|Home | About | Journals | Submit | Contact Us | Français|
In vivo targeting of dendritic cells (DC) represents an attractive alternative to currently applied ex vivo DC-based genetic tumor vaccination protocols. Finding the optimal vector for in vivo targeting of DC is important for such strategies. We therefore tested a panel of subgroup C/B chimeric and fiber-modified adenoviruses (Ad) for their relative capacity to transduce human DC. We made use of in vitro generated Langerhans Cells (LC) as well as of ex vivo human skin and melanoma-draining lymph node (LN) derived DC. Of the tested viruses the C/B-chimeric Ad5/3 virus most efficiently transduced in vitro generated LC. In addition, Ad5/3 preferentially targeted mature myeloid DC from human skin and draining LN and transduced them at significantly higher frequencies than Ad5. In addition, Ad5/3 was more specific for mature human skin-derived CD1a+ CD83+ DC than the previously reported DC-transducing C/B-chimeric vector Ad5/35, infecting less bystander cells. It was previously reported that Ad5/3 transduced human monocyte-derived DC by binding to the B7 molecules CD80 and CD86. High-efficiency transduction of mature skin-derived DC was similarly shown to be mediated through binding to CD80/CD86 and not to interfere with subsequent T cell priming. We conclude that Ad5/3, in combination with DC-activating adjuvants, represents a promising therapeutic tool for the in vivo transduction of mature DC, and may be less likely to induce unwanted side effects such as immune tolerance through the infection of non-professional antigen-presenting cells.
Dendritic cells (DC) are a promising target for gene therapeutic strategies to induce antitumor responses, as DC initiate immune responses by presenting tumor associated antigens (TAA) to naive T cells. Therapeutic strategies that target DC should aim for high transduction efficiency and specificity, as optimal TAA delivery to DC is critical for the therapeutic efficacy of DC-based vaccination strategies. Adenovirus (Ad)-based therapies are ideal for in vivo targeting and infection of DC, as adenoviruses can infect non-replicating cells and can be genetically modified to express a TAA of choice. It is well known that the commonly used subgroup C adenovirus serotype 5 (Ad5) is of limited utility for transduction of human DC, both in vitro and in vivo, as DC lack expression of the adenovirus binding receptor CAR (Coxsackie-and Adenovirus Receptor). Other Ad types without pre-existent immunity and with a natural tropism for DC-associated molecules may be more readily translated to the clinic as vaccination vehicles, without a further need for genetic modification of their fiber knob to enable DC targeting. 1,2,3 We previously showed more effective skin DC targeting using the subgroup B2 virus Ad35 or a chimeric Ad5/35 virus containing the Ad5 fiber and Ad35 knob.4 This virus enters cells via CD46 5, 6, which is a complement regulatory molecule, present on many cells including DC.7 Although Ad5/35 was shown to be superior over Ad5 in terms of DC transduction, a more specific and truly DC-targeted virus might be preferable, as CD46 is expressed ubiquitously in the skin. Infection, and subsequent antigen presentation on MHC class I molecules by other cells than professional antigen-presenting cells (APC), might result in sub-efficient T cell activation, or even T cell anergy or tolerance induction, as a strong co-stimulatory signal via CD80 and CD86 is required to fully activate T cells. We therefore aimed to find a more DC-specific adenovirus with high-efficiency infection characteristics for future in vivo DC-targeting strategies. Short et al. previously reported on the binding of the subgroup B1 adenovirus Ad3 to the co-stimulatory molecules CD80 (B7.1) and CD86 (B7.2) on human monocyte-derived DC (MoDC).8 Since expression of these markers is abundant on activated DC, we set out to investigate whether a virus containing the Ad3 knob domain would effectively transduce human skin DC in situ.
The skin presents a readily accessible and highly efficient conduit for the delivery of Ad-based vaccines and subsequent activation of specific T cells.4,9 Specialized DC subsets line both the epidermis (i.e. Langerhans Cells [LC]) and the dermis, which, upon transduction and appropriate activation, can induce TAA-specific T cell responses in the draining lymph nodes (LN).10 A panel of genetically modified adenoviruses was therefore analyzed for their relative capacity to infect in vitro cultured LC and interstitial (dermal) DC. Amongst these were three different subgroup C/B chimeras, i.e. Ad5/3, Ad5/11 and Ad5/35 as well as three fiber modified viruses i.e. Ad5.RGD 11, Ad5.pK7 and Ad5.RGD.pK7. 12 The RGD motif binds integrins and the pK7 polylysine motif can bind heparin sulfate containing receptors. 12 Replication deficient Ad5 virus was employed for comparison. The most efficient in vitro transduction was observed with Ad5/3 and Ad5.RGD, in addition to the previously reported Ad5/35. Ad5/3, and not Ad5.RGD, was found to enhance transduction of DC in the context of intact skin in human explant cultures. Conflicting data were previously reported on the tropism of Ad3 (a B1-type virus): whether it binds cells primarily via the costimulatory molecules CD80 and CD86 8 or, like Ad35 (a B2-type virus) and other type B viruses, via CD46.5, 13,14, 15,16 Here we show that Ad5/3 is a high-efficiency DC-transducing adenovirus that binds DC through the B7-family members both in MoDC and human skin-migrated DC without interfering with their function. Moreover, the ability of Ad5/3 to selectively target mature human myeloid DC in the context of skin and in melanoma skin-draining LN, confirms its possible utility as an in vivo DC-targeted tumor vaccine vector.
