By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Dendritic cells (DCs) are the key orchestrators of the adaptive immune system. DCs have an outstanding ability to capture, process and present antigens to activate naive T cells. They also have the ability to regulate the nature of the T cell response by providing appropriate costimulatory signals that dictate either immunogenic or tolerogenic T cell stimulation. These unique features make targeted manipulation of DCs an attractive approach for modulating immune responses against cancer.
Ex vivo modified DC-based vaccines have been used in the clinic for the immunotherapy of cancer, so far with only limited success . The production of these DC vaccines usually entails differentiating DCs from individual patients' progenitor cells and their subsequent loading with tumor-associated antigens (TAAs), followed by appropriate maturation induction and their injection back into the patients in the hope of inducing an effective anti-tumor immune response . Limited clinical efficacy observed so far may be a result of sub-optimal ex vivo DC activation/maturation (required for effective T-effector cell priming and differentiation) and limited migration of the administered DCs; typically less than 2% of the injected DC number reach the vaccination site-draining lymph nodes (LNs) where T cell activation should occur [1, 2]. Even though further optimization may be expected to enhance the efficacy of these DC vaccines, this approach remains laborious and expensive. Moreover, inter-individual variability in the quality of the vaccines is inherent to such an autologous cell-based approach and complicates the interpretation of clinical data. An attractive alternative approach is to load the resident DCs in vivo by targeted delivery of TAAs and activating them simultaneously, thereby avoiding the cumbersome isolation and culture of DC precursor cells. This also serves to exploit the natural homing ability to LNs of activated DCs in situ for effective antigen-specific T cell activation.
The outcome of the immune response induced by targeting of antigens to DCs depends on many factors, including the specific receptor targeted, the expression level of the targeted receptor and its distribution among different DC subsets, the presence or absence of co-administered adjuvant, and the antigen delivery system used . The specific receptor of choice should have the following characteristics: (i) the target should be selectively or highly expressed on DCs; (ii) the targeted antigen should be effectively internalized, processed and presented to antigen-specific T cells, ideally to both CD4+ T-helper cells (Th) and CD8+ cytotoxic T lymphocytes (CTL); and (iii) the targeted delivery of antigen should preferably lead to selective activation of the targeted DC, bypassing the need for co-administration of an adjuvant to induce an optimal anti-tumor immune response. A variety of receptors expressed on DCs have been exploited for targeting antigens to DCs in vivo, including major histocompatibility protein (MHC) II, CD11c, CD36, DEC205, DCIR2, Dectin-1/2, CD80/86, F4/80-like receptor, CIRE and mannose receptor [3-5].
Configurations used for the targeted delivery of antigens in vivo include: (i) peptides or proteins conjugated to receptor ligands or monoclonal antibodies specific for receptors expressed on the DC surface; (ii) micro- or nanoparticles loaded or coated with peptides; and (iii) viral or nonviral vectors carrying TAA encoding genes.
Among the above systems, the natural ability of viral vectors to efficiently infect target cells and utilize their cellular machinery for the production of virally encoded antigens, makes them very attractive vaccine vehicles . In addition, viral vectors also are very effective because they mimic a natural infection and trigger the right kind of innate immune response, which is a precondition for the development of an effective Th1 response to the vaccine-encoded antigen [7, 8]. Some viruses even exhibit a natural tropism for DCs, making them pre-eminently suitable vaccine carriers for DC-targeted vaccination . Finally, viral transduction results in durable endogenous antigen expression, which facilitates MHC class I presentation after proteasome-mediated degradation and also continuous presentation of antigen to specific CTLs over prolonged periods of time, hence making attenuated, nonreplicating viruses attractive vaccine carriers for cancer immunotherapy [10, 11]. By contrast, nonviral systems often need to be specifically designed and optimized to achieve the above listed advantageous natural traits of viral vectors .
Adenoviral (Ad) vectors as vehicles for targeted gene delivery
Ad vectors have several characteristics that make them particularly useful as vehicles for DC-targeted gene delivery [12-14]. They are non-enveloped viruses that can be stably packaged as lyophilized preparations in vials or capsules. Ad DNA does not integrate into host chromosomal DNA; Ad-mediated transduction thus poses minimal risk for insertional mutagenesis . Ad vectors are easily grown to high titers, exhibit low cytotoxicity, have a large cloning capacity and are efficient in gene transfer, even to nondividing target cells. The latter in particular is relevant to DC targeting because DCs residing in peripheral tissues have reached end-stage differentiation and lost their proliferative capacity. The most commonly used vectors are derived from Ad type-5 (Ad5) and have shown their potential to stimulate a potent humoral and cellular immune response to transgene encoded antigens [16, 17]. Although notable toxicities and reduced efficacy have been associated with pre-existing immunity to the most common Ad vectors infecting humans , Ad vector-based vaccines and therapeutics are still considered to be among the safest and most effective, leading to their continued application in developing improved treatments for a diverse array of diseases [19-22]. An exceptional ability of Ad vectors to generate cellular immunity against recombinant antigens in humans  makes them preferred antigen carriers for cancer vaccination.
