DC are the sentinel cells of the immune system. They sample their environment, acquire and integrate information and guide the adaptive immune response 1, 2. The outcome of presentation of their antigenic cargo to T cells depends on the context in which the antigen was acquired. DC have various receptors that alert them to microbial intruders or tissue damage. Self-antigen presented by DC, in the absence of these “danger” signals, generally results in tolerance and serves as a mechanism to limit auto-reactivity. However, antigen presented in the context of such danger signals results in a vigorous response that establishes immunity. The recognition of microbial patterns can result in DC maturation, which in turn directly affects DC functions: Immature DC efficiently take up antigen but are poor at activating T cells, whereas mature DC shut down antigen acquisition but up-regulate antigen presentation and the co-stimulatory machinery required for efficiently priming T cells. A complication to this general picture of DC biology is the fact that DC are not homogenous. Rather, there is a network of DC subtypes differing in location, developmental history and specialised functions 3, 4. These different DC subtypes can directly determine the type of immune response generated 3, 4.
As the key players orchestrating immunity, DC are a natural focus for immune therapy. In the clinical setting, DC from various progenitor cells have been expanded in culture, loaded with tumour antigens and then re-administered into cancer patients. Such immunisation strategies have produced some limited successes but so far have not emerged as an effective means of treating cancer 5. Even if this approach does eventually prove effective, it remains a complex and expensive procedure. An alternative approach of wider application involves directly delivering antigen to DC in vivo and thereby manipulating the immune response 5. The aim might be to induce tolerance to the antigen and prevent autoimmunity or transplant rejection, or to induce CTL responses to a tumour or viral infection, or to improve efficiency of an antibody-dependent vaccine. Most of these DC-targeting studies have been conducted in mice where mAb against different DC cell surface molecules are used to deliver antigen to the DC.
This review considers various murine models of targeting antigen to DC, and the factors that have been found to determine whether effective tolerance, cellular immunity or humoral immunity is obtained. These mouse models raise issues that need to be considered before the approach can be applied in human clinical trials.
Which molecule should be used to target antigens to DC?
The first issue is whether a surface molecule needs to be DC-specific to serve as a target for effective antigen delivery. In fact, there are few if any DC-specific surface molecules. Even CD11c, the DC hallmark molecule, is expressed on macrophages 6, NK cells 7 and activated CD8 T cells 8. Similarly, DEC-205 (CD205) is present in cells other than DC, including thymic epithelial cells, and at lower levels, on B cells 9, 10. However, both CD11c 11, 12 and, particularly, DEC-205 13, 14, have been used successfully to target DC in mice to induce immunity. In addition, CD36 15, the mannose receptor 16 and LOX-1 17, which are all expressed on multiple cell types as well as DC, have been found to be effective target molecules for eliciting anti-tumour responses. Lew 18 has demonstrated that target molecules that allow accumulation of antigen within the lymph nodes serves to enhance immune responses. DC are strategically placed within lymphoid tissue and are efficient at antigen uptake, and so may have selective access to any antigen that arrives close to the right “address”.
Nevertheless, selective targeting of antigen to DC alone should have a number of advantages. Most obviously, the efficiency of the targeting antigen may be reduced if other cells “mop up” the mAb carrying the antigen cargo. Thus, the efficiency of only tiny amounts of mAb against Clec9A in targeting antigens and enhancing immune responses 19 may be the consequence of the relative restriction of this molecule to the DC surface. In addition, the acquisition of targeted antigen by other cell types may directly affect the immune response generated. For example, B cells can induce tolerance 20 and macrophages may rapidly degrade antigen and reduce its effectiveness 21.
Since the DC subsets differ in specialised functions, there may be a need to target antigen to a molecule selectively expressed by one particular DC subtype; some examples of this are discussed later. There should also be an advantage in targeting a molecule that is an effective endocytic receptor. In addition, it is clear that different endocytic receptors can shuttle antigen into different processing pathways 22. A comparison of the efficiency of targeting different cell surface molecules has revealed some hierarchical capacity to promote MHC-class-I- or II-restricted responses, suggesting that different molecules may preferentially be equipped to elicit CD4 or CD8 T-cell responses 11, 23–25. Ultimately, the nature of the molecule targeted, the DC subset expressing that molecule, other cell types expressing the targeted molecule and even the type of mAb used to target may all impact the immune outcome. How these theoretical concerns play out may often have to be determined empirically.
