Dartmouth Engineer - The Magazine of Thayer School of EngineeringDartmouth Engineer - The Magazine of Thayer School of Engineering

Cracking the Immune System Code

As she tries to assess immune responses to HIV vaccines, Professor Margaret Ackerman is working toward a grand unified theory of antibody activity.

By Michael Blanding
Photographs by John Sherman

Margaret Ackerman
DECODER: Margaret Ackerman wants to know how vaccine components educate the immune system.

Inside each of our bodies is an arsenal of defenders just waiting for the right attacker to show up. Our front-line troops are legions of T cells, white blood cells on patrol looking for cells that are infected with foreign pathogens, binding to them, and killing them. Behind the lines, other blood cells called B cells produce antibodies, proteins that bind to foreign invaders and either neutralize them or act as a beacon to recruit other immune cells to come and attack.

Our bodies have millions of these B cells, each producing its own unique antibodies, which together are capable of recognizing any infection we might encounter. “Protein engineers would call this a library,” says Margaret Ackerman, an assistant professor of engineering at Thayer. “Any time you are infected, your immune system samples thousands of different antibody proteins and identifies those that bind best to the target pathogen.”

While the body is capable of doing this by itself in response to most infections, scientists have found a way to jumpstart the process with vaccines to help the body sort through its own vast libraries in order to produce the right antibodies to fight off invaders it has yet to see. That’s done by using a weakened, inactive, or dead form of a pathogen or sometimes by using unique protein components, or antigens, from pathogens. “By exposing your T cells and B cells to weak, dead, or component antigens of a pathogen, they can learn to recognize a real infection,” says Ackerman. Once exposed, the body maintains populations of T cells and antibodies, each with specific orders to seek out and destroy a specific virus or bacteria.

The system works incredibly well—until it doesn’t. While most foreign invaders spur this natural response in the body, some diseases fail to muster the right troops in our bloodstream. In the cases of flu and malaria, for example, the pathogens are so variable, with so many unique strains, that the body can’t know in advance what will hit it—or in the case of flu, an annual vaccine is needed to help battle each year’s particular strain.

Then there are other diseases, like HIV, the virus that causes AIDS, that just seem to thwart the body’s natural immune response. For reasons scientists don’t completely understand, the body’s white blood cells seem helpless in the face of HIV’s attack, failing to generate the right T cells and antibodies to effectively target the disease. And if the body can’t do it on its own, that makes traditional ways of developing immunity through vaccination more problematic.

“If your immune system can’t naturally induce a protective response, then building a vaccine to do so is a fundamentally more challenging goal,” says Ackerman, who has dedicated her lab to achieving exactly that goal. In combination with a team of collaborators, her research group’s secret weapon is a unique platform to evaluate antibodies in much greater detail than has been possible previously, which is then coupled to complex computer algorithms to determine how best to make vaccines better.

“It’s sort of looking under the hood of your immune system,” Ackerman says. “We are looking at the thousands of different variants of antibodies your B cells have produced in response to vaccination and weeding through that data to learn the features of potent immune responses in order make better vaccines.” Using these techniques, her research may contribute in the not-too-distant future to one of the holy grails in modern medicine—a vaccine that could provide immunity against HIV.


Ackerman was trained as a protein engineer at the Massachusetts Institute of Technology in Dane Wittrup’s lab in the departments of chemical engineering and biological engineering. There she learned to evolve antibodies in the lab to change or enhance their function. “It’s just crazy cool to evolve a protein in a test tube and to make a new molecule that can do something that doesn’t exist in nature,” she says, “or take a natural protein and make it ten times or a thousand times or a million times more effective.”

As intriguing at that sounds, that type of engineering is actually quite routine. (“Undergrads do this,” Ackerman says.) But it’s one thing to develop an antibody in a lab; it’s another to create one in the human body. “How can you get people with different immune systems and different antibody libraries to each mount a response that is protective?” she asks.

That’s even trickier in cases like HIV, where the body has such a limited capacity to do that on its own. The HIV-positive subjects who have developed protective antibodies have tended to do so over long periods of time after being exposed to different strains of the virus. Those antibodies belong to a class called broadly neutralizing antibodies, which work by binding to the HIV virus and rendering it inactive. But candidate vaccines have not been able to induce the immune system to robustly generate those types of antibodies.