PBMC from healthy donors were isolated from buffycoats obtained from Sanquin blood supply foundation (Amsterdam, The Netherlands) by density gradient centrifugation over Hipaque lymphoprep (Nycomed AS, Oslo, Norway). Monocytes were obtained by plastic adherence and were cultured into MoDC in Iscove’s modified Dulbecco’s medium (IMDM) (Lonza, Verviers, Belgium), supplemented with 10% fetal calf serum (FCS) (HyClone), 100 IU/ml sodium-penicillin, 100μg/ml streptomycin, 2 mM L-glutamine, 50 μM β-mercaptoethanol (2ME) (complete IMDM) and 100ng/ml recombinant human granulocyte colony stimulating factor (rhGM-CSF) (Sagramostim, Berlex) and 10ng/ml interleukin-4 (IL-4))(R&D systems) for 5-6 days.
The acute myeloid leukaemia cell line MUTZ3 was cultured in Minimum essential medium-α (MEM-α) (Lonza, Verviers, Belgium) containing 20% FCS, 100 IU/ml sodiumpenicillin, 100μg/ml streptomycin, 2 mM L-glutamine and 50 μM 2ME and 10% conditioned medium of the 5637 renal cell carcinoma cell line (MUTZ3 routine medium) in 12-well plates (Co-star) at a concentration of 0.2 million cells/ml and were passaged twice weekly as described.17 MUTZ3-LC were cultured by seeding MUTZ3 precursor cells in 12-well plates at 0.1 million cells/ml in MEM-α supplemented with 20% FCS, 100 IU/ml sodium-penicillin, 100μg/ml streptomycin, 2 mM L-glutamine and 50 μM 2ME (complete MEM-α). Differentiation medium contained 100ng/ml rhGM-CSF, 10ng/ml TGFβ1 (Biovision, Mountain View, CA) and 120 IU/ml TNFα (Miltenyi Biotec) and cells were differentiated into immature LC for 9-10 days.18 Immature MoDC and MUTZ3-LC were matured by adding a maturation cocktail containing 2400 IU/ml TNFα, 100ng/ml IL-6 (R&D systems), 25ng/ml IL-1β (Miltenyi Biotec) and 1μg/ml prostaglandin E2 (PGE2) (Sigma Aldrich) for 2 days.
HEK-293 cells (Microbix, Toronto, Canada) and QBI-293A cells (Qbiogene, Heidelberg, Germany) were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Lonza, Verviers, Belgium), supplemented with 10% FCS and 100 IU/ml sodium-penicillin, 100μg/ml streptomycin, 2 mM L-glutamine.
Human skin tissue was obtained after informed consent from patients undergoing corrective breast or abdominal plastic surgery at the VU University medical center (Amsterdam, The Netherlands), the Tergooi hospital (Hilversum, The Netherlands) or the Slotervaart hospital (Amsterdam, The Netherlands), following hospital guidelines. Human melanoma lymph node material was obtained with informed consent from clinically Stage I/II patients undergoing the sentinel lymph node procedure at the VU university medical center, without interference with pathological diagnostics and following hospital guide lines.
All luciferase encoding Ad5.RGD, Ad5.pK7, Ad5.RGD.pK7, Ad5, Ad5/3, Ad5/11 and Ad5/35 vectors were E1- and E3-deleted 12 [Murakami et al., manuscript submitted for publication]. Ad5/35.eGFP was a kind gift from Prof. A. Lieber (University of Washington, Seatlle, USA) and was described previously.19 Ad5.eGFP was cloned by homologues recombination of the linearized pAdtrack.CMV shuttle vector with pAdEasy transfected BJ5183 cells. Correct homologous recombination was analyzed by PacI enzyme digestion. Virus was propagated in either HEK-293 or QBI-293A cells. For Ad5/3.eGFP production, homologous recombination of linearized pAdTrack.CMV was performed in BJ5183 cells co-transfected with pAdEasy 5/3 (a kind gift of Prof. M Yamamoto, DHGT, Birmingham, AL, USA). This homologous recombination gave rise to an eGFP-expressing, Ad5-based virus with the knob-domain of Ad3. Generated vectors were analyzed for transgene expression, amplified in T175-cm2 flasks, purified by double CsCl gradient ultracentrifugation, and dialyzed against PBS containing 5% sucrose. Viral titers were determined by spectrophotometry (Viral particles (Vp)/ml)20 and functional titers in infectious units (iu) were determined by limiting dilution titration following standard techniques on HEK-293 cells.