One overall consideration is that information from the mouse-targeting model should eventually be translatable to human clinical application. Our own earlier studies targeting the seven-transmembrane molecule FIRE on mouse DC were of limited use in these terms, since the human gene contains a stop codon and the molecule is never expressed on the surface of human DC 26, 27. In contrast, current targeting studies from our group 19 and others 28 on the C-type lectin Clec9A, selectively expressed by CD8+DC, are much more likely to translate to human application, since the equivalent human molecule is also selectively expressed on human BDCA-3+ DC, the proposed equivalent of the murine CD8+DC lineage 29.
Can targeting different DC subsets direct different types of immune responses?
Though all DC share the capacity to take up and present antigen, the various DC subsets have functional specialisations. Plasmacytoid DC are proficient producers of type 1 IFN, but their more restricted capacity to activate T cells remains controversial 3. By contrast, conventional DC are efficient at activating naïve T cells, and are especially effective in their mature or activated state. In the mouse, subsets of lymphoid tissue-resident conventional DC can be segregated based on their expression of CD8α. The CD8+DC are especially equipped for “cross-presentation” of exogenous antigen on MHC class I 30–32. In addition, they are the major producers of IL-12 33, 34 and thus promote Th1 responses 35. By contrast, the CD8−DC are better at MHC class II presentation, especially when antigen levels are limiting 31, 32 and have been suggested to drive Th2 type responses 35, 36.
Thus, it seems likely that targeting different DC subsets could result in different types of immune responses. Indeed, under activating conditions, when CD8−DC and CD8+DC were separately targeted, using the anti-DCIR-2 mAb (33D1) and anti-DEC-205 mAb (NLDC145), respectively, the targeted CD8−DC were more effective at inducing OVA-specific CD4 T-cell responses, while the CD8+DC were more effective at producing OVA-specific CD8 T-cell responses 24. Similarly, targeting a CD8−DC subset using anti-Dectin-1 mAb induced stronger OVA-specific CD4 T-cell responses but weaker OVA-specific CD8 T-cell responses than targeting CD8+DC with anti-DEC-205 mAb 25. In agreement with the notion that CD8−DC are particularly efficient at MHC class II presentation and driving Th cell differentiation, a Th-dependent humoral response was induced by targeting the surface molecules CIRE (mDC-SIGN) and FIRE, molecules that are selectively expressed by CD8−DC 37.
However, it is important to note that CD8+DC are nevertheless capable of inducing effective CD4 T-cell responses. CD4 T-cell proliferation and Th-dependent humoral responses were obtained by selectively targeting antigen to CD8+DC with anti-Clec9A 19. Interestingly, the mechanism of CD4 T-cell activation was found to be different when the Leishmania major LACK protein was targeted to CD8+DC using anti-DEC-205 mAb, compared with targeting CD8−DC using anti-DCIR2 mAb (33D1) 38. Targeting the CD8+DC in this way induced transgenic LACK-specific CD4 T cells to produce IFN-γ exclusively, whereas targeting the CD8−DC induced both IFN-γ and IL-4. The biased Th response translated into a biased humoral response, where targeting CD8+DC resulted in the exclusive production of an IgG2a isotype response, while targeting CD8−DC resulted in a mixed isotype response (IgG1, 2a and 2b) 38. The underlying mechanism governing the induction of IFN-γ production by the CD4 T cells also seemed to differ; the production of IFN-γ induced by the CD8+DC was IL-12p40-independent but CD70-dependent; conversely, targeted CD8−DC induced IFN-γ in an IL-12p40-dependent but CD70-independent manner 38.
Overall, targeting different DC subsets may prove useful in protection against particular infections where disease resistance requires a specific type of immune response. However, the extent to which targeting a molecule on a particular DC subset will reflect the immunological bias of that subset, as opposed to overriding it, will depend on many factors, including the antigen dose delivered, and will need to be established on a case-by-case basis.
Can tolerance be induced by targeting antigen to DC?
The notion that targeting antigen to steady-state DC in the absence of DC-activating agents or adjuvants may result in tolerance was first introduced by Finkelman 39 using the anti-DCIR2 mAb 33D1. However, the largest body of work exploring targeting of DC for the induction of tolerance comes from the laboratories of Steinman and Nussenzweig using DEC-205 as the target molecule 13, 14, 40. DEC-205 is a multi-lectin endocytic receptor preferentially expressed by CD8+DC in the spleen, but also expressed by Langerhans' cells, dermal DC and interstitial DC in lymph nodes. The anti-DEC-205 mAb has been used to deliver a multitude of antigens to DC. While delivering DEC-205 targeted antigen in conjugation with DC activation/maturation agents (such as anti-CD40 mAb and polyI:C) results in potent cellular and humoral immune responses 13, 38, 41–47, targeting DEC-205 in the absence of such maturation signals results in T-cell tolerance (Fig. 1).
