emerged from studies of healthy volunteers who have been immunized with DCs charged with model antigens, KLH protein and influenza virus matrix peptide. Mature DCs elicit a polarized Th1 type of CD4 T-cell response, only a single injection being required. Vaccination with antigen-bearing DCs also markedly improves the functional affinity of CD8+ T cells. These findings are important in the context of immunotherapy because Th1 cells are more efficient helper cells in experimental models of viral infection and tumors, while high-affinity CTLs should improve recognition of tumor-derived peptides. An important caution also has surfaced when DCs are not adequately differentiated. Immature DCs can silence adaptive T-cell responses, e.g., by inducing IL-10–producing, T-reg cells.

DC-based active immunization does not result in major short-term toxicity in healthy subjects or cancer patients. Tumor-specific T-cell responses have been induced and detected in fresh blood specimens without the need for restimulation in vitro. However, the immune responses observed in the first protocols are still smaller than those seen naturally in acute viral infections. Occasional clinical regressions have been noted in these initial feasibility studies, particularly in melanomas, pediatric tumors, lymphomas, prostate cancers and renal cell cancers. This information, coupled with progress in DC biology, suggests many ways to improve the efficacy of this new therapy. Some relevant topics include antigen loading and DC maturation procedures, frequency and route of DC injection, efficiency of DC homing to lymphoid tissues and their longevity once there and the role of distinct DC subsets. A valuable positive control in active immunization protocols is to include an aliquot of DCs pulsed with a viral peptide to verify that the DCs and the patient's immune system are measurably competent.

Most of the first vaccination protocols have used DCs charged with synthetic, HLA-binding tumor peptides. Many methods are in place to allow DCs to process a broader array of peptides appropriate for the patient's MHC haplotype. These include transfection (RNA, DNA and viral vectors), select pathways of adsorptive endocytosis, exosomes and tumor cell fusion. An intriguing new approach involves the processing of whole tumor cells. Immature DCs are able to internalize tumor cells, including melanoma, EBV lymphoma and prostate carcinoma cell lines. Following maturation, the processing of tumor cells leads to presentation of multiple epitopes on both MHC class I and II products of DCs.

Basic issues in human cancer immunology can be investigated with ex vivo–based DC immunization. Some topics for the near future are the extent to which tumor-specific tolerance is an obstacle to immune therapy and the need to mobilize several adaptive and innate mechanisms in tandem, including Th1 type CD4+ helper T, NK and NKT cells. Progress is also being made in manipulating DCs directly in vivo. This will help to design vaccine adjuvants that improve presentation of tumor cells in patients. Although tumors challenge the immune system in formidable ways relative to infectious diseases, it is now feasible to use active immunotherapy and DC biology to manipulate and study the human response to cancer.


CTL, cytolytic T lymphocyte; DC, dendritic cell; EBV, Epstein-Barr virus; KAR, killer activating receptor; KIR, killer inhibitory receptor; KLH, keyhole limpet hemocyanin; MAb, monoclonal antibody; NK, natural killer; NKT cell, natural killer T cell; ODN, oligodeoxynucleotide; TRAIL, TNF-related apoptosis-inducing ligand; TRANCE, TNF-related activation-induced cytokine; T-reg, T-regulatory.


Antigen-presenting DCs activate several cell types (Fig. 1) that can resist tumors.1, 2, 3, 4, 5 We will discuss one way in which to exploit and study DC function in the setting of human cancer: generation of large numbers of DCs from a patient ex vivo, charging cells with tumor antigens and then readministering them to enhance cell-mediated resistance. This approach may appear more complicated than direct immunostimulation of patients. However, the ex vivo method ensures access of tumor antigens to the DCs and provides a means to manipulate the quality of the DCs, 2 critical control points in the active immunization process.

Figure 1.

Distinct cell-mediated resistance mechanisms against cancer. Four types of lymphocyte are able to kill transformed human cells directly and make abundant IFN-γ. Additional resistance mechanisms exist in the immune system, e.g., those mediated by antibodies and phagocytes. Each cell in this diagram has a distinct mechanism of antigen recognition, allowing the cells to work together to increase resistance to tumors. CD8+ CTLs recognize tumor peptides presented on MHC class I products and then kill the targets by a rapid granzyme-dependent mechanism. However, tumors often escape CTLs by dampening class I presentation. CD4+ T cells recognize tumor peptides presented on MHC class II products and kill targets through a slower, fasL-dependent mechanism. However, most tumors do not express MHC class II. Nevertheless, CD4+ T cells can recognize antigen on other antigen-presenting cells in the tumor, thereby recruiting phagocytes with antitumor potential, at least in mice.278 Tumor-specific CD4+ T cells can also activate DCs via CD40 and improve presentation to CD8+ T cells. NK cells have a 2-stage recognition mechanism to kill targets. First, KIRs263 recognize MHC class I. If targets lack MHC class I264 (Karre's “missing self” hypothesis289), inhibition ceases and a second set of natural cytotoxicity receptors or KARs264 (ligands unknown) takes over, allowing lysis of tumors that have subverted MHC I presentation. NK cells also express Fcγ receptors, to kill opsonized targets. NKT cells express select T-cell receptors for lipids and glycolipids presented on CD1d. NKT cells can have striking effects on a spectrum of mouse tumors in vivo273 and human cell lines.276 The interaction of these lymphocytes with DCs allows DCs to produce cytokines, especially large amounts of IL-12.

Mature DCs are the final immunostimulatory stage of DC differentiation. In mice, mature DCs are more immunogenic than their immature counterparts.6, 7, 8, 9 During maturation, DCs regulate several valuable components of the immunization process (Fig. 2). These include higher levels of MHC–peptide complexes,9, 10, 11 numerous lymphocyte costimulatory molecules (e.g., cytokines and members of the B7 family),12, 13 important TNF and TNF-receptor molecules (e.g., CD40, 4-1BB-L and TRANCE receptor)14, 15, 16 and many chemokines and chemokine receptors that help attract T cells and guide DCs to lymphoid tissues.17 DC family maturation is induced by many stimuli, ligation of TNF and Toll receptors being the best studied. DC maturation also can be blocked by several mechanisms, such as ligation of the thrombospondin receptor CD4718 and tumor-derived cytokines like IL-10.19, 20

Figure 2.