Many tumor cells express antigens that provide potential targets for therapeutic vaccine strategies to boost natural immune-mediated tumor rejection . The identification and characterization of genes encoding such cancer rejection antigens have opened new ways for the development of cancer vaccines. Many recombinant Ad vectors encoding TAAs have been generated for cancer immunotherapy purposes . Limited clinical experience has shown recombinant Ad vaccination in cancer patients to be safe, although convincing evidence of its clinical efficacy is still lacking .
The Ad5 fiber knob domain mediates binding to the coxsackie-adenovirus receptor (CAR) on the target cell surface of Ad5-susceptible cells (Figure 1). Subsequent interaction of the Ad penton base (i.e. where the protruding fibers connect to the viral capsid) with integrins on the cell surface facilitates internalization of the virion. However, DCs are resistant to Ad5 infection as a result of their lack of expression of the primary Ad5 docking receptor CAR, presenting a challenge for effective transduction of DCs by Ad5 . Direct in vivo administration of untargeted Ad5 may thus result in cytopathic effects as a result of ectopic gene transfer to CAR expressing bystander cells rather than DCs. Moreover, subsequent antigen presentation by these transduced nonprofessional antigen-presenting cells (APCs) may lead to suboptimal T cell activation, or even tolerance induction . Systemic utility of Ad5 vectors is also limited as a result of the high prevalence (in 50–90% of the population) of pre-existing immunity towards the vector . This pre-existent immunity may also limit the vaccination efficacy of Ad5 vectors in prime-boost protocols. To avoid these side-effects and to improve therapeutic efficacy, the natural tropism of the vector can be ablated and altered to improve selective tropism to DCs. In general, either a ‘two-component’ or a ‘single-component’ approach has been used to achieve these goals (Figure 2) . In the two-component approach, a bispecific adaptor molecule is complexed with the Ad5 vector to simultaneously block native receptor binding and retarget Ad5 vectors to receptors expressed on the target cell of interest. In the ‘single-component’ approach, the Ad5 vector is genetically-modified to ablate native receptor binding and to re-target the vector to a cell receptor of interest. The main advantage of the two-component approach is its great flexibility because vectors carrying different genes of interest can be easily coupled to these adaptors. On the other hand, single-component systems offer advantages for standardized manufacturing, providing greater potential for commercial distribution of a therapeutic vector . We have explored both approaches in the retargeting of Ad5 vectors to the CD40 surface receptor on skin DCs in an endeavor to translate an intradermal (i.d.) DC-targeted vaccination approach to the clinic (Figure 2).
CD40 as a receptor for simultaneous antigen targeting and activation of DCs
CD40 is a cell surface receptor that belongs to the tumor necrosis factor (TNF)-receptor family. It was first identified and functionally characterized on B lymphocytes [30, 31]. Later reports revealed that the expression of CD40 was not restricted to B cells but, instead, it was also expressed on monocytes, DCs and on nonhematopoietic cells, including keratinocytes, fibroblasts and neurons, as well as endothelial, epithelial and tumor cells [32-38]. CD40 receptor on DCs plays a crucial role in the maturation of these cells into fully competent antigen-presenting cells and the generation of long-term CD8+ effector T cell mediated immunity . The interaction of CD40 receptor on B cells with CD40L is critical for the induction of a humoral immune response by promoting the proliferation and differentiation of B cells into immunoglobulin-producing plasma cells . On nonhematopoietic cells, such as endothelial cells, CD40 is involved in the amplification and regulation of inflammatory responses. During a normal T cell response, CD40 on DCs is engaged by CD40L, which is transiently expressed on activated Th cells . CD40–CD40L interaction is important for activation of DCs both under normal, as well as under regulatory T cell (Treg) induced suppressive conditions . The importance of DC activation via CD40 ligation has been demonstrated in a number of studies, where DC targeted antibody-antigen fusion products induced tolerance to the antigen unless an agonistic anti-CD40 monoclonal antibody (mAb) was coadministered [43-45]. Hence, agonistic mAb against CD40 or recombinant trimerized CD40L is administered in combination with the antigen to induce memory and protective CD4+ and CD8+ T cell responses [31, 46, 47]. We and others have pursued the strategy to combine antigen delivery to DCs and their simultaneous activation through CD40 [41, 48, 49]. Although CD40 expression is not restricted to DCs, its DC activation properties nevertheless make it a very attractive targeting motif. Moreover, the route of administration may determine whether relative over-expression of CD40 on DCs provides a sufficient window to achieve their selective targeting. Our data obtained in human skin  and in human skin-draining LNs  certainly indicate this to be the case for the i.d. route of administration, clearly demonstrating that any binding that might occur to other cells (e.g. endothelial cells in the dermis or B cells in the LN) does not interfere with efficient vaccine delivery to DCs and subsequent CTL activation. Of note, the expression of other DC targeting motifs, pursued for vaccination purposes, is often also not restricted to DCs. A case in point is DC-SIGN, with confirmed expression on macrophages in human LNs, despite its often perceived and expounded specificity for DCs . Also, many of these alternative targets (including DC-SIGN) lack the added advantage of mediating DC activation. Therefore, we and others have designed and tested vaccine constructs that agonistically bind to CD40 for targeted delivery of antigens to DCs and simultaneous maturation induction.
