New Insights on Adenovirus as Vaccine Vectors

Abstract

Adenovirus (Ad) vectors were initially developed for treatment of genetic diseases. Their usefulness for permanent gene replacement was limited by their high immunogenicity, which resulted in rapid elimination of transduced cells through induction of T and B cells to antigens of Ad and the transgene product. The very trait that excluded their use for sustained treatment of genetic diseases made them highly attractive as vaccine carriers. Recently though results showed that Ad vectors based on common human serotypes, such as serotype 5, may not be ideal as vaccine carriers. A recently conducted phase 2b trial, termed STEP trial, with an AdHu5-based vaccine expressing antigens of human immunodeficiency virus 1 (HIV-1) not only showed lack of efficacy in spite of the vaccine's immunogenicity, but also suggested an increased trend for HIV acquisition in individuals that had circulating AdHu5 neutralizing antibodies prior to vaccination. Alternative serotypes from humans or nonhuman primates (NHPs), to which most humans lack pre-existing immunity, have been vectored and may circumvent the problems encountered with the use of AdHu5 vectors in humans. In summary, although Ad vectors have seen their share of setbacks in recent years, they remain viable tools for prevention or treatment of a multitude of diseases.

Introduction

Adenovirus (Ad) vectors were developed to replace genes in inborn errors of metabolism. Enthusiasm toward the use of first-generation Ad vectors in gene replacement therapy diminished because they not only failed to affect sustained gene transfer, but also resulted in significant toxicity and in the death of an individual.^1,2,3 Due to their aptitude for inducing potent innate and adoptive immune responses, Ad vectors have been and are being explored as vaccine carriers.^4,5 Till recently, replication-defective Ad vectors of the human serotype 5 (AdHu5) were heralded as the most promising vaccine platform for antigens of human immunodeficiency virus (HIV) 1.⁴ However, they failed to meet expectations and in a large-scale clinical trial, termed STEP trial, not only showed lack of efficacy, but appeared to cause harm by slightly increasing rates of HIV-1 acquisition in individuals with pre-existing neutralizing antibodies to AdHu5.^6,7 The underlying mechanisms by which AdHu5 vaccination cause a potentially transient increase in susceptibility to HIV-1 remain unknown. Although the STEP trial was not a success in its ultimate goal to protect against HIV-1, it was a success in its impeccable execution and as such will provide guidance on future vaccine efforts, which at least for HIV-1 are shifting to Ad vectors derived from rare human serotypes⁸ or from serotypes derived from nonhuman primates (NHPs).⁹

Here, we briefly review the different applications of Ad vectors and the approaches that are being taken to improve their performance.

Ad Classification, Genetic Organization, and Structure

Ads have been isolated from multiple species including primates, bovines, fowls, reptiles, and frogs. Human Ads have been classified into 51 immunologically distinct serotypes, which are divided into 6 subgroups, i.e., subgroups A–F. Ads commonly used as gene transfer vehicles are human serotypes (AdHu) 4 (E), 5 (C), 26 (D), 35 (B), 41 (F), and 48 (D), and chimpanzee serotypes (AdC) 3, 6, 7, 63, and 68. Most AdC viruses are phylogenetically grouped within human Ads of family E.¹⁰ In humans, family E contains only one serotype, i.e., 4, which is cross-neutralized by antibodies to AdC63. Family E viruses thus have evolved into several serotypes in chimpanzees, but into only one serotype in humans. This in turn indicates family E Ads originated in chimpanzees and that AdHu4 may formerly have been an AdC virus that successfully crossed into humans. A uniform nomenclature for chimpanzee-derived Ad vectors referred to here as AdC viruses is not yet available. For clarification, Table 1 shows a list of AdC viruses and their different names.

Table 1.

