Randy Buckner

Randy Buckner Stephanie Mitchell/Harvard News Office

Randy Buckner is widely known for his innovative work on the functional neuroimaging of memory. A Howard Hughes Investigator, Buckner is a professor of psychology and member of the Center for Brain Science at Harvard University, and is associate director for the Martinos Center for Biomedical Imaging at Massachusetts General Hospital/Harvard Medical School. Previously on the faculty at Washington University in St. Louis, Missouri, he has collaborated extensively with John Morris and other members of the Alzheimer Disease Research Center there to study the overlap between normal memory networks and Alzheimer disease pathology. ARF talked to Buckner about the sharing of new imaging tracers, the default network, and other hot topics in AD brain imaging.

Q&A

ARF: Randy, you're a leader in functional imaging and the field of structure and function both in memory and in Alzheimer disease. Let's talk about how imaging studies that you and others have done over the past few years are affecting our views of memory and AD. But first, I'm curious about how you got started in imaging research?

RB: My interest in imaging stems back to when I was an undergraduate at Washington University. Some of the first human imaging studies were being conducted in St. Louis when I was a student by Marc Raichle, Steve Petersen, and colleagues. I was aware of the imaging methods as an undergraduate because it happened near me and because of a few lucky opportunities. Steve Petersen came and gave a guest lecture in my neuropsychology class and showed us the brain imaging maps that were part of his seminal studies of language. It was one of the most remarkable things I had seen.

I was interested in memory because of my coursework. I wanted to understand how we can think back to something we saw earlier or remember something from last week, and also why we forget. Here was a set of tools that allowed us to visualize mental activity in the human brain, and I became excited about the opportunity such methods presented for memory research. So I went over and talked to Marc and Steve. They encouraged me to come work in the imaging center and teamed me up with a then-young master's degree student, Jeff Ojemann, to work on a project studying memory in collaboration with Larry Squire at the University of California-San Diego. Their enthusiasm and support set the stage for the rest of my scientific career.

ARF: They were using PET technology. To ground us, can you run through the basic types of imaging that get applied in AD these days and explain their different uses?

RB: One main method that is important to the understanding of AD is structural magnetic resonance imaging, or structural MRI, which is basically a detailed picture of the brain's anatomy. We have known for a long time that regions of the brain involved in memory, including the hippocampus, are involved in AD, and over the last decade, some robust methods to measure the structure of the hippocampus have been developed by key individuals like Ron Killiany at Boston University and Cliff Jack at the Mayo Clinic in Rochester, Minnesota. Their work has shown that during the early stages of Alzheimer disease, this structure is particularly sensitive to atrophy. You can, in fact, predict progression of AD based on measuring the volume of the hippocampus.

Another important method is a variant of positron emission tomography, or PET, that measures brain metabolism—these are the FDG-PET methods used by people like Eric Reiman at the Banner Institute. These measures of brain metabolism are not structural but look at the overall metabolic rate in different brain regions. With FDG-PET, specific regions, including temporal parietal cortex, show reduced metabolism in the earlier stages of AD. In fact, individuals with genetic risk for AD show reduced metabolism in these areas years prior to AD diagnosis.

A third method, functional MRI, is used to measure brain activity. You can make brain activity measurements during many kinds of mental tasks. For example, you can place a person in the scanner, and taking advantage of indirect measures of neural activity that can be seen with MRI, you can make measurements of the regions that person uses during memory and attention and language. This has been done in healthy young adults to understand memory systems of the brain. Recently our laboratory and others, including Brad Dickerson's here at MGH, Reisa Sperling's at the Brigham and Women's Hospital, and Susan Bookheimer's at UCLA, have shown that in the earlier stages of AD, one can see functional changes that suggest the brain is compensating for disease processes.

A very recent advancement has come in the form of molecular imaging based on PET. You notice that I am repeating the names of some of the methods. These technologies—PET, MRI—each can be used in multiple ways to glean different aspects of brain function. I'm just giving you a laundry list of how they are being used for AD. PET molecular methods, spearheaded by Bill Klunk and Chet Mathis at the University of Pittsburgh, have recently allowed us to actually see the amyloid plaques that are an important correlate of AD, and this is revolutionizing the field.

So MRI methods help us see brain structure with increasing detail as these methods are advancing. PET methods visualize reduced metabolism in the brain in the earliest stages of AD, even in people genetically at risk. fMRI measures brain activity, and you can see changes in early AD that perhaps reflect compensation for the stress of the disease on the system. Perhaps the most exciting recent developments have come in molecular PET methods that visualize amyloid plaques in living people.