DC maturation. Antigen-capturing, immature DCs have many uptake receptors, including FcγR and receptors for cells dying by apoptosis and necrosis. Immature DCs have numerous endocytic compartments rich in MHC class II and other molecules, like the invariant chain and HLA-DM, to help load MHC II with peptides.11 Shortly after receiving a maturation stimulus (*), DCs produce IL-12 and other cytokines and chemokines80, 83 and start to form large amounts of MHC–peptide complexes clustered with CD86 costimulators.12 These move to the surface at the same time that DCs express new chemokine receptors,17, 290, 291 upregulate many costimulators like CD40 and CD5813, 86 and display markers of unknown function, such as surface CD83292 and intracellular DC-LAMP.293

Each lymphocyte subset in Figure 1 has distinct antigen-recognition mechanisms and functions. As a result, their action in concert has greater potential to resist tumors. CD4+ and CD8+ T cells are the adaptive arms of cell-mediated immunity, differentiating upon antigen encounter to produce cytokines and lytic products, expanding clonally and establishing memory. T cells recognize peptides derived from proteins within tumors and can resist the growth of experimental tumors, even eradicating established cancers.21, 22 NK and NKT cells have innate functions, already prepared to kill and produce cytokines upon tumor recognition. These cells can recognize targets lacking MHC class I products (NK) as well as glycolipids (NKT), thus providing ways to kill tumors that, by one mechanism or another, fail to present peptide antigens to T cells. NK and NKT cells grow upon exposure to cytokines like IL-2, but they are not known to establish memory with expanded cell numbers or improved function. DCs have several mechanisms to link innate and adaptive components of the immune response, especially their rapid production of large amounts of cytokines like IL-1223 and IFNs α and γ.24, 25

Active immunization with DCs has the potential to mobilize in patients each of the cellular resistance mechanisms in Figure 1. In addition to their tumor lytic capacities, these cells can produce large amounts of IFN-γ. This cytokine upregulates expression of MHC products on tumors so that the presentation of peptides is enhanced. IFN-γ also has antiangiogenic effects.26, 27 Tumor-specific CD4+ IFN-γ–secreting cells even eliminate MHC class II-negative cancers in mice, presumably when activated by antigen-presenting cells in the tumor milieu.28 Impressively, IFN-γ-deficient but otherwise immunocompetent mice show decreased tumor surveillance to spontaneous tumor development29 and to chemical carcinogens.30

Admittedly, tumors are much more formidable opponents than most infections, which are the established targets for cell-mediated immunity.31 Tumors may tolerize reactive T cells in the same way that self tissues maintain tolerance against autoimmunity. Tumors typically dampen the expression MHC class I,32 the antigen-presenting molecules for CD8+ CTLs and valuable tumor antigens may not be processed and presented.33 Tumors may block the development of immunity more globally by inhibiting the differentiation of active DCs.20 Tumors develop antiapoptotic mechanisms, where apoptosis is the basis for tumor killing by immune cells. Apoptosis also provides a means for DCs to capture and process antigens from tumors, as we will discuss. Despite these challenges, several developments in lymphocyte and DC biology are intensifying their use in vaccine and therapeutic approaches to human cancer. For one thing, antigens have now been identified in many tumors;34, 35, 36 for another, sensitive and quantitative assays are available to monitor the immune response.

Antigens and assays are not enough, however, as illustrated by the demands of developing an HIV-1 vaccine. HIV-1 genes, proteins and assays have long been available, but strong T-cell immunity to HIV-1 has been difficult to induce. Instead, control of cell-mediated resistance is much more complex and, in particular, requires enhancers or adjuvants. DCs are “nature's adjuvants”, capable of processing complex antigens like whole tumor cells and then delivering the resulting MHC–peptide complexes in such a way that strong cell-mediated immunity ensues.1, 2, 3, 4, 5 Intriguingly, certain DC subsets or states of maturation can additionally control self-tolerance and immune regulation.37, 38, 39 Therefore, DCs can enhance or dampen antigen-specific responses.

We will review 5 features of this new use of DCs to manipulate and study the human immune response to cancer antigens: (i) the control of the quality of the T-cell response by DCs, (ii) areas of DC biology to consider in the near future to improve control of the human immune system, (iii) approaches to the loading of DCs with a broad array of tumor antigens, (iv) initial results from DC vaccination studies in human cancer and (v) some important unknowns in cancer immunology that can be pursued with DCs as adjuvants.


The first clinical studies of DC therapy were performed in the setting of advanced cancer. Several of the initial studies reported occasional regressions of metastatic lesions following DC vaccination.40, 41, 42 There were few side effects of repeated injections of autologous DCs, especially in the report from Schuler's group,42, 43 who first used potent mature DCs in patients with stage IV melanoma. They found that the mature, immunostimulatory stage of DC differentiation was nontoxic but at the same time could expand tumor antigen-specific immunity.42, 43 Being aware of the lack of overt toxicity, we used DCs to immunize healthy volunteers to model foreign antigens. Such protocols would allow key functions of DCs and T cells to be assessed in the absence of potentially confounding features of tumor-bearing patients.

Antigen-bearing DCs reliably and rapidly expanded antigen-specific T-cell immunity in healthy individuals.44 Standard T-cell proliferation assays on blood mononuclear cells from the vaccinees showed that CD4+ T cells were rapidly primed to the protein, KLH and boosted to tetanus toxoid. In both ELISPOT (IFN-γ secretion) and cytotoxicity (51Cr-release) assays, CD8+ T cells specific for an influenza viral peptide were also increased with DC-based immunization. These immune responses were detected in fresh blood samples, whereas prior tumor vaccinations required prolonged tissue culture assays to detect T-cell immunity.45, 46 The capacity of DCs to expand multiple arms of the human T-cell response is valuable in light of evidence that resistance to experimental tumors is enhanced when both CD4+ and CD8+ T cells are engaged.47, 48 For example, tumor-specific CD4+ T cells express CD40L more effectively than CD8+ T cells, leading to activation of DCs via CD40. This in turn improves their antigen presentation to CD8+ killer cells49, 50, 51 and leads to production of valuable cytokines such as IL-12 and IFN-γ, to mobilize NK cells. Subsequent studies in small groups of healthy individuals have reported 3 surprising findings (discussed below) that reveal the capacity of DCs to control critical qualitative features of the human immune response.

Mature DCs polarize CD4+ T cells toward the Th1 phenotype

In 4/4 volunteers tested, a single dose of mature, KLH-pulsed DCs induced specific CD4+ T cells polarized to produce IFN-γ but not IL-4.39 This polarization to Th1 is potentially an important component of effective host resistance. In mice, T cells specific for the LCMV nucleoprotein were polarized ex vivo with IL-12 or IL-4 to produce virus-specific, IFN-γ–producing Th1 and IL-4/ IL-5–producing Th2 subsets, respectively.52 Upon adoptive transfer to naive animals, both Th1 and Th2 cells helped B cells to produce neutralizing antibody but Th1 cells were decidedly more protective in T cell–dependent immunity. In an analogous fashion, ovalbumin-specific Th1 and Th2 T-cell lines were produced and used to provide resistance to tumors bearing ovalbumin as a surrogate tumor antigen. Again, both Th1 and Th2 cells resisted the tumor but only the Th1 cells helped to establish memory for tumor-specific CTL responses.53 Analogous Tc1 and Tc2 subsets of CD8+ CTLs also were produced against an influenza viral antigen. The former IFN-γ–producing CD8+ T cells were more protective in an influenza model.54 Other studies indicated that IFN-γ–producing CD4+ T cells are better at protecting mice against chronic viral infection.55, 56

Th1 cells may be more valuable because they home better to sites of inflammation57 due to different chemokine receptors.58 For example, DCs induce Th1 cells expressing CXCR6 receptors, which mediate migration to inflammatory foci.59 Th1 cells may activate DCs better to sustain CD8 T-cell function,49, 50, 51 especially in the setting of a persistent viral or tumor antigen load. Th1 cells exert fasL-dependent cytotoxicity.60, 61 As mentioned, IFN-γ has antiangiogenic effects26, 27 and is able to mediate tumor regression even when the CD4+ helper cell recognizes MHC II–peptide on adjacent antigen-presenting cells.28 Therefore, it will be important to pay attention to the distinction between Th1 and Th2 when exploiting DCs to bring about stronger cell-mediated immunity in humans.

