|Home | About | Journals | Submit | Contact Us | Français|
A major obstacle to the use of adenovirus vectors derived from common human serotypes, such as human adenovirus 5 (AdHu5), is the high prevalence of virus-neutralizing antibodies in the human population. We previously constructed a variant of chimpanzee adenovirus 68 (AdC68) that maintained the fundamental properties of the carrier but was serologically distinct from AdC68 and resisted neutralization by AdC68 antibodies. In the present study, we tested whether this modified vector, termed AdCDQ, could induce transgene product-specific CD8+ T cells in mice with preexisting neutralizing antibody to wild-type AdC68. Contrary to our expectation, the data show conclusively that antibodies that fail to neutralize the AdCDQ mutant vector in vitro nevertheless impair the vector's capacity to transduce cells and to stimulate a transgene product-specific CD8+ T-cell response in vivo. The results thus suggest that in vitro neutralization assays may not reliably predict the effects of virus-specific antibodies on adenovirus vectors in vivo.
Adenovirus (Ad) vectors are effective at inducing potent CD8+ T-cell responses to immunogens. In animal models, Ad vectors encoding antigens of simian immunodeficiency virus (SIV) and human immunodeficiency virus (HIV), used in combination with plasmid-based DNA vectors, generate CD8+ T-cell responses that attenuate infection by SIV (9) and by HIV-SIV chimeras (16). In humans, Ad vectors derived from human serotype 5 (AdHu5) are immunogenic and are well tolerated at immunogenic doses; however, in a recent clinical trial, an AdHu5-based HIV-1 vaccine failed to prevent (and may have facilitated) infection (1a). It is not clear whether CD8+ T-cell responses will be sufficient to prevent or control HIV infection and disease. However, it seems likely that the induction of effective immune responses against HIV will require multiple doses of antigen, with a priming dose followed by one or more booster immunizations. Prime-boost regimens based on the sequential use of DNA and AdHu5 vectors are being tested clinically, and regimens involving the sequential administration of serologically distinct Ad vectors are being explored in preclinical animal models (1, 5, 8, 9).
One major obstacle to the use of vectors derived from AdHu5 and other common human serotypes is the high prevalence of virus-neutralizing antibodies (VNAs) in humans. Preexisting VNAs to the vaccine carrier prevent the vector from transducing target cells, which reduces the amount of vaccine antigen that can be produced and dampens the resultant adaptive immune responses (2, 3, 12). Approximately 40 to 45% of the U.S. population has VNAs to AdHu5, and seroprevalence rates are even higher in Asia and Africa (6, 24).
We developed vectors derived from chimpanzee Ads to which humans lack preexisting immunity. When tested in a rodent model, one such vector, AdC68, induces potent transgene product-specific CD8+ T-cell responses that can be increased by booster immunizations with serologically distinct Ad vectors (3, 19, 23). However, because the use of multiple serotypes in a prime-boost regimen may prove cumbersome in clinical applications, we have attempted to modify the major neutralizing binding sites within the AdC68 capsid. It has been suggested that the binding sites for Ad-neutralizing antibodies preside primarily within the major capsid protein hexon (4, 10, 14, 15, 17). We defined a single hexon surface loop as the major neutralization site on AdC68 and showed that a mutant vector, AdCDQ, which incorporates a 3-amino-acid mutation within this loop, resists in vitro neutralization by polyclonal antisera obtained from animals immunized against AdC68 (10). Because it is serologically distinct from its parent vector, we expected that AdCDQ could be used in combination with AdC68 in an effective prime-boost regimen.
In the present study, we tested whether the AdCDQ vector induces a transgene product-specific CD8+ T-cell response in mice with preexisting neutralizing antibody to wild-type AdC68. Contrary to our expectation, the data show conclusively that antibodies that fail to neutralize the AdCDQ vector in vitro nevertheless impair the vector's capacity to transduce cells and to stimulate a transgene product-specific CD8+ T-cell response in vivo. The results thus suggest that in vitro neutralization assays may not reliably predict the effects of virus-specific antibodies on Ad vectors in vivo.
