By Rachael Stickland & Nils Muhlert
Professor Helen Mayberg is a pioneer of neuroimaging and neurostimulation for depression. As a behavioral Neurologist she has helped to identify the brain circuits implicated in mood disorders, and then developed and refined effective treatments based on deep brain stimulation. She is a member of the National Academy of Medicine, The American Academy of Arts and Sciences and the National Academy of Inventors. As a founder member of OHBM we found out about her work, her experiences of seeing impact statements become reality and about holding on to the ‘OHBM train’.
Nils Muhlert (NM): I'm joined today with Professor Helen Mayberg, who is a professor of Neurology at Mount Sinai, as part of the OHBM oral history initiative. First, can you tell us how and why you became interested in neuroimaging?
Helen Mayberg (HM): I was a neurology resident in the early 80s. Imaging was in its infancy. In medical school, in the late 70s, we had our first CT scanner. MRI was relatively new during my residency training at Columbia, and it was an important diagnostic tool. I planned to train in behavioral neurology in Boston with Norman Geschwind. But in my last year of residency, he passed away suddenly, so I needed a change of plan.
My change in direction to imaging as a focus for further training was the result of a lucky coincidence. One of the first year neurosurgery residents on my team had just come to New York from Baltimore. He had written one of the first papers characterizing opiate receptor subtypes in the brain, and told me about this new research imaging method being developed at Hopkins where you could image chemistry in living people using positron emission tomography. I had always been interested in neurochemical abnormalities in psychiatric disorders but there was no way to study that directly in humans. Despite my intense interest in severe mental illnesses, I didn’t find the training in psychiatry to be a good fit for me, so I pursued my interest in behavior via neurology training. It wasn’t a perfect fit, but neurology seemed a better choice for clinical training than psychiatry where I just never felt comfortable with their vernacular. While it was still a stretch to understand chemical mechanisms of behavioral disorders it did provide foundation for thinking about structure-function relationships in the brain, an approach that really wasn't applied yet to psychiatric syndromes. So suddenly, here was this new scanner that might provide a way to do what I wanted: assess regional chemical abnormalities in patients with mental illness.
I started as a research fellow at Hopkins in the Nuclear Medicine Department, learning PET scanning in 1985. I was learning basic methods to map and model various neuroreceptor systems - mostly opiate receptors, but with some projects involving dopamine and importantly serotonin systems. The lab I was a part of did little to no behavioral mapping studies; it was a dedicated chemical neuroimaging lab. There was very little work on blood flow or glucose metabolism except as ancillary maps for receptor studies. At the same time, in LA, John Mazziotta, and Mike Phelps were working with glucose metabolism to map abnormalities in a variety of neurological disorders. At Wash-U, Marcus Raichle, Peter Fox, Mark Mintun and their colleagues were developing methods for behavioral mapping using cerebral blood flow. Nora Volkow was using various methods at the Brookhaven Labs. There was a relatively small group of teams with PET scanners worldwide that developed specific niches of expertise using this technology.
Whilst working within a chemical mapping group, I was reading about other methods and as my work evolved it was clear that my questions required tools that I didn’t have access to. Because there were so few groups, we were a very small and interactive community. Before OHBM, this type of research was discussed at the Cerebral Blood Flow and Metabolism meetings. The neurologists, like John Mazziotta, Marcus Raichle and Richard Frackowiak, would go to the Neurology annual meeting, and we’d have our own imaging sessions there. You’d also see different imagers at various meetings--Society for Neuroscience, American College of Neuropsychopharmacology, Society for Nuclear Medicine. The PET community was tight and continued to grow but it was relatively insular as there were few research PET scanners as you needed a cyclotron.
There was a point where the questions I wanted to ask about depression required methods that were not the focus of our lab. Peter Fox was setting up a new research imaging center in San Antonio; I had met him at a Cerebral Blood Flow meeting and heard what he was up to. It was a method that I didn't know but that would provide a set of tools to take my depression research in a new and potentially interesting direction.