Figure 1. Possible results of antigen targeting to DC in the presence or absence of adjuvants. Antigen is linked to mAb recognising various DC surface molecules. Immature (quiescent) or mature (activated) CD8+DC and CD8-DC are depicted using confocal images provided by Dr J. Villadangos. Scenario 1: Antigen delivery in the presence of activation/maturation stimuli results in antigen presentation and activation of T cells leading to immunity. CD8+DC are better equipped for MHC class I presentation and CD8 T cell activation whereas, the CD8-DC are superior at MHC class II presentation and activation of CD4 T cells. Scenario 2: Antigen delivery in the absence of an activation/ maturation stimuli results in antigen presentation that induces only transient T cell proliferation, which is then followed by T cell deletion, anergy and the generation of Treg cells. Scenario 3: Antigen delivery in the absence of additional activation/maturation stimuli results in antigen presentation that selectively activates CD4 T helper cells and B cells, but not CD8 T cell effectors. Recent data discussed provides evidence for scenario 3, although it is not clear whether the targeting mAb cause some activation of the DC that finally present the antigen.
Download figure to PowerPoint
Delivering antigen to quiescent DC via DC-205 initially results in MHC class I and MHC class II presentation and transient antigen-specific T-cell proliferation 14, 40. However, antigen-specific CD8 T cells are subsequently deleted and the mice became unresponsive to re-challenge even with adjuvant 40. Similarly, antigen-specific CD4 T cells are deleted, or rendered unresponsive 14. Furthermore, Treg cells are induced 48–50. Multiple injections of antigen-conjugated anti-DEC-205 results in increased expression of Foxp3, CTLA-4, IL-10 and TGF-β by antigen-specific CD4 T cells 51. The ability to induce tolerance via targeting antigen to DEC-205 has been tested in experimental models of type 1 diabetes 51, 52. Both the onset of this autoimmune disease 51 and its progression once initiated 52, have been prevented. The mechanism of tolerance induction involved both the generation of Treg and the deletion of auto-reactive CD8 T cells. These model studies provide hope that autoimmune responses can be specifically switched off by appropriate DC targeting. Similar to the effect of targeting antigen to CD8+DEC-205+DC in the absence of maturation/activation stimuli, targeting antigen to DCIR2+CD8−DC results in tolerance 39. Interestingly, targeting both CD8+DEC-205+DC and DCIR2+CD8−DC results in the induction of Treg cells, though the underlying mechanisms differ 50. Furthermore, CD8+DEC-205+DC preferentially induce Treg from adoptively transferred antigen-specific CD4 T cells, whereas DCIR2+CD8−DC are superior at stimulating natural Foxp3+ Treg 50. A better understanding of the role of DC subsets in the induction of tolerance should lead to the design of better therapeutic strategies.
Are adjuvants required to obtain immune responses to DC-targeting antigens?
The case study of DC targeting with anti-DEC-205 supports the notion that the activation state of the DC will determine the immune response generated: quiescent (immature) DC tolerise, whereas activated (mature) DC immunise. But how well does this generalisation hold and does the molecule targeted affect the outcome? There are now many examples of effective immune responses to DC-targeted antigens delivered in the absence of DC-activating agents. Avoiding the requirement for adjuvants would be a major advantage in clinical applications of DC targeting. These studies can be segregated into those measuring humoral responses and those measuring cellular responses.
In 1987 Carayanniotis and Barber 53 used anti-class II MHC mAb to target antigen to APC and reported that antibody responses could be induced in the absence of adjuvant; this procedure clearly targeted B cells as well as DC. Subsequent studies showed that targeting many other molecules found on DC could induce potent antibody responses without using adjuvants; these included MHC class I 1 FcγRII 54, CD45, CD45RA, DCIR-2, CD4 55 and the mannose receptor 16 (Fig. 1).