Two immunoglobulin-based vaccine constructs were designed by Schjetne et al.  that agonistically targeted multiple myeloma and B cell lymphoma antigens to the CD40 receptor on murine APCs. The first construct consisted of a recombinant mAb with V regions specific for CD40 and its C region containing a defined T cell epitope. The other was a homodimer, each chain of which was composed of a single-chain variable region fragment (scFv) targeting CD40, a dimerization motif, and an antigenic unit. Both constructs bound CD40, induced the maturation of DCs, and enhanced primary and memory T cell responses. Intramuscular delivery of these vaccine constructs, in the form of naked DNA, induced T cell responses specific for the MHC class II-restricted epitopes, as well as antibody responses, and also protected mice from myeloma and lymphoma growth. The observed effects were attributed to both targeted delivery of antigen to APCs for presentation on MHC class II molecules and simultaneous activation of the targeted APC. Because CD8+ CTLs have the ability to specifically recognize and kill tumor cells, recent efforts in the development of anti-cancer vaccines have mainly focused on enhancing the induction of tumor-specific CD8+ T lymphocyte responses . We have achieved this through CD40-mediated delivery of Ad5 vectors to DCs in situ. Over the past decade, we have explored different CD40-targeted Ad5 (CD40-Ad) configurations for their efficacy in terms of (in situ) DC targeting and transduction, of subsequent T cell activation, and of tumor protection (Figure 2). Below, the different CD40-Ad configurations are described and pre-clinical evidence of their efficacy is discussed, highlighting their pros and cons with respect to clinical translation. Together, these observations illustrate the pre-clinical triumphs and failures that line the long and winding road to clinical translation.
Re-targeting Ad5 vectors to CD40 using bispecific mAb or scFv adaptors
We first prepared and tested a chemically conjugated bispecific antibody construct consisting of a Fab fragment binding the Ad5 fiber knob in a neutralizing fashion (1D6.14) and an agonistically binding anti-CD40 mAb (G28-5) . Retargeting of Ad5 to CD40 using this bispecific antibody construct dramatically enhanced gene transfer to monocyte derived DCs (MoDCs) and also induced both their phenotypic and functional maturation . In this first study, functional activation was demonstrated by increased T cell stimulation in an allogeneic mixed leukocyte reaction and by enhanced IL-12p70 release. When using a chemical Ab conjugate that incorporates the murine CD40 (mCD40) targeting mAb FGK45, we found that this particular targeting approach also significantly enhanced gene transfer and induced the phenotypic maturation of murine bone marrow-derived DCs (BMDCs) . Subcutaneous (s.c.) injection of BMDCs that were transduced with a CD40-targeted Ad5 vector encoding the viral E7 oncogene derived from human papillomavirus (HPV) type-16 resulted in superior protection against the outgrowth of HPV-16-transformed tumor cells in vivo . Of note, this protection was shown to depend on CD8+ T cells.
To further demonstrate the clinical utility of CD40-Ad vectors, we deemed it essential to assess their performance in the context of three-dimensional human tissue and to test their DC targeting ability under these physiological and clinically highly relevant circumstances. We therefore demonstrated the utility of this approach for in situ gene transfer to DCs through i.d. injection of CD40-Ad, employing the bispecific mAb conjugate in a human skin explant model (Figure 3A) . CD40 targeting of Ad5 vectors enhanced selective transduction of human CD1a+ DCs located in the dermis of cultured skin explants and also increased and stabilized their maturation status upon migration. Indeed, we observed that, after CD40-mediated i.d. delivery in skin explant tissue, the absolute number of enhanced green fluorescent protein (eGFP)+ transduced DCs present in migratory populations was significantly increased (Figure 3B). Importantly, these transduced and phenotypically mature (i.e. activated) DCs demonstrated a superior capacity to stimulate antigen-specific CD8+ memory T cells, employing the Haeminfluenza matrix protein M1 as a model antigen.