In 1991, Peter invited me to be a founding member of this fledgling new center, so I moved from Baltimore to San Antonio in Texas. I was part of his team, setting up a research lab as part of the brain imaging group. We were very small, and focused on PET scanning but I was no longer doing any neuroreceptor mapping studies. By this point, I'd moved to studying depression exclusively. I worked as a clinical neurologist but did my research with collaborators in psychiatry using imaging.
In the early to mid nineties, The Research Imaging Center (RIC) was host to the original Human Brain Mapping workshops. As part of these workshops, there were ongoing discussions of how to develop common platforms to share data. The RIC team started compiling spreadsheets of coordinates of brain activation findings from the literature. These were the early days of imaging meta-analyses, done by hand. While the work was slightly peripheral to my own studies, I couldn't help but become involved. So I received an education in the world of mapping beyond my own area and saw a style of thinking that was way ahead of its time. We now take so much for granted with our contemporary approach to big data and data sharing. It was laborious work in those days.
NM: It sounds like a very exciting time - certainly within that lab.
HM: It’s funny. For every scientist you can never know when you're in the middle of something important, if anything you're doing will have long-term traction. We delude ourselves, write grants, make statements about the potential impact and how significant we think our work is. In fact, it's only when looking backwards that we can actually see how it all evolved. It's hard sometimes to be reminded that during my time in medical school, CT scanning was new. We studied dead brains, we had anatomical atlases, we had white matter maps from studies in nonhuman primates that we used to mentalize how brain regions were connected to each other. We mentalized a connectome in our head, by piecing together these various studies. We didn’t yet have access to multimodal mapping. I don’t think I could have even conceptualized such methods.
What was a paradigm shift of imaging, was that you could directly test your hypotheses rather than simply make interpretative inferences from pathology or animal models of behavior. Not only did technology allow visualization of the brain in action, but with time the choice of methods greatly expanded. My neurology professor in medical school said: “pick a topic, not a method.” At that time I was learning PET scanning. He said, “you'll reach a point where your current methodology no longer allows you to answer your question. So you’ll learn new methods and tailor your questions as methods evolve.” As a clinician and not a technical or methods developer, that was incredibly important advice - don’t just learn a method for method’s sake, but learn a method in service to your clinical question. That's been my approach since then. So what I know the most about, PET scanning, is something I don't even do much anymore. But the use of imaging as my experimental method has never changed, I have just learned to adopt new imaging tools to best address my next depression study. One can now pick and choose, and with that range of choices you can really go deep to answer your own questions.
NM: On that note, one highlight of your career is that you’ve helped identify the role of Brodmann area 25 in basic drives affected in people with depression. When did you start focusing on this area and how did that come about? Was it as a result of these meta-analyses?
HM: We didn’t go looking for it. We just followed the data and there it was. But it wasn’t on our radar with our early studies, which started with examining post-synaptic dopamine and serotonin receptor in post-stroke depression patients. These studies were complemented by studies of opiate receptor changes following electroconvulsive therapy as a model of epilepsy, to studying resting state abnormalities in basal ganglia disorder patients with and without depression. Our goal was to test the hypothesis that, regardless of etiology, there was a common set of regions affected in patients with depression. We were working to define a depression circuit. As a common pattern of limbic-cortical abnormalities emerged, we felt it was reasonable to move to study primary depression where clinical heterogeneity was well described. I presented that set of findings in 1989 at one of the Cerebral Blood Flow & Metabolism meetings. We did the next natural experiment: how the abnormalities change with treatment.
For the most part, we found what everybody else was finding in depression, low metabolism in the frontal lobes. When we treated people, the frontal activity normalized--it increased. By using the statistical methods Peter Fox had brought to San Antonio from Wash U-- change distribution analysis-- combined with new computer algorithms and higher resolution scans, we could further examine the whole brain instead of predefined regions of interest. I remember analyzing a specific set of data of depressed patients studied before and after successful antidepressant treatment. I was looking at the statistical change maps and I figured there must be some sort of misregistration artifact. I kept looking at the pictures. It was midnight and, all of a sudden, I realized the ventral parts of the brain were showing decreased activity while dorsal parts of the brain were showing increases. I squinted my eyes and looked at what I was seeing: there were brain regions that weren't abnormal at baseline that showed decreasing activity as people got better. When I looked up the brain regions, I found we were in the subgenual cingulate. I had my Talairach atlas, and I'm looking up with the ruler where I am - it was all done by hand, and I thought ‘what the hell is this BA25?’
