New Approaches in Cancer Biology | Charles River · Cancer and viruses use various strategies to evade the immune system. One of these is by evolving antigens that are invisible to

New Approaches in Cancer Biology

22 | IN TRANSLATION

How did you and your lab develop an interest in cancer vaccines?We started off doing very basic immunology about 30 years ago. We made a discovery that helped us predict the location of T-cell epitopes in protein sequences, and that led us into vaccine research. Using our methodologies, we identified some of the first helper T-cell epitopes from HIV protein in the late 80s.

Eventually we got interested in cancer vaccines—because I’m in the cancer institute—and a lot of the same principles

can be applied to cancer that can be applied to chronic virus infection like HIV. For example: HIV and cancer both evade the immune system, they can suppress the immune system, there’s a need for a therapeutic vaccine, not just a prophylactic vaccine, and so forth.

How do you go about developing a therapeutic vaccine against a chronic virus like HIV or against cancer?Most viral vaccines that are currently licensed today are based on the virus itself. And that’s fine for viruses in

Bedside to Bench in OncologyCancer and viruses use various strategies to evade the immune system. One of these is by evolving antigens that are invisible to immune surveillance. In his lab at the National Cancer Institute, Jay Berzofsky develops cancer vaccines based on a technique called epitope enhancement, which modifies tumor antigens to make them strongly bind immune cells and trigger a potent immune response. He is currently testing a vaccine based on this approach in prostate cancer patients, as well as a more traditional vaccine targeted to HER2-positive tumors.

Jay Berzofsky, MD, PhDQ&AChief, Vaccine Branch; Senior Investigator Head, Molecular Immunogenetics and Vaccine Research Section, National Cancer Institute, NIH

Q&A | JAY BERZOFSKY | 23

which infection leads to long-lasting protective immunity—if you get measles or mumps or something like that, if you survive the infection, then you usually have life-long immunity that protects you against future exposures. But a virus like HIV, you get infected and you never reject the virus.

So in cases of chronic viral infection, the virus itself is not an adequate vaccine. We need to do better. And one of the ways we can do better is to take those viral antigens and improve on them, because they have already escaped immune recognition. Similarly, cancers that become clinically evident are ones that have escaped immune surveillance, because the immune system tries to eliminate cancers. In fact, it’s now been shown that probably many cancers start to develop and you never even know they’re there because the immune system gets rid of them. So the ones that we see clinically are ones that have already escaped the immune system. The tumor antigens or the viral antigens that we see are ones that have somehow avoided immune recognition.

We developed an approach called epitope enhancement where we modify the sequence of the epitope. An epitope is a region of the antigen that’s recog-nized by an antibody or by a T-cell. If we can identify the epitopes that the T-cells recognize and then make them stronger, we can make a better vaccine.

How is this approach applied to cancer?The prostate cancer vaccine that we have in clinical trials was developed by taking a prostate cancer antigen called TARP that our collaborator, Ira Pastan, discovered, and first searching the TARP amino acid sequence for epitopes that bind to HLA-A2, which is the most common human class 1 HLA molecule.

We identified the peptide from the TARP protein that would bind HLA-A2 and then we modified those to increase

binding to HLA-A2. We tested a number of possible modifications using HLA-A2 transgenic mice that we could immunize. And then we went from the mouse model to humans, where we took T-cells from HLA-A2 positive patients and then tried to induce an in vitro T-cell response, and tested them for their ability to kill human cancer cells that express both the TARP antigen and HLA-A2. And we found the peptides that work best at doing that. Then based on the fact that we could induce human T-cells in vitro that would kill human tumor cells, we were able to get FDA and IRB permission to do a human clinical trial. We actually showed preliminary evidence for clinical benefit in stage zero prostate cancer patients, where almost three-quarters of the patients had a decrease in tumor growth rate at one year after immunization.

Your lab is also working on a vaccine for HER2-positive cancers. What is the principle behind that vaccine?HER2 is a driver tumor antigen— it actually causes the cancer. It’s expressed in about a quarter of breast cancer patients and a smaller fraction of other cancer types. And it’s also unusual in that HER2 is expressed on the surface of the cancer cell, so it’s accessible to antibodies as well as T-cells. So we made this vaccine and we immunized HER2 transgenic mice. By the age of 25 weeks these mice get 10 independent tumors in all 10 mammary glands, so it’s really an inexorable process driving these tumors.

If we immunized them at a young enough age, we could actually prevent the cancer. We then tried to make a vaccine to treat cancer in a model, using the tumors from those mice in ordinary mice. And in that setting, we could allow the tumors to grow to a fairly large size—up to 2 centimeters—and still immunize the mice and completely cure the tumors.