Despite the promising pre-clinical results obtained with this bispecific antibody-based Ad5 targeting approach, its clinical translation was impeded by the inherent variability of the chemical antibody conjugation process, resulting in low yields of functionally active bispecific molecules. To overcome this limitation, we constructed a well-defined molecular fusion protein between soluble CAR (sCAR) and a scFv of the G28-5 mAb . By binding and neutralizing the Ad5 fiber knob with the CAR portion of the fusion protein and agonistically binding CD40 through the G28-5 scFv, significantly enhanced transduction of DCs from both rhesus monkey and human origin was achieved . Similarly, a bispecific scFv fusion protein was constructed from the cDNAs of an anti-Ad fiber knob scFv (S11) and an anti-CD40 G28-5 scFV . Retargeting of Ad vector to CD40 using this bispecific scFv fusion protein significantly enhanced the transduction efficiency of MoDCs and also their ability to functionally activate Haeminfluenza M1 specific memory CD8+ T cells isolated from the peripheral blood of healthy donors.
Despite their proven functionality, the clinical development of these molecular fusion proteins was abandoned as the general consensus at the time was that rapid clinical translation required a ‘clean’ single-component vaccine, which would greatly simplify quality assurance/quality control criteria for vaccine production according to clinical-grade Good Manufacturing Practices (cGMP). Moreover, it was also assumed to facilitate safety assessment in a single Phase-I clinical trial. Single-component Ad-based DC-targeted vaccines required the large-scale production of genetically altered Ad vectors with ablated natural tropism, consisting of retargeted fibers. This proved to be a considerable technical challenge.
Re-targeting Ad5 vectors to CD40 by genetic modification
Early efforts to genetically retarget Ad vectors mostly focused on alteration of the tropism of the virus through incorporation of target receptor-specific ligands into either the carboxy terminus or HI loop of the fiber. These sites tolerated the insertion of only small linear ligand sequences without substantially affecting the infectivity and propagation of the resulting retargeted vector . Insertion of complex, folded and, consequently, more selective ligands appeared to disturb trimerization of the Ad fiber and prevent subsequent incorporation of the fiber into the adenovirus capsid . To overcome this limitation, a fiber replacement strategy for the generation of tropism-modified Ad5 was considered. In this strategy, the fiber knob domain, alone or together with the shaft domain, was replaced with a combination of two heterologous domains; one of which functioned as a receptor-binding ligand and the other of which supported the functional structure of the entire protein chimera. We confirmed the feasibility of this approach by replacing the Ad5 fiber protein with a protein chimera comprising an N-terminal portion of the Ad5 fiber, genetically fused with the phage T4 fibritin, which contained a trimerization motif and thus facilitated the naturally trimerized structure of the Ad5 fiber (Figure 4) . Targeting of the resultant rAd5 vector to an artificial receptor was achieved by incorporation of C-terminal peptide ligands into the protein chimera. Subsequently, we used the same strategy to create rAd5 vectors retargeted to human CD40 by incorporating the functional TNF-like domain of hCD40L into a chimeric Ad fiber consisting of the Ad5 fiber capsid-anchoring N-terminus and trimerized phage T4 fibritin (Figure 4C) . Despite the considerable size of the CD40L domain and its complex tertiary structure, both components of the targeting protein, the CD40L domain and the Fiber-Fibritin(FF) backbone, folded properly, thereby making the entire chimera fully functional. The incorporation of the chimeric fiber into the Ad virion did not affect the functional structure of its CD40-binding component, resulting in a vector (Ad5.FFCD40L) that used CD40 as a surrogate receptor for cell entry. However, when comparing the CD40-targeted Ad with untargeted Ad containing wild-type (wt) fibers, the CD40-targeted Ad showed significantly reduced transduction efficiency of target cells expressing both CAR and CD40 at high levels . Experiments with radiolabeled CD40-targeted Ad and untargeted Ad nevertheless revealed that both viruses were equally efficient in binding to target cells. This result led us to hypothesize that complete deletion of the fiber in CD40-targeted Ad affected its ability to infect target cells by interfering with a downstream process from primary binding to the cell surface. The construction of a mosaic version of CD40-targeted Ad which, in addition to phage T4 fibritin, and the human CD40L chimeric fiber, also contained a full size wt Ad5 fiber engineered to lack CAR-binding ability (by deletion of the CAR-binding TAYT sequence) enhanced its ability to infect target cells and facilitated the large-scale production of a stably CD40-retargeted rAd5 vector (i.e. Ad5.FFCD40L[F5dTAYT]) . Subsequent comparison of the CD40-targeted mosaic Ad with untargeted Ad5 showed the superior transduction efficacy of the mosaic vector on human MoDCs in vitro . To determine the potential of the Ad5.FFCD40L[F5dTAYT] vectors to effect in situ targeted antigen expression in DCs, we constructed a CD40-targeted mosaic Ad encoding both GFP and the carcino-embryonic antigen (CEA) . The application of this vector in the stringent human skin explant model demonstrated its utility for in vivo selective transduction of DCs and induction of their activation and migration .