Nomenclature of adenovirus isolated from chimpanzee

Ads are nonenveloped and simple double-stranded DNA viruses. The Ad genome size is ~30–35 kilobases and contains five segments that encode early gene products, i.e., E1a, E1b, E2a, E2b, E3, and E4, and five segments that encode late gene products, i.e., L1–L5. E1, E2, and E4 gene products have regulatory functions that allow for transcription and translation of the late genes^11,12,13 and are indispensable for viral replication. E3 gene products subvert immune responses by affecting antigen presentation, cytokine and apoptosis pathways^14,15 and are not needed for viral replication. The Ad capsid is composed of 20 triangular facets of an icosahedron.¹⁶ Each icosahedron contains 12 copies of hexon trimers. The fiber proteins, which also form trimers, are inserted at the 12 vertices into the pentameric penton bases. The hexon, the most abundant of the capsid proteins, has a shaft that is highly conserved between different serotypes of Ads, and a tower that contains a number of loops that show variability between different serotypes.¹⁰ Most neutralizing antibodies to Ad are directed against the variable hexon loops. The Ad fiber is composed of a thin N-terminal tail, a shaft, and a knob domain. The shaft is composed of variable numbers of β-strand repeats. The numbers of repeats largely determine the length of the shaft, which varies between Ads from different families.¹⁷ The knob contains loops that are lettered A–J. The fiber knob loops bind to cellular receptors. The A–B fiber knob loops of most Ad bind to the coxsackie Ad receptor (CAR),¹⁸ which after birth is expressed mainly on epithelial cells.¹⁹ Other Ads bind to CD46,²⁰ a complement regulatory protein that is expressed on most cells in humans and NHPs, but only in testes and sperm in mice.²¹ Cell entry is mediated upon binding of a secondary receptor, which in some Ad viruses such as AdHu5, is provided by an Arg-Gly-Asp motif within the penton base that binds to vβ3 or vβ5 integrins.²² The Ad genome encodes several additional minor capsid proteins, termed proteins IIIa, VI, VIII, and IX, which have been reviewed previously.²³

Immune Responses to Ads

Natural infections with Ads induce antibodies. Neutralizing antibodies, directed mainly against the hexon loops, are serotype-specific. They affect the efficacy of Ad vector gene transfer by blocking cell transduction.^24,25,26 This in turn reduces transgene product expression. Measures of neutralizing antibodies have been used to estimate prevalence rates of Ad infections. Neutralizing antibodies to AdHu5 virus, the basis for the most commonly used Ad-based gene transfer vehicles, are present in about 40–45% of adults in the United States and in up to 90% of humans residing in sub-Saharan Africa.²⁷ The prevalence of neutralizing antibodies to other serotypes also varies.^28,29 More recent efforts have focused on developing vectors based on so-called rare serotypes, such as AdHu26, 35, or 48, to which most humans do not have neutralizing antibodies.^30,31 Notwithstanding, it should be pointed out that the seroprevalence rates of neutralizing antibodies to different Ad serotypes vary considerably between geographic regions, so the serotypes that may be rare in the United States can be quite common in other regions.²⁹ Neutralizing antibody titers are typically determined in tissue culture by neutralization of vectors expressing a reporter protein. A recent publication suggests that such assays may not faithfully recapitulate neutralization in vivo.³² Non-neutralizing binding antibodies to Ads crossreact among different serotypes (Z.Q. Xiang and H.C.J. Ertl, unpublished results).

T-cell responses to natural infections with Ads have been studied less vigorously. CD4⁺ and CD8⁺ T-cell responses have been demonstrated in humans,^33,34,35 and as confirmed by studies in rodents, these cells appear to crossreact between multiple serotypes,²⁵ which is to be expected considering the high sequence homology of Ad gene products. Several groups are currently assessing T-cell responses to Ad antigens in human subjects, and most of these data have been presented at conferences but not yet been published. Results indicate that most human adults have circulating Ad-specific T cells. One publication reported the presence of Ad-specific T cells in human gut tissue and also showed that circulating Ad-specific T cells are mainly of the central memory CD4⁺ T-cell subset, producing predominantly tumor-necrosis factor-α.³⁶ Other groups reported discordant results, and one publication showed that humans have both circulating CD4⁺ and CD8⁺ T cells in blood that predominantly produce interferon γ.⁷ The most sensitive method to measure T-cell responses in outbred populations is based on intracellular cytokine staining with concomitant staining for cell surface markers that identify T-cell subsets and their differentiation status and functions. This assay, which commonly uses antibodies with up to 16 different dyes, is technically difficult, which may in part explain the contradicting results. Also, the outcome of such experiments is influenced by the amount of antigen used for in vitro stimulation of the T cells and this amount varied between the different reports.^7,36