Full thickness skin layers (2-3mm), consisting of dermis and epidermis, were collected from healthy donor skin with a dermatome (Aesculap, Tuttlingen, Germany). Skin was incubated in 50ng/ml dispase II (Roche, Penzbeg, Germany) for 20 minutes at 37°C, 5% CO2 after which the epidermal and dermal layers were separated with tweezers. Epidermal sheets were incubated in 0.05% trypsine at 37°C for 15 min and resuspended in IMDM + 10% FCS. Dermal skin sheets were incubated in collagenase (200mg collagenase diluted in 26.6ml PBS and 6.6ml dispase II) for 90 minutes at 37°C. Single cell suspensions were made by filtering twice through 100μm sterile filters. Viable cells were counted by trypan blue exclusion. Skin single cell suspensions, taken up in IMDM with 5% human pooled serum (HPS), 100 IU/ml sodium-penicillin, 100μg/ml streptomycin and 2 mM L-glutamine, were activated with 100ng/ml GM-CSF and 10ng/ml IL-4 in T75 culture flasks for 24 hours prior to Ad infection.
Human sentinel LN single cell suspensions were prepared as described previously.21 In short, LN were cut in half. From one half, 10 scrapes were made with a scalpel, rinsing the blade after each scrape in a tube filled with 15ml of 0.1% DNase and 0.14% collagenase A (Roche Nederland B.V., Woerden, The Netherlands). The cell suspension was then transferred to a sterile flask in a total volume of 30ml DNase/collagenase, after which the suspension was incubated in a 37°C water bath on a magnetic stirrer for 45 minutes. The cell suspension was filtered through a 100μm cell sterile cell strainer and the flask and strainer were rinsed with 20ml IMDM with 10% FCS. Cells were centrifuged at 1500rpm, 4°C for 5 minutes, washed with 10ml IMDM with 10% FCS and taken up in 3ml IMDM with 10% FCS for viable cell count by trypan blue exclusion.
DC were phenotyped using FITC- / PE- or APC-conjugated monoclonal antibodies (Mabs) directed against: CD1a (1:25), DC-SIGN (1:10), CD86 (1:25), CD80 (1:25), CD40 (1:10) (PharMingen, San Diego, CA), CD14 (1:25), HLA-DR (1:25) (BD Biosciences, San Jose, CA), Langerin (CD207), CD83 (1:10) (Immunotech, Marseille, France) and CD34 (1:10) (Sanquin, Amsterdam, The Netherlands); 2.5 to 5·104 cells were washed in FACS buffer (PBS supplemented with 0.1% BSA and 0.02% NaN3) and incubated with specific or corresponding control Mabs for 30 minutes at 4°C. The LN single cell suspension were stained with CD123-PerCP_Cy5 (1:10), CD11c-APC (1:10), CD14-PerCP_Cy5 (1:10), CD1a-PE (1:25), CD19-PE (1:25), CD3-APC (1:10) (BD Biosciences, San Jose, CA) and BDCA2-PE (1:25) (Miltenyi, Bergisch Gladbach, Germany) antibodies. Cells (1-2.105) were analyzed on a FACS-Calibur flow cytometer (Becton and Dickinson, San Jose, CA) equipped with CellQuest analysis software.
In vitro generated MoDC, MUTZ3-LC, or GM-CSF + IL-4 activated skin single cell suspensions were in vitro transduced with the indicated adenoviruses at a MOI of 1000 viral particles (Vp) or 10-100 infectious units for the titration studies, as indicated. 0.05-0.1 million cells were generally used per transduction and transductions were performed in duplicate or triplicate. 250μl of cell suspension in serum free DMEM/ F12 (Cambrex, Verviers, Belguim), supplemented with 100 IU/ml sodium-penicillin, 100μg/ml streptomycin and 2 mM L-glutamine was seeded in either 24-well plates (Co-star) or 4PP-tubes (Greiner). 250μl of Ad-suspension in serum free DMEM/F12 was added to the cells and plates or tubes were incubated at 37°C and 5% CO2 for 3 hours. 1ml of complete culture medium, depending on the cell type, was added after these 3 hours. For luciferase transgene read-outs, cells were harvested 24 hours after transduction and pellets of 0.1 million cells were lysed in 100μl 1× reporter lysis buffer (Promega, Madison, WI) and frozen at -80°C. Lysates were thawed and 10μl lysate was added to 50μl luciferase substrate (Promega, Madison, WI) and luciferase activity was measured on a luminometer. LN single cell suspensions were transduced with a MOI of 1000 (Vp) (based on approx. 5% DC (including myeloid and plasmacytoid subsets) within the total cell population). 1-2 million cells were transduced, as described above. For eGFP read-out, cells were harvested 48 hours after transduction and were analyzed for eGFP expression in conjunction with PE-, PerCP_Cy5 or APC-labeled markers on a FACS-Calibur flowcytometer as described above.
