Immunity's Yin and Yang

A successful vaccine must first avoid being eliminated by pre-existing immunity before it can promote a protective immune response

By Philip Cohen, PhD

In the hunt for better vaccines, researchers are engineering viruses and bacteria into harmless vectors, delivery vehicles for genes from pathogens, to safely stimulate protective immunity. But as researchers attempt to domesticate these microbes to deliver genes for vaccines, possible immune reactions against those vectors become important obstacles to overcome.

Consider, for example, human adenovirus serotype 5 (Ad5), a naturally-occurring virus that has been engineered so that it carries HIV genes, creating an AIDS vaccine vector. Results from a preliminary human clinical trial of such a vaccine by Merck were very encouraging; the vaccine elicited the strongest cellular immune responses yet seen for an AIDS vaccine.

But since the natural form of Ad5 regularly infects humans, causing a severe version of the common cold, experts were worried that many people may already have "pre-existing immunity"—antigen-specific antibodies and immune cells—able to attack the Ad5 vaccine vector. And that worry was confirmed. In clinical trials of an early version of the vaccine containing a gene for the HIV Gag protein, the response to the vaccine was blunted in people even with moderate levels (a titer over 1:200) of pre-existing anti-Ad5 antibodies. In the US and Europe about one-third of the population has pre-existing immunity able to significantly reduce vaccine efficacy—and in developing countries that could be as high as 80% of the population. The specter of pre-existing immunity cast a shadow over this otherwise promising trial.

That's why last fall's announcement by Merck was as welcome as it was unexpected. It turned out that a higher dose of a newer version of the vaccine carrying three HIV genes elicited detectable immunity in a broader spectrum of volunteers. "We were seeing immune responses in 60-70% of people, including those with high levels of pre-existing immunity," says Robin Isaacs, executive director of HIV vaccine clinical research at Merck. "We don't really know why, but it suggests that Ad5 vaccination could be useful for more people."

The Merck Ad5 story may be a small triumph in the larger ongoing battle against pre-existing immunity that vaccine designers face. But researchers have been busy investigating many other strategies to help vectors escape immune destruction so that they can do their job.

The problem of pre-existing immunity isn't unique to Ad5 or even to vaccine design. GlaxoSmithKline, for example, recently announced plans to develop an HIV/AIDS vaccine based on an existing measles vaccine, against which many people have childhood immunity. The same issue has arisen for various vectors for vaccines and gene therapy based on poliovirus or BCG (a bacterial strain used as a vaccine against tuberculosis) and adeno-associated virus (AAV).

But while Ad5 is not unique in presenting the challenge of pre-existing immunity, it provides an illuminating example. It is a popular vector—it's now being explored as a vaccine vector for HIV, Ebola virus, anthrax, and the SARS coronavirus, as well as many applications in gene therapy—and many strategies to overcome pre-existing immunity to this vector are being developed.

The attraction of Ad5 lies partly in its ability to infect many types of cells, both actively dividing and non-dividing. It is easy to make the virus defective in replication, and it doesn't integrate into chromosomes, addressing two clinical safety issues. Genes incorporated into the virus are expressed at high levels, an important advantage for presentation of vaccine antigens and therapeutic proteins. The virus is also easy to grow in large scale tissue culture. For vaccinologists, its best feature is that it is a natural born stimulator of immunity—a single administration induces strong antibody and CD8+ T cell responses. And it can also induce immunity after being delivered under the skin, into muscle, into the blood, orally, or intranasally.

Some important details of how pre-existing immunity interferes with subsequent use of Ad5 have also been worked out in animal models. Inoculation of the virus elicits strong neutralizing antibodies against viral capsid components, the most potent of which are against the adenovirus hexon protein. Antibodies block the free virus, stopping its entry into cells and the expression of its genes, and therefore subsequent antigen presentation. Any viral particles that manage to pass the antibody gauntlet can still be destroyed by Ad5-specific CD8+ cytotoxic T lymphocytes (CTLs). As a result of this double hurdle, few viral particles can sneak through, generally not enough to kick start a strong immune response against vaccine components or to boost previous vaccinations with the same vector.

Isaacs says the problems with pre-existing immunity to Ad5 were known when the Merck team began to build their HIV vaccine vector. It was no great surprise when these researchers found that in people with antibody levels over 1:200, they saw a poor cellular immune response, barely above background, even at the highest dose of 1x10¹¹ infectious particles. "Increasing the vaccine dose seemed to overcome, at least partially, the impact of pre-existing immunity to Ad5," says Isaacs. "But the response rates were still much worse than in those subjects with low, less than [1:200], Ad5 antibody titers."

