Two years ago, I interviewed Alex Kostic, who was then a postdoctoral fellow at the Broad Institute of MIT and Harvard exploring the microbiome’s connection to type 1 diabetes. His work studying children in Finland and parts of neighboring Russia showed that the microbiomes of children with type 1 diabetes were drastically different from the microbiomes of those without the disease. Now Kostic is running his own lab at the Joslin Diabetes Center in Boston, and investigating questions such as whether or not the changes in the microbiome are causing disease or are merely a symptom of it. He is also looking at the microbiota of the Joslin medalists—those who have lived with type 1 diabetes for more than fifty years. About 20-30 percent of those medalists still produce a trace amount of insulin, and Kostic is trying to understand whether that insulin production can be explained by differences in those medalists’ microbiomes.
Jessica Dunne, director of discovery research at JDRF, which is funding Kostic’s study of the medalists commented on the new lens that Kostic is bringing to the study of T1D. “We’re often thinking about how the microbiome is affecting the immune system. He took a different tack that we haven’t seen anyone take: what’s the role of the microbiome on beta cells? To me, it’s a completely novel approach; it’s very out of the box thinking in terms of how the microbiome can affect residual insulin production in type 1.” The question is, she continued, “can we reawaken those sleeping beta cells by modifying the microbiome?” Dr. George King, who oversees Kostic’s lab as Joslin’s chief scientific officer, noted how integral the microbiome is now thought to be in terms of the development of type 1—and therefore how important it is to understand it through the specific lens of this disease. “A lot of immune cells are located in the gut, and they can affect the rest of the body with a variety of problems. We believe this is potentially one of the major reasons why people develop type 1 diabetes,” King said. King points out that Kostic’s “beautiful studies” on the Finland and Russia populations helped clarify this association, and that with his training in immunology, bioinformatics, and computational biology Kostic’s lab is well positioned to further the understanding of the interplay between the microbiome and T1D.
I spoke with Kostic in his lab earlier this fall. Toward the end of the conversation, we were joined by Jacob Luber, a PhD student in the Bioinformatics and Integrative Genomics Program at Harvard Medical School who pairs computational analysis with new sequencing technologies to study the microbiome. He also has type 1, so the work of the Kostic lab is of both personal and professional interest.
It seems like the fact that you now have a lab at Joslin is a signal of how seriously the type 1 diabetes community is taking the microbiome as an approach to the disease.
Yes, exactly. I think over time there will be more labs that focus exclusively on this connection, because it’s so rich and so dense, both the basic science—understanding basic autoimmunity and how the microbiome plays a role—and the very strong clinical promise of being able to change your microbiota and potentially see therapeutic results. So I think it’s a very promising area, and as more research comes out the more promising it looks.
Can you tell me about some of your research on the microbiome and type 1 diabetes in mice? What are you finding out?
I am a really big fan of the mouse models, because while it’s not the same as human type 1 diabetes for sure, as far as mouse models go, the NOD [non-obese diabetic] mouse is as good of a model of human disease as it gets. There are about fifty genes that contribute to diabetes in the mouse, and almost all those are shared with human genetics. Just like in the human, it’s spontaneous. And it doesn’t happen to all the mice—it depends on the facility where they are. At Joslin, something like sixty percent of female NOD mice will develop diabetes, and about ten percent of the males.
Is that because the genetics of the mice differ?
When we’ve studied the microbiome in human cohorts [in Finland and neighboring Karelia], we’ve seen this very pronounced drop in diversity, which happens about a year prior to diagnosis with type 1 diabetes. For me one of the most frustrating things about human cohorts is that you can’t make a direct inference about causality. Because all you have at the end of the day is an association. You know that the microbiome changes prior to onset of disease, but you can’t say that this change in the microbiome is causing disease. And so that’s why as a next step the mouse models are so important.
What we’re doing is introducing the microbiota from individuals who remain diabetes free and introducing microbiota from individuals who develop disease, and testing whether this changes the incidence in the mice. And then the critical thing is being able to follow these mice over time. The nice thing is, the time frame is so short. At ten weeks we can do an MRI scan of the pancreas, and at that point we can make a very good prediction about whether that mouse is going to develop diabetes or not.
What do you see in the pancreas?
First you inject magnetic nanoparticles into the blood and they accumulate in the pancreas. If the vasculature is more permeable to the magnetic nanoparticles, then this is a strong predictor that that mouse is going to develop type 1 diabetes way before the onset of symptoms. You can show it in humans as well, and this seems to be a promising kind of early diagnosis.
What will you do with the information?
We’re trying to take what we’ve learned from the human cohorts and dive into the mechanisms in much more detail, using mice. We’re trying to close that loop of causality and answer whether these changes in the microbiome are actually causing disease or whether they might just be nothing more than diagnostic factors. If it is causal, then we’d look at whether changing certain aspects of the microbiome can stem the development of the disease.
So what would be the next step in testing if this goes the way you think it might?
Then there are a few different scenarios. In other diseases fecal transplants are a viable treatment and work well, but I don’t think that’s ever going to be a good idea in type 1 diabetes because the risk of developing the disease is just so low, according to genetics. Even if you have the highest genetic risk the chance of developing the disease is maybe five percent.