For the CD80 and CD86 blocking experiments, mature MoDC or migrated skin DC were incubated with 10 μg/ml control-Ig or CTLA4-Ig (kind gifts from Dr. L. Chen, Ann Arbor, MI) for 1 hour on ice. Cells were washed to remove excess fusion-proteins and were transduced with Ad5/35.eGFP or Ad5/3.eGFP (MOI 1000 (Vp)) as described above.
6mm human skin biopsies, pre-activated with GM-CSF and IL-4, were injected with 1*109 Vp of Ad5.eGFP, Ad5/35.eGFP or Ad5/3.eGFP in a total volume of 10μl as described previously.22 Cells were allowed to migrate from the biopsies for 48 hours, after which they were harvested, quantified by trypan blue exclusion and flowcount fluorospheres, and analyzed for eGFP expression and DC marker expression by flow cytometry.
1·102 – 3·104 migrated skin DC, in situ or in vitro transduced with Ad5.eGFP, Ad5/35.eGFP or Ad5/3.eGFP, were co-cultured with 1·105 peripheral blood lymphocytes (PBL) for 4 days in 96-wells plates in IMDM containing 10% HPS, 100 IU/ml sodium-penicillin, 100μg/ml streptomycin and 2 mM L-glutamine and 50 μM 2ME. To determine the effect of CD80/CD86 blockade on T cell proliferation in an allogeneic MLR, mature MoDC were incubated with medium, control-Ig –or CTLA4-Ig fusion protein for 30 minutes prior to T cell co-culture. At day 4, 2.5μCi/ml [3H]-thymidine (6.7Ci/mmol, MP Biomedicals, Irvine, CA) was added per well for 16 hours. Plates were harvested onto glass fiber filtermats (Packard Insturments, Groningen, The Netherlands) using a Skatron cell harvester (Skatron Instruments, Norway), and [3H]-thymidine incorporation was quantified using a Topcount NXT Microbetacounter (Packard, Meriden, CT).
Immature MoDC were matured with a maturation cocktail containing 2400 IU/ml TNFα, 100ng/ml IL-6 and 25ng/ml IL-1β for 8 hours before Ad transduction (MOI 100 based on IU). For IL-10 and IL-12 secretion, 40,000 control or Ad-transduced DC were co-cultured with 40,000 irradiated CD40L-expressing J558 cells. Supernatants were harvested 24h after co-culture and were analyzed for IL-12 and IL-10 [IL-10 ELISA kit, Sanquin, The Netherlands] secretion by ELISA following manufacturer’s guidelines or as described previously.18
The paired two-tailed student’s T-test was used for statistical analysis, except for analysis of DC-targeting indices, where a non-parametric analysis was used instead (two-tailed Wilcoxin) due to a lack of a normal distribution. Differences were considered statistically significant when p<0.05.
Seven replication-deficient adenoviruses (Ad) with embedded distinct tropism profiles were analyzed for their capacity to infect LC, as the major skin-resident DC subset. To this end we employed immature and mature in vitro cultured LC derived from the CD34+ acute myeloid leukemia-derived cell line MUTZ3. 17, 23 The viruses tested in comparison to the subgroup C virus Ad5 (Ad5) included three subgroup C/B chimeras, i.e. Ad5/3, Ad5/11, Ad5/35 and three fiber-modified viruses, i.e. Ad5.RGD, Ad5.pK7 and Ad5.RGD.pK7. 12 All viruses contained a luciferase reporter gene under a cytomegalovirus (CMV) promoter. The Ad5.RGD, Ad5.pK7 and Ad5.RGD.pK7 viruses in addition contained a GFP reporter gene driven by a CMV promoter. To compare the data with the C/B chimeras, luciferase expression was used as reporter assay. An isogenic Ad5 with double-reporter genes was included as a control for the RGD and pK7 viruses (data not shown).
The type C/B chimeras gave higher transgene expression levels than Ad5 in LC, with the highest levels in mature LC (mLC) (Fig. 1A). Ad5/3 infection was specifically more effective in mLC than in iLC (p < 0.002 for iLC vs mLC and p < 0.003 compared to mLC transduction levels by Ad5/11 or Ad5/35). Viruses containing the RGD motif only showed enhanced transduction efficiency compared to Ad5 in immature LC (iLC) (p < 0.004). No advantage was observed with the pK7 insert alone or in combination with RGD, compared to Ad5 infectivity of iLC or mLC. Similar results were obtained with in vitro cultured interstitial DC (IDC) generated from human monocytes (data not shown). Based on the fact that they induced significantly higher transgene levels compared to Ad5, Ad5.RGD and Ad5/3 were selected for further studies.
