The Realm of Facts:
A Conversation with Nathaniel Heintz

A member of the National Academy of Sciences and former investigator of the Howard Hughes Medical Institute, Dr. Heintz has devoted his career to elucidating a fundamental problem: how the molecular identity of cells relates to health and disease. His laboratory has pioneered groundbreaking methods for profiling different cell populations in the brain, revealing new layers of complexity in the nervous system’s structure and organization. That work has enabled neuroscientists to understand the brain as an ecosystem of highly specialized cell types, laying the foundation for the development of equally specialized therapeutic strategies against neurodegeneration.

In this conversation, Dr. Heintz reflects on the evolution of his scientific vision, the challenges of translating biological insight into treatment, and his cautious optimism about the adoption of artificial intelligence in the lab. He also discusses the importance of human brain research and the need for grounded, testable hypotheses as the key to understanding Alzheimer’s disease—and ultimately finding a cure.

You are a biologist and a neuroscientist. Which one came first?

It feels like I was always interested in science, though it’s hard to know what a scientist actually does when you’re young—I certainly didn’t. My first real scientific experience was with electron microscopy, which I discovered in college. All of a sudden, I was able to see biological structures at extraordinary magnification—thirty thousand times or so. And that completely changes the way you think.

When you look at something like the mammalian intestine, or the brain, or any complicated organ at that scale, you realize how incredibly beautiful and organized these things are. You understand that beauty is not something confined to the shape of a flower: it’s part of how cells themselves come together. That made a deep impression, drawing me toward the study of how these structures are built, what genes are involved, how cells and tissues develop their particular forms. I also became interested in how cells divide and proliferate, and how molecular biology controls those processes.

Would you say that experience put you on a path to your current research?

In some ways it did. I spent the first ten years of my career at Rockefeller studying cell division, cancer, and the biology of cell proliferation. I wanted to understand the regulation of division at a fundamental level, as well as how those processes go wrong in disease. But once the key mechanisms were established, I found that my deepest curiosity ultimately lay somewhere else.

I began to shift toward neuroscience when I became interested in the cell types that make up the nervous system. I was fascinated by the fact that the brain requires an almost unimaginably complex set of events to develop and function; and it was obvious that hundreds or even thousands of distinct cell types were necessary to enable those events. Yet nobody actually knew what distinguished different cell types in the brain at a molecular level—not even how many genes were expressed in each one. So, I decided to investigate those differences and what they meant for the function of different cells. That’s how biology eventually led me to neuroscience.

Was that also how you became interested in exploring the therapeutic applications of your research?

Yes—as we discovered the properties of different cell types, my team and I also started thinking about how they might be targeted in therapy. We developed a technique called FANSseq, which uses fluorescent protein markers to identify different kinds of cells and visualize how they’re organized in the brain. We also focused on translational profiling, which allowed us to see exactly what genes are expressed and what proteins are produced by specific cell types. And that changed everything.

We learned that every neuron expresses roughly twelve thousand genes—about half the genome—and that several hundred of those genes, in their particular combinations, are highly specific to different types of neurons, determining their different functions. This implied something truly stunning: namely, that it didn’t make much sense to talk about “neurons” as though they were all fundamentally the same. In reality there are hundreds, maybe thousands of different neuronal types in the brain. They differ in structure, molecular composition, and response properties. And that means that they can be targeted pharmacologically with a high degree of specificity.

That insight led us to work on circuit-based therapeutics, collaborating with Biotech companies to identify new drug targets for neurodegenerative disease. One of these drugs is now in phase 3 clinical trials for Parkinson’s disease, which shows that this kind of work has serious potential for connecting basic science to real-world therapeutic advancement.

How has that work coalesced around Alzheimer’s specifically?

When Paul Greengard was alive, Alzheimer’s was mostly his domain here at Rockefeller. My team was working on other neurodegenerative diseases, like Parkinson’s and Huntington’s, and what eventually got me involved was the realization that both in those diseases and in Alzheimer’s disease, neurons don’t all die at the same time. Some neurons are vulnerable to neurodegeneration from very early on, whereas others—sometimes right next to them—are more resilient. That immediately raises the question of what it is about certain cells that exposes them to disease, and what protects their neighbors.

These are late-onset diseases. That means that they unfold over decades in the human brain, often for a long time before any visible symptoms appear. So, to understand what makes certain neurons more vulnerable, it’s essential to work directly with human brain samples—and the last fifteen years have made an incredible difference for our ability to do that. Using the techniques we developed, my team and I have refined methods for profiling tens of thousands of neurons of the same specific type at a time. This allows us to examine brain tissue donated by Alzheimer’s patients in great detail, and to piece together the molecular disturbances that characterize the earliest stages of the disease in each cell type.

Did you work on that with Greengard before succeeding him as director of the Fisher Center Lab?

Paul and I were friends for many years and our scientific partnership began much earlier, through the study of the two major cell types in the corpus striatum and basal ganglia. Paul wanted to characterize each cell type molecularly, and my lab’s techniques were perfect for doing that. It was a great collaboration, full of fun and energy. It was also an opportunity to refine some ideas about developing new drugs for neurodegenerative disease. So, to some extent, Paul helped me explore the link between brain biology and therapeutics.