A grant proposal was subsequently submitted to the NIH/Rapid Application of Innovative Drugs programme to support cGMP production of this vector (without the eGFP sequence) for safety and feasibility testing in a Phase-I clinical trial of patients with advanced colon cancer. Although the proposal received a favorable review and was deemed ‘scientifically fundable’, the NIH Biological Resources Branch Oversight Committee insisted on removal of the Ad5 dTAYT fibers from the mosaic construct because the ratio of incorporated F5dTAYT and FFCD40L fibers could not be controlled during the production process, thereby complicating standardization and quality control of the final product. A poor performance of the subsequently constructed Ad5.FFCD40L-CEA vector in the human skin explant model, coupled to cGMP upscaling issues involving initial wt fiber complementation in a two-tiered 293-based system of vector propagation and production, ultimately precluded the clinical development of this chimeric Ad vector.
In a final attempt to circumvent the need for wt fiber complementation during the production of a retargeted Ad vector, we explored a new strategy based on replacement of the Ad fiber molecule with a reovirus σ1-based chimeric attachment protein, consisting of the N-terminal domain of the Ad fiber fused to the T(ii) domain of σ1 for proper trimerization , and incorporated human CD40L as a targeting moiety. We constructed an expression plasmid encoding a fusion protein of the Ad tail and σ1 T(ii) domains, linked to human CD40L. Upon expression, this chimeric attachment protein efficiently trimerized and retained the capacity to specifically bind CD40; moreover, reconstitution of fully functional infectious viral particles was observed upon transfection of this CD40-retargeted fiber in 293 containing the doubly ablated Ad5 genome (i.e. with deletion of both the CAR- and integrin-binding sequences) (D. Oosterhoff, F.H. Schagen, B.N. Hangalapura, P.G.J.T.B. Wijnands, D.T. Curiel, V.W. van Beusechem, and T.D. de Gruijl, unpublished data). Nevertheless, this σ1-based approach would still require the cGMP development of a CD40-expressing cell line for clinical-grade propagation and large-scale production of the CD40-retargeted Ad vector. Because this was deemed cost-inhibitive to such an extent that it would have precluded clinical translation within the foreseeable future, this approach was abandoned. Indeed, a different attitude was gaining the upper hand in the field of translational cancer studies, favoring the accelerated clinical testing of combinations of therapies, and thus generating a renewed interest in the two-component approach to CD40-targeted Ad vaccination.
CD40L-sCAR adaptor proteins: the way to go?
Most Ad vectors bind to CAR via the knob domain of their fiber proteins. Three CAR molecules can bind per Ad5 fiber knob . Based on this understanding, we developed a novel CAR-containing adaptor molecule that assumes a trimeric conformation via trimerization motifs derived from a 71-amino acid fragment of the bacteriophage T4 fibritin protein . The trimeric structure of CAR resulted in efficient binding to and neutralization of the rAd5 fiber knob. To facilitate efficient transduction and activation of murine DCs, we designed and produced a recombinant fusion protein, CFmCD40L, that contained the ectodomain of CAR genetically linked via a trimerization motif to the extracellular domain of mouse CD40L . Incorporation of the trimerization motif served two purposes: (i) it increased the Ad fiber knob binding avidity  and (ii) it promoted and stabilized a native trimeric conformation of CD40L, which is necessary for efficient CD40 binding and function . The CFmCD40L adaptor was produced in mammalian cells to benefit from post-translational glycosylation. The produced adaptor demonstrated strong affinity for both Ad5 fiber knob and mouse (and human) CD40, as well as its use to retarget Ad5 enhanced transduction and activation of both murine and human DCs [51, 63]. This compatibility with both mCD40 and hCD40 is explained by the high homology of the incorporated murine TNF-like CD40L domain with its human counterpart . Subcutaneous injection of this CD40-Ad configuration, encoding the β-gal model antigen, induced superior antigen-specific Th and CTL responses in Balb/c mice  and i.d. injection of CD40-Ad encoding SARS-CoV NP resulted in a more Th1-skewed antigen-specific immune response in a Balb/c infectious mouse model . Similarly, stronger T cell and immunoglobulin G2 antibody responses (indicative of Th1 skewing) were recently observed against the human tumor antigen CEA upon i.d. delivery of a CD40-targeted Ad5 vaccine, incorporating the same CD40L domain, in a canine model .