Actually, the Talairach atlas I had misidentified it. So people thought for many years that it wasn't really in Area 25, and I say look I'm just following Talairach. I went to see who else had seen anything in this region. I found changes in this region in a study by Jose Pardo on mood induction. We tried looking at correlations, to see which part of this multi-node network went with which symptoms of depression, and I couldn't separate out the mood from the attention symptoms with the data I had. Peter suggested “Let's do a blood flow scan following a mood induction.” The intention was that if we induced a negative mood in healthy volunteers we would determine if you could dissociate the presumed limbic-emotional regions from the cognitive cortical regions in this presumed depression network. To our surprise, mood induction did indeed reveal limbic activations and they were in Area 25, but it also decreased activity in the prefrontal cortex--the same regions identified in the depressed patients. Area 25 and the prefrontal cortex were inversely correlated with each other in both experiments--depression recovery mapped over six weeks and sad mood induction over 2 minutes with the magnitude of behavioral changes correlating with both regions. In essence, our hypothesis was just wrong; we couldn’t induce solely change in limbic regions by focusing on mood. The two systems, limbic and cortical, could not be separated. Obviously a simplistic notion if viewed through today’s use of graph theory and dynamical modeling approaches to time series data. But at that time, these simple experiments using blood flow and glucose metabolism PET gave us one of our most important insights--these regions were yolked and worked as a synchronized limbic-cortical circuit to mediate the interaction of mood and cognition.
This pattern of reciprocal changes involving midline and lateral cortex regions was new. Today, we would look at this pattern and immediately see the default mode and executive networks displaying their typical anti-corrrelation with each other. But then, in the mid 90s, that concept was just developing. We looked at it thinking ‘what are these regions and what do they do?’
Area 25 had very little written about it. It was described in the animal literature as a visceral motor outflow area and not necessarily a mood area. You could even find references to its homologue in lizards, as it's a very old, highly conserved, part of the brain. I would get into fights with rodent anatomists by asking about the rodent equivalent: “is it infralimbic? Is it prelimbic?” Lots of opinions as to whether or not it is even a good idea to attempt to match rodent and human prefrontal cortices, if one is really interested in studying depression, a uniquely human clinical construct. That was sort of a turning point; if I wanted answers I needed to really learn to read the tract tracing studies done in nonhuman primates and learn the connections between regions by looking at combinations of anterograde and retrograde studies. Little did I know that I was laying the foundation for future work that would rely on maps of structural connectivity defined using DTI.
NM: And then you later moved into intervention studies, where you used targeted deep brain stimulation (DBS) of region BA25 to see how it affects symptoms. What was that like - setting that up and seeing the results from those first studies?
HM: In all honesty, I became an interventionist almost by accident. I wasn’t a trialist; I merely used treatments as probes to better understand depression and treatment mechanisms. I spent the first 20 years basically trying to prove that depression was a circuit disorder, first by identifying the nodes, and then the connections and making inferences about causal relationships using changes with various kinds of treatments. There became a point in Toronto where, findings in Area 25 were so consistent across all of our treatment studies, that we hypothesized that if you didn't downregulate this region then people didn't get better. It seemed to be really at the center of the antidepressant treatment response.
The idea to target Area 25 with DBS for treatment resistant patients was highly influenced by the neurosurgical literature and the evolution of ablation to DBS for Parkinson’s disease. The leading theory about DBS mechanisms at the time posited that high frequency stimulation resulted in a local depolarization block. As we had consistently demonstrated that effective antidepressants decreased or blocked activity in Area 25 and if you couldn’t block it you didn’t get better. We followed that logic to hypothesize that if you can't talk or drug or shock it down, maybe you could block it with targeted stimulation delivered very precisely at this node in the network.