And that is the holy grail of cancer vaccines. Many people have immunized mice and then given them the tumor, and they can reject it, but if you let the tumor get established the vaccine doesn’t do any good. This is a vaccine that could actually cure large, established tumors, so we were very excited, and we were able to then get approval to make a human version that expresses the human HER2 and immunize cancer patients. We have a clinical trial, which is also showing some promising results in several kinds of cancer.

We could allow the tumors to grow to a fairly large size...and still immunize the mice and completely cure the tumors.”

24 | IN TRANSLATION

Q&A Francesco Marincola, MD, FACSDistinguished Research Fellow in Immuno-Oncology, AbbVie

A Unifying Model to Explain Primary and Compensatory Immune Resilience of CancerCancer immunotherapy has made great strides by disabling the defenses that some tumors use to avoid being attacked by the immune system. But many tumors do not possess these mechanisms to hold the immune system at bay. Instead, they evade detection by avoiding interactions with the host. At AbbVie, Francesco Marincola and his colleagues are investigating the mechanisms that these tumors use to hide themselves, and how treatments might induce them to reveal their presence so that immunotherapy drugs might become effective against them.

Could you give a brief introduction to the work you plan to discuss and why it is important? My position here in Abbvie is in the discovery wing. My work aims to under - stand the mechanisms of immune resistance in cancer. If you really think about it, there are two kinds of immune resistance in cancer. One is what I call compensatory immune resistance, which is in tumors that are immunogenic. You have a lot of immune-regulatory mechanisms that allow them to survive in spite of their immunogenicity.

But the other tumors are completely silent. There is nothing. They are not stimulating immune response. That’s also why there is no immune-regulatory mechanism, because evolutionarily they don’t need that.

While most work is being done in trying to increase the response to immunogenic tumors, very little is being done to try to work on the immune-silent ones—what I call primary immune resistance, or even primary immune ignorance.

I’m trying to understand how you can turn a non-immunogenic tumor into an immunogenic one that could be respon-sive to immunotherapy. We are working very hard on this and we are using different approaches. Most of them are discovery driven, like bioinformatics approaches or open access data, internal data and other ways to rapidly understand basic questions related to this problem.

You have been involved in translational research and have used animal models to study tumor immunology. Based on your experience, what do you think are the biggest drawbacks of animal models of human diseases? Yes, so I’ve been spending a lot of time thinking about that. I think that animal models are crucial at this point because for screening of immunogenicity, particularly in the silent tumors, to test whether you can disrupt their behavior,

Q&A | FRANCESCO MARINCOLA | 25

making them immunogenic—you have to have a very broad and efficient system to screen a lot of therapeutic approaches. So, I think the identification of mouse models that are representative of the immuno-silent tumors is critical. The question is, I don’t know if such models are there, because I believe most animal models are representative of the immunogenic tumor type.

How are you going to identify silent tumor animal models? We believe that one important step is to identify the genetics—to characterize the genetics of tumors that are immuno-silent, so that you can model the animal tumors in a way that they reproduce the mechanism that allows this tumor to grow silently, stealthily, without any immunogenicity. This way, you may have a relevant model.

Overall, I think there are ways by which we can start focusing on models that actually represent human biology for screening because I think the current ones are very good for mechanistic studies, but I don’t think are necessarily representative of what happens in humans.

Do you think computer and math-ematical modeling approaches could help in cancer research?They will. I think those models are just as good as knowledge of the biology. So if you can find ways to train them, in a

way that they are truly representative of what happens biologically, you can then even maybe develop some kind of machine-learning mechanism that can give you or propose potential targets and ways to inhibit certain pathways.

But, again, always depending on how much you understand the biology to start with. In my group I have a lot of bioinformatics people, but I also have a lot of biologists that can actually interpret the data and train the data.

What do you think could be the single change that would have the greatest impact on our approach to modeling diseases for drug discovery? I think the animal models are pretty good already to test immunogenic-type tumors. So, focusing on the silent ones, I think the major thing we should figure out is, what are the most common pathways. And there are already candidates. There are plenty of theories and models. A lot of people are describing the PI3K pathway, the beta-catenin story and the MAP kinases. And so there are pathways.

The question is, can you identify some common denominator downstream? Because I don’t think targeting mutations is very useful, because there are too many. I think it would be more useful to identify downstream

molecules in the pathway and then identify mouse models that can be representative of that and then screen like crazy on those. Do radiation, pathway inhibitors, chemotherapy, you name it. Any ways to see, if you disrupt this, can you change this tumor, maybe turn it into an immunogenic one?

