An Immunological Rationale for Vaccines
A recent meeting brought immunologists and vaccinologists together for the first time to discuss rational vaccine development
By Andreas von Bubnoff
“We really don’t know how to make vaccines in a predictable way. It’s still a little bit of black magic.” That was one of the messages Tachi Yamada, president of the global healthprogram at the Bill & Melinda Gates Foundation, had for the attendees at this year’s Keystone Symposium on Immunological Mechanisms of Vaccination, held Oct. 27-Nov. 1 in Seattle, Washington.
Bali Pulendran, one of the conference organizers and a professor of immunology at Emory University, agreed and said that a more rational approach is needed for future vaccine development using insights from immunology. “What’s happened in the past is that most vaccines have been made empirically without a real immunological rationale,” Pulendran said. “That’s worked very well for many vaccines like smallpox, yellow fever, and so on, but increasingly with the more difficult vaccines like HIV, TB, [and] malaria, that’s not going to work because the immunological issues are much more complex.”
Therefore, one goal of this year’s conference was to bring vaccinologists and immunologists together, Pulendran said. To encourage dialogue between the two groups, the organizers intentionally mixed talks on the two research areas in the same sessions. An unusually diverse array of topics were covered at the meeting, including the development and mechanism of action of adjuvants, systems biology approaches to studying vaccination, and the immunological mechanisms that give insights as to why some HIV candidate vaccines might have failed and how they could be improved.
By most accounts, the pairing of vaccinologists and immunologists made for a successful meeting. In his closing remarks, conference co-organizer Rino Rappuoli, a vaccinologist who is head of research at Novartis Vaccines and Diagnostics, said he hopes this won’t be the last time the two groups interact. “In order to approach the problems of the future, we need to do better than nature,” he said. “So far we managed to do as good as nature. If we want to do better than nature, we need to get together.”
Short and sweet
Perhaps a good example of how basic immunology can inform vaccinology was a talk by Surojit Sarkar, an assistant professor of immunology and infectious diseases at Pennsylvania State University. He reported on experiments in mice that suggest that the adenovirus serotype 5 (Ad5)-based vaccine candidate (MRKAd5) developed by Merck may not have induced a protective CD8+ T-cell response in the Phase IIb STEP trial because the CD8+ T cells it induced were dysfunctional and exhausted.
The conventional wisdom for traditional vaccinology, Sarkar said, is that antigen needs to be expressed continuously for a vaccine to induce a strong and lasting immunological memory. But Sarkar’s lab, working with the lab of Vandana Kalia, also at Penn State, found evidence that actually, antigen should only be around for a limited time during the initial few days when the antigen induces naive CD8+ T cells to become effector CD8+ T cells, which can kill the pathogen. But for these effector T cells to then develop into efficient memory T cells, antigen must be absent, he said. Otherwise, the memory CD8+ T cells will be exhausted and dysfunctional. Persistent low levels of antigen, he said, might explain why MRKAd5 did not induce a functionally optimal CD8+ T-cell response. “If you are making a good vaccine, make it short and make it sweet,” Sarkar said.
In mice, Sarkar and colleagues measured the quantity and the quality of the CD8+ T-cell response to acute viral infections that are naturally cleared by the immune system, such as infection with the Armstrong strain of lymphocytic choriomeningitis virus (LCMV) and vaccinia virus (VV). Because such infections are naturally cleared, the immune response to them “must be doing something right,” Sarkar said. “We take that as a gold standard.” The researchers then compared this to the CD8+ T-cell response to Merck’s Ad5 vector used in the STEP trial carrying an LCMV antigen.
They found that the quality of the Ad5-induced memory CD8+ T cells was worse than that of the LCMV-induced cells in that they were shorter lived, less likely to produce several different functional anti-viral cytokines, and less able to proliferate and induce protection in response to a secondary pathogen challenge. Their gene expression signature was similar to dysfunctional, exhausted CD8+ T cells observed in chronically infected people. “This signature is enriched only in dysfunctional and exhausted T cells that see longer duration of antigen,” Sarkar said. Consistent with this, preliminary analyses of human samples from the STEP trial also show that the CD8+ T cells induced in the vaccinees are not very likely to produce several different cytokines, Sarkar said.