To obtain definitive pre-clinical proof-of-concept for this CD40-Ad configuration in terms of breaking immune tolerance against TAAs and therapeutic anti-cancer efficacy, ex vivo human CTL priming and in vivo CD40-Ad vaccination studies were performed. Their size (approximately 80 nm) predicts that, upon i.d. delivery, Ad5 vaccines will not only target dermally located DCs, but also will rapidly drain through lymphatics to LNs, consequently targeting LN-resident DC subsets. It is therefore important to note that our recent studies have demonstrated the enhanced ex vivo transduction of conventional and plasmacytoid DC subsets by CD40-targeted Ad (by use of the CFmCD40L adaptor) in suspensions of human melanoma-draining sentinel LNs (SLN) (Figure 3C) . Moreover, CD40 targeting of Ad5 encoding the melanoma TAA MART-1, enhanced the expansion of functional MART-1-specific CD8+ T cells from these SLN (Figure 3D) with concomitant decreases in both CD4:CD8 T cell ratios and proliferation rates of immunosuppressive CD4+CD25hiFoxP3+ regulatory T cells (Tregs). Additional in vitro studies revealed that CD40-mediated transduction and activation of MoDCs with CD40-Ad5-MART-1 significantly enhanced their priming efficiency of fully functional CD8+ effector CTLs that responded to the HLA-A2-restricted MART-126–35 L epitope with high avidity. Importantly, these CD40-Ad primed CTLs were of significantly higher avidity than CTLs primed by MART-126–35 L peptide-loaded MoDCs and efficiently lysed HLA-A2+MART-1+ melanoma cells. In addition, we recently obtained evidence that i.d. delivery of a CD40-targeted Ad vector encoding the full-length melanoma antigen gp100, can efficiently induce antigen-specific CTLs in mice and significantly retard the subcutaneous growth of highly aggressive and poorly immunogenic B16F10 murine melanomas in a therapeutic setting .
Together, these data provide clear in vitro and in vivo evidence for the enhanced efficacy of Ad vaccines encoding full-length tumor antigens when they are targeted through the CFm40L adaptor protein to CD40 on DCs in skin and skin-draining LNs. We now aim to translate this approach to the clinic. The envisioned approach is to recruit activated Langerhans cells (LCs) and dermal DCs in the skin and selectively transduce these (and possibly DCs residing in the vaccination site-draining LNs)  in situ with CD40 targeted Ad vector encoding full-length TAAs. This would result in a clinically applicable and highly flexible two-component DC-targeted Ad vector configuration, allowing vaccination with different TAAs encoding Ad vectors simultaneously, depending on antigen expression profiles of the primary or metastatic tumors.
The choice of relevant pre-clinical models to evaluate in vivo efficacy of CD40-targeted Ad vectors
In vaccine development, pre-clinical studies bridge the gap between the design of candidate vaccines and their clinical development. Among the vast number of pre-clinical models available, choosing the right model to evaluate the efficacy and mode of action of a candidate vaccine is always a very crucial step in vaccine development and eventual successful clinical application. In the case of CD40-targeted Ad vaccine development, we utilized the following model systems: (i) in vitro human and murine DC culture models; (ii) a human full-thickness skin explant model; (iii) a human ex vivo melanoma-draining LN model; (iv) in vivo murine melanoma and infectious models; and (v) an in vivo canine immunization model. These models allowed us to investigate the key immune and non-immune events that affect the mounting of an effective anti-tumor immune response in vivo.