Ads induce potent inflammatory responses, in part due to the activity of structural viral proteins. Activation of innate responses appears to involve several pathways, including at least two toll-like receptors, i.e., toll-like receptor 2 and toll-like receptor 9, as well as type 1 interferon.^37,38,39 In addition, the Ad DNA is recognized in the cytosol by NALT3, a NOD-like receptor family member, which in turn triggers a pro-inflammatory cytokine response.⁴⁰ Overall, Ad viruses and vector are recognized by innate sensors through multiple pathways that all result in activation of cytokine and chemokine release, which in turn impose dose-limiting toxicity for the use of Ad vectors in humans.⁴¹

Types of Ad Used as Vectors

Early Ad gene transfer vectors were deleted in the E3 domain, a deletion that does not render the virus replication-defective.⁴² Most currently used vectors are deleted in E1 and E3 (Table 2). The E1 deletion renders the virus replication-defective, whereas the E3 deletion may affect the vectors' interaction with the host's immune system. This effect is expected to be insignificant in E1-deleted Ad vectors due to lack of transcription of the E3 gene product. Nevertheless, E3 deletions are useful to increase the packaging capacity of Ad vectors. Although E3 is not needed for viral replication, E1 is essential and has to be transcomplemented in suitable packaging cell lines in order to allow viral propagation. Two packaging cell lines are commonly used to grow E1-deleted Ad vectors: (i) human embryonic kidney (HEK) 293 cells contain a large segment of the AdHu5 genome including the E1 domain⁴³ and (ii) PerC6 cells contain only the E1 genes of AdHu5 virus.⁴⁴ Outgrowth of replication-competent Ads is common in HEK 293 cells due to homologous recombination between the E1 flanking regions in the vector and the viral sequences present in HEK 293 cells. PerC6 cells do not allow for homologous recombination and are therefore superior for production of replication-defective AdHu5 vectors—unfortunately, this cell line is not available to academic investigators. Serotypes other than AdHu5 can be grown on HEK 293 or PerC6 cells.^44,45 For some serotypes, this requires the replacement of the open reading frame 6 of E4 with that of AdHu5.⁴⁶ For some applications, only E3 or E1 and E3 deleted vectors are used. To avoid outgrowth of replication-competent adenoviruses on HEK 293 packaging cells, E1 deleted vectors with further deletions in E4 are being tested as vaccine vectors.⁴⁷ Additional deletions including so-called fully gutted vectors have been explored for gene transfer.⁴⁸ As a rule, most additional deletions in early genes reduce the immune response to the Ad vectors and prolong transgene product expression.⁴⁸ Fully gutted Ad vectors, in which all of the viral coding sequences have been removed, have shown reduced toxicity caused by innate immune responses and achieved sustained gene transfer in immunocompetent rodents.⁴⁹ Nevertheless, even fully gutted Ad vectors can elicit immune responses to the transgene product.⁵⁰ Due to cross-priming, a process in which antigen degraded by other types of cells is acquired and presented by professional antigen-presenting cells,⁵¹ one would expect that large doses of gutted Ad vectors would also elicit T-cell responses to structural proteins of the vector particles, especially in humans with pre-existing immunological memory to Ad viruses.⁵²

Table 2.