To study the percentage of infected DC in conjunction with DC surface marker expression, enhanced GFP (eGFP)-encoding Ad5 and Ad5/3 viruses were constructed for further studies. Ad5.RGD and Ad5/3 were analyzed for their capacity to target and transduce human skin DC in situ compared to Ad5 control viruses. Intra-dermal (i.d.) injection of Ad5.RGD did not enhance the percentage of GFP-positive migrated human skin DC compared to injection of Ad5. In contrast, i.d. injection of Ad5/3 resulted in a marked enhancement of the transduction efficiency of migrated human skin DC compared to Ad5. Figure 1B shows the transduction efficiency ratios for in situ skin DC transduction relative to Ad5 for both Ad5.RGD and Ad5/3 (n=3). In light of its superior performance Ad5/3 was further investigated. A titration with eGFP-encoding Ad5 and Ad5/3 (MOI 10-1000) on mature human MoDC showed that also at lower viral concentrations, Ad5/3 was superior over wild type Ad5 (Figure 1C) in terms of DC transduction.
Pre-conditioning of human skin with GM-CSF and IL-4 leads to strong activation of skin resident DC. 24 Since the in vitro infection data showed enhanced luciferase expression levels in mature LC compared to immature LC after transduction with Ad5/3 (Fig. 1A), human abdominal skin was pre-conditioned with the cytokines GM-CSF and IL-4 prior to Ad injection. For the in vitro Ad transduction of migrated human skin DC, in situ GM-CSF and IL-4 activated and subsequently migrated cells over a period of two days, were harvested and transduced in vitro with Ad5 or Ad5/3. The transduction efficiencies, represented by the percentage of eGFP-positive cells are shown in figure 2A. Ad5/3 transduced significantly more of the total (p < 0.04), CD1a+ (p< 0.03) and CD83+-i.e. mature- (p < 0.04) skin-emigrated DC as compared to Ad5. As a consequence of the maturation achieved through GM-CSF and IL-4 injection, no further activation of the DC was detectable upon transduction with either of the adenoviruses (data not shown). Enhanced in situ transduction capacity by Ad5/3 over Ad5 was observed, as shown in figure 1B. To more closely look at the subsets of DC transduced by Ad5/3 during in situ targeting of human skin explants, we gated either on the total DC population based on scatter profiles, the cells expressing the DC marker CD1a, or the maturation marker CD83 (Fig. 2B). Additional gating was performed on CD80 and CD86 expressing cells (data not shown). Significantly more eGFP+ cells were present after Ad5/3 injection as compared to Ad5 injection within all groups (p < 0.05). Figure 2C shows the transgene expression levels in the in situ-transduced migrated skin DC. Although not significantly so (most likely due to inter-experimental variation), transduction of human skin DC with Ad5/3 led to considerably increased transgene expression levels as compared to Ad5 (n=3).
As Ad5/3 was reported to bind the B7 family members CD80 and CD86 on in vitro generated human MoDC 8,25, but has also been reported to bind CD4614,16, we decided to examine the targeted surface molecules by Ad5/3 on human skin-DC. MoDC and skin-emigrated DC were incubated with a control-Ig or CTLA4-Ig fusion protein, the latter to prevent CD80 and CD86 binding by the virus. As a control, Ad5/35 was taken along, as this subgroup B virus is believed to transduce MoDC through binding of CD46 26 -although binding to CD80 and CD86 has also been reported.25 Figure 3A shows control-Ig and CTLA4-Ig binding to mature MoDC. All CD80 molecules were blocked after CTLA4-Ig incubation of MoDC, as a PE-labeled anti-CD80 antibody was unable to bind the cells (Fig. 3B). In contrast, binding of a PE-labeled anti-CD86 antibody was reduced, but not completely blocked after CTLA4-Ig treatment (Fig. 3B). In the MoDC cultures, blocking of CD80 and CD86 binding sites did not affect Ad5/35 transduction of the cells (91% vs. 93% of the cells were transduced after control-Ig or CTLA4-Ig pre-incubation, respectively) (Fig. 3C). Among the skin-emigrated DC, some Ad5/35 transduction was blocked (20% reduction) by pre-incubating the cells with CTLA4-Ig (transduction efficiency was decreased from 84% to 68% after control-Ig or CTLA4-Ig incubation, respectively). Ad5/3 infection was reduced from 61% to 18% in control-Ig vs. CTLA4-Ig treated MoDC (70% reduction) (Fig. 3C) and from 83% to 30% in control-Ig vs. CTLA4-Ig treated skin DC (64% reduction) (Fig. 3C). These data clearly show that Ad5/3 primarily uses CD80 and CD86 to bind human DC, whereas Ad5/35 does not.
Since Ad5/3, unlike Ad5/35, was shown to bind to the DC-specific markers CD80 and CD86 in order to transduce human skin DC (Fig. 3C), we analyzed whether this influenced the selectivity of these viruses for mature DC within human skin single cell suspension. Single cell suspensions were made from dermal or epidermal sheets to investigate the transduction specificity of Ad5, Ad5/3 and Ad5/35 for the DC populations within human skin. Figure 4A displays the percentages and expression levels of Langerin, CD80, CD86 and CD83 on CD1a+ DC within epidermal- and dermal single cell suspensions, twenty-four hours after GM-CSF and IL-4 treatment. These GM-CSF + IL-4 activated, total cell suspensions were in vitro transduced with the different Ads.