Mature DCs induce functionally superior CD8+ T cells

DCs also play a role in the establishment and maintenance of T-cell memory.62, 63, 64 This in part may be due to production of IL-15,65, 66, 67 a cytokine that sustains memory CD8+ T cells.68 When 3 volunteers were given a second booster of mature DCs pulsed with an influenza virus peptide, the functional affinity of the memory cells was greatly increased.69 The boosted CD8+ T cells could be activated with much smaller doses of viral peptide, 10 to 100 times lower. This finding has parallels in mice, where higher-affinity T cells develop upon multiple immunizations.70, 71 Improved functional affinity of tumor-specific CD8+ T cells could prove vital for the success of immune therapy because tumors may express only small amounts of some antigenic peptides. In adoptive transfer models, higher-affinity T cells can elicit superior protection against tumor and viral infection.72, 73 The near future would benefit from further analysis of the functional affinity of tumor-reactive T cells.

Immature DCs induce regulatory, IL-10–producing CD4+ and CD8+ T cells

A fascinating development comes from 2 reports, one with CD438 and one with CD839 human T cells, that immature DCs are not simply ignored by the immune system. This had been assumed in prior mouse studies reporting that immature DCs were decidedly less immunogenic. Indeed, immature DCs are able to silence an immune response and this is associated with the formation of T-reg cells. T-reg cells produce the suppressive cytokine IL-10, but a cell contact-dependent mechanism also appears to be needed for full regulatory function.74 The new human studies show that immature DCs can elicit antigen-specific T-reg cells and that CD8+ T cells are subject to regulation and silencing much like CD4+ T cells. Jonuleit et al.38 studied immature DCs and T-reg cells in vitro. The induced T-reg cells severely dampened Th1 cells activated by mature DCs. We worked in vivo in healthy volunteers,39 injecting immature DCs pulsed with influenza virus peptide. Upon immature DC immunization, the peptide-specific, IFN-γ–secreting CD8+ T cells fell substantially and did not return to preimmunization values for more than 3 months. For at least 1 month after immunization, peptide-specific, IL-10–producing CD8+ T cells were detected. In tissue culture, antigen-specific (MHC I tetramer binding) T cells were evident, but these could not form virus-specific CTLs when boosted with mature DCs. Current studies indicate that CD8+ T cells, after immunization with immature DCs, inhibit IFN-γ–secreting CD8+ T cells in preimmunization and recovery blood samples. This silencing of effector T cells by immature DCs may provide a way to dampen unwanted T-cell responses, as in autoimmunity and transplantation. However, in the context of immunotherapy, induction of T-reg cells would be anathema.


Since mature DCs charged with MHC class I and II binding peptides are able to elicit specific T-cell responses in humans, the stage is set to interface research on antigen-pulsed DCs with additional pertinent areas in DC biology (Table I). There is a potential methodologic gap and variable, which is to quantify the amount and longevity of the MHC–peptide complexes being displayed by the DCs under different conditions of antigen loading. Perhaps this can be explored by developing reagents that specifically recognize complexes of MHC products and a tumor peptide, e.g., soluble T-cell receptors75 or MAbs.9, 76 Although T cells begin to respond at much lower doses of ligand than can be detected by MAbs, specific antibodies allow direct assessment of the level of antigen presentation (MHC–peptide complexes) at the single-cell level. High levels of MHC peptide may prove valuable for skewing the immune response toward Th1 and for increasing the longevity of the immune stimulus. Soluble T-cell receptors and MAbs also help to monitor the distribution of MHC–peptide complexes on DCs. In one system, where hen egg lysozyme was processed onto mouse MHC class II, the MHC–peptide complexes formed clusters that included the CD86 costimulatory molecule at the DC surface. It is postulated that the coclustering of MHC–peptide and CD86 allows DCs to set up effective immunologic synapses for signaling naive T cells and initiating immunity.12

Table I. Some Parameters to Improve the Efficacy of DCs
Maturation stimulus and maturation state of DCs
Migration of injected DCs to draining lymph nodes
Life span of DCs once in the lymph node
Frequency, route and dose of DC injections

State of DC maturation

As mentioned above, mature DCs are more immunogenic in mice. There is preliminary evidence in cancer patients that even when DCs are directly injected into the node, mature DCs induce stronger T-cell responses than immature DCs.77 As discussed above, DCs can rapidly polarize the human immune response to the Th1 type. DCs can produce high levels of the Th1-inducing cytokine IL-12 in vivo,23 as well as the IL-23 homologue,78 especially when they are exposed to additional cytokines such as GM-CSF and IL-479, 80 and select chemokines.81 However, prolonged culture of DCs in a CD40L maturation stimulus “exhausts” much of the capacity of the DCs to produce IL-12 and induce Th1 cells in culture.80, 82, 83 These tissue culture data contrast with in vivo observations that mature DCs effectively elicit Th1 responses39 as well as in vitro findings that CD40L DCs are valuable for expanding melanoma-specific CD8+ CTLs and Th1 CD4+ cells from tumor-infiltrating lymphocytes.84 Perhaps the loss of high level IL-12 production during the later stages of maturation primarily dampens the capacity of DCs to influence other cells, like NK cells, in a paracrine fashion. Sufficient amounts of IL-12 or IL-23 may still be produced to act locally to polarize T cells toward the Th1 type.

The B7 family of costimulatory molecules85 also is regulated during maturation and influences the production of Th1 and Th2 cytokines. For example, CD86 is upregulated markedly,86 and newly expressed CD86 is expressed in clusters with MHC–peptide molecules.12 In contrast, mature DCs may downregulate the ICOS ligand, which polarizes toward potentially undesirable Th2 cells.87 Mature DCs also express the B7-DC/PD-L2 molecule. This molecule may enhance Th1 cytokine production by naive T cells88 and suppress cytokine formation by activated T cells.89 B7-H3 is yet another B7 homologue, cloned from DCs, which also costimulates IFN-γ production.90 When MAbs to this intricate B7 family become available in the near future, it will be valuable to monitor the expression of B7 family members during DC maturation and vaccination studies.

Although DCs most likely should be given a maturation stimulus prior to their use in immunotherapy, the time of stimulation in vitro needs evaluation. For example, it might be important to test DCs that have been matured for shorter times than the 1- to 3-day periods used to date. On the one hand, prolonged maturation may diminish IL-12 production (“exhaustion”), but on the other hand, the capacity of DCs to resist the immunosuppressive effects of IL-10 may be enhanced.91

Efficiency of DC migration

The efficiency with which DCs home to the draining lymph node and their longevity once they reach the node are research areas for improving the efficacy of current active immunization protocols. In mice, only small numbers of the injected DC inoculum are found in the draining lymph node,63 where the immune response begins.92, 93 Recovery of injected DCs is increased if the DCs are treated with viability-enhancing CD40L or TRANCE/RANKL prior to injection.63 Nevertheless, fewer than 10% of injected DCs can be identified in the draining lymph node. CD40 may be essential for at least epidermal DCs to migrate to lymph nodes.94 In humans, migration efficacy can be followed with DCs labeled with 111indium.95 The near future will hopefully use tracers of this kind to determine the efficacy with which DCs leave the injection site and home to lymphoid organs. In this way, alternative methods of DC injection and conditioning of the injection sites could be evaluated.