In fact, only four vaccines have ever been tested in large-scale human efficacy trials to see if they can drive development of antibodies or T cells that will make vaccinated subjects immune to HIV. Of those four vaccine trials, two appeared harmful to patients, one had no effect, and one was marginally effective. In that last case, however, the vaccine, called RV 144, seemed to function in a surprising way.

Ackerman was a postdoc at the Ragon Institute of Massachusetts General Hospital, MIT, and Harvard researching the immune response to HIV when she heard the results of the RV 144 vaccine trial. Carried out in Thailand, the trial involved more than 16,000 people who were given the vaccine or a placebo and then monitored for HIV infection. The vaccine, consisting of an inert form of a bird virus engineered to express several HIV genes followed by a mix of two HIV protein antigens, was controversial at the time, criticized for the amount of time and money spent on such an experimental treatment. “Prominent scientists wrote a letter to Science questioning whether the study was a waste of resources,” Ackerman recalls.

It was a surprise, then, when results released in September 2009 showed that the vaccine provided 60 percent protection against the disease after a year and 30 percent after two years. When blood samples were tested for the presence of several common antibody types, however, the broadly neutralizing antibodies weren’t among them. “Scientists thought we knew how a vaccine was going to work. It was either going to induce broadly neutralizing antibodies or a really potent T-cell response,” says Ackerman. “This vaccine appeared to work, but it didn’t do either of those things very well. So the question was, what did it do?”

One clue was the presence of different types of antibodies in people who were protected compared to those who weren’t. These so-called binding antibodies can attach themselves to a specific site on a pathogen and act as a molecular beacon that recruits other immune cells to eliminate pathogens from the body. Some of these cells, such as neutrophils, can trap pathogens, while others “eat” the infected cells or circulating virus. In other cases, the antibodies initiate a chain reaction known as a complement cascade, which punches holes in infected cells, killing them.

The possible activities of these binding antibodies are complex. Each antibody possesses a unique means to recognize the pathogen and a unique ability to summon innate immune effector cell power. Collectively, these antibodies program pathogen recognition and provide instructions about what should be done with the tagged cell or virus.

These instructions can have a wide range of possible outcomes. “There are order-of-magnitude differences in how antibodies recruit different innate immune cells,” says Ackerman.

In the Thai trial, however, researchers performed a restricted analysis of correlates of infection risk, limiting both the information they could obtain about what was going on after exposure to the virus and their ability to solve the riddle of how these antibodies might provide protection from the disease.


In the research she began at Ragon, Ackerman has taken the exact opposite approach. “We said we are going to measure as many different things as we possibly can, and then we are going to use data-driven discovery,” she says. “When you limit the questions you ask, you limit the answers you can get. Instead of pre-ordaining our measurements, we’ll let the analytical models tell us which of those measurements or which combinations of measurements are associated with protection.”


With her postdoc advisor, Galit Alter, Ackerman developed a set of procedures that would analyze antibodies in incredible detail. (While there hasn’t been another large-scale efficacy trial since the Thai trial, multiple HIV vaccines tested in monkeys or humans have generated a range of antibody responses.) Alter and Ackerman’s tests would not only examine pathogen recognition and inactivation, but also more globally evaluate the types and activities of the antibodies, including which immune cells they recruited and how they affected the pathogen.

Such cell-based assays, however, are still difficult and time-consuming, requiring a large number of human immune cells to test, and therefore limiting the number of antibodies Alter and Ackerman could test. After coming to Dartmouth in 2011, Ackerman called on her molecular engineering background to create faster and more efficient means of analyzing antibody responses. “I wanted to step away from more complex assays and enable better understanding at the molecular level of how the antibody binds to viral antigens and how the innate immune system recognized the antibody,” she says.

Using a boxy, plastic instrument vaguely reminiscent of an all-in-one printer or a Star Wars droid, researchers in Ackerman’s lab test antibodies for their ability to recognize up to 500 different variants of viral proteins in a single assay. They do this by attaching each protein to a tiny bead called a microsphere, which fluoresces a different color and intensity when a laser is shined across it. When the beads are incubated with samples from vaccine recipients, each of the antibodies in the sample can bind to its respective protein target, and another fluorescent measurement can be made to determine how prevalent each antibody specificity is and how well each antibody type interacts with the various antibody receptors expressed on innate immune cells. “Instead of just saying how many antibodies are there and what they recognize, we look at how well those pathogen-specific antibodies interact with the innate immune antibody receptors,” Ackerman says. That information, in turn, can offer clues about how effective antibodies are in recruiting immune cells to attack the virus.