Although human and murine in vitro DC culture models provide an easily accessible system to obtain proof of principle in terms of DC transduction and activation, the more stringent human skin explant model provides clinically relevant information about the performance of human DC-targeted vaccines in an intact tissue context. Furthermore, increased appreciation for species-specific differences in DC subset activities, such as poor cross-priming activity by murine LCs in contrast to superior cross-priming and induction of CD8+ T cells by human LCs, supports studies utilizing the human skin explant model [68, 69]. Selective and efficient in situ LC and dermal DC transduction by CD40-targeted Ad was demonstrated in this organotypic model in the presence of all epidermal and dermal cellular and matrix components, as would be the case upon in vivo dermal delivery. Moreover, this model provided vital proof of increased DC maturation induction, functional DC migration and transgenic antigen-specific CD8+ T cell stimulation by migrated DCs upon i.d. injection of CD40-targeted Ad5 (Figures 3A and 3B) . All these events are vital steps towards efficient anti-tumor T cell priming and, ultimately, clinical efficacy. Similarly, ex vivo suspensions of melanoma SLN were utilized to demonstrate selective DC transduction in LN, as well as DC maturation and enhanced activation and expansion of TAA-specific T cells residing in the SLN (Figures 3C and 3D) . Together, these two human ex vivo models represent the two vital in vivo compartments where immunological priming is instigated upon i.d. delivery of a vaccine: the skin with its DCs and dense lymphatic network and the draining LNs where the migrated DCs encounter naive T cells and where specific T cell priming and expansion takes place. We feel that these human organotypic/ex vivo models are vital to predict how the targeted virus will behave in tissue microenvironments in a clinical setting. This is important because oft-used mouse models differ considerably from human skin in anatomy, cutaneous lymphatic drainage patterns and DC subset definitions . For example, in contrast to human skin [50, 71], we failed to demonstrate the recruitment of LCs and dermal DCs to the dermal vaccination site in murine skin by pre-treatment with either cytokines [e.g. granulocyte macrophage-colony-stimulating factor (GM-CSF)] or immunostimulatory Toll-like receptor ligands (e.g. CpG and lipopolysaccharide) (B.N. Hangalapura, P.G.J.T.B. Wijnands, J. de Groot, D. Oosterhoff, R.J. Scheper, T.D. de Gruijl, unpublished data). This recruitment in human skin provided a ‘reservoir’ of migratory LCs/DCs that could be efficiently targeted by intradermally delivered CD40-targeted Ad5  but, in the much thinner murine dermis, we could not reproduce this observation, neither ex vivo, nor in vivo. Nevertheless, i.d. administration of CD40-targeted Ad5 in the murine B16 melanoma model provided evidence for the transduction and involvement of skin DCs in the observed anti-tumor responses and revealed DC targeting in the vaccination site-draining LNs (Figure 5) . Interestingly, i.d. injection of untargeted Ad5 resulted in higher total transgene expression levels in the LNs (a clear sign of successful ablation of the natural Ad5 tropism upon CD40-retargeting), which was mostly attributable to transduction of macrophage-like cells in the marginal sinuses (Figures 5A and 5B). By contrast, CD40-targeted Ad5 preferentially transduced CD11c+ DCs in the paracortical areas (Figures 5C and 5D). Obviously, this anatomical information could not have been obtained from the transduction experiments on human ex vivo SLN-derived single-cell suspensions. Murine in vivo data also revealed efficient induction of TAA-specific CTLs and prophylactic and therapeutic anti-tumor efficacy of CD40-targeted Ad5 . Anatomically, canine skin and its lymphatics more closely resemble human skin and dogs may therefore provide a more relevant transitional animal model for pre-clinical validation of i.d. administered DC-targeted Ad vaccines. Similar to humans, canines rapidly recruit DCs and myeloid cells to the dermis, locally, at a site of i.d. injection of GM-CSF  (see also L. Timares Ph.D., ALB. Smith, Ph.D., E. Thacker, ALD. D.T. Curiel, Ph.D., unpublished data). We recently showed the utility of a canine model for immunological efficacy assessment of an i.d. injected CD40-targeted Ad vaccine encoding the TAA CEA . CD40 targeting of the Ad5-based vaccine resulted in enhanced proliferative and serological responses against CEA and, in addition, pointed to the induction of a type-1 skewed immune response. Because, with respect to histology, molecular make-up and biological behaviour, canine tumors closely resemble human tumors , and because dogs, in contrast to experimentally available mouse tumor models, are outbred, canine tumor models may provide a highly relevant setting in which to pre-clinically test the anti-tumor efficacy of DC-targeted vaccines such as CD40-targeted Ad5.
In conclusion, in obtaining pre-clinical proof of efficacy of i.d. delivered DC-targeted vaccines, the limitations of murine models should be recognized, specifically differences in anatomy and DC subset definition and functionality. Complementary human ex vivo models and more relevant transitional-translational animal models (e.g. canine) should therefore also be considered for pre-clinical validation (Figure 6).