Female 6- to 8-week-old BALB/c mice were purchased from the Jackson Laboratory (Bar Harbor, ME) or Ace Animals, Inc. (Boyertown, PA), and housed at the Animal Facility of the Wistar Institute (Philadelphia, PA). All experiments were performed according to institutionally approved protocols. BALB/c mice deficient in the Fc receptor common gamma subunit (FcRγnl) (18) were purchased from Taconic (Hudson, NY).
E1/E3-deleted AdHu5 and E1-deleted AdC68 recombinant viruses carrying the gag gene of HIV-1 clade B or green fluorescent protein (GFP) were constructed and purified, titers were determined and quality was controlled as described previously (10).
4C1, a murine immunoglobulin G2a (IgG2a) monoclonal antibody (MAb) that neutralizes AdC68; 4D1, an IgG2a MAb that binds AdC68 without neutralizing the virus; and 502, an IgG2a MAb specific for rabies virus glycoprotein, were prepared as ascites in SCID mice. Antibody concentrations were determined with an enzyme-linked immunosorbent assay (ELISA) kit for IgG (ZeptoMetrix). A polyclonal antiserum to AdC68 was prepared from pooled sera of rhesus macaques that had each been immunized once intramuscularly (i.m.) with 2 × 1011 virus particles (vp) of AdC68.
BALB/c mice were injected in the muscle of the lower leg (i.m.) with Ad vectors diluted in 100 μl of phosphate-buffered saline (PBS). All vector doses were adjusted to the number of vp. In the prime-boost experiment, groups of mice were immunized i.m. on day 0 and boosted on day 60. For passive antibody transfer, mice were injected via the tail vein either with 50 μg of MAb or with 100 μl AdC68-specific macaque serum (with a neutralizing titer of 1:5,000) diluted in PBS to a final volume of 200 μl.
Single-cell suspensions of splenic lymphocytes were prepared as described previously (11) and washed with Leibovitch's L-15 medium (Mediatech, Inc., Herndon, VA). Red blood cells were lysed with ACK lysing buffer (Invitrogen, Carlsbad, CA) for 5 min on ice. To examine HIV-1 Gag-specific gamma interferon (IFN-γ)-secreting CD8+ T-cell frequencies, lymphocytes were incubated (1 × 106 cells/sample) with Gag peptide (AMQMLKETI) for 5 h at 37°C with 5% CO2. Control cells were stimulated with an irrelevant peptide. Cells were surface stained with an anti-CD8 antibody conjugated to fluorescein isothiocyanate and then fixed and permeabilized with Cytofix/Cytoperm (BD Pharmingen, San Jose, CA) for intracellular staining with an antibody to mouse IFN-γ conjugated to phycoerytherin. Flow cytometry analysis of the cells was performed with Beckman-Coulter (Fullerton, CA) XL flow cytometers at the Wistar Institute Flow Cytometry Core Facility (Philadelphia, PA); data were analyzed with WinMDI 2.8 (Howard Scripps Institute, La Jolla, CA) or FlowJo 7.1.1 (Tree Star, Inc., Ashland, OR). Statistical analysis to calculate P values was performed with Intercooled Stata 8.2 (StataCorp LP, College Station, TX). All antibodies were purchased from BD Pharmingen (San Jose, CA) unless otherwise noted.
Mice injected with polyclonal serum or MAbs were bled for measurements of neutralizing titers before immunization with Ad vectors. Blood was collected from the retroorbital plexus or the mandibular vein. Serum was separated and heat inactivated at 56°C for 30 min.
Replication-deficient wild-type or mutant Ad encoding GFP (6 × 107 particles in 500 μl Dulbecco's modified Eagle's medium-10% fetal calf serum) was mixed with serial dilutions of heat-inactivated sera and incubated for 1 h at 37°C in a 24-well plate in the presence of 5% CO2. The virus-MAb mixtures were transferred into a 24-well plate containing confluent monolayers of HEK 293 cells (2.5 × 105 cells/well) and incubated for 48 h. Cells were harvested with trypsin-EDTA and analyzed by flow cytometry to quantitate GFP expression.