Unlike the kind of assays used in the Thai trial, which looked at a handful of different antibody types, this multiplexed assay can generate a tremendous amount of data about the wide variety of different antibodies active in a single sample. “For any one of those serum samples, we might have a thousand different characteristics about the pathogen-specific antibodies that are present,” says Ackerman. “It’s much more data than has been typically generated for evaluating vaccines.”

In order to make sense of all of that information, Ackerman runs the data through machine learning software that can help make broad predictions. “Just like Amazon learns from other people’s choices or Netflix learns from other people’s viewing habits to make suggestions for you, we weed through our thousands of data points and see how well the data we collected can predict or differentiate between subjects that were protected versus those that were not,” she says.

That level of prediction is key, says computational biologist Chris Bailey-Kellogg, a Dartmouth computer science professor who collaborates with Ackerman to analyze the data. “One way she is changing the field is not just to say what’s different, but also to make predictions about what is going to happen,” he says. “If Netflix only told you are different from someone else, that’s not as useful as telling you what movie you might like.”

Even the same vaccine antigens can produce dramatically different results depending on how they are administered to patients (for example, intravenously or nasally), how often they are given, and what kind of other chemicals, called adjuvants, are added to boost their effectiveness. By feeding all of this data into their computational models, Ackerman and her collaborators can better predict the ideal combination of factors that will cause a vaccine to work at its best. They are then able to go back to scientists testing vaccines and give them valuable predictions with which to tweak their next vaccine. “It’s not just trial and error anymore,” says Bailey-Kellogg. “You are not just trying a vaccine and seeing if it worked. You can see what vaccine is working and why.”

Ackerman and her collaborators recently looked at a National Institutes of Health (NIH) study of nearly 100 monkeys, one of the largest non-human primate vaccine studies ever done. “When you have this many animals, and this many outcomes, you have to sift through the data to figure out what’s important,” says study leader Mario Roederer, a senior investigator at NIH’s Vaccine Research Center. “There are a ton of different ways of modifying the vaccine, and they generate different qualities of the antibodies,” he says. “What her work does is identify what flavor of antibody we want to develop and which vaccine is much more likely to develop that antibody.”

Not only is Ackerman able to do this with individual vaccine trials, but because she and her collaborators are testing multiple vaccines from different studies, they are also able to pool data to synthesize results to look for patterns of effectiveness. So far, many of the analyses her lab has made seem to bear out the Thai trial findings—that the key to protection may not be a broadly neutralizing antibody but, rather, a combination of binding antibodies sending signals to the immune cells. “It really is a swarm of antibodies working in concert,” she says.

While the samples Ackerman’s lab has been able to study so far come from small-scale human and non-human primate trials, she is looking forward to the next efficacy trial currently underway in South Africa, which will use a variation of the same vaccine used in the Thai trial, RV 144, modified for strains of the virus prevalent in Africa. “The entire HIV research community is keen to see if the vaccine will protect in another subject population,” says Ackerman. She hopes that when results are released in two years, she will be able to analyze samples to examine exactly what type of antibodies the vaccine induced—whether it proves effective or not.

More than that, however, Ackerman hopes her research can ultimately help unlock the secrets of exactly how antibodies function in all of their complexity. “We want to have a grand unified theory of antibody activity,” she says, “so we can say, if you want an antibody that is really great at recruiting a neutrophil, it should have this type of profile, or if you want an antibody that initiates the complement cascade really well, you want that type of antibody.” Having that kind of information could help crack the immune system code, allowing scientists to create vaccines to fight not only HIV, but potentially a large number of other diseases, including influenza, malaria, and even cancer.

Beyond this understanding, immunologists would like to do more than see the effect of their candidate vaccines on B cells in terms of the antibodies they produce. They want to know the details of how this process takes place. “It’s sort of the black box of immunology,” says Ackerman. “When we evaluate a vaccine, we know what we are injecting and we know something about the antibody or T-cell responses raised in subjects, but we’d also like to know how exactly the vaccine components educated the immune system to produce which types of responses.”

Call it designer vaccine development. “By understanding exactly how vaccines function,” Ackerman says, “we can better understand how to engineer them to allow our bodies to do the work fighting off the diseases they can’t combat well enough on their own.”

—Michael Blanding is a Boston-based writer and the author of The Map Thief.

Categories: Features

Tags: engineering in medicine, faculty, innovation, research

comments powered by Disqus