We have obtained potent Th dependent-antibody responses without adjuvants by targeting molecules such as CIRE (mDC-SIGN) and FIRE 37, found on CD8−DC, and by targeting Clec9A found on CD8+DC 19. Accordingly, it appears to be the nature of the molecule targeted, rather than the DC subset, that determines this outcome. Other DC molecules targeted have not given strong antibody responses unless adjuvants were provided; these include Dectin-1 25, DCIR2 39 and notably DEC-205 37, 41. Targeting CD11c has given contradictory outcomes, including antibody responses without adjuvant 12, 39 and lack of any antibody responses 55; the different methodologies used presumably contributed to the disparate results.
Many of the studies where adjuvant was not required failed to exclude inadvertent exposure to activation agents: traces amount of endotoxin within the antibody-antigen preparations, or exposure to pathogens in facilities that were not specifically –pathogen-free, might have provided DC activation signals. However, the recent potent antibody responses obtained by targeting antigens with Clec9A 19, and the earlier responses to FIRE (Caminschi, unpublished data), were also obtained in MyD88−/− TRIF−/− gene-deficient mice, which are unable to respond to LPS or other TLR ligands. In some cases the molecule targeted itself might give an activation signal to the DC 56. However, in the case of targeting Clec9A, there is neither evidence of generalised DC activation nor of activation responses by the targeted DC population 19, 28. It appears that, if appropriately targeted, even steady-state DC can generate Th cells that are effective in promoting antibody responses.
In contrast to the humoral responses, cellular immunity and, particularly, the generation of CTL from the endogenous repertoire usually requires the administration of an adjuvant or DC activation agent along with the targeted antigen. This was well established for DEC-205 13, 14, 40, but also applied to targeting to DCIR2 24, 38, Dectin-1 25, mannose receptor 16, CD11c and MHC class II 11. The contrast with humoral responses was particularly striking with targeting Clec9A, which gave potent antibody responses without adjuvants, but required DC activation agents (polyI:C, anti-CD40, LPS and CpG) before CTL could be generated 28 (Caminschi, unpublished data) and harnessed to eradicate tumours 28.
The one strikingly different result was the targeting of antigens to CD36, a multi-ligand scavenger receptor that is expressed on CD8+DC but also on a variety of other hemapoietic cells. Such targeting in the absence of adjuvant resulted in activation and expansion of CD8 T cells, which in contrast to the finding with DEC-205 were not deleted 15. The cellular immune response so generated was able to prevent tumour growth. However, it remains unclear whether endogenous CTL were generated and if so whether these, as opposed to CD4 T cells, contributed to the elimination of tumour cells. Furthermore, confirmation of the original finding in MyD88−/−TRIF−/− mice would strengthen the conclusion that targeting CD36 in the absence adjuvants (i.e. TLR ligands) induces cellular immunity.
What are the optimal conditions for enhancing CTL production by DC targeting?
Immune destruction of tumours and resistance to many viral infections requires CTL as effector cells. Theoretically this should be best achieved by targeting antigens to a DC subset efficient at uptake of exogenous antigens and cross-presentation on MHC class I for effective CD8 T-cell priming. The specialised “cross-presenting” CD8+DC subset in the mouse is such a target, and targeting DEC-205 or Clec9A should accomplish this; there is evidence that CLEC9A marks an equivalent human DC subtype 19, 28, 57. Using an endocytic receptor that selectively channels antigen into the MHC class I presentation pathway should also be advantageous 22. Targeting antigen to DEC-205 on human-monocyte-derived DC promotes cross-presentation 58. However, effective antigen presentation to CD8 T cells alone may be insufficient, as CD8 T cells have been reported to require CD4 T-cell help for effective memory formation 59, 60 and in some cases even for primary responses 61. To ensure optimal killer T-cell immunity, CD8 T cells need to be differentiated into effector CTL and memory T cells, and CD4 T cells need to be driven to effective helpers rather than into Treg. Most studies indicate this is best accomplished by providing additional adjuvants 11, 13, 16, 28, which cause simultaneous DC activation. One well-established factor for CTL development is IL-12p70 62 and so the production of IL12p70 by CD8+DC when activated is one of the reasons why simultaneous DC activation may be required for the induction of an effective CTL response.
What are the optimal conditions for enhancing antibody responses by DC targeting?