But what about the people who have antibodies but haven’t yet developed the disease?
That’s a population where maybe it does become a reasonable question. But the thing is, and maybe this is because I’m a scientist and not a clinician, but transferring a whole microbiota as a black box and not understanding anything about how it works or what it does, with the risk of potentially transmitting infections and other things unknown, isn’t so palatable to me. Maybe in the future it could become a reasonable path. But I think it’s a question of refinement. One level is giving a whole microbiota transplant. The next step is probiotic cocktails—finding which bacteria are really key in preventing onset of disease, and that involves understanding some of the mechanisms that go along with the disease. And this is the trajectory of the science over the years. Ideally when you understand the mechanisms you can start understanding the small molecules which are naturally produced by the microbiota. These small molecules interfere with our immune system and possibly with other aspects of physiology that prevent diabetes, at least in the mouse. So understanding what those small molecules are mean that you can deliver them in a much safer way in the form of very specific therapies.
When we talked before you were talking more in terms of prevention. Are you now also thinking in terms of therapies?
Yes, I am a little bit more optimistic about therapies than I was two years ago. Although having said that, I guess right now, I think the first major hurdle is understanding the triggers and making some progress in understanding those mechanisms between the microbiome and the immune cells. Understanding exactly what those interactions are is what’s going to give us hope of a much more specific therapy that would lead to reversal. But to be completely honest, I don’t have any sense of what that would be exactly. I just do think it’s possible down the line.
You talked before about the way science is funded in the U.S., in short cycles of grants, which can make it hard to develop a cohort to study over time. Will your association with the Joslin clinic affect the type and length of studies you’re able to design?
Yes, and that’s one of the really exciting things about the clinical research center at Joslin. I think it’s the biggest center in New England and one of the biggest in the country, so that means there are a lot of new patients coming in who for the most part will help promote the research however they can. With JDRF funding we’re studying a really interesting question in the medalists—people who have had type 1 diabetes for over 50 years. For some unknown reason, a subset of the medalists, maybe twenty to thirty percent, are C-peptide positive, meaning they produce some residual amount of insulin. This is work that’s been done for many decades by Susan Bonner-Weir, who’s just upstairs, and is a remarkable world expert on the beta cell and understands so much about its biology. The interesting thing about the C-peptide positivity is that it waxes and wanes; it’s there in some people for a few months, and then it goes away. We’re trying to understand whether that could be something that could be explained by changes in the microbiota over time.
So what we’re doing in our lab is sequencing the stool samples from the medalists, and we take the blood samples after an immune tolerance test to see which ones are C-peptide positive and which ones are C-peptide negative. We’re trying to understand whether there are any differences at all in the microbiota of these individuals, and if so, very specifically we’re looking for this one gene [that we think] causes the beta cells to expand, and whether there’s a difference in abundance between these populations. The hypothesis is that we might be able to see a difference either in the abundance of the gene or in the type of that gene that can promote beta cell expansion. So now, in addition to getting this cohort of the medalists, we’re doing in vitro and in vivo assays where we’re putting this gene’s protein onto cultured rat pancreas cells, which really neatly form islets in culture, and we want to see whether we can change the amount of beta cell expansion. We’re doing the same thing in vivo in live mice, and trying to see whether there’s evidence of this in the medalists as well. It’s definitely a high-risk project, because there’s a good chance we’re not going to find anything. But if we do find some evidence, then this would be a perfect example of a rational probiotic where we use something that we know enhances beta-cell expansion. So it could be a direct therapy; it could be something that we could give to people who currently have diabetes.
So are you also comparing the microbiomes as a whole in the medalists?
Yes, we’re also trying to discover other differences in the microbiome that may exist between these groups, that may also serve to explain why one group is C-peptide positive and the other is not. The long-term strategy for me is to make the discoveries in the mouse models and then once we have those to start generating the cohorts here at Joslin to start testing what we’ve found.
When we were talking before I think you said that the study of the microbiome using modern techniques really started less than a decade ago. It seems like we’re really still at the beginning stages of understanding it, because it’s so enormous and there are so many different aspects to it.
Exactly. It’s really early, and we’re really far away from being able to understand the whole system. If you use the human genome—which was sequenced around 2003—as an analogy, there you have 20,000 genes. And right now, 20,000 genes is not that hard of a problem. More or less every single one of those genes has been studied in detail by some lab or other around the world. The big challenge with the microbiome is that it’s estimated to be more than 10 million different genes, not in each person but across the human collective gut microbiome. In one person the number of genes might be around 100,000. And so the subset of those genes that we have any information about is less than ten percent. That really limits what we can do with them because they haven’t been studied. The only genes we know well are the ones that are shared with the common lab bacteria that people have studied for decades: bacteria like E. coli and Bacillus subtilis. But most bacteria in the human gut are not culturable, or at least are really hard to culture, and haven’t been studied, and that means that we just don’t know what their functions are. So that’s one of the challenges in the microbiome going forward.