Everything I knew about connectivity (even though at that point, there were no tractography tools available to us, so implied connectivity) was that if you downregulated a region such as BA25, maybe you would also get disinhibition of regions it was connected to. The DBS technology at this point, in 2002 or so, was well established and readily available. I had a surgeon that was willing to test my hypothesis. It was actually very much an imaging-driven idea. If I hadn't been doing imaging, would I have even thought about it? Probably not, but I was in the right place at the right time. We had the maps that pointed to a putative DBS target for treatment resistant depression, a surgeon with extensive experience with DBS for Parkinson’s disease and a team of investigators willing to learn about DBS and manage this group of extremely ill patients with this novel intervention. It was in some ways a natural next step for our ongoing studies. So that's why we did it - because we could. But the logic was basically built on that first mood induction depression recovery finding.
NM: It's as we discussed before - the impact statement becoming true over time, where you think ‘what areas are involved?’, then ‘what can you do about it?’ And here you've got a great example. You were involved in the creation of OHBM. What was your role?
HM: Well, mainly I was involved because I showed up. I attended the first meeting in Paris, which was a natural extension of the Cerebral Blood Flow meetings I had been attending since starting my post-doc in PET imaging. With time, I became an officer; I was elected secretary in 2000, and served from 2000 to 2003. It's interesting that many of the originators of OHBM were clinician-scientists. Several of the key drivers - Mazziotta, Fox, and Evans all in North America - had a grant together, and joined forces with many key thought leaders and teams worldwide to make it happen. A shared vision. Again, being in San Antonio with Peter, I had a front row seat to the evolution of the organization. Timing and opportunity are a common theme here.
When I think back, how could anybody not participate? It was happening all around me. So you get on the train and hold on and see where the journey takes you. We all had a ringside seat, and saw an idea grow and mature. Like any diverse scientific community, building an infrastructure that requires not just expertise but buy-in and cooperativity is challenging. But like any democracy, there was a lot of trial and error and compromise seeing what worked, what the community, the stakeholders wanted; it evolved by taking great ideas and giving them space to evolve and mature. What was great was it was very inclusive - methodologists, clinicians, statisticians, engineers, all topics, all scan types, multimodal approaches, new science, courses, and great opportunities for networking. The multinational and multidisciplinary collaboration that established OHBM has continued to define it and foster its unique position among imaging meetings.
NM: And what have you found most rewarding about your experience in holding on to that train with OHBM over the years?
HM: Well, I've had the opportunity to collaborate with people world-wide and adopt a multimodal imaging approach to our team’s clinically-oriented research questions. Maximizing use of novel technologies is at the core of our work--with critical reliance on state-of-the-art engineering and statistics. OHBM is where I can always count on seeing the newest technical and analytic advances and where discussion is scholarly and collegial. Our own work is quite iterative, so it’s useful to see a new method used by others before jumping in ourselves. OHBM provides an important sounding board for our ideas and I have always found the meetings personally and scientifically rewarding.
OHBM has evolved beyond anything any of us could have imagined. Technological advances have been the critical catalyst but applications of the technologies have been important drivers. Perhaps I am biased, but imaging in one way or another has been at the center of many of the advances in neuroscience over the last 50 years.
NM: And what do you see as the most promising things that are coming out now?
HM: Like with anything, progress is not linear. Sometimes it seems like it's three steps forward, and then one steps backward or sometimes even sideways. I'm reminded of one of the first imaging meetings I attended prior to OHBM where we would sit and listen to thought leaders debate the advantages of their particular methodology. It was a curious sort of testosterone storm of statistical one-upmanship. It was as though one method had to defeat all others.
It has been fascinating to be part of our maturation as a field. Where our focus is on matching methods and technologies to a specific category of question rather than assuming one size fits all. How could it be otherwise? That's the natural evolution of any field… the first thing is you don't believe it, the second thing is it's obvious, then it evolves to be much richer because everybody starts to dig into working out the details.