There are many circumstances where candidate antigens for antibody-dependent vaccines provide only partial protection because of low antibody responses. Targeting such antigen to DC might greatly increase their effectiveness. The assumption here is that Th cells are a limiting factor and that targeting antigen to DC enhances Th cell production. There are many reports of DC targeting enhancing CD4 T-cell responses 11, 14, 16, 19, 24, 25, 38, 40. Direct demonstration that the enhanced antibody production depends on such T cells has been provided for targeting Clec9A; no antibody response was obtained by targeting DC in T-cell-deficient mice 19. In addition, no response was obtained in mice lacking MHC class II (Caminschi et al., unpublished data). Since, the CD8−DC subset is generally more adept at activating CD4 T cells, targeting CD8−DC would seem an effective strategy; indeed enhanced antibody responses have been obtained by targeting CD8−DC 25, 37, 39. In this context, it is surprising that the highest antibody responses we have obtained are by targeting antigens to Clec9A 19, which is selectively expressed by CD8+DC rather than CD8−DC. However, CD8+DC have a striking ability to collect and reprocess antigens from dead or dying cells 32, 63, and it seems likely that Clec9A is involved in this process (Lahoud, unpublished data; Reis e Sousa, personal communication). The overall efficiency of antigen uptake and processing via this route may override any bias away from MHC class II presentation.
There are obvious advantages in a vaccine that does not require adjuvants and their associated side effects. It is unclear why the antibody response produced by targeting certain DC surface molecules, such as Clec9A, does not require additional DC-activating agents. The targeted molecule may transmit subtle signals, short of overt DC activation, that allow production of appropriate cytokines or that improve processing towards MHC class II presentation. Alternatively, processing and presentation may be so efficient that even the levels of co-stimulatory molecules on steady-state quiescent DC are adequate for activation of Th cells, if not for CTL production. The factors that promote the development of Th cells specialising in supporting B-cell antibody production are still being characterised 64–66. Whether Clec9A-targeted DC provide the appropriate soluble factors, or whether they more directly instruct Th cells for optimal B-cell activation, remains unclear and is the subject of further investigation.
However, there are several factors that need to be investigated before using this approach to improve antibody responses mediated by human vaccines. Although good and prolonged primary antibody responses can be obtained by targeting DC molecules like Clec9A 19, it is not yet clear whether they can mediate effective long-term memory, a requirement for most vaccines. Generation of effective immunological memory might require adjuvants in the initial immunisation. In addition, it is not yet clear how the B cells finally responsible for the antibody production should encounter the antigen for optimal responses. If the only limiting factor is at the Th-cell level, B-cell encounter with free antigen, not presented by DC, might suffice. However, the possibility that the targeted DC itself takes up and presents antigen to both the B cells and the Th cells must be considered 67. This could explain the high efficiency of antibody responses for some targeting strategies.
Will targeting antigen to DC be effective in humans?
The results in mouse models make it an attractive concept to immunise patients using DC-targeting antibodies that carry an antigen cargo. Studies attempting to predict the clinical feasibility of this approach have already been conducted, using as various human DC target molecules including DEC-205, DC-SIGN, DCIR and the mannose receptor 68–73. In primates, administration of anti-DC-SIGN mAb was shown to target APC 68. The in vivo targeting capacity of DC-SIGN was assessed in “humanised mice”, that is mice reconstituted with human immune cells. These humanised mice were immunised with anti-DC-SIGN mAb carrying either tetanus toxoid peptides or keyhole limpet hemocyanin (KLH) antigen and were found to have enhanced T-cell responses 74. Similarly, in another humanised mouse model, the anti-DEC-205 mAb was used to deliver EBV nuclear antigen 1 to DC, resulting in moderate induction of IFN-γ-producing T cells 70.
Human DC, targeted in vitro with anti-DC-SIGN coupled with KLH induced expansion of KLH-specific T cells 69. Similarly, even low doses of anti-DEC-205 mAb carrying the HIV p24 gag protein induced CD8 T cells from HIV patients to proliferate and produce IFN-γ in vitro58. The mannose receptor was also successfully used to deliver antigen to human DC in vitro: anti-mannose receptor mAb fused to melanoma antigen (pmel17) or an oncofetal antigen (human chorionic gonadotropin β subunit) induced both antigen-specific cytotoxic and Th responses 72, 73. Of note, Bozzacco et al. compared DEC-205, DC-SIGN and the mannose receptor regarding their ability to target HIV p24 gag protein to DC and elicit memory T-cell responses. They found that although all mAb bound their molecules comparably on monocyte-derived DC and were internalised, anti-DEC-205 reliably induced more IFN-γ-producing CD8 T cells 58. Whether these in vitro results will predict in vivo outcomes is unclear. These results are very promising, but determination of the human target molecule that is most efficient at delivering antigen to DC and promoting immunity will need to be assessed in clinical trials.