To test the effect of plasma on antibody-mediated neutralization, virus (2 × 107 particles) was incubated with medium alone (200 μl modified Eagle's medium-5% fetal calf serum) or with MAb 4C1, 4D1, or 502 (10 μg/ml) with or without 10% mouse serum or 10% fresh-frozen mouse plasma (Innovative Research, Novi, MI). After 1 h at 37°C, virus-MAb mixtures were transferred into HeLa monolayers in a 96-well plate (2 × 104 cells/well), incubated for 24 h, and then examined in an inverted fluorescence microscope.
Mice were injected i.m. into the lower leg with AdC68 or AdCDQ vectors (1 × 1011 vp) encoding GFP. Mice were euthanized 24 h later, and the injected leg was removed and illuminated with an Illumatool lighting system (Lightools Research). Photographs were taken with a Kodak DCS14N digital camera with a 60-mm Micro Nikkor lens (Nikon). Images were captured as raw files and converted to TIFF files using Kodak DCS Photodesk software.
Popliteal lymph nodes were removed 24 h after injection of the GFP vector. The single-cell suspension was passed through a 70-μm nylon cell strainer (BD Falcon) and washed with L-15 medium (Mediatech). A total of 1.5 × 106 cells/well were plated in a 96-well plate, Fc receptors were blocked by treatment with 50 μl purified anti-mouse CD16/CD32 (BD Pharmingen) for 30 min at 4°C, and cells were then stained with allophycocyanin-labeled anti-mouse CD11c, a dendritic cell-specific marker (BD Pharmingen), for 30 min at 4°C. The cells were washed and suspended in PBS, and flow cytometry was performed using a Cyan-LX flow cytometer (DakoCytomation).
AdC68 or AdCDQ vectors were diluted in 100 mM sodium carbonate (pH 9.5) and immobilized on a 96-well plate (5 × 109 particles/well) by overnight incubation at 4°C. The plates were washed four times with PBS and then blocked for 1 h at room temperature in 3% bovine serum albumin-PBS. The plate-immobilized virus was incubated with hexon-neutralizing antibody 4C1 or with control antibody 4D1 (nonneutralizing hexon binding antibody) or 502 (directed against rabies virus) starting at 0.1 μg antibody/well and continuing with twofold serial dilutions. Bound MAb was detected with horseradish peroxidase-conjugated goat antibody to mouse IgG2a (Santa Cruz). Plates were developed with 3,5,3′,5′-tetramethylbenzidine at room temperature for 5 min. The reaction was stopped with o-phosphoric acid, and the absorbance was read at 450 nm.
Previously, we demonstrated that AdCDQ, which differs from AdC68 by 3 amino acids within the hexon, escaped neutralization by a panel of MAbs specific for the hexon of wild-type AdC68. The mutant also showed reduced susceptibility to neutralization by high-titer AdC68-specific polyclonal antisera from rabbits and rhesus macaques (10). We have now tested whether the modified vector, AdCDQ, induces transgene-specific CD8+ T-cell responses in mice previously exposed to wild-type AdC68 and whether the AdCDQ vector can be used in combination with wild-type AdC68 in a prime-boost regimen.
We immunized naïve mice with either AdC68 or AdCDQ vectors encoding HIV-1 Gag and, 10 days later, measured the frequencies of Gag-specific splenic CD8+ T cells by intracellular cytokine staining. Across a range of doses (109 to 1011 vp), both vectors induced Gag-specific CD8+ T cells with similar efficiencies (Fig. (Fig.11).