Examples of adenovirus vectors used for vaccination

More recent efforts have focused on modifying rather than deleting genes of Ad vectors. Structural gene products of Ad vectors have been modified to improve their performance for a specific application. Modifications of the hexon protein of an AdC vector by a point mutation in the loop that carries the dominant neutralizing antibody epitope was shown to reduce binding of serotype-specific antibodies in vitro, but still resulted in vector neutralization in vivo.³² Replacement of all of the hexon loops of AdHu5 with those of AdHu48 reduced binding of AdHu5 neutralizing antibodies, and allowed for the effective use of this vector as a vaccine carrier in animals with pre-existing neutralizing antibodies to AdHu5.⁵³ Replacement of the fiber knob of a CD46-binding AdHu35 vector with the fiber knob of the CAR-binding AdHu5 virus changes the tropism of the vector and increases transgene product–specific immune responses.⁵⁴ Reversely, replacing the AdHu5 fiber with that of AdHu3, another CD46-binding virus, broadens the tropism of the chimeric AdHu5/3 vectors to CD46-expressing cells.⁵⁵ Insertion of an Arg-Gly-Asp sequence, which binds to integrins, into the H–I loops of the fiber knob of a CAR-binding vector also increases the tropism of such vectors.⁵⁶ Similar effects were seen upon incorporation of a polylysine motif into fiber, which allows binding to heparan sulfate receptors.⁵⁷

AdHu5 vectors have a high tropism for liver where most vector particles are taken up and destroyed by Kupffer cells.⁵⁸ A number of modifications have been attempted to reduce the high liver tropism of AdHu5 vectors. Some reported that simultaneous deletions of CAR-binding sequences in fiber and the integrin-binding sequences in penton greatly reduce liver transduction.^59,60 However, other groups failed to confirm these results.⁶¹ Subsequent studies showed that CAR is not used for liver transduction, but the coagulation factor (F)X binds to the AdHu5 hexon and that this complex is readily taken up by hepatocytes due to a heparin-binding motif in the FX serine protease domain.^62,63 Mutation of a putative heparan sulfate proteoglycan—binding motif in the fiber shaft of AdHu5 vector was shown to reduce levels of liver transduction in animals.⁶⁴

Biomedical Applications for Ad Vectors

Gene transfer by replication-defective Ad vectors has mainly been explored for gene replacement and vaccination.

Gene replacement

Ad vectors were initially developed for gene replacement therapy, where they effected cures in immunocompromised animal models but failed in immunocompetent hosts including humans, as adaptive immune responses to the antigens of Ad, most notably, CD8⁺ T cells, rapidly eliminated the transduced cells.⁶⁵ Ad vectors showed significant toxicity in human volunteers.⁶⁶ A number of second-generation Ad vectors were tested subsequently to circumvent vector elimination by the immune system.^67,68 The immune system, which has evolved over the millennia to fight off pathogens such as viruses, cannot be fooled that easily. As detailed above, Ads are ubiquitous viruses that infect all humans. They also persist⁶⁹ and one would expect that they are periodically reactivated—otherwise, evolution would not have favored that trait. Repeated infection, combined with repeated reactivation of a persisting virus, will not only augment immune responses but also maintain T cells at a fairly active stage (Tables 3 and 4). Avoiding such potent responses even temporarily by the use of immunosuppressants designed to prevent T-cell proliferation rather than effector functions will not be possible without causing irreversible harm to the patients' immune system.

Table 3.

Examples of structural modifications on adenovirus vectors

Table 4.

Examples of nonstructural modifications on adenovirus vectors

Ad vectors for vaccination

Ad vectors have been tested extensively as vaccine carriers. For vaccination, most vectors are deleted in E1 or E1/E3. Initially, most vaccines were based on AdHu5 virus.^4,70 In preclinical models, these vectors were shown to induce potent transgene product–specific T- and B-cell responses.^71,72 T cells were mainly of the CD8⁺ T-cell subset, although low CD4⁺ T-cell responses of the T-helper (Th) cell type 1 were also induced. However, responses were clearly reduced by pre-existing neutralizing antibodies to AdHu5 virus.^25,27 In addition, repeated immunizations with AdHu5 vectors given in a prime boost regimen were not overly effective.⁹ T-cell responses were remarkably sustained and failed to contract after the initial effector phase. This was linked to the vectors' ability to persist in a transcriptionally active form in T cells.⁷³