Antigen encoded by the Ad5 vector seemed to be expressed much longer in the Ad5-vaccinated mice than viral antigens are expressed in the LCMV-infected mice. And when Sarkar and colleagues removed the Ad5-induced CD8+ T cells from antigen exposure after one week (by placing them into naive, unvaccinated mice), the cells developed into better quality CD8+ memory T cells. “We think that the antigen persistence in Ad5 is [one of the reasons] the quality of Ad5-induced memory differentiation is impaired,” Sarkar said. “You need to give that effector cell a long time of rest so that it can recharge to become a good memory cell.”
To develop a vaccine that can induce a better quality CD8+ T-cell response, Sarkar and colleagues are now experimenting in mice with a modified Ad5 vector, the expression of which can be turned off on demand by certain bio-molecular agents that can theoretically be given as a pill, such as the antibiotic Doxycyclin. “If that works, we may try to test it in nonhuman primates,” Sarkar said.
Adjuvants, an important ingredient added to many vaccines to enhance the immune response, were discussed by several speakers at the conference. Although adjuvants are used in many existing vaccines, researchers still don’t understand how most of them work. Alum, for example, which consists of insoluble aluminum salts, was discovered about 80 years ago and is used in three quarters of all our childhood vaccines, according to Stephanie Eisenbarth, an instructor at Yale University who recently started her own lab there. However, “we still don’t know why it’s a good adjuvant,” said Philippa Marrack, an immunologist who also studies alum at National Jewish Health, a nonprofit hospital in Denver. Marrack’s lab has looked at many of the cell types that are attracted by alum injection to see if they are required for its adjuvant effect. “We have done experiments in which we eliminate each of those cells that are showing up separately and [it] has no effect on the adjuvant activity,” Marrack said. Perhaps, she said, alum acts in so many different ways simultaneously that it’s hard to understand which one is important.
One hint as to how alum might enhance immune responses comes from a finding by Eisenbarth in collaboration with Richard Flavell, also at Yale University. They found that in mice, the induction of immune responses by alum requires activation of an intracellular sensor called NLRP3, which is part of a so-called inflammasome, a multiprotein complex which activates inflammatory processes after the detection of pathogens or cellular stress (Nature 453, 1122, 2008). However, Marrack said her lab has been unable to confirm these results. While Marrack confirmed that alum activates NLRP3 in mice, she could not confirm that this activation is required for the induction of immune responses by alum in mice.
Another clue as to how alum might work came in a talk by Ken Ishii, a professor at Osaka University, who suggested that part of the effects of alum might come from the fact that it kills cells, which results in release of DNA. Some of the adjuvant effects could then come from the DNA.
Alum is believed to mostly induce CD4+ T-cell and B-cell immune responses. Hana Golding, a principal investigator at the US Food and Drug Administration (FDA), reported that when it comes to the quality of the humoral immune response to flu vaccine, alum is not the best choice as an adjuvant. Golding said that the oil-in-water adjuvant MF59, not alum, can direct the human antibody response induced by inactivated H5N1 avian flu vaccine to epitope targets in the HA1 part of the viral hemagglutinin protein, which are more important for protection than the epitope targets in the HA2 part (Sci. Transl. Med. 2, 15ra5, 2010). She said that unpublished data suggest that MF59 has a similar effect in people vaccinated with pandemic H1N1 vaccine.