We then tested the effect of preexisting immunity to AdC68 or AdCDQ on the responses to immunization with AdC68- or AdCDQ-Gag vectors. Mice were exposed to AdC68 or AdCDQ vectors encoding GFP (1011 vp) and then immunized 2 weeks later with AdC68- or AdCDQ-Gag (1011 vp); frequencies of splenic Gag-specific CD8+ T cells were measured at 10 days and 3 weeks after immunization. In animals that had been exposed to wild-type AdC68, AdCDQ-Gag failed to induce an anti-Gag CD8+ T-cell response, as measured at 10 days (Fig. (Fig.2A)2A) or 3 weeks (not shown). Similarly, AdC68 did not induce responses in mice preexposed to AdCDQ.
To test the efficiency of AdCDQ in a prime-boost regimen, mice were given a priming dose of AdC68-Gag (1011 vp or no priming dose) and boosted with AdC68-, AdHu5-, or AdCDQ-Gag (1011 vp) 60 days later. Splenocytes were harvested, and Gag-specific CD8+ T-cell responses were measured at day 70 and day 120. Although a pronounced increase in Gag-specific CD8+ T-cell frequency was seen in mice boosted with the AdHu5 vector, booster immunizations with the AdC68-Gag or the AdCDQ-Gag vectors were markedly less effective (Fig. (Fig.2B2B).
The results described above suggested that preexisting immunity to AdC68 inhibited effective immunization with an AdCDQ vector. To determine if antibodies contributed to the inhibition we observed, we conducted passive transfer experiments with AdC68-specific antiserum from rhesus macaques. We previously found that this serum neutralized AdC68 in vitro but was 32-fold less active in neutralizing AdCDQ (10). Mice were injected with the AdC68-immune serum or a control serum sample from naïve macaques. Twenty-four hours after passive transfer, mice were bled to determine neutralization titers and immunized with AdC68- or AdCDQ-Gag. Sera from the recipient mice neutralized AdC68 at dilutions of 1:160 to 1:320 but failed to neutralize AdCDQ (Fig. (Fig.3A).3A). Mice injected with nonimmune serum showed no neutralizing activity against either virus. As expected, Gag-specific CD8+ T-cell responses induced by the AdC68-Gag vector were inhibited by the anti-AdC68 antibodies. Unexpectedly, the AdC68-specific antibodies inhibited responses by the AdCDQ-Gag vector even though the antibodies failed to neutralize AdCDQ in vitro (Fig. (Fig.3B3B).
Although the AdCDQ mutation ablates the major neutralizing epitope within the AdC68 hexon, polyclonal anti-AdC68 serum is likely to contain neutralizing antibodies directed against multiple capsid proteins, and these may have interfered with immunization by AdCDQ. To examine the effect of an antibody specific for the hexon neutralization site, we measured Gag responses after passive transfer of a hexon-specific MAb (4C1). We previously found that 4C1 neutralizes Ad68 in vitro but has no activity against AdCDQ, even at high concentrations (10). Mice were injected intravenously with 50 μg of either 4C1, a nonneutralizing AdC68 hexon-specific MAb (4D1), or a control MAb (502) directed against the rabies virus glycoprotein; all of the antibodies were of the IgG2a isotype. Mice were bled 24 h later for measurements of neutralizing titers and then immunized with AdC68- or AdCDQ-Gag.
As expected, sera from mice injected with 4C1 neutralized AdC68 but not AdCDQ (Fig. (Fig.4A).4A). CD8+ T-cell responses elicited by both AdC68 and AdCDQ were not affected by the control antibodies (4D1 and 502) (Fig. (Fig.4B).4B). However, 4C1 inhibited Gag-specific CD8+ T-cell responses induced by both AdC68- and AdCDQ-Gag. Thus, a MAb that fails to neutralize AdCDQ in vitro nevertheless inhibited the immunogenicity of this vector in vivo.