Transduction efficiencies were low for Ad5, whereas both Ad5/3 and Ad5/35 transduced CD1a+ DC within the epidermal or dermal cell populations at significantly higher efficiencies than Ad5 (Fig. 4B, p<0.05 for both Ad5/3 and Ad5/35 vs. Ad5), in keeping with our previous observations for Ad35.4 While Ad5/35 in turn was more efficient in transducing LC and dermal DC than Ad5/3, it was also less specific for mature skin DC than Ad5/3, as significantly more CD1a-CD83- bystander cells were infected by Ad5/35 (p < 0.001 or p < 0.01 for CD1a+ and CD1a- dermal and epidermal cells, respectively) (Fig. 4B). As there are far more CD1a- or CD83- cells within human skin compared to CD1a+ or CD83+ DC, the high transduction efficiency of Ad5/35 for these bystander cells makes this virus less selective for DC. In figure 4C the DC-targeting index (% transduced cells of the CD1a+ and/or CD83+ population divided by % transduced cells of the CD1a-CD83- population) is shown for the three viruses within the epidermal and dermal single-cell suspensions. CD1a+ DC-targeting with Ad5/3 was more efficient in the dermal than in the epidermal cell suspensions (DC-targeting index: p < 0.04 compared to Ad5 or Ad5/35).
As the size of adenoviruses (approx. 80 nm) may allow for direct diffusion to draining LN upon i.d. injection27, DC selectivity and transduction efficiency were also determined for Ad5, Ad5/3 and Ad5/35 in ex vivo LN suspensions. We made use of first-line skin tumor-draining LN (so-called Sentinel Lymph Nodes [SLN]), from clinically Stage I/II melanoma patients. This is a particularly relevant model in view of our intent to use these Ad vectors as tumor vaccine vehicles. We previously showed CD11c+ resident and CD1a+ skin-derived myeloid DC subsets within these melanoma SLN to have a mature CD83+ phenotype28,29 and we therefore hypothesized them to be susceptible for Ad5/3 and Ad5/35 infection. Five different leukocyte subsets were analyzed, i.e. CD11chiCD14- LN-resident myeloid DC 29, skin-derived CD1a+ myeloid DC, CD123+BDCA2+ plasmacytoid DC (pDC), CD19+ B cells, and CD3+ T cells. Figure 5a shows typical dot plots (left) of the three different DC subsets within the LN and the gates used for flow cytometric marker analysis. The histogram plots show CD83, CD86 and CD80 expression levels within the CD11chiCD14- and CD1a+myeloid DC populations and the CD123+BDCA2+ pDC population. The median fluorescence indices are listed in the upper right corners. The percentage of eGFP+ cells within the CD11chiCD14-, CD1a+, CD123+BDCA2+, CD19+ and CD3+ subsets are shown in figure 5b (n=3 for the myeloid DC subsets and n=2 for pDC, T- and B-cells). Ad5/3 was significantly better than Ad5 in transducing the myeloid DC subsets (p < 0.03 and p < 0.01 for CD11chiCD14- and CD1a+ DC, respectively). Ad5/35 transduced significantly more CD11chiCD14- cells as compared to Ad5 (p < 0.01) or Ad5/3 (p < 0.02) and more CD1a+ DC than Ad5 (p < 0.002). Ad5/35 transduced slightly more pDC compared to Ad5 or Ad5/3. In the two donors tested, hardly any B- or T cells were transduced (Fig. 4B). We showed that the Ad5/3 required binding to the co-stimulatory molecules CD80 and CD86 on human skin DC in order to transduce the cells (Fig. 3C).8 In line with this and in conjunction with higher expression levels of CD86 and CD80 on CD1a+ DC as compared to CD11chiCD14- DC present within the LN (Fig. 5A), there was a significantly better transduction of CD1a+ DC by Ad5/3 (p<0.02) and virtually no transduction of the CD80-CD86- immature pDC (Fig. 5A and 5B). Of note, both Ad5/3 and Ad5/35 transduction of CD1a+ DC within the SLN resulted in significantly enhanced transgene expression levels as compared to Ad5-mediated transduction (Fig. 5C).