DC migration is an elegant area of DC biology and ostensibly one of the key control points in the efficacy of immune activation. DC migration is regulated through multidrug resistance receptors, which pump select compounds like cysteinyl leukotrienes.96, 97 These in turn modulate the responses of chemokine receptors, especially CCR7, on DCs. The respective chemokines (CCL19 or MIP-3β and CCL21 or SLC), which are made by the lymphatics and select lymph node cells, including DCs themselves, influence DC movement and other functions.98, 99 The efficacy of injected antigen-bearing DCs therefore may require that the investigator learn to control the availability of these leukotrienes and chemokines. We feel that current methods of active DC-based immunization do not optimize either DC migration into lymphatics or their life span upon reaching the lymph node. These functions, if improved, could markedly increase the efficacy of DCs.

Dose, frequency and route of DC injection

DC dose may influence the quality of the T-cell response since, at least in vitro, a low dose of DCs,100 like a low dose of antigen,101, 102 can polarize toward Th2. The injected dose also could be too large for the DCs to gain access to the lymphatics or, as mentioned above, the injection site may need to be further conditioned to allow efficient homing in vivo. Another concern with some of the initial DC vaccine studies is that a rapid injection schedule induced active CTLs, which then killed the booster doses of DCs and reduced efficacy.103 The immunogenicity of DCs after different routes of administration in humans also needs to be compared. Some data suggest that the s.c. and i.d. routes lead to greater nodal migration over the i.v. route104, 105 and improved Th1 polarization.106 Direct injection into the lymph node is thought to facilitate the access of DCs to T cells,41 yet this also needs to be proven. A new approach is to directly inject DCs into tumor deposits,107 as a way of initiating the afferent limb of immunity in vivo. Interestingly, some DCs have the capacity to directly kill tumor targets, e.g.,via TNF family members like TRAIL.108, 109, 110, 111 The near future would benefit from randomized 2-arm studies to document responses under defined conditions of DC dose, frequency and route.

Sources and subsets of DCs

Most studies to date have used monocyte-derived DCs. These offer the advantages of high yield, purity and feasibility, especially when leukapheresis is used to collect the starting monocytes.111, 112 It currently is feasible to obtain from patients with advanced melanoma 500 million mature DCs (current protocols use approx. 5 million DCs per vaccination). Monocyte-derived DCs can be successfully cryopreserved, even after the DCs have been loaded with antigen, though this needs more testing. The Dendreon firm's methodology (Seattle, WA), in contrast, depends on a small fraction of DCs that are normally present in blood after 1 to 2 days in culture.113, 114 The vaccine then uses the equivalent of an entire leukapheresis for each immunization. The alternative approach of differentiating large numbers of monocytes to DCs requires exogenous cytokines and longer culture periods (6 days or more), but the yield of potent DCs is much larger and the cells are more homogenous. In current practice, a single leukapheresis provides sufficient monocyte-derived DCs for dozens of vaccinations. It also may be feasible to cryopreserve the antigen-bearing mature cells, greatly facilitating vaccination and therapy.111, 112

DC maturation from CD34+ progenitors also is being characterized.115 A clinical study with CD34-derived DCs has shown immune and some clinical efficacy in the short term.116 These cells consist of 2 subsets, termed Langerhans cells or epidermal DCs and monocyte-derived or interstitial (dermal) DCs.117, 118, 119 The 2 DC subsets differ in markers and some functions. For example, Langerhans cells are ineffective at stimulating B-cell development directly,120 a function that may require the BAFF/BlyS TNF family molecule on monocyte-derived DCs. CD34+-derived DCs may have improved efficacy at eliciting CTLs.121, 122 Cells with some of the features of Langerhans cells could also be derived from monocytes, TGFβ being a key determinant.123, 124 It should be possible in the near future to look more closely at the immunogenicity of Langerhans cells relative to other forms of DCs.

Yet another DC subset is the plasmacytoid DC or preDC2.125 These CD11c-low precursors have some very different features relative to monocyte precursors or preDC1. Plasmacytoid cells respond to IL-3 and express CD62L or L-selectin, thereby homing directly to T-cell areas via high endothelial venules.126 Monocytes respond to GM-CSF and IL-4, lack CD62L and probably must traverse the tissues and afferent lymphatics127, 128 before accessing the T-cell areas. The 2 DC subsets also differ in their expression of Toll receptors; e.g., plasmacytoid cells express TLR7 and TLR9 and monocyte-derived DCs express TLR2 and TLR4.128a Monocyte-derived DCs selectively express CD1 antigen-presenting molecules. They can produce high amounts of IL-12 upon stimulation with CD40L, whereas plasmacytoid DCs produce large amounts of IFN-α with certain enveloped viruses.24, 126 IFN-α might serve to mature monocyte-derived DCs129 and to activate NK cells. Mature monocyte-derived and plasmacytoid-derived DCs have the potential to induce either Th1 or Th2 CD4+ T-cell responses,100, 130, 131 though intriguingly, the Th2 polarizing capacity of mature plasmacytoid DCs does not appear to require IL-4, as is usually the case for Th2 differentiation.130 The numbers of plasmacytoid DCs can be increased in blood with flt-3L or G-CSF.132, 133 Methods have been developed to produce larger numbers of human plasmacytoid cells from CD34+ progenitors in culture.134, 135 As a result, it may be feasible to begin studies of the biologic properties of plasmacytoid DCs in vivo.


Defined tumor peptides have been and will be invaluable GMP-quality reagents for DC-based manipulation of the human immune system. The sequence of the peptides can also be altered relative to the naturally occurring tumor peptide, to enhance their affinity for the presenting MHC molecule or the affinity of the MHC peptide for the T-cell receptor.75 Nonetheless, there are significant limitations to the use of peptides as the source of tumor antigen for vaccination and therapy. First, a few peptides are only a small part of the antigenic makeup of a tumor. Cancer cells are much too formidable to combat with limited weaponry. For example, one would expect tumors to more readily escape the immune pressure exerted by a narrow array of peptide-specific T cells. Loading DCs with a diverse or broad spectrum of tumor antigens would lessen the chances for escape. Second, because of MHC restriction, chemists must tailor each synthetic peptide to the patient's MHC haplotype, whereas DCs efficiently carry out the task, using complex protein mixtures and tumor cells as the starting material. Thus, DCs efficiently extract the peptides that load the patient's MHC products and the resulting MHC–peptide complexes can be quite stable at the cell surface with half-times >24 hr.10, 136 Third, during DC transport of MHC–peptide complexes to the cell surface, i.e., from newly processed proteins, there is coassembly of the complexes with valuable stimulators, such as CD86,12 helping to set up what is termed the “immunologic synapse”. The latter is a close contact region between the antigen-presenting cell and T cell. In the synapse are found distinct molecular couples with similar sizes, spanning about 135 Å, such as MHC–peptide coupled with the T-cell receptor, CD86 with CD28 and CD58 with CD2. This clustering and coassembly likely improves the efficacy of the signaling process. In the near future, many strategies to allow DCs to process multiple tumor antigens will be pursued (Table II).