With those results in hand, Merck began the Phase I safety trial of the next version of the vaccine, a trivalent mixture of vectors containing genes for three different HIV proteins (Gag, Pol and Nef). The 118 volunteers included people with a broad range of anti-Ad5 antibodies, up past 1:4000. But when the company started to recruit the 1500 volunteers for the larger Phase IIb trial of this vaccine, they decided to exclude people with pre-existing immunity above 1:200 since their previous work suggested the vaccine would have little or no effect in this population.

That's when Merck got a pleasant surprise. Analysis of data from the Phase I trial of the trivalent vaccine showed that it behaved quite differently from the version carrying a single HIV gene. They saw that 60-70% of people had a detectable cellular response to HIV proteins based on IFN-g ELISPOT assays, including people with levels of antibody 10 to 20-fold higher than the 1:200 cut off. Based on that new data, Merck, in collaboration with the HIV Vaccine Trials Network (HVTN) and the NIH Division of AIDS, decided to double the size of the Phase IIb trial, with the second 1500 drawing from individuals with a large range of pre-existing anti-Ad5 specific antibodies over 1:200. And at the recent 13th Conference on Retroviruses and Opportunistic Infections (CROI), Larry Corey, Principal Investigator of the HVTN, announced that they will partner with Merck on a second Phase IIb trial of the trivalent vaccine in South Africa that will include volunteers with anti-Ad5 antibodies.

So why was the second version of the vaccine more immunogenic? There are two obvious possibilities, according to Isaacs. There could be some immunological synergy when the three HIV proteins are presented together, rather than one at a time. The dose of vector may also have been crucial. The monovalent dose was 1x10¹⁰ particles while the trivalent dose was 3 times higher, containing 1x10¹⁰ particles of each component. That extra dose of vector may have been just sufficient to allow more viral particles to bypass antibodies at the site of injection, invade cells and express their HIV genetic cargo. "My guess is that both factors count, your gene inserts and the dose," says Gary Nabel, director of the Vaccine Research Center (VRC) at the National Institutes of Allergy and Infectious Diseases. In the VRC's clinical trials of Ad5 vectors, Nabel says his team has also noticed that certain immunogens, such as the HIV Env protein, are particularly good at getting around pre-existing immunity compared to others like Gag. "Even with antibodies around, a good immunogen gives more punch for the same amount of protein," he says.

Whether similar strategies could help overcome pre-existing immunity for other Ad5 vaccines or vectors isn't clear. But the HIV community is eager to see if larger trials confirm the result and show these vaccines to be effective at preventing infection. Meanwhile Merck, the VRC, and other research groups are exploring other ways to tackle pre-existing immunity.

One approach taken by Nabel and Dan Barouch at Beth Israel Deaconess Medical Center in Boston and their colleagues is to attempt to lower the threshold of Ad5 vector particles needed for immunogenicity in the face of pre-existing immunity. This team investigated sensitizing the immune system of mice to HIV proteins by injecting them with plasmid DNA vaccines, either alone or in conjunction with adjuvants such as the cytokines GM-CSF and MIP-1a. In naïve mice primed with a DNA vaccine, a dose of 10⁶ particles of Ad5 carrying the gp120 gene was enough to boost cellular immunity so that 25% of CD8+ T cells were antigen-specific. In mice that were pre-immunized with Ad5, a similar cellular response required a dose of 10⁹ particles. If Ad5-exposed mice first received a prime of a cytokine-adjuvanted DNA vaccine, a subsequent dose of 10⁷ Ad5 particles gave the same cellular response, suggesting that pre-existing immunity was partly overcome.

These researchers suggested that the DNA priming elicited the production of antigen-specific memory T cells that then readily responded to smaller doses of the same antigen presented in the Ad5 vector (J. Virol. 77, 8729, 2003). This strategy is now being tested in human trials by the VRC group. In data presented at meetings from trials VRC009 and VRC010, Nabel has reported that the CD8+ T-cell response from the DNA prime/Ad5 boost can be 10- to 100-fold higher than for the DNA or Ad alone. In Merck trials, however, DNA/Ad5 was found to be no better than Ad5 alone, suggesting that the composition of the vector and vaccination protocol may be crucial for this effect.