Expressing conserved antigens of HIV-1 for induction of CD8⁺ T-cell responses, which are assumed to provide some protection against the virus, early phase clinical trials with an AdHu5 vaccine yielded sufficiently promising results in volunteers at low risk for HIV-1 acquisition to allow for two large phase 2b trials in humans at high risk for HIV-1 acquisition. Two trials, called STEP and Phambili trials, were powered to assess the effect of vaccination on rate of acquisition of HIV-1 and on viral set points. The vaccine expressed several antigens of HIV-1 to induce T cells and was given three times to volunteers that had either no or moderate-to-high titers of neutralizing antibodies to AdHu5. The STEP trial was stopped and unblinded before enrollment had been completed because an interim analysis showed lack of efficacy and even more worrisome trend toward increased HIV-1 acquisition in vaccine recipients. A more detailed analysis of the data gathered before unblinding showed that in the STEP trial, only men had become infected in the vaccine group. Infection rates were slightly higher in vaccinated males with moderate-to-high pre-existing neutralizing antibody titers to AdHu5 than in the corresponding control group.⁶ The acquisition rate of HIV-1 was identical among men who were seronegative for AdHu5 before the trial. Further analysis has shown that the trend for higher HIV-1 acquisition rates in the vaccinated group mainly affected uncircumcised men with moderate-to-high pre-existing Abs to AdHu5.⁶ Additional data have been gathered from these cohorts since unblinding of the trials, which suggest that the observed trend of increased HIV-1 acquisition may have been transient and could no longer be confirmed in follow-up studies. Notwithstanding, it has to be stressed that acquisition trends, after the trial was stopped, have to be interpreted with extreme caution, as the unblinding of the trial and the high publicity of the trial result may well have affected the behavior of the vaccine recipients.

An assessment of the vaccine-induced immune responses showed that, as expected, the vaccine stimulated HIV-1–specific CD8⁺ and CD4⁺ T-cell responses.⁷ Responses were slightly lower in individuals with pre-existing neutralizing antibodies to the vaccine carrier. There was no striking difference in the functionality of the T-cell responses in individuals with or without pre-existing neutralizing antibodies to AdHu5, nor was there a clear difference in other T-cell markers used to determine the differentiation status of T cells. There were also no differences upon comparison of preinfection samples from vaccinated cases or noncases. None of the results of these thorough studies could explain the increased acquisition rates, leading to the formulation of several hypotheses. Some argued that the increase in rates of HIV-1 acquisition was unrelated to the vaccine. Others postulated that antibodies to the vaccine carrier affected vector targeting and hence antigen presentation, in a way that favored the induction of a Th2 response to antigens of HIV-1; or that CD4⁺ T cells to antigens of Ad induced at very high frequencies in Ad seropositive individuals served as target cells for HIV-1 infection. Available data support neither of these hypotheses.

HIV-1 differs from other pathogens by primarily infecting and replicating in activated CD4⁺ T cells. CD4⁺ T cells have to be induced by vaccines to promote activation of CD8⁺ T cells and B cells. Other vaccine platforms, such as protein vaccines or poxvirus vectors that induce CD4⁺ T cells, were not found to increase HIV-1 acquisition rates in clinical efficacy trials. These vaccines induce T cells that rapidly transition into central memory cells after an initial effector phase, unlike Ad vectors, which stimulate a protracted effector/effector memory phase due to low-level persistence of the viral vector.⁷³ Whether the prolonged effector/effector memory phase was causative for the increased susceptibility to HIV-1 infection remains unknown.