Table II. Vehicles for Loading DCs with Diverse Tumor-Associated Antigens
Tumor-derived RNA
DNA for tumor antigens
Viral vectors
Receptor-mediated uptake of proteins
Dead or dying tumor cells and tumor cell lines
Fusions between DCs and tumor cells

Tumor-derived RNA

Gilboa's lab developed the principle that transfection with tumor RNA could lead to the presentation of a wide array of tumor peptides.137 The RNA can be used directly or after amplification from small amounts of starting material, potentially from a few tumor cells in a biopsy. RNA is transfected with138 or without139 lipofection, depending on the laboratory. These RNA approaches have the potential to elicit immune responses to proteins uniquely expressed in a patient's tumor.

Plasmid DNA for tumor antigens

Plasmid DNA, usually containing viral promoters, has been used extensively in the development of DNA vaccines.140 The weight of current evidence indicates that DCs are important intermediaries in the vaccine response.141, 142, 143, 144 In mice, small numbers of DCs can be transduced directly by the DNA vaccine. DCs additionally might be able to cross-present protein expressed by other cells. Previously, it was difficult to transfect DNA into DCs with any efficiency, but a newly described cationic CL22 peptide carrier greatly improves transfection.145 Transfected DCs present the corresponding antigens to human T cells in culture and to mouse T cells in vivo, including induction of protective immunity against experimental tumors. Plasmid DNA approaches have the potential to include sequences other than those encoding the tumor antigen, e.g., sequences that lead to better antigen processing and T-cell costimulation.146, 147, 148

Viral vectors

By administering DCs infected with a viral vector, the virus is hidden from potentially neutralizing antibodies. High-efficiency gene expression in DCs should be both feasible and useful. The methodology has been reviewed149 and is only briefly considered here. Adenoviral vectors efficiently express proteins in immature DCs, which express relevant virus receptors like the αvβ5 integrin. Infection does not directly result in large perturbations of the DCs, which can then be matured with standard stimuli to carry out normal functions such as IL-12 production,150 antigen presentation to CD8+ T cells151 and antitumor immunity in mice.152, 153 Poxvirus vectors, especially avipox and vaccinia, readily express recombinant protein in a sizable fraction of monocyte-derived DCs. The infection is abortive, with only early viral gene products being expressed; and when immature DCs are infected, infection is followed by either overt cytotoxicity154 or a block in maturation.155 Nevertheless, efficient presentation of the recombinant gene is observed in culture,154, 156 possibly through cross-presentation of dying infected DCs by other noninfected DCs.157, 158 Even UV-inactivated, recombinant vaccinia is presented on MHC class I products of DCs.154 Several other vectors, with excellent potential for the genetic modification of DCs, have also been described, such as retroviruses,159 lentiviruses160 including lentiviruses pseudotyped with the VSV viral envelope161 and influenza.162 The near future should see experiments using viral vectors to assess the value of DCs as adjuvants. In other words, protocols have been approved in monkeys and humans to compare the efficacy of poorly replicating vectors given directly in a standard way (s.c., i.d., i.m.) or after first being used to modify DCs ex vivo prior to injection.

The vectors summarized above are used to infect DCs ex vivo. Some viral envelopes may also prove useful to directly and selectively target DCs in vivo. The HIV-1 envelope may target DCs via a lectin called DC-SIGN that is abundant on DCs,163 the LCMV virus targets mouse DCs via α-dystroglycan164 and the Dengue virus targets human LCs and monocyte derived DCs,165 though the receptor is not yet identified. In other words, viral envelopes provide potential DC targeting strategies, especially in the future when DCs will be manipulated directly in vivo.

Receptor-mediated uptake of proteins by DCs

DCs have a number of mechanisms to enhance uptake and presentation to CD4+ T cells on MHC class II. These include macropinocytosis, where specific aquaporins appear to be essential for the enormous associated influx and efflux of fluid;166 phagocytosis of organisms167 and dead or dying cells,136, 168 where many different receptors may contribute;169 and adsorptive uptake via clathrin-coated vesicles, such as mannose receptor–mediated uptake of certain glycoproteins.170 Uptake by fluid phase pinocytosis and phagocytosis typically ceases upon full maturation of DCs, but adsorptive pinocytosis via clathrin-coated vesicles can persist.171 In most cases, when loading DCs with antigens for immunotherapy, the immature cell is the best, while the mature cell is optimal for presenting processed MHC–peptide complexes to T cells.

One DC receptor, DEC-205, greatly enhances the efficiency of MHC class II–mediated presentation after the adsorptive uptake step. DEC-205 does not recycle through peripheral endosomes, as is typical for most adsorptive receptors. Instead, the cytosolic domain is able to move the receptor more deeply into the cell through MHC II+ late endosomes or lysosomes. This unusual traffic pattern is associated with a 10- to 100-fold increase in presentation of ligands relative to the classical macrophage mannose receptor, which recycles through peripheral endosomes.172

Several pathways target the MHC class I compartments of DCs, which is remarkable because in most cells MHC I molecules are loaded from newly synthesized proteins. For some of these “exogenous” pathways to MHC class I, specific receptors are known, such as the Fcγ receptor for immune complexes173 and a glycolipid for the B subunit of Shiga toxin.174 Other intensifiers of MHC class I presentation in vitro include exosomes,175 liposomes,176 hepatitis B virus,177, 178 proteins coupled to the HIV-1 tat protein179 and dying cells (see below). Heat shock proteins, particularly gp96 or hsp70, can deliver peptides to MHC class I in vivo.180, 181 Mouse DCs exhibit saturable uptake of gp96182, 183 possibly via CD91, the receptor for α2-macroglobulin–protease complexes184 (but other receptors may exist). Heat shock proteins are therefore being considered as adjuvants for peptide delivery to DCs and even for modifying other aspects of DC function such as maturation.185, 186 As in the case of viral vectors, the ex vivo use of heat shock proteins allows control of the DC targeting step as well as DC maturation.

Dead or dying tumor cells and cell lines

There is an intriguing and potentially powerful way to improve the diversity of tumor-derived peptides that are presented by DCs. This is to allow DCs to phagocytose tumor cells or their fragments, whereupon tumor cell antigens are cross-presented by the MHC products of the DCs, both MHC classes I and II. Cross-presentation by DCs was first described with viral peptides from apoptotic, infected cells.168, 187 It is now evident that tumor-associated peptides also can be presented and from just 1 to 2 dead cells per DC.157, 187–192 Both apoptotic and necrotic types of cell death can be followed by cross-presentation.189, 193 In some instances, necrotic cells also mature the DCs194 and release gp96.195

A priori, there are pros and cons to the use of tumor cells as the source of antigen. These will require direct research to resolve. One concern is that dead cells load the DCs with numerous self peptides, capable of eliciting autoimmunity.196, 197 However, it has been proposed that the critical normal function of cross-presentation of dying cells by DCs is to induce tolerance to self peptides.198 According to this proposal, most patients already are tolerant to many self peptides that DCs are able to process. Another concern is that DCs may process antigens onto MHC class I via immunoproteasomes199 and the resulting peptides may thereby differ from those expressed by tumors.200 This problem may also be obviated if IFN-γ from immune T cells upregulates the immunoproteasome in the tumor.