Successful immunization with an Ad vaccine vector requires the transduction of target cells and the expression of the immunogenic transgene product. To determine if MAb 4C1 reduces transgene product expression by the AdCDQ vector in vivo, we repeated the passive transfer described above and, 24 h later (a period sufficient to permit antibody equilibration within tissues), injected the mice i.m. with AdC68 or AdCDQ vectors expressing GFP. Twenty-four hours after injection, we isolated and illuminated the injected leg and measured GFP fluorescence. In mice passively exposed to 4C1, the level of GFP expression induced by both AdC68 and AdCDQ was much lower than it was in mice exposed to control MAb 502 (Fig. (Fig.5A5A).
We also measured GFP expression in dendritic cells from lymph nodes draining the immunization site. Twenty-four hours after immunization, the popliteal lymph nodes were isolated, single-cell suspensions were stained with an antibody specific for the dendritic cell marker CD11c, and the number of GFP-positive dendritic cells was determined. Consistent with GFP expression levels in the leg, the level of dendritic cell GFP expression, induced by both AdC68 and AdCDQ vectors, was reduced in mice passively exposed to 4C1 (Fig. (Fig.5B).5B). These results suggest that 4C1, a MAb that does not neutralize AdCDQ in vitro, inhibits AdCDQ-induced immune responses in vivo by blocking transgene delivery to target cells.
Coagulation factors have recently been found to bind to the Ad hexon and to influence the tropism of virus in vivo (7, 21). We considered the possibility that virus interactions with coagulation factors present in murine serum might contribute to the inhibitory effect of MAb 4C1 that we observed in vivo but not in vitro. To test this, we measured levels of neutralization of virus in the presence of fresh-frozen murine plasma. AdC68 and AdCDQ were incubated with 4C1, with nonneutralizing MAb 4D1, or with control MAb 502 in the presence or absence of murine plasma or murine serum (which has been depleted of clotting factors), and the transduction capacity was then measured. As we previously observed, MAb 4C1 markedly inhibited transduction by AdC68 but had no effect on AdCDQ (Fig. (Fig.6).6). Although transduction by both AdCDQ (Fig. (Fig.6)6) and AdC68 (not shown) was partially inhibited by plasma, the same effect was observed in the absence of antibody or in the presence of control MAb 502. Thus, the effect of plasma cannot explain the antibody-dependent inhibition of AdCDQ-mediated gene delivery which we observed in mice.
We previously found that the CDQ mutation in hexon eliminated the susceptibility to in vitro neutralization by 4C1, but we had not tested whether the MAb still bound to AdCDQ. As measured by ELISA, both 4C1 and the nonneutralizing antibody 4D1 bound to both AdC68 and AdCDQ (Fig. (Fig.7).7). However, AdC68 was bound more avidly by 4C1 than by 4DF1, and conversely, AdCDQ was bound more avidly by 4D1 than by 4C1. The results indicate that the CDQ mutation decreases, but does not ablate, the recognition of AdC68 hexon by 4C1.
We considered the possibility that although 4C1 does not have a direct neutralizing effect on AdCDQ, 4C1 might bind to virus in vivo and facilitate its uptake and destruction by phagocytic cells expressing Fc receptors. To test this, we used mice deficient in the Fc receptor common gamma subunit (FcRγnl) (18), which is essential for the expression of both FcγRI (the high-affinity IgG receptor, which binds monomeric IgG) and FcγRIII (a low-affinity receptor for IgG-antigen complexes) and for the phagocytic function of FcγRII. FcRγnl mice are defective in antibody-dependent phagocytosis as well as antibody-mediated cellular cytotoxicity.
FcRγnl and control mice were treated intravenously with neutralizing MAb 4C1 or with control MAb 502, and after 24 h, to permit antibody equilibration, mice were injected i.m. with AdC68 or AdCDQ. Twenty-four hours after injection, we isolated and illuminated the injected leg and measured GFP fluorescence. In both FcRγnl and control mice passively exposed to 4C1, the level of GFP expression induced by both AdC68 and AdCDQ was much lower than it was in mice exposed to control MAb 502 (Fig. 8A and B). The expression of the GFP transgene in local dendritic cells, induced by both AdC68 and AdCDQ vectors, was inhibited by MAb 4C1 in FcRγnl mice just as it was in control mice (Fig. 8C and D). These results indicate that the inhibition of transgene expression by MAb 4C1 does not depend on the interaction with Fc receptors in vivo.