ARF: How about diffusion tensor imaging?

RB: One of our other methods I forgot to mention. Thank you for the reminder. We and many other laboratories use diffusion tensor imaging, or DTI, as well. This is a powerful method. If you tweak an MRI scanner in a different way, you can measure water diffusion, which is very fast along white matter pathways and axons. The brain is built primarily of the cell bodies of the neurons, the axons are the cables that connect them, and the white matter pathways are the bundles of these axons. They look white because many of them have a fatty coating around them for insulation.

With diffusion tensor MRI, you can look at the integrity of the white matter pathways in the human brain. There are robust changes in aging and also in AD, and these methods are exquisitely sensitive to see the disruption of white matter pathways.

ARF: With imaging, what is the earliest thing you can see that's useful for telling if someone is going down the path to AD?

RB: Atrophy of the hippocampus, changes in metabolism as measured by FDG-PET, and amyloid imaging, though amyloid imaging is the newest approach and needs still to be formally validated.

ARF: Will any of those be useful for measuring therapeutic efficacy?

RB: Almost certainly. They all could potentially be useful and are all being explored as biomarkers for drug efforts or more generally for other kinds of treatment procedures and preventive methods.

Looking to the future, I imagine that a number of different imaging approaches may be important for measuring the presence and progression of AD and other degenerative diseases. Diseases of aging are complex processes that progress gradually and influence different aspects of neurobiology at different stages of their progression. New brain imaging methods are enabling visualization of many distinct facets of disease processes. I believe we are witnessing the evolution of a set of methods that can potentially image many, if not most, of the key brain events associated with degenerative disease, tau pathology being probably the most difficult of them. If somebody asks, what is the palette of measurements one would ideally want to be able to disentangle what changes come first in the brain to cause dementia, which come second, etc., I think we are on the verge of having all of them.

Let's put this in context of where we are today. As a field, we have a long history of structural brain imaging and imaging brain metabolism using FDG-PET. Amyloid imaging made a big breakthrough recently. Now there are potential markers of inflammation, potential markers of microglia activity, which are the first responders when there's brain damage; there are potential markers of apoptosis (cell death), potential markers of neurogenesis (cell birth). Once we have that constellation of imaging methods, and they are accessible, then I think the next decade will bring a much clearer understanding of the course of brain degenerative diseases.

ARF: Are you working on any particular imaging techniques?

RB: I have been spending a great deal of my recent efforts on functional imaging techniques that measure brain activity at rest, as has been shown to be important to Alzheimer disease by my colleague Cindy Lustig at the University of Michigan and Michael Greicius at Stanford University.

Beyond my laboratory, many exciting new methods are being developed by chemists. New probe development such as those mentioned earlier that target apoptosis and neurogenesis are very important to our field. While I personally do not do probe development, I benefit from advances as does everybody in neuroimaging. The development, availability, and accessibility of probes drive a lot of the field. These are hard studies to do. They are expensive, and sometimes as these probes develop, they are not disseminated in ways that the many research programs that could make discoveries can actually do so. Opportunities are sometimes lost.

One success story is PIB amyloid imaging. It has been remarkable how amyloid imaging has disseminated to so many laboratories. There are a number of reasons for that, two being Bill Klunk and Chet Mathis, who have been remarkably generous and helpful. It is important for the broader neuroimaging community to understand how these new markers could be used, and also their limitations. Such understanding is happening rapidly for PIB, and I think that is going to speed science.

ARF: You published work recently that combined images from a lot of these techniques to make a hypothesis about the connection between Alzheimer pathology and the default network (see ARF related news story and ARF news story). That got a lot of praise from peers, but it was complicated. Tell us a little about the idea of the default network.

RB: Yes, that was a complex paper that encompassed many years of data collection and methods development. While the data sets are quite complicated, the principles that emerge from them are actually pretty simple. For a long time, my laboratory has been trying to understand neural pathways involving normal memory and normal brain function in healthy, young adults doing memory tasks and in other states. We have also simultaneously been using all of the brain imaging methods that we can to understand the cascades and disruption within the early stages of AD. The research is motivated by the idea that if we understand normal memory and brain function, and also what becomes disrupted in AD, the two lines of scientific exploration will eventually inform each other.