The major potential with the use of tumor cells as the source of antigen is to allow DCs to be charged with a wide array of tumor peptides and on several molecules: MHC I, II and possibly CD1 (Fig. 1). Allogeneic tumors may suffice in instances where shared antigens are expressed frequently in a given tumor type, e.g., melanocyte differentiation antigens in melanoma and cancer testis antigens in a wide spectrum of malignancies. Cancer testis antigens are essentially tumor-specific shared antigens, because they are only found in one normal tissue, the testis, which does not express MHC products.201 It is now clear that DCs can be used to expand immunity to cancer testis antigens in patients.43, 116 Processing of allogeneic cells may even amplify the helper component of the immune response, because allogeneic MHC products can themselves be processed onto the MHC molecules of DCs.136 Alternatively, the use of autologous tumor may ensure the presentation of critical peptides derived from tumor-specific mutations, including mutations that likely contribute to oncogenesis.202, 203, 204 The near future should test in the clinic whether cross-presentation can meet the desired goal of improving the breadth of antitumor immunity. Laboratory and clinical studies should be intensified to increase tumor processing by maturing DCs in situ.

Fusions between DCs and tumor cells

Another strategy to deliver several tumor antigens to DCs is to fuse the 2 cell types, creating heterokaryons. Fused mixtures of tumor cells and DCs (even allogeneic DCs) enhance resistance to mouse mammary tumors, human ovarian cancer and advanced renal cancer.205, 206, 207 These results are puzzling because heterokaryons between different cell types, here DCs and tumor cells, normally are unstable. Also in human studies,207 allogeneic DCs were used, so most immune T cells would have to be primed to antigens presented on the self MHC molecules of the tumor cells.


Exosomes are small, membrane-bound vesicles either released from tumor cells and presented by DCs208 or released by DCs that have processed tumor cells.209 Exosomes contain MHC class I and II products, costimulatory molecules like CD86 and other cell-derived products such as heat shock proteins and the protein lactadherin.210 A potential limitation of this and other approaches requiring autologous tumor is that it may not be feasible to obtain large amounts of tumor cells from most patients. It is envisaged that exosomes could be developed into a standard form of tumor antigen, which would be captured and presented by DCs in vivo or by ex vivo–derived DCs.209 This will require methods to charge the exosomes with tumor-derived antigenic peptides and to ensure the maturation of DCs within the cancer-bearing patient.


The primary goal in these studies is to use ex vivo antigen-charged DCs to amplify the afferent limb of the immune system and thereby bypass potential defects in the presentation of tumor antigens in vivo. For example, tumors may be ignored by DCs, tumors may suppress DC function, or the cancer cells might be presented by immature DCs to induce T-reg cells. The initial DC studies reviewed here were designed to demonstrate feasibility, i.e., a lack of acute toxicity and responses to MHC class I– restricted tumor peptides. One omission from many of the first protocols is a simultaneous test with control (e.g., viral) peptides, to certify that the DCs were immunogenic for CD8+ T-cell responses and that the patient's immune system was generally competent to respond. In this regard, use of an aliquot of DCs pulsed with a viral peptide, as discussed above, would have provided pertinent information. Many of the first human protocols also were designed with DCs that were likely to have been immature, which may be suboptimal in many regards (see above). Despite this and even though basic variables like cell dose and route of injection were in the beginning uncharted territory, antigen-bearing DCs can enhance immunity in humans and induce clinical regression in some patients with advanced cancer. The published studies reviewed here used short-term vaccinations carried out over a few months. Studies with a longer follow-up or with revaccination strategies are under way211 to test the potential of antigen-bearing DCs to stabilize metastatic disease and even induce complete remissions.


Melanoma is the best-studied tumor from an immunologic perspective. Several melanoma antigens have been defined, especially from work with tumor-infiltrating lymphocytes,35 SEREX or antibody-based approaches212 and limiting dilution cloning from the blood of melanoma patients, especially with a favorable clinical outcome.213 Some of these antigens are also expressed on normal tissues (differentiation antigens shared with melanocytes); therefore, the possibility of tolerance to these antigens exists. Other antigens, such as cancer testis antigens, are more restricted to the tumor.212 As mentioned, cancer testis antigens are expressed in many tumors and the testis. At least one of these antigens, NY-ESO-1, is known to be immunogenic in the setting of cancer.214 At a minimum, cancer testis antigens serve as a valuable means to monitor T-cell immunity during DC-based vaccination protocols.

Several DC-based studies have been carried out in stage IV melanoma.41–43, 116, 215, 216 Two reports have shown an important principle, that even patients with advanced cancer typically are competent to mount CD8+ T-cell responses to MHC I–restricted viral peptide (influenza matrix), pulsed on DCs as a control antigen.43, 116 CD4+ T-cell responses to KLH or tetanus toxoid as control antigens pulsed on DCs have also been demonstrated in several studies.42, 116 Additional DCs were pulsed with melanoma peptides, including MAGE-3, tyrosinase, gp100 and Melan A/MART1. Immune responses to the peptides were detectable using ELISPOT assays, suggesting that tolerance against these antigens is partial. However, the responses still need to be assessed by other criteria, especially the capacity of the expanded T cells to kill melanoma targets.

Several intriguing findings have been made on 18 stage IV melanoma patients injected s.c. with CD34-derived DCs pulsed with tumor peptides.116 The vaccinations induced antigen-specific T cells (ELISPOT assay) that could be measured directly in fresh blood and further expanded in vitro in a 1-week recall assay, much as was reported with monocyte-derived DCs.43 In 2 patients who had some evidence of vitiligo prior to DC vaccination, the vitiligo flared markedly, suggesting that immune effectors could recognize targets displaying naturally processed peptide. The preliminary data from the 18 patients suggest a correlation between immune responses and early clinical outcome. The study also demonstrates that DCs can expand immunity simultaneously to multiple epitopes, which may be important for protective immunity.

DC biology should therefore be pursued in melanoma, to identify requirements for more robust immune responses and to achieve further clinical efficacy. For example, it will be important to use DCs exposed to more complex antigens delivered by viral vectors and dying cells and to examine the effects of more prolonged immunization protocols. In the search for stronger immunity, we include quantitative and qualitative aspects. For example, can tumor-specific CD4 and CD8 T cells be induced at comparably abundant levels to those seen in acute viral infections and can these T cells kill transformed targets?

Prostate cancer

In most of the published studies, tissue-specific self-antigens (PSA, PSMA, PAP) have been targeted, using either blood DCs or monocyte-derived DCs.217, 218, 219, 220, 221, 222 Some reductions in serum PSA, a prostate-specific antigen the levels of which likely reflect tumor burden, were observed in 17/66 hormone-refractory cancer patients injected with monocyte-derived DCs pulsed with PSMA peptides.219, 221 In 2 studies, injection of blood-derived DCs pulsed with a GM-CSF–PAP fusion protein led to >50% reduction in serum PSA in 3/13223 and 3/31222 patients. In these studies, the evidence for immunogenicity of DCs was based on antigen-specific proliferative responses after vaccination, but these were only detected in up to a third of patients in some,217, 222, 223 but not all,224 studies. Prostate cancer offers 2 key challenges for research in DC therapy: the availability of serum markers to follow tumor cell burden and defined prostate-restricted antigens and carcinoma lines to monitor the T-cell immune response.