It also turns out that Ad5 vector delivered mucosally sometimes performs far better in the context of pre-existing immunity than when delivered systemically. A report from Hildegund Ertl, James Wilson, and their colleagues at the Wistar Institute and University of Pennsylvania analyzed the effects of route of vaccine administration on the development of antibodies against rabies glycoprotein when its gene was delivered in adenovirus vectors. Mice exposed to Ad5 (but not the rabies glycoprotein gene) were subsequently given Ad5 vector containing the glycoprotein gene either orally or as an intramuscular injection. As expected, the development of antigen-specific rabies protein antibodies elicited by the injected vector was highly blunted by pre-existing Ad5 immunity. But the response of the animals to the same rabies protein vector delivered orally was unaffected by previous Ad5 exposure of the animals (J. Virol. 77, 10780, 2003). Later experiments showed the ability of orally-delivered vector to overcome pre-existing immunity was dose-dependent: rabies protein antibody induction was reduced by about half at a dose of 2x10⁵ particles, but not affected in animals given a dose of 2x10⁶ vector particles. The reason why ingestion of the virus overcomes pre-existing immunity isn't clear, but the authors suggest that the vast number of cells in the gut that carry the coxsackie Ad receptor (CAR), used by Ad5 to invade cells, may make neutralization by antibodies difficult.

But this trick doesn't appear to work for every mucosal route or every antigen. Nabel's group assessed the ability of intranasally-delivered Ad5 to overcome pre-existing immunity. The nasal route is of interest to researchers working on sexually-transmitted diseases because application at this site can result in strong secretion of protective antibodies in the genital mucosa. However these researchers found that mice previously exposed to Ad5 had reduced antigen-specific antibody titers when later given an intranasal dose of an Ad5 vector carrying genes for HIV Gag, Pol, and Env proteins compared to animals with no previous Ad5 immunity given the same vector. They also found that intranasal delivery resulted in infection of the nervous system through the olfactory bulb, suggesting this method may raise safety concerns about potential neurotoxicity (J. Virol. 77, 10078, 2003).

Another strategy to help Ad5 vectors evade immunity is to use chemical tricks to hide the virus. Suresh Mittal's team at Purdue University has encapsulated Ad5 inside alginate microparticles as a way to shield vector proteins from neutralizing antibodies and immune cells. The particles are small enough (5 to 10 micrometers) to be taken up by antigen presenting cells such as macrophages and dendritic cells. In their mouse study, they looked for gene expression of a bacterial b-galactosidase gene carried in the vector in the trachea and lungs after inoculating either naïve animals or those that had received one or two previous injections of Ad5. For the unencapsulated vector, gene expression dropped dramatically with one or two previous Ad5 immunizations, to about one-third and one-eighth of that in naïve animals, respectively. In contrast, encapsulation of the virus preserved at least 75% of the expression level in animals with no prior exposure to Ad5. The absolute level of gene expression in naïve animals was initially 50% lower for the encapsulated vector, however, which could lower the efficacy of an encapsulated vaccine (Gene Ther. 9, 1722, 2002).

Chemistry has also been used to alter the Ad5 vector surface by linking various activated forms of polyethylene glycol (PEG) molecules. Wilson's team found that these modifications did not affect the ability of "pegylated" virus to invade cells, but it did reduce the effect of antibodies and immune cells raised against the unmodified virus. Interestingly, the addition of PEG to the surface doesn't make the virus invisible to antibody recognition. If a vector treated with the same version of a pegylated virus is administered again, the second inoculation is blunted by pre-existing immunity. But PEG appears to disguise the virus by hiding epitopes recognized on the native virus and creating new ones, suggesting that by shifting PEG chemistries, the same vector could be used multiple times in the same animal or person (Hum. Gene Ther. 13, 1887, 2002).

Another strategy that has emerged is to use genetic engineering to alter Ad5 vectors to escape pre-existing immunity by giving the virus a new immunological face. The most successful approach to date has been to incorporate the major structural component of the Ad5 viral capsid, the hexon protein, from different serotypes of adenovirus. But researchers at Merck have encountered at least two problems with this approach. First, presumably due to structural constraints of the viral capsid, the majority of these chimeric viruses can't replicate. And even though this approach allows the virus to overcome anti-Ad5 antibodies, CTLs reactive to other Ad5 proteins blunt the response (Hum. Gene Ther. 13, 311, 2002). At CROI, Barouch presented promising data on a chimeric Ad5 vector in which only the seven short hexon hypervariable regions of Ad5 were exchanged from human adenovirus serotype 48 (see CROI covers advancements from start to finish). This chimeric vector replicated well in complementing cell lines and effectively evaded anti-Ad5 immunity in both mice and rhesus monkeys.

Researchers have also begun developing new vectors based on one of the other 50 or so known human serotypes of adenovirus. Merck researchers are working on a vector based on adenovirus serotype 6 (Ad6). According to data presented at the AIDS Vaccine 2005 conference, pre-existing immunity against Ad6 is significantly lower than for Ad5: 40% of people in Europe and 35% in US had a titer above 1:200 for neutralizing antibodies against Ad5, while only about 7% and 17% of the same populations have neutralizing antibodies against Ad6. Merck has not yet published data on the immunogenicity of Ad6 vectors.