We have been progressing steadily across these two somewhat independent lines of research. Over time, they converged in a very interesting way that is reported in the paper. And the convergence is not what we expected. To tell the story of our findings and make clear their implications, I have to start at the beginning and discuss what has come to be known as the brain's default mode.

The basic observation is this: If you place somebody in the scanner and image their brain while they are musing to themselves or doing a passive task that is not causing them to engage the external world, a specific brain activity pattern will emerge. The activity pattern is distributed across the brain prominently, including memory systems and frontal systems. We refer to the specific regions as being part of a default mode or default network because it is this network of brain areas that we default to using when we are not otherwise engaged in a task. We have known for a long time—from dozens of studies, many laboratories, both PET and functional MRI —the default network tends to become active when people are just musing to themselves.

This is an extraordinarily interesting phenomenon. It suggests that while there are differences across individuals, and there are different things that we do when we are in our passive states, there still are commonalities in brain activity. We think we might be getting a serendipitous glimpse of the kind of brain activity that goes on much of the time as humans go through their day. In fact, it is quite possible, but unproven, that this is what our brains do most of the time.

The other part of the story emerges from our studies trying to understand memory pathways that link to the hippocampal memory system. We recently described a novel brain pathway that we think represents an important component of the memory system that interconnects with the hippocampus. The critical piece is that the cortical brain areas in this newly observed memory pathway overlap with the brain areas we default to when we muse to ourselves. It raises the possibility that, when we go into these passive states, we are actually defaulting to processes that heavily use memory systems. We do not know why that is. And I do not know if it is a conscious choice. Perhaps we are thinking to ourselves in ways that include memories. Alternatively, there may exist spontaneous replay of events in our brain that heavily relies on memory systems. Some work in rats by Matt Wilson and colleagues at MIT suggests that there is a great deal of spontaneous brain activity that replays recent events in ways that may consolidate information in memory systems.

So we have a default brain network that most of us use most of the time, or in passive states, and we know this common default network includes memory systems.

ARF: How does that relate to AD?

RB: The development of the PIB amyloid imaging method brought that relationship out. Human amyloid imaging was an important breakthrough because for the first time, it allows us to see the amyloid depositions across the living human brain in people in the early stages of AD.

What we observed was that the pathways you or I [i.e. people in mid-life] use during our default activity, which include memory systems, look very similar to where amyloid deposition occurs in 80 year olds. This leads to a very simple hypothesis: The activity states we are using much of the time throughout our lives, and the metabolic milieu they encourage, may set the stage for the eventual deposition of amyloid that relates to the toxicity in AD.

This observation surprised us, particularly because of prior emphasis on the medial temporal lobe, which is clearly part of this network. So we revisited the question of whether only the medial temporal lobe atrophies in early stages of AD. We asked if the cortical regions involved with the default network, and the regions where we see amyloid deposition, such as the precuneus and posterior cingulate, are also disrupted early in AD. And from the PET literature on metabolism, we knew they might be.

Specifically, these cortical regions show reduced metabolism in early AD. However, we do not typically think of them as atrophying early on in AD. So my long-time collaborator Avi Snyder of Washington University worked with us to develop a new set of methods to measure atrophy across the entire brain in AD in an unbiased way; that is, we did not pick a region like the hippocampus and ask, is there atrophy there? We developed methods that allowed us to just let the data tell us where the atrophy is across the whole brain. And while the hippocampus was shrinking prominently, so were many of these regions in the default network that are used during memory processes and show amyloid deposition.

We do not know if this is a complete picture. Probably not. But the data we put together suggests a novel way to view the conditions preceding AD. The data suggest that throughout our lives, there are prominent activity states in our brains that probably promote conducive metabolic conditions for eventual amyloid deposition and the toxicity in AD. When that happens, you see atrophy and metabolic disruption and eventually the clinical impairment that we know as Alzheimer's dementia.

As an aside, all of the results you see in this paper were realized because Bill Klunk and I were in the same session in a meeting at Bar Harbor, Maine. I was presenting on the brain pathways used in normal memory and the default network, and Bill Klunk showed the complete images of the amyloid deposition in the human brain in AD, and they looked nearly the same. That was an "aha!" moment for me when I realized a relationship I had simply not anticipated. Brain amyloid formed in AD in the same (or at least very similar) areas involved in the default mode. In fact, several other scientists such as Mark Mintun at Washington University who were aware of the default network also noticed the correspondence. That observation in Bar Harbor inspired me to go home to St. Louis and piece together this puzzle. We are not sure the puzzle is complete but the observations are clearly worth further study.