Lymphoma and myeloma

The target selected in published studies for both tumors is a tumor-specific antigen, the idiotype of the monoclonal Ig expressed as a surface receptor or secreted by the tumor.40, 225–227 In some studies, patients also received soluble antigen,40 complicating evaluation of the response to the idiotype-pulsed DCs. Occasional tumor regressions were observed in some patients with lymphoma.228 Both low-grade, non-Hodgkin's lymphoma and myeloma respond to, but are not cured by, standard chemotherapy and may therefore be amenable to vaccination in the setting of minimal residual disease.

Virus-associated malignancies

In virus-associated malignancies, virus-derived antigens that are foreign to the immune system serve as attractive targets for immune therapy. The challenge here may therefore lie not with tolerance or avidity of the T-cell repertoire but with viral mechanisms of immune escape. Examples of such viruses and associated malignancies include hepatitis C and hepatoma, human papillomavirus and cervical cancer, EBV and lymphoma or nasopharyngeal cancer. Vaccination approaches have begun to target cervical cancer caused by human papillomaviruses,229, 230 but the immunity induced with current non-DC adjuvants appears weak, requiring weeks of cell culture to become manifest.46 For EBV, it is clear that DCs present several latency antigens (EBNA1, LMP1, LMP2 and EBNA3) to CD4 and/or CD8 T cells, using many different sources of antigen, i.e., peptides, proteins encoded by recombinant viral vectors and dead EBV-transformed B cells.157, 189, 231, 232

Viruses associated with human malignancy pose 2 other challenges for immunotherapy. First, mucosal immunization may be valuable. Hepatitis C, EBV and human papillomaviruses are transmitted via mucosal surfaces. DCs can be visualized beneath the antigen-transporting epithelium of mucosa-associated lymphoid tissues.233, 234 The CCR6 chemokine receptor, initially studied in the setting of epidermal DCs,235 now appears to be important for the capacity of DCs to initiate immunity to antigens deposited in mucosal sites.236 Second, the immune defenses of individuals who are infected but resist the transforming virus can be investigated. EBV is an important example since most adults are able to contain the virus at low levels throughout life and do not develop EBV-associated cancer. Healthy adults reliably recognize the EBNA1 antigen,157 essential for maintenance of EBV in growing cells. The CD4+ EBNA1 response is of the protective Th1 type,232 one of the only examples of a naturally polarized Th1 response in humans. These observations suggest that CD4+ T-cell immunity of the Th1 type is a valuable component of a protective response to a cancer-causing virus.

Renal cancer

Renal cancer is resistant to chemotherapy but can occasionally respond to biologicals such as IL-2. DCs have been charged with crude lysates of autologous tumors,237 but a provocative study used fusions of allogeneic DCs and patient-derived renal cancer cells.207 Of 17 patients, 6 (30%) underwent striking regressions of their cancers. The underlying mechanisms are not yet clear, but the allogeneic DCs may have induced IL-2 and in turn NK cells, to which renal carcinoma cells can be sensitive.

Other tumors

Immunotherapy directed against other tumors broadly falls into 2 categories: one is to target defined antigens such as overexpressed self antigens (Her2/neu, CEA, muc-1)238, 239 and mutant oncoproteins (bcr-abl, p53, mutant ras) and another is to use crude antigenic preparations derived from autologous tumor lysates,240 peptides eluted with acid from cells,241 tumor-derived RNA or heat shock proteins. Injection of tumor lysate-pulsed DCs led to marked tumor regression in a child with metastatic fibrosarcoma and substantial tumor burden.240

Immune monitoring

Active immunization with DCs provides an opportunity to test hypotheses on the immune response to cancer and to link these with biologic outcomes. Detailed immune monitoring will be valuable, especially information on the quality of the immune response, as mentioned above. In this way, the efficacy of immunization can be improved and immune correlates can emerge. As mentioned, more routine incorporation of control antigens (e.g., viral CD8+ T-cell epitopes) in the vaccine will help to strengthen the trial design by providing information on the competence of the injected DCs and the immune system of the patient. A DC vaccine study in melanoma suggests a correlation between the development of antigen-specific T-cell responses and a favorable early clinical outcome.116

It is now feasible to assess active immunotherapy in a quantitative way. The best-studied ligands for antigen-specific T cells are fluorescent MHC I–peptide tetramers,214, 242 but MHC II243 and CD1d244, 245 complexes are being developed. ELISPOT assays provide a means to obtain quantitative data on cytokine-producing cells and in some instances, the use of DCs as antigen-presenting cells can improve assay sensitivity.246, 247, 248 FACS assays for intracellular cytokines, following stimulation with tumor antigens or combined with surface binding of MHC tetramers, should be informative. In several in vivo situations, antigen-specific but dysfunctional CD8+ T cells can develop, i.e., cells that fail to express cytokines and cytolytic function.249 In summary, it is becoming more feasible to monitor tumor-specific T-cell numbers and function, i.e., the quantity and quality of the immune response. Several potentially critical, qualitative aspects can be monitored, such as the functional avidity of the T cells, Th1 and Th2 patterns of cytokine secretion, regulatory T cells, the capacity of effectors to recognize and kill tumor targets and the longevity of immune memory.

Immune monitoring is typically carried out on blood samples, but it will be valuable to relate these measurements to the tumor microenvironment. Nonetheless, the circulation should reflect the pool of active effectors and memory cells that are available to tissue depots.

In the near future, 2 other issues also need to be considered. One is comparison of DC-based immunization with other adjuvants. It is possible that the efficacy of chemical adjuvants,250e.g., GM-CSF, Quil A and CpG ODNs, depends on their capacity to mobilize and mature DCs. CpG ODNs are of special interest since this class of adjuvant matures DCs and polarizes T cells toward Th1.251, 252, 253, 254, 255 Theoretically, the ex vivo approach provides greater control over antigen presentation and DC dose, but it is more cumbersome. It is therefore essential to establish that these potential advantages of immunizing with ex vivo–activated DCs lead to superior immunologic outcomes. A second issue, which several groups are now addressing, is whether the immune or clinical responses observed in the early studies are durable and whether it is necessary to give booster injections with antigen-charged DCs or antigen in other adjuvants to maintain these responses.


DC-based immunization allows the investigator to evaluate a number of critical questions centered about cancer-specific T cells34 and cancer vaccines.146 Previously, these were difficult to address because of a lack of an efficient way to immunize T cells, even though many pertinent antigens and assays had been defined.

To what extent is the immune system tolerized or ignorant of tumor antigens?