Because immunotherapy is shown to give long-term responses. If you just target tumors only by using inhibition of pathways, what happens, they come back right away with other compensatory mechanisms. So it doesn’t seem to work, just doing pathway inhibition. But maybe you can open a window during pathway inhibition or other anti-cancer treatment to make them immunogenic, and then you can have the immuno-therapy that may give longer-term effects. We don’t know if it’s the case, but it’s something that I think would be by far the most important thing to test.

While most work is being done in trying to increase the response to immunogenic tumors, very little is being done to try to work on the immune-silent ones.”

28 | IN TRANSLATION

Q&A David M. Sabatini, MD, PhDWhitehead Institute for Biomedical Research; Investigator, Howard Hughes Medical Institute; Senior Associate Member, Broad Institute; Professor of Biology, MIT; Member, David H. Koch Institute for Integrative Cancer Research at MIT

Genetic Screens to Study Cancer BiologyDr. Sabatini studies the regulation of growth in cells and its implications for processes such as cancer, diabetes and aging. He did pioneering research in elucidating the target of rapamycin (TOR) pathway, a central controller of cell growth. More recently he has developed genetic screening methods that use CRISPR and other methods to understand the metabolism of cancer and other cells. In his address he will discuss how large-scale CRISPR-based screens could be useful in discovering targets for cancer treatment.

Could you give us a broad overview of your work and the subject of your presentation?My lab is mostly focusing on growth and metabolism; how organisms basically accumulate mass and how they regulate their metabolism.

This is a very broad area and can encompass many, many different things. We tend to focus a lot on growth regulation by a pathway called the mTOR pathway. And we also focus quite a bit on the metabolism of cancer cells. How do cancer cells get the appropriate metabolites to grow and to divide?

In that work, we try to develop tech-nologies as we need them. One of the

things that we worked on, over the last five or six years, has been the development of somatic cell genetics largely based on CRISPR-Cas9 based screens. We did some of the first genome-wide screens in this space where we could look at the function of every single gene in a process—for example the survival of a cancer cell.

What I'll talk about at the meeting is our efforts to do very large-scale CRISPR-based screens in cancer cells and to use that understanding to potentially find new targets for treatment of cancer. I'll also talk about the difference between growing cells in culture versus in vivo, and also point out differences between mice and

Q&A | DAVID M. SABATINI | 29

humans in terms of their metabolism and how this may have impact on these kinds of studies.

Based on your experience, what do you think have been some of the biggest drawbacks of animal models of human diseases? And how can we develop more accurate and more predictive animal models?Well, first of all animals are not humans, right? There are some fundamental differences, both in the genes that they express and how they regulate them. And in certain systems, like the immune system, clearly there are quite a few differences. The differences that we've focused on and have tried to understand a bit better are the metabolic differ-ences. There are differences between the metabolism of human cells and other animals, and even between the metabolism of humans and most mammals. For example, there are some genes that we and higher primates have lost and other mammals have not.

I think the challenge is really to be aware that the metabolism of animals is different and to take that into account. There are areas in metabolism, like in nucleic acid metabolism or lipid meta-bolism, where there are quite important differences, which may eventually affect your experimental outcomes.

So I think that's a caveat to always keep in mind, that these differences can have impacts and that perhaps one has to think about different models to use to compare between them, but also maybe to manipulate the models to mimic the human situation. I think once one is aware of the genetic differences, one can certainly manipulate animals to be more similar to the human situation. You can also engineer the animal to have the metabolism that a human might. That can be tricky though, because sometimes the animal won’t survive under that situation.

But I would say in general, by knowing the genes that underlie whatever you are studying through these comprehen-sive studies, you can then go into an animal and try to make it as similar as possible to the human situation.

How does your work help in this context?In the past we developed something called cell-based microarrays for doing library screens. We now do everything in pooled fashion, where everything is barcoded.

I think that what's valuable about these very comprehensive large-scale screens is they give you a very complete picture. And then that lets you make quite informed decisions as you go into the animal model, and comparing the

human genome to that model, really help you to be smart about how you design your experiment.

What single change do you think would have the greatest impact on our approach to modeling diseases for drug discovery? You know one of the beauties of animal models is that you have basically isogenic animals. You’re basically working with clones of each other. They are heavily inbred. The challenge of that is that it captures relatively little of the diversity of that animal and by extension perhaps little of the diversity of humans. And so one thing that’s going to be important to incorporate into animal studies is to capture more diversity, to therefore have more chances of under-standing when you go into humans, which are inherently much more diverse in their genetic background.

Additionally, I think you know one of the challenges always is that one can understand a mechanism in quite a deep way, but then once you go into the organism and that mechanism participates in many, many different tissues, doing many different things, and then we have a drug that would act on all of these tissues—the story becomes more complicated. That, I think is a major challenge.