Last year John Cirrito working in Dave Holtzman's laboratory at Washington University showed through an elegant series of studies that stimulating brain activity increases amyloid-β levels in living transgenic mice. Yet another piece in the puzzle.

ARF: How about the "use-it-or-lose-it" idea? You are saying the opposite—these are areas that we use a lot, and they might be the first to go in AD.

RB: This is a complex question. All of the data I know of show that if you stay cognitively engaged, mentally active, that's good and will delay the onset of AD symptoms and mitigate cognitive decline. Everybody agrees on that. But the idea of use-it-or-lose-it works very differently from the kind of hypothesis we are putting forward, and I do not know how the two relate.

One way I find useful to think about the relationship of our observations to the protective effects of cognitive activity is to remind myself that we do not catch AD from a pathogen such as a virus, so at some level we know that AD pathology comes from normal brain processes. If you live long enough, these processes seem to have a higher likelihood of building up. From this perspective, it is not surprising that if disease processes do not affect the brain uniformly, the disease processes may relate to how we use brain areas during normal function, that is, to their activity, their metabolism levels throughout life. In fact, if the use-it-or-lose-it idea was not so entrenched in our thinking to explain the protective effects of education and cognitive activity, I wonder if people might see our hypothesis as intuitive. We are proposing a very specific cascade that leads to AD from brain activity and metabolism, and even potentially the origins of its special distribution.

The other interesting implication of the work is that it suggests a mechanism for why AD affects memory. Perhaps it is not random happenstance or a special vulnerability of these structures involved with memory. It may be more to do with how they are used throughout life. That is a different kind of explanation than people have put forth before.

It is also possible that the use-it-or-lose-it idea captures the manifestation of the disease in terms of clinical impairment and cognitive decline, but is not speaking directly to its pathology. Perhaps staying cognitively active, which is clearly a good thing, is not truly protecting one against the underlying disease, but rather helping one to cope with it as it occurs and progresses. I wonder if use-it-or-lose-it speaks more to the ability to hide the cognitive symptoms of the disease than to the underlying progression of disease pathology.

ARF: You've worked on neural compensation. Does it support that idea?

RB: We have identified functional mechanisms that we think are compensatory in aging and AD, as have others, but we are not certain yet how they link to factors associated with cognitive reserve, such as those studied by Yaakov Stern of Columbia University. We have some unpublished data that make me think that it is worth exploring whether these other phenomena capture more of a use-it-and-hide-it phenomenon rather than fundamental differences in the course of the disease process itself.

ARF: Can you tell us what that is?

RB: The work, which is by a talented graduate student, Anthony Fotenos, is presently in a paper under revision. We hope to be able to share the results with everyone soon. To foreshadow the results, his observations and several other observations in the literature suggest that many individuals harbor undetected AD, and education may help one to hide the symptoms of undetected AD for longer.

ARF: By undetected AD, you mean clinically?

RB: Yes. These are people in whom you cannot detect any notable cognitive deficits yet show plaque deposition using amyloid imaging.

ARF: You talked about early AD. If you could make a time-lapse movie imaging a normal person's brain as he or she ages, and then develops AD, what would you see?

RB: Great question. With the caveat that this is speculation, I think as you watch people throughout life you would see all sorts of dynamic transitions in brain activity as they do things—as they sleep, they wake, they engage in this activity or that activity. In the aggregate, you would see on average higher activity in a specific set of brain regions that have become known as the default network. The function of that network is still a mystery, but I think in aggregate over time you would see this network active much of the time and more so than other brain networks, and you would see this preferential activity from a very young age.

If you could shrink yourself down and watch all these events at the molecular level, you would see neurons firing more often in these default network areas than other areas, but firing everywhere at different times. In these regions within the default network, along with this activity of firing a lot, you would also see numerous metabolic events that are in response to this firing, and you would see them as setting up the efficient milieu for normal function throughout life. You would see increased amyloid-β, because there are hints of that from work such as John Cirrito et al.. You would see that as all of these metabolic events are going on during normal function, amyloid-β is increasing. As you watched these normal activity and metabolism events over the course of a lifetime, you would start to see things change and pathological forms of these proteins evolve. The process would be gradual and creep up. You would start to see that these high concentrations of amyloid-β took on forms that were toxic and interacting with other neurobiological cascades such as those we have been referring to. And you would, late in the process, start to see deposits of amyloid in these regions in the form of plaques as brain cells were being killed. Most of those would be cells in the very regions of this default network that are used in aggregate so much of the time. Then you would watch clinical impairment occur.