It is possible that a tumor behaves in the same way as self tissue, leading to tolerance or avoiding autoimmunity. Several potential mechanisms underlie peripheral self tolerance: ignorance,256 clonal deletion and anergy257, 258, 259, 260 as well as T-reg cells.74 These active tolerance mechanisms may not be induced directly by the tumor but, instead, can be mediated by bone marrow–derived cells261 (perhaps immature DCs), as is the case with certain relatively abundant self antigens.257, 258, 259, 260 Initial data from patients with stage IV melanoma indicate that tolerance is incomplete. When these patients were immunized to DCs pulsed with peptides shared with normal melanocytes, most could expand peptide-specific CD8+ T cells.42, 43, 116 It remains to be shown that these T cells are functionally efficient, i.e., able to kill tumor targets, but as mentioned above, some signs of short-term clinical efficacy have been observed.116 Tumors may actively promote “clonal ignorance” by blocking DC maturation and the presentation of antigens.20, 262 If cancers do block DC maturation in situ, then active immunization with ex vivo–derived DCs could obviate this potentially critical deficit. Fortunately, it is not difficult to prepare potent immunostimulatory DCs from patients with advanced cancer.

What are the properties of tumor-reactive T cells once induced?

If tolerance does not prove to be a major or consistent obstacle to active immunotherapy, we can begin to study the properties of tumor-reactive cells in situ, an area full of unknowns. Are T cells able to home to the tumor, or must local inflammation be generated and tumor-suppressor factors blocked? Is there a critical tumor mass that can be handled by immune T cells since it is now directly evident that larger experimental tumors resist CD8+ killer cells without tolerizing them?21 If T cells access a cancer, has antigen presentation by the tumor cells been blocked, e.g., by dampening expression of MHC class I?32 Will we need to improve the efficacy of tumor-specific CD8+ T cells through the coordinate mobilization of Th1-type helpers and innate NK and NKT cells?

Can NK cells be mobilized to kill tumors that have dampened MHC I presentation?

Tumors frequently downregulate presentation on MHC class I, which are the ligands for KIRs on NK cells.263 Once inhibition by KIRs is dampened, NK cells should be allowed to kill tumor targets lacking MHC class I. However, tumor cells must also express the ligands for natural cytotoxicity or KARs,264 but their biochemical identification is still in the early stages. Alternatively, tumors must be coated with antibodies to activate Fc receptors on NK cells (Fig. 1). Several tumors, e.g., melanoma, myeloma and lymphoma, are targets for NK cells. It would be valuable to determine if DCs can be used to mobilize NK cells and increase resistance of these patients.

Several initial pieces of information are in place with respect to DCs and NK cells. DCs can be targets for NK cells265 and, therefore, should also have ligands for activating NK receptors. DCs express high levels of requisite costimulators for NK cells, e.g., CD48 that costimulates NK cells via 2B4.266, 267 DCs can directly activate NK cells in mice, probably via a contact-dependent mechanism.268 DCs can release large amounts of the key cytokines, e.g., IL-12 and IFN-γ, which improve NK function. IL-2 induced from T cells by DCs also could expand NK numbers. DC–T-cell aggregates have been observed in tumors,269 and if this leads to local release of IL-2, the growth of NK cells would ensue. Therefore, even though NK cells are part of the innate immune response, possibly they can be mobilized by DCs to enhance tumor cell killing and to avoid key tumor escape mechanisms.

Can NKT cells be mobilized to kill tumors and further activate innate and adaptive immunity?

NKT cells are a newly recognized and potentially powerful link between innate and adaptive immunity, which is controlled by DCs.270 A synthetic α-galactosyl ceramide can be used as a glycolipid mimic to activate NKT cells.271, 272 After triggering by α-galactosyl ceramide, especially on DCs, activated NKT cells kill a wide spectrum of established tumors in mice,273 most likely by recognizing some endogenous tumor glycolipid presented on CD1d molecules.274 The DC–NKT interaction also generates large amounts of IFNs and IL-12,273, 275 which activate other cells, especially NK cells. NKT cells directly kill many human tumor lines in culture but not mitogen-activated lymphocytes.276 Possibly, authentic human tumors express endogenous ligands, e.g., glycerophospholipids,274 better than nonmalignant cells.273, 276 Therefore, as in the case of NK cells, mobilization of NKT cells could valuably expand (through release of cytokines) and buttress (through recognition of MHC class I-low tumor cells) the adaptive T-cell response.

How important are CD4+ T helpers, including Th1 polarized cells, for the efficacy of tumor-specific CD8 T cells?

Helper cells play a critical role in experimental tumor immunity, even when tumors lack MHC class II.27, 28, 47, 48, 277–280 The requisite helpers develop after cross-presentation of tumor cells by DCs. Indirect evidence for the presence of tumor-specific CD4+ helper cells comes from the presence of antibodies to tumor antigens. However, antibody formation can be promoted by Th2-type cells and, as discussed earlier, Th2 cells may not be optimal for adaptive and innate resistance. The ways in which Th1 CD4+ T cells amplify tumor resistance are being investigated. IFN-γ upregulates immunoproteasome activity and expression of MHC I on tumor targets. CD40L on CD4+ T cells can improve many aspects of DC function, including production of IL-1565, 66 to sustain memory cells.68 CD40-independent helper mechanisms also exist.281, 282 The same scenario is proposed for chronic viral infections; i.e., CD4+ Th1 helpers are important for resistance55 and, at least in EBV infection, these helpers might be induced by cross-presentation of transformed cells by DCs.157 Helper epitopes enhance human CD8+ T-cell responses in patients to the HER-2/neu antigen.283

Can DCs and cell-mediated resistance be manipulated directly in vivo?

There have been advances in manipulating the DC system in vivo, without having first to prepare and manipulate the cells ex vivo. One such advance is the use of either FLT-3L or G-CSF to mobilize DC precursors in large numbers.132, 133 The DCs used are not fully differentiated or mature. Immature DCs are advantageous for antigen uptake and targeting, but maturation stimuli will also be necessary for immune efficacy. A second advance relates to the control of DC maturation directly in situ. When used in low doses, CpG ODNs provide a nontoxic means for maturing DCs, including IL-12 production.251–253, 255, 284–288 There is no information comparing the number of mature DCs mobilized by current adjuvants relative to the number mobilized through ex vivo DC-immunization methods. It is to be kept in mind that DC subsets may differ in their content of specific Toll receptors, through which ODNs function. Nonetheless, the combination of antigens plus CpG ODNs may directly allow more mature DCs to present antigens in vivo.

DC approaches to immunotherapy will likely benefit from coupling to other therapies, immune-based and otherwise. Tumor-specific MAbs might enhance tumor cell uptake and presentation by DCs, as might chemotherapy and antibody–toxin conjugates. After one learns to immunize patients to produce more tumor-reactive T cells, hopefully at levels comparable to those seen in successful resistance to viruses and other infections, other approaches (chemotherapy and radiotherapy) may reduce tumor burden or act synergistically with the DCs, e.g., to ensure that the immune cells access the tumor bed. In other words, active immunization enhances the formation of tumor-reactive T cells, but ancillary therapies should improve their efficacy once induced.


We are grateful to many colleagues for their comments, Drs. A. Granelli-Piperno, I. Mellman, C. Münz, G. Schuler and J.W. Young. This work was supported in part by grants from the National Institutes of Health: K23-CA81138, PO1-CA84512 and MO-RR00102 to the Rockefeller General Clinical Research Center.