Contrary to the predominant belief that alum primarily induces CD4+ T- and B-cell responses, Marrack reported that, at least in mice, injection of alum together with ovalbumin (OVA; the main protein found in egg white) as an antigen can induce OVA-specific CD8+ memory T cells. “[This] was an unexpected result, an accident actually,” Marrack said. These CD8+ memory cells are not very good killers, but when alum is given in combination with monophosphoryl lipid A (MPL), a detoxified form of bacterial lipopolysaccharides (LPS) that is used as an adjuvant in licensed vaccines, the killing capacity of these induced CD8+ T cells can be improved. Further experiments showed that mice vaccinated with flu vaccine containing a small part of an internal protein of the influenza virus were better protected against flu challenge eight weeks later when alum and MPL had been added to the vaccine than with alum addition alone. Because the vaccine in these experiments contains an internal viral protein inaccessible to antibodies, Marrack said this protection is likely due to the CD8+ T-cell response. Therefore, adding alum and MPL to human flu vaccines wouldn’t necessarily prevent flu infection, Marrack said, but would make people less sick and therefore less likely to infect others.
Not everyone was convinced, however, that Marrack’s finding in mice is relevant to humans. The induction of a CD8+ T-cell response by alum is not found in humans, Rappuoli said, adding that often, T-cell responses are reported from studies in mice without understanding whether the findings are relevant for humans. This is an example of the disconnect that exists between immunology and vaccinology that he hopes the dialogue started at the meeting can solve.
Developing new adjuvants
While alum and MPL are the only two adjuvants that are added to licensed vaccines in the US, Norman Baylor of the FDA said new adjuvants are becoming necessary because as newer vaccines are purified to make them safer and less reactogenic, the natural adjuvants they contain are removed. “The irony is that when we were using whole cell products we had natural adjuvants,” Baylor said. “We purified these products and now we need the adjuvants back so we’ve got to make new adjuvants and then we face potential safety issues all over again.”
Steven Reed, director of research and development at the Seattle-based non-profit Infectious Disease Research Institute, reported on the status of the development of a new adjuvant called GLA, a synthetic glycolipid based on MPL. A synthetic compound has the advantage that it can be better purified. “Having something pure allows you to produce your product at a lower cost, keep the dose very low because you are only injecting what’s active, and to control your manufacturing process,” he said. Reed and colleagues chose to base their synthetic compound on MPL because MPL is already used in licensed vaccines, including a hepatitis B vaccine and the human papillomavirus vaccine Cervarix, both manufactured by GlaxoSmithKline. Reed said MPL has been shown to reduce the number of doses required for the hepatitis B vaccine, and with Cervarix, MPL broadens the immune response to include antibodies against serotypes not included in the vaccine.
Because MPL is a glycolipid and activates the Toll-like receptor (TLR) 4 pathway, Reed and colleagues identified GLA by screening a library of synthetic glycolipids for their ability to activate the TLR4 pathway in cultured human dendritic cells and macrophages, antigen presenting cells that are crucial for generating the humoral and cellular immune response by activating CD4+ and CD8+ T cells. While alum tends to primarily induce humoral immune responses, GLA induces or increases both humoral and cellular immune responses, according to Reed. And in mice and nonhuman primates, GLA is safe and can broaden the immune response to flu strains not contained in the flu vaccine.
Results from a Phase I clinical trial also suggest that GLA is safe when added to injectable influenza vaccines that contain inactivated flu proteins, said Reed. Another Phase I clinical trial showed “very promising” dose sparing effects, which means that less vaccine may be necessary to achieve the same immune response, Reed said. Next year, Reed and colleagues are planning to test GLA in clinical trials of HIV vaccine candidates.
In Seattle, researchers reported that the expression of interferon (IFN)-induced genes can predict vaccine effects. Rafick Sékaly, chief scientific officer and co-director of the Vaccine & Gene Therapy Institute in Florida, in collaboration with Louis Picker, a professor at Oregon Health & Science University, vaccinated rhesus macaques with different live-attenuated vaccines, including simian immunodeficiency virus (SIV) mac239∆nef, and challenged them intravenously a year later with SIVmac251. When the researchers checked gene expression nine days after the initial vaccination they found that the macaques that later turned out to be best protected had a high IFN response shortly after vaccination, but a low IFN response after challenge, whereas macaques that were not protected showed the opposite profile.