It was previously shown that CD46 surface expression was down-regulated upon Ad35 binding and did not recover for up to 96 hours after removal of the virus.30 If a similar downregulation of CD80/CD86 occurs or if CD80/CD86 are effectively blocked upon Ad5/3 binding, this may confound T cell activation by Ad5/3-targeted DC and disqualify this virus for vaccination purposes. Therefore, an allogeneic mixed leukocyte reaction (MLR) was performed with migrated, in situ transduced skin DC to show that binding of Ad5/3 did not interfere with the capacity of the DC to induce T cell proliferation. Figure 6A shows that i.d. transduction with Ad5 or Ad5/3 did not reduce the expression levels of either CD80 or CD86 on subsequently human skin-emigrated DC (48h after Ad injection). In addition, there was no difference in DC-induced T cell proliferation using these DC in an allogeneic MLR (Fig. 6B: left) or indeed, using skin-emigrated DC which were transduced in vitro with Ad and added to MLR cultures 4h later (data not shown). To prove that this model of T cell allostimulation is sensitive for CD80/CD86 blockade, a control experiment was performed with mature MoDC pre-incubated with control-Ig or CTLA4-Ig fusion proteins prior to the onset of the MLR. As shown in Figure 6B (right), T cell proliferation was indeed hampered when the costimulatory molecules were blocked with the CTLA4-Ig fusion protein.
Finally, we also investigated the effect of the different Ad types on cytokine release by DC. Supernatants were harvested from skin-emigrated DC two days after i.d. Ad transduction or from MoDC, 24 hours after transduction. No significant alterations in IL12p70, TNFα, IL-1β, IL-6, IL-8 or IL-10 levels were observed in a flow cytometric cytokine bead array, although there was a reduction in IL-10 levels after Ad5/3 transduction in both skin DC and MoDC samples (data not shown), similar to our previous findings for Ad5/35.4 Additionally, no significant differences in IL-10 or IL-12 secretion were observed upon Ad5/3 transduction combined with CD40 ligation of the DC (Fig. 6C). Thus, transduction of DC by Ad5/3 through binding of CD80 or CD86 does not hamper DC-induced T cell activation or DC cytokine release.
The aim of this study was to examine the applicability of Ad5/3 as a human cutaneous DC-targeting vector in situ for immunotherapeutic vaccination purposes. Similar as in human MoDC 8, Ad5/3 specifically bound human emigrated skin DC through interaction with the B7-molecules CD80 and CD86 and was capable of efficient transduction of mature human skin IDC and LC both in vitro and in situ and of human LN DC ex vivo. Whereas contradictory reports can be found in literature with respect to the attachment receptors of subgroup B viruses,6,14-16,25,26 our data show that Ad5/3, and not Ad5/35, primarily uses CD80 and CD86 molecules to bind human DC. This preference of Ad5/3 for CD80 and CD86 can explain why it is more specific for mature DC. We can however not conclude that interaction of Ad5/3 with CD46 (or another surface molecule) is not involved in Ad5/3-based DC transduction, as the blockade of CD86 expression by the CTLA4-Ig was not complete. As different CD46 isotypes have been characterized,31 it could be that Ad3, which is a subgroup B1 virus, and Ad35, which belongs to subgroup B2, bind different CD46 isotypes and that the isotype present on human DC is preferentially bound by Ad35.
Unlike Ad5/3, the fiber-modified Ad5.RGD, which displayed significantly enhanced DC transduction over Ad5 in vitro, performed far less efficiently in the ex vivo skin explant system, showing no benefit over the use of unmodified Ad5. Possibly, the majority of i.d. injected Ad5.RGD viral particles transduced keratinocytes and fibroblasts within the explant32, reducing the chances of effective DC transduction. We conclude therefore that viruses with sole RGD-modifications are not suitable for human DC-targeting strategies based on intradermal delivery.
Comparing Ad5/3 with the previously identified DC-targeting adenovirus Ad5/35 4 and with Ad5 for selective targeting of DC, Ad5/3 displayed enhanced specificity for DC, as Ad5/3 transduced less CD1a-CD83- non-DC bystander cells within human epidermal and dermal single cell suspensions. This might reduce the chances of any untoward side effects associated with bystander infection, such as cytopathic effects on keratinocytes or stromal cells, or suboptimal T cell activation or tolerance induction through transduction of non-professional antigen-presenting cells. This hypothesis should however be tested in the context of an antigen-specific response.
Intradermally delivered Ad viruses may also drain directly to LN and infect DC populations there. The observed DC transduction efficiencies for Ad5/3 in melanoma-draining SLN correlated with the detected CD80/86 expression levels on the respective LN-DC subsets: high on CD1a+ DC (which were transduced at a three-fold higher efficiency than the CD11chi DC), intermediate on the CD11chiCD14- DC, and absent from pDC. As the levels of these important co-stimulatory molecules likely correspond to T cell stimulatory capacity, this may point to an attractive potential of Ad5/3 viruses to target the most potent T cell stimulatory DC subsets. Unlike the CD11chiCD14- DC, which may mostly consist of LN-resident DC with an unknown capacity for T cell activation, we previously showed that CD1a+ DC in these melanoma-draining LN most likely correspond to LC and dermal DC migrating from the skin.21 Upon i.d. GM-CSF administration, frequencies of these CD1a+CD83+ DC in the melanoma SLN correlated to melanoma-specific CD8+ T cell rates33 and we recently confirmed the capacity of both mature CD1a+ epidermis- and dermis-emigrated DC to prime CD8+ T cells specifically recognizing the melanoma-associated antigen MART-1 in vitro.34 Thus, the preferential transduction of this skin-derived DC subset by Ad5/3 may be conducive to the subsequent priming of TAA-specific T cells. Ad5/3 did not transduce B- or T cells in the LN samples, neither did it show enhanced transduction of activated T-or B cells in human skin samples or tonsil material (data not shown), confirming its relative specificity for DC and precluding the possibility of B cell-mediated antigen presentation which was previously implicated in CD8+ T cell tolerance induction.35 In conclusion, our observation of preferential and enhanced CD1a+ DC transduction (with high transgene expression levels) by Ad5/3 in melanoma-conditioned SLN bodes well for its possible use as vaccine vehicle in cancer patients, but awaits further confirmation by T cell priming/activation experiments with melanoma antigen encoding Ad5/3 vectors.