Another human adenovirus, serotype 35 (Ad35), has also emerged as a leading candidate for a vaccine vector. Prevalence of pre-existing antibodies against Ad35 is lower than Ad5 in every population yet examined. For Ad35, neutralizing antibodies were present in less than 10% of populations in Europe, US or Asia and in approximately 20-30% of African populations, while prevalence of Ad5 antibodies in the same four populations are 50, 30, 40, and 90% respectively (J. Virol. 77, 8263, 2003; AIDS 18, 1213, 2004). And a group at the University of Pittsburgh School of Medicine has reported that even when Ad35 antibodies are present, they are usually at low titers (Clin. Diagn. Immunol. 11, 351, 2004). Barouch, working with the Dutch company Crucell, has shown in mice that Ad5 pre-existing immunity does not significantly suppress antigen-specific cellular responses in mice to the SIV Gag gene carried in an Ad35 vector (J. Immunol. 172, 6290, 2004). An HIV/AIDS vaccine based on the Ad35 vector is being developed in a collaboration between IAVI and the Crucell group.

So far, though, these rarer adenovirus serotypes don't live up to Ad5 in every regard. "The rare human Ad serotypes have proven less immunogenic than Ad5 in both mice and monkey studies to date," says Barouch. "Whether this is also true in humans is an empiric question." Barouch's team has also experimented with engineering Ad35 to be more immunogenic by replacing a section of the capsid fiber protein, the "knob," with that of Ad5. This chimeric virus uses the CAR receptor to invade cells instead of CD46, which is the natural receptor for Ad35. This small change results in enhanced immunogenicity of Ad35 in mice and monkeys. But this improvement also had its costs, as it rendered the virus much less stable (J. Virol. 79, 14161, 2005).

The search for other immunologically-distinct alternatives to adenovirus has sent researchers screening viruses isolated from other species. The most closely related of these are the nine serotypes of chimpanzee adenoviruses. Serotypes 6 and 7 (Pan 6 and Pan 7; the vectors are referred to as AdC6 and AdC7) are being actively developed as HIV/AIDS vaccine vectors based on work from the Ertl and Wilson labs. They found that humans rarely have neutralizing antibodies to these viruses and vectors based on them are not affected by high levels of pre-existing immunity to Ad5 or for the other chimpanzee adenovirus in mice (J. Virol. 75, 11603, 2001). These vectors can also induce high levels of antigen-specific cellular immunity even against the backdrop pre-existing immunity to Ad5 (J. Immunol. 170, 1416, 2003).

One of the potential benefits of having a number of vectors from different adenovirus serotypes available is that they can be used in prime boost protocols. For a chimpanzee vector carrying a truncated gene for the HIV Gag protein, these researchers found that priming mice with AdC6 followed by a boost from AdC68 (based on chimpanzee adenovirus serotype 68) resulted in antigen-specific CD8+ T cells expanding to a frequency of 40% of total (J. Immunol. 171, 6774, 2003). The order of vector usage seems to be important for both the size and character of the immune response. A prime-boost-boost immunization of four Chinese rhesus macaques with AdC7, AdC6, and Ad5 Gag-vectors achieved a sequential improvement of Gag-specific antibodies in four Chinese macaques, while their use in the opposite order was far less effective. However, the Ad5, AdC6, and AdC7 order appeared superior for the induction of Gag-specific CD8+ T cell responses (J. Virol. 78, 7392, 2004). "The downside of these viruses is that you are using something new and the experience in humans is much less, which sets the regulatory hurdles higher," says Ertl. IAVI has started a collaboration with GlaxoSmithKline to produce HIV/AIDS vaccine chimp adenovectors that were licensed from the University of Pennsylvania.

Beyond chimpanzees there is a menagerie of adenovirus serotypes from other species: 10 from cattle, 2 from dogs, 7 from sheep, 5 from pigs, and at least 4 from birds. Some of these are already being investigated for use in veterinary vaccines against, for example, bovine herpesvirus, canine distemper virus, and classical swine fever virus. The ovine adenovirus serotype 7 (OAd7) has also been tested in mice as a vector for a hepatitis C virus vaccine and found to be effective at eliciting IFN-g secreting T-lymphocytes even in animals with antibodies against Ad5 (Vaccine 22, 2717, 2004).

One irony about overcoming pre-existing immunity is that if one of these new vectors is actually incorporated into an effective vaccine, it could quickly become a victim of its own success. "If you vaccinate everyone in a population with an HIV adenovirus vaccine, you can't turn around and use the same vector against malaria," points out Ertl. That's why the battle against pre-existing immunity isn't just the search for one perfect vector for each vaccine application, but the hunt for a whole toolkit of related approaches that overcome pre-existing immunity time after time. "Once we have something that works," says Nabel, "there's no telling how many ways we'll want to use it in the future."