In the experiments described here, we found a lack of correlation between the inhibitory effects of virus-specific antibody in vitro and in vivo. We found that although both polyclonal antibodies and MAbs failed to neutralize the AdCDQ vector in a standard in vitro assay, they nonetheless prevented effective immunization with that vector in vivo.
The detrimental effect of AdHu5-specific VNAs on transgene product-specific B-and T-cell responses has been well documented in experimental animal models (2, 3, 12). Consistent with this, in a recent clinical trial of an AdHu5-based HIV-1 vaccine (the STEP Trial), levels of CD8+ T-cell responses to HIV-1 antigens (as measured by IFN-γ enzyme-linked immunospot assays) were reduced in subjects with high levels of AdHu5-specific VNAs (9a). Furthermore, although there did not appear to be a specific correlation between infection and a low-level CD8+ T-cell response, there was a possible trend toward increased infection in vaccinees with higher levels of preexisting VNAs (1a). The results of the STEP trial are not yet understood, and it remains to be determined whether vaccine-induced CD8+ T cells can ameliorate HIV infection in humans. Nonetheless, it is clear that Ad-specific neutralizing antibodies present a barrier to immunization with Ad vectors.
To circumvent the problem of preexisting VNAs, a number of investigators have developed chimeric AdHu5 vectors in which critical portions of the hexon are replaced by hexon sequences from other Ad serotypes and have found that such chimeras escape neutralization by AdHu5-induced VNAs (12, 14, 15). An AdHu5 chimera in which exposed hexon loops were replaced by those of AdHu48 was recently demonstrated to escape neutralization by AdHu5-specific antibody and to provide effective immunization in the face of preexisting immunity to AdHu5 (12). However, our results suggest that in vitro assays may not reliably predict the effects of antiviral antibodies in vivo.
Viral neutralization escape mutants selected by growth in monoclonal neutralizing antibodies often have single-site mutations that significantly reduce (but do not necessarily eliminate) the binding affinity of individual antibodies. To create the AdCDQ vector, we replaced 3 amino acids within a hexon surface loop recognized by a panel of MAbs including MAb 4C1 (10). We found that 4C1, despite its inability to neutralize AdCDQ in vivo, bound to AdCDQ in an ELISA (although its avidity appeared to be significantly reduced). Interestingly, MAb 4D1, which recognizes a different epitope within hexon, bound efficiently in the ELISA but did not neutralize virus in vitro or inhibit virus in vivo.
The mechanism by which Ads are neutralized is not certain: antibodies may aggregate virions, block interactions with cell surface receptors, or interfere with postentry events in virus replication (20, 22). Even less is known about antibody-mediated effects in vivo. We found that a MAb that does not neutralize AdCDQ in vitro nonetheless prevented transgene delivery and the induction of transgene-specific CD8+ T-cell responses in vivo. Neutralization of AdCDQ in vitro was not potentiated by murine serum or plasma, nor was inhibition in vivo attenuated by the elimination of Fc receptor function.
We do not know the mechanism by which anti-AdC68 antibody inhibits effective immunization with the AdCDQ vector, but it is clear that antibody prevented the transduction of both muscle cells and dendritic cells at the site of immunization. The in vivo environment is much more complex than that encountered in a neutralization assay. Our data suggest that in vitro neutralization assays, despite their convenience and widespread use, cannot replace in vivo trials for the assessment of new viral vectors.
We thank Roger Burnett and the members of the Ertl laboratory for their helpful discussions and advice. We also thank the Flow Cytometry Core of the Wistar Institute and the Vector Core of the University of Pennsylvania.
This work was supported by grants from the National Institutes of Health (grants P01AI52271 and R21AI060434) and the Pennsylvania Department of Health.
Published ahead of print on 11 March 2009.