Right now, I think we're going through a stage where there's so much data that we don't know how to parse it. We're at a time where doing experiments that people care about is expensive and hard. Early on, the focus was on ensuring the methods were valid and reliable. Like any broad field, people have different interests. I am grateful to know that there are people pushing the limits of the technologies and those using it to understand basic principles of brain function; the big data consortiums with multimodal data archives for general use are priceless resources for the community. As a depression researcher, I want to exploit the technology; ‘If I take on learning a new method, I need to decide if it's worth it.’ Then, ‘How is it going to help us test our next set of hypotheses?’ I don't think it's just because you're a physician that you want to do that. Everything is hard and time consuming - how do you have meaning in the way you spend your time, scientifically?
We're back to a plateau where we're “fighting” about ‘what's the right way to do it? Do we believe anything we know? Is it all an artifact? How do we replicate?’. We're learning that the brain is very adaptive. And when we think we control an experiment, we don't control it as well as we thought we did.
We're looking for big signals. We've gone from working with deforming the brain into a common space (which was to increase signal to noise and you didn't believe it until you saw it robustly across subjects) to trying to understand inter-individual variability (if you don't understand the individual, you know nothing). You can get dizzy, realizing these are natural evolutions. Everybody's right, just not at the same time.
This is the beauty of the OHBM culture: you can develop tools to answer the question in the way you want to. So for me, as a very specific example, we guessed based on a blob on a PET scan, where Area 25 was, stimulated with an invasive implant in that approximate spot, and made people better. It worked. And then we’ve spent the next 15 years trying to figure out what exactly we did, how to do it better, and why it worked.
Imaging has remained a key method towards these goals. For instance, Kisueng Choi in the lab had identified the critical white matter bundles that mediate the DBS treatment effects and developed tractography methods to reliably define the optimal surgical target in any individual patient. I love and use the data from the human connectome to test ideas, but at some point, we need to make a decision about an individual patient's brain. Where is the spot? Can it be visualized reliably? Can we hit it with millimeter precision? I’ve got to make a map for the surgeon to put the electrode where we say with a high level of accuracy. And then be able to show prospectively, that what we wanted to do, is in fact what we did.
I am envious, as we all are, of the amazing advances in circuit mapping techniques using cell-specific labeling, such as optogenetics, and CLARITY. While I can learn a lot from these exquisitely detailed maps in rodents and more recently non-human primates, I also need reliable lower resolution methods because I don't have the luxury of single cell stimulation. We inject a big amount of current into a pretty sizable brain area that contains a mix of many cell types and passing fibers. It remains a real mystery how such nonspecific stimulations work but it does. Obviously, more advanced methods will evolve. But for now, we work with what we have. That said, we’re always looking to see what new methods people are developing. When I was a kid, I used to hang out with my uncle who was a biochemist and nuclear medicine physician in the pre-PET era. He used low resolution detectors to measure radioactively tagged chemicals in the brain often injected through the ventricle during surgery and scanned later. The images were horrible, like looking at a fuzzy bowl of soup. You could measure changes in brain concentration of various compounds, but without the spatial detail. Still, it was really amazing we could do it at all. Now we're working to improve on 0.8 mm isotropic voxels. Looking to push the envelope further. All of this change in less than 50 years; I am just sorry my uncle missed all of this. He would have loved it.
NM: I could almost end there but one last question about your personal experiences attending OHBM. Are there any moments that stand out for you?
HM: There are so many. To hear the giants of imaging give the Talairach lecture is always a thrill. But I think like any meeting it’s the camaraderie, the openness of students who read your papers and want to get your opinion on their work; to both meet old and new heroes and to maintain relationships with colleagues over 30 years; to be able to sit down and talk or just hang out. I always enjoyed the grandeur of the big lectures and the rigor of the science as well. But it has always been the relaxed atmosphere that catalyzes new ideas and new collaborations. I can remember meeting the Oxford team -- Paul Matthews and Heidi Johansen-Berg and that short chat changed the course of my fledgling tractography work. I remember making sure to arrange meetings around world cup matches; Resting up for the dance party night; figuring out the train to Sendai. Just so many big and small wonderful memories.
NM: Professor Mayberg, I'd like to thank you very much for joining us. It was a fascinating insight into your experiences with OHBM.
HM: Well, thanks so much for including me. It's really an honor.