If Paul was here today, he’d probably suggest that I continue to investigate the role of amyloid-beta in Alzheimer’s disease, as he did. I don’t disagree—that remains crucial. But there are plenty of people who are already studying that, and I think the field also needs to focus on earlier events in the disease, which take place long before the formation of amyloid plaques and other pathologies. Our goal is to explore the applicability of the experimental frameworks that my team has developed for other neurodegenerative diseases to the study of Alzheimer’s. That allows us to introduce an approach that has already proven very effective for the study of neurodegeneration.

How would you characterize your current view of Alzheimer’s from that vantage?

My view of Alzheimer’s centers on the elements I’ve discussed: the vulnerability of certain brain cells to the disease, the molecular disturbances through which that vulnerability manifests, and how early those disturbances can be detected and targeted for treatment. Amyloid plaques and tau tangles are an important part of Alzheimer’s pathology. But the question we need to answer, from a biological standpoint, is why plaques and tangles form in the first place. What happens in the cell so that the genes it expresses, the proteins it produces, and the molecular pathways it uses change as the disease emerges?

This perspective orients all our work toward potential therapeutic applications. Right now, for instance, we’re putting together our findings across hundreds of thousands of brain cells, to identify which molecular pathways are perturbed in the earliest stages of the disease. That will enable us to understand how those pathways might be regulated, or even restored. For Huntington’s disease, this approach has allowed us identify molecules that could be targeted pharmacologically. And for Alzheimer’s, we have identified a key pathway and we have a lot of data supporting a hypothesis for how it might be regulated. It will take some time to validate that hypothesis, but I think the data are compelling.

Do you foresee challenges to that progress?

You’re always going to face challenges when dealing with human brain samples. As I said, human brain tissue is invaluable for this kind of research—there’s no other way of observing how the disease develops in humans—but it’s also extraordinarily difficult to work with. You get a very limited amount of usable tissue from each sample, first of all—and within that tissue, only a small subset of cells are the ones needed for our studies. That material disappears very quickly once you start working with it, because you can’t use the same sample too many times. So, a very large number of experiments must be performed with a very limited amount of material.

The challenges are also the reason this work is essential, however. Precisely because the material is difficult to handle, every good dataset extracted from it is extremely valuable. The biology of Alzheimer’s disease is highly specific to the human brain, and so the insights you gain from studying it this way are more relevant than anything you’re able to get in other ways. Although it’s taxing, we have to learn to work under these constraints. It’s just part of how the field will continue to make progress.

Has artificial intelligence been of any help for navigating those difficulties?

It has. We’ve initiated a number of collaborations with the Fisher AI Platform here at Rockefeller, and one of them helped us refine a new imaging technique that was used for a study recently published in Neuron. It’s clear that AI is working well for the study of brain anatomy, therefore, and I’d say that in general it’s very good at identifying small differences. When you’re able to look at two situations and see specifically why one is pathological and the other isn’t, then the question becomes an operative one—how do I fix this situation?

Where I am less certain is in the interpretation of data to form hypotheses. AI may be able to analyze large datasets in useful ways, but from what we’ve seen so far, you can’t simply feed it the data and trust whatever answer comes out. Data quality, format, and their biological meaning are way too variable, so you need trained researchers to finalize the results.  In this sense, AI reminds me of excessively theoretical approaches to science. The idea of relying on large conceptual frameworks to explain what you’re studying has never convinced me—not only because biological problems are way too complicated, but also because I don’t think it’s necessary. With Alzheimer’s, for instance, you don’t need to understand the entire operation of the brain to come up with options for treatment.

The basic problem, as I said, is identifying the differences between a normal situation and a pathological one. That is a tractable scientific question, which can be answered by testing hypotheses until the mechanism you want to fix comes into view. Biology is complex, but it is not mysterious. It belongs to the realm of facts. And facts require a nuanced understanding that cannot be delegated to theory or technology because it ultimately comes down to human insight and expertise.

Where do you see the field going in the next few years?

I think the future of Alzheimer’s research depends on two things. One is openness to multiple ideas. If you gathered a group of Alzheimer’s researchers around this table right now, they would propose four or five different hypotheses for tackling different aspects of the disease. Alzheimer’s is too difficult for the field to rely uniquely on any one approach, unless and until that approach clearly yields the best therapy.

The second thing is being realistic about what we can do—which could mean being more optimistic. It’s important to remember that effective therapies don’t have to correct the primary pathology. Sometimes, you can do something highly beneficial by stimulating or modulating a brain circuit in a way that improves symptoms, even without directly impacting the primary disease. A good example of this is dopamine therapy for Parkinson’s disease, which doesn’t reverse the degenerative process but reduces symptoms like tremors and stiffness, improving a patient’s quality of life.

That’s an important lesson. If the way in which we study Alzheimer’s leads to therapies that provide relief, that matters enormously. It doesn’t solve everything. But if you can make a damaged nervous system work better, that deserves to be pursued as rigorously as any direct disease modifying strategy. Because it may buy time. It may protect things like a patient’s dignity. And it may help us understand the disease better. That may not sound very glamorous, but it’s how science operates. And in a field like neurodegeneration, where everything’s at stake—literally—for all of us, I think being realistic matters a great deal.