IFN-inducible genes can also predict adverse events. I-Ming Wang, an associate scientific director at Merck Research Laboratories, found that genes induced by IFN can predict adverse events in response to a vaccine. —AvB
The systems approach
Systems biology is becoming increasingly important for characterizing the effect of vaccinations on the immune system, and at the conference several speakers described projects using this approach. In an update on his experiments using a systems biology approach to study the immune response to yellow fever and flu vaccines (see A Systems Approach to Understanding Vaccines, IAVI Report, July-Aug. 2010), Pulendran showed that microarray data can reveal quite unexpected functional links between the immune response and other biological processes.
Pulendran’s previous study of the immune response to the yellow fever vaccine found that a gene called EIF2AK4or GCN2 was correlated with the magnitude of the later adaptive CD8+ T-cell response to the vaccine (Nat. Immunol. 10, 116, 2009). Pulendran and colleagues have now shown that GCN2 knockout mice had reduced CD8+ T-cell responses to yellow fever vaccine, suggesting this gene was required for the adaptive immune response. While it was known that the protein encoded by GCN2 is phosphorylated in response to amino acid starvation of cells in response to stress, which results in a shutdown of translation, the link of GCN2 to the adaptive immune response was not known before, Pulendran said.
Microarray analysis of the response to the flu vaccine revealed another unexpected functional link of a gene to the immune response. The analysis identified a gene, CaM Kinase 4, the expression of which just three days after vaccination was negatively correlated with the HA antibody titers four weeks after vaccination. This was confirmed in CaM Kinase 4 knockout mice, which showed strongly increased antibody titers in response to a flu shot. “[This shows] that by combining rigorous human experiments with mechanistic animal model experiments, you can learn really new biology,” Pulendran said.
Ronald Germain, director of the program in systems immunology and infectious disease modeling at the National Institute of Allergy and Infectious Diseases, and colleagues have also been using a systems biology approach to look at the response to flu vaccination. Germain, who is also an associate director at the Center for Human Immunology, Autoimmunity and Inflammation (CHI), which was recently established by the US National Institutes of Health, presented an update on a CHI project that is characterizing, in unprecedented detail, the immune response of 160 people before and after vaccination with the killed version of the seasonal flu virus and the H1N1 swine flu virus (see A Systems Approach to Understanding Vaccines, IAVI Report, July-Aug. 2010). The project collects data prior to vaccination to establish baseline measurements and help define the normal so-called “immunome,” and post-vaccination data on day one for the innate response, day seven for the adaptive response, and day 70 for the memory response.
Among other things, the analyses so far include measurements of genome-wide gene expression and more than 60 cytokines in blood. Preliminary results from over 60 of the individuals suggest that many gene expression changes occur as early as one day after vaccination, some of which are changes in innate immune response genes. But Germain said one potential concern is that at least some of these early gene expression responses one day after vaccination could simply be due to stress related to the injection rather than the vaccine itself. To address this problem, he plans to measure responses to a mock injection.
Germain also mentioned another challenge with gene expression analyses that was repeatedly discussed at the conference. When researchers use microarrays to measure the expression of thousands of genes in mixed cell populations, the data of gene expression changes might simply reflect changes in the abundance of certain cell types and not cell type specific gene expression changes. “This comes up again and again and again,” Germain said.
One way to address this problem is to separate the cell types first and then do gene expression analyses. However, Germain said, this can be expensive if it has to be done for many samples, and the number of available cells can also be limiting for some subjects. Less expensive bioinformatic tools are being used to extract the cell type specific data from microarray measurements, Germain said, but experiments in influenza-infected mice suggest that this approach can miss important gene expression changes that reflect disease state. “Cell sorting may be critical in some cases,” Germain said.