The long and winding road that leads to the clinic's door
CD40-targeted Ad vaccines provide a novel approach to cancer immunotherapy. Convincing evidence obtained over the past decade in a number of ex vivo human and in vivo animal models (for a comprehensive overview, see Figure 6) shows the efficacy of this DC-targeted vaccination approach in terms of immune activation and tumor protection. These pre-clinical studies thus provide a strong rationale for its clinical application. Presently, we are aiming to produce a cGMP-grade ‘two-component’ configuration: Ad5 encoding the melanoma antigen MART-1, and the sCAR-hCD40L (CFhCD40L) adaptor molecule. This particular configuration for targeted modification of DCs in vivo carries several advantages: (i) It offers great flexibility over single-component approaches, ultimately enabling selection of Ad5 vectors carrying specific TAA genes based on TAA expression profiles in the tumor(s) of the patient to be vaccinated. Selected vector(s) can then be complexed with the adaptor molecule before intradermal injection. (ii) cGMP grade Ad5 vectors encoding various clinically relevant TAAs such as MART-1, gp100 and prostate-specific antigen have already been used in phase I/II clinical trials and were demonstrated to be safe, making them obvious candidates for combining with the CD40-targeting adaptor [12, 20, 74]. (iii) The availability of cGMP-grade Ad5 vectors and relatively easy and straightforward procedures associated with the production of the adaptor protein, CFhCD40L, should accelerate the clinical translation of this approach. To minimize the presence of either free adaptor protein or noncomplexed vectors in the vaccine preparations (potentially resulting in a poor vaccine efficacy by inadequate activation and/or transduction of DCs), the optimal adaptor-to-vector ratio should be determined by careful cross-titration before testing the transduction efficiency of monocyte derived-, dermal and SLN DCs. Importantly, in vivo biodistribution and anti-tumor studies have clearly demonstrated the stability of the adaptor-complexed Ad5 vectors, most notably by the impressive ablation of the natural Ad5 tropism (Figure 5) [65, 67].
An often voiced concern in relation to Ad5-based vaccines is pre-existent and induced Ad5 specific seroreactivity that might neutralize the virus and interfere with its efficacy . In this regard, it is encouraging to note that sCAR-FFCD40L-complexed Ad5 retained its ability to infect human DCs in vitro in the presence of Ad5-neutralizing antibodies . Moreover, it led to the induction of reduced levels of Ad5-specific antibodies in vivo, at the same time as ensuring long-lasting type-1 cell-mediated responses against the transgene product .
We are currently considering to vaccinate advanced-stage melanoma patients with a CD40-targeted Ad5 vector encoding MART-1 and to evaluate its clinical performance in a small-scale exploratory Phase-0 trial. Such a human micro-dosing study would adhere to FDA guidelines for Investigational New Drug studies, released in 2006 . Because MART-1 encoding Ad vectors have already been shown to be clinically safe [12, 20], we would propose to administer these Ad5 vectors i.d., in a CFhCD40L-complexed configuration in low sub-toxic doses (e.g. 108–109 viral particles) and check for: (i) targeted transduction of DCs in skin and vaccination-site draining LNs; (ii) maturation of transduced DCs from skin and vaccination site-draining LNs; and (iii) induction and expansion of Mart-1 specific CTLs in the vaccination site-draining LNs and in peripheral blood. The necessary monitoring techniques have been developed in the context of previous clinical studies on the immunomodulation of melanoma SLNs [71, 77, 78]. Such a study would provide vital confirmation that the CD40-targeted Ad5 vaccine behaves in the same way in patient as observed in our pre-clinical models. After thus providing a clinical proof-of-concept, we could move forward and establish its clinical efficacy in an extended Phase-I/II trial.
Finally, after a decade of pre-clinical development, clinical translation of CD40-targeted Ad5 vaccines appears to be within reach. Although we assumed we had reached this point before, the many twists and turns in the road, past different CD40-targeted Ad5 configurations and experimental models (Figure 6), have helped to refine and strengthen the conceptual basis for the application of our current two-component configuration. As always in translational science, it has led us back to the clinic's door.
The authors would like to acknowledge Bryan Tillman, Alexander Pereboev, Hidde Haisma, Winald Gerritsen, Nikolay Korokhov, Joana Brandão, Erin Thacker, Victor van Beusechem, Jan de Groot, Bruce Smith, Saskia Santegoets, Fons van den Eertwegh, Natalya Belousova, Pepijn Wijnands, Joel Glasgow, Victor Krasnykh, Sinéad Lougheed, Joanne Douglas, Rieneke van de Ven, Sylvia Luykx-de Bakker, Tyler Curiel, Bob Pinedo and all the other colleagues who, over the years, have contributed in one way or another to the work described in this review. This work was supported by grants from The Netherlands Organization for Scientific Research (NWO) 901-10-116 and 917-56-321, from the Dutch Cancer Society (KWF-grant VU2005-3284), by The University of Alabama Skin Diseases Research Center (NIH P30 AR-05-0948-05), and by NIH/NIAID grants 7R33 AI076096-05 & 04S1. The authors declare that there are no conflicts of interest