High transduction efficiency and Ad-binding after retargeting should not interfere with DC functionality. Whereas CD46 surface expression was shown to be down-regulated upon binding of Ad35,30 transduction with Ad5/3 did not reduce expression levels of CD80 or CD86. Ad5/3 transduction had no effect on the capacity of the DC to induce T cell proliferation in an allogeneic MLR, whereas CD80/CD86 blockade did, nor on the secreted cytokine profile of the infected cells. These data thus indicate that Ad5/3 can be used to target mature human DC in situ, while preserving the capacity of the DC to migrate and subsequently stimulate T cells.
As Ad5/3 transduced less bystander cells in the skin compared to Ad5/35, there may be a reduced likelihood of causing unwanted side effects, but whether in a therapeutic setting one would benefit more from a virus with somewhat higher efficiency but lower specificity (Ad5/35) or vice versa (Ad5/3) remains to be further examined in relevant in vivo models. Based on data by Ophorst et al., these comparative in vivo studies, or indeed clinical trials, could best be carried out with Ad35- and Ad3-based vaccines rather than the respective Ad5 Fib35 or Fib3 chimeras. They found that an Ad5/35 chimera could not circumvent Ad5-directed pre-existing immunity in vivo in BALB/c mice or non-human primates,36 whereas there was a far lower prevalence of anti-group B immunity than anti-Ad5 immunity and Ad35-based therapeutic vaccines were not affected by pre-existing anti-Ad5 immunity.37,38 Wu et al. showed that replacing the Ad5 hexon with an Ad3 hexon already prevented neutralization in mice with pre-existing anti-Ad5 immunity.39 In addition, it would be interesting to analyze whether Ad3-based vaccines are hampered by anti-Ad35 immunity and vice versa. If there is no efficient cross-neutralization, Ad35 and Ad3 could be used in a prime-boost therapeutic strategy in hosts with pre-existing anti-Ad5 immunity. Noteworthy, a prime-boost strategy with Ad5 and Ad35 alone already dramatically enhanced the generation of antigen-specific CD8+ T cells in mice.37 Similar to CD80/86-targeted Ad5/3, we previously showed the ability of CD40-targeted Ad5 (CD40-Ad) to target mature DC in the context of human skin.22 With a panel of different, efficiently DC-targeting adenoviruses, e.g. Ad35, Ad3 and CD40-Ad, it may become feasible to design sequentially varying prime-boost regimens that facilitate the generation of an even more robust antigen-specific T cell response while circumventing anti-Ad immunity. Additionally, combining these different DC-targeted viruses in one vaccine might enhance vaccine efficacy as more than one DC-receptor can be targeted simultaneously. Also, with the Ad3 knob having a natural DC-targeting ability without the need for further genetic modification, the generation of molecularly defined, clinical-grade batches might be more feasible than with genetically modified DC-targeted viruses (e.g. a previously reported CD40-targeted fiber mosaic variant40). While more research is warranted, these prime/boost or combinatorial Ad vaccine options could certainly benefit future immunotherapeutic strategies.
The authors would like to thank Dr. Hongju Wu (UAB) for providing the Ad5.RGD, Ad5.pK7 and Ad5.RGD.pK7 viruses, Dr. Martijn Bijker for testing Ad5.RGD in the human skin explant model and Sinead Lougheed for technical assistance with the cytokine bead array. The authors would further like to thank the plastic surgery units of the VU medical center in Amsterdam, the Tergooi hospital in Hilversum and the Slotervaart hospital in Amsterdam for supplying the human skin specimens.
This study was supported by a grant from the Royal Dutch Cancer Society (Koningin Wilhelmina Fonds-KWF) to R.J.S., G.L.S. and T.D.G. (KWF2003-2830), a grant from the Netherlands Organization for Scientific Research (NWO) to T.D.G. (NWO VIDI grant 917-56-32), a grant from the National Institute of Health (NIH) to D.T.C. (NIH 5R01CA083821) and travel grants from the KWF, NWO and the Dutch Society for Immunology (NvvI) to R.V.
Financial Disclosure: All authors have declared there are no financial conflicts of interest in regards to this work.