The STEP trial results may have consequences for the use of Ad vectors as vaccine carriers in general. Without a firm understanding of the underlying pathways that led to the trend of increased HIV-1 acquisition in vaccinated Ad seropositive individuals, AdHu5 vectors may no longer be used as vaccine carriers for HIV-1, nor may they be used in postpubescent juveniles or adults as vaccine carriers for other pathogens. Replication-competent Ad vectors⁷⁴ are being developed as vaccine carriers for HIV-1 antigens, intended for oral delivery in enteric-coated capsules. Immunity to these vectors based on AdHu4 and AdHu7 is common in humans, but oral delivery may circumvent the effect of pre-existing neutralizing antibodies.⁷⁵ On the other hand, induction of CD4⁺ T cells within the intestine, the preferred targets of early HIV-1 replication, may pose risks.

Lack of efficacy of the AdHu5 vaccine in the STEP trial in individuals without pre-existing neutralizing antibodies may reflect that T cells cannot protect against HIV-1. Alternatively, the vaccine may not have achieved responses that were of sufficient magnitude and/or quality to increase resistance against HIV-1 infection or spread. The latter is supported by the finding that vaccination of Ad seronegative elite controllers, which in general mount higher CD8⁺ T-cell responses to some antigens of HIV-1, appeared to provide some protection to this small subgroup, as they developed significantly lower viral loads upon infection compared to nonvaccinated elite controllers. With the caveat that the trial had not been designed for such a subgroup analysis, the results are nevertheless intriguing, as they may suggest that a vaccine to HIV-1 that achieves a more potent T-cell response than the STEP trial vaccine may be efficacious. The repeated use of a viral vector for boosting of insert-specific immune responses is relatively ineffective, and one would expect, as has been shown in preclinical studies, that prime boost regimens combining different serotypes of Ad vectors or Ad vectors with other vaccine platforms, such as poxvirus vectors, would induce higher frequencies of T cells, which may suffice to induce protective immunity.^76,77,78,79

Efforts to develop HIV-1 vaccines based on different serotypes of Ad continue. In NHPs, the STEP trial vaccine induced protection against challenge with a pathogenic simian immunodeficiency virus/HIV chimera called SHIV89.6P, which is now considered a “soft” challenge virus, but failed to protect against an infection with simian immunodeficiency virus. Ad vaccine regimens based on the sequential use of different serotypes, which induced more potent T-cell responses in NHPs and provided protection against simian immunodeficiency virus, are now undergoing or scheduled for early stage clinical testing.⁸

Ad vector vaccines are being developed for pathogens other than HIV-1 as well as for cancer and degenerative diseases. Pathogens targeted by Ad vectors include parasites such as Plasmodium falciparum,⁸⁰ leishmania,⁸¹ bacteria such as Mycobacterium tuberculosis⁸² or pneumococcus,⁸³ and viruses such as dengue virus,⁸⁴ Japanese⁸⁵ or Venezuelan encephalitis virus,⁸⁶ rabies virus,⁷² influenza virus,⁸⁷ or hepatitis viruses.^88,89 Tumor antigens that have been tested include oncoproteins of human papillomavirus,⁹⁰ tumor-associated antigens of melanoma such as MART-1⁹¹ or carcinoembryonic antigen.⁹² The amyloid β protein is being tested to reduce the amyloid load in Alzheimer disease.⁹³ Many investigators are still focusing on AdHu5 vectors, although these vectors are expected to perform suboptimally in humans with high titers of AdHu5-specific neutralizing antibodies. Other vaccine efforts are shifting to rare human serotypes or AdC vectors. The immunogenicity of Ad vectors of either serotype is being augmented by coexpression of antigens with immunomodulators such as cytokines⁹⁴ or inhibitors of immunosuppressive pathways⁹⁰ or by retargeting through structural modifications of capsid proteins.⁹⁵

Summary

Although, in our minds, Ad vectors should not be pursued for permanent replacement of genes, their use for vaccination should be explored further. New serotypes to which most humans lack neutralizing antibodies have been vectored and are expected to circumvent problems encountered with pre-existing neutralizing antibodies in humans. The effect of pre-existing memory T cells that crossreact between different serotypes on Ad-mediated gene transfer remains to be investigated in more depth. Advances in our understanding of the basic biology of Ad viruses and the structure of their main capsid antigens now allows for targeted modifications that are expected to further improve the performance of these highly versatile vectors.