ARF: Where is the hippocampus in that movie?

RB: It is part of the default network. The default network is a widely distributed network that prominently includes the hippocampus. Michael Greicius of Stanford and recently Justin Vincent in my laboratory showed that activity in the default network correlates with activity in the hippocampus (Vincent et al., 2006). In our studies, Justin made functional brain imaging measurements over time and measured the spontaneous activity in the hippocampus as it waxed and waned. He then observed that the activity in the hippocampus is correlated with that of the default network. On average, when activity in the hippocampus is up, the default network is up; when it goes down, activity in the default network goes down.

The observation that the hippocampus is part of a widely distributed functional network raises an intriguing possibility regarding the origins of hippocampal dysfunction in AD. This is subtle. We know that the hippocampal formation is affected early in the disease, and the regions that appear to be connected functionally, and likely anatomically, seem also to be disrupted. If I had to make a guess, based on the extensive data from prior studies of pathology, I would probably lean towards the possibility that the hippocampus itself is the primary site of pathological effect in AD, and these other connected brain areas are secondarily affected. However, our data are raising an alterative possibility. When you look at the activity and metabolism throughout life, you realize that the hippocampus and the correlated regions of the default network may be linked together, as part of a distributed network. So you might imagine that these distributed regions are affected early and are the primary site of the pathology, and that they converge upon the hippocampus. That is, it is not that the hippocampus is disrupted and then the broader brain network becomes affected as a byproduct; rather, all these regions in the network may by affected together at the earliest stages of the disease. It is even possible that the cortical regions are affected by certain events before the hippocampus.

From the work of Brad Hyman and Joel Price and others, we know there is cell loss very early on in entorhinal cortex. We badly need more detailed anatomic work on these other distributed cortical regions such as the precuneus and posterior cortical regions in parietal cortex. There are existing studies of the broad distributions of plaques and tangles in the brain. What I am suggesting is that it may be worth revisiting some of these earlier data and also conducting new studies that specifically target detailed analysis of pathological changes in the areas recently emphasized by the imaging results.

ARF: We do not know their anatomy?

RB: They have been studied, and some of the findings, such as where there is amyloid burden, converged with our ideas. Other findings, such as the tangle distributions, do not obviously converge and point more to origin in the hippocampal formation as do the seminal studies of cell loss. I do not know how to reconcile this complex set of observations.

ARF: Some investigators report a hyperactivation of hippocampus early in AD. Have you looked at that?

RB: That is a very interesting observation. Brad Dickerson, Reisa Sperling, and colleagues showed that in people with mild cognitive impairment, many of whom have early AD, you see increased hippocampal activation using functional MRI. Susan Burkheimer and colleagues at UCLA also reported this same general observation and noted that activity increases might be used as a stress test. And Katharina Henke in Bern, Switzerland, reported a case of fMRI hyperactivity in a presymptomatic FAD case (Mondadori et al., 2006). As the system is being disrupted, it might actually compensate to ramp up, much as if your muscles in your arms weaken, you might strain harder to be able to lift things initially. Eventually, your arm becomes so weak that it fails and you can't lift things, but in the first initial phase, when you're starting to see strain in the system because of this disease process, the system might respond by actually increasing its activity, and functional imaging picks that up.

We see a related phenomenon in frontal cortex. Even in advanced aging, these higher-order brain regions seem to be more engaged during tasks, as if as the system becomes challenged, rather than receive these changes passively, the brain compensates and recruits more resources to accomplish the same task. Tasks that were easy may become difficult, but nonetheless, we have enough reserve to mitigate against at least these mild changes that we all experience.

ARF: Do you see tau in your imaginary movie?

RB: We do not have markers for tangles. There's exceptionally important work by the Braaks and others, Brad Hyman, Joel Price, and John Trojanowski, who have looked at the details of tau pathology. There is a lot of debate about which pathological correlates most relate to clinical impairment, and their order in terms of their distributions across the brain. Our data do not speak to this question which is better answered by these other experts. I suspect that any complete understanding of AD is no doubt going to explain both the tau pathology and amyloid-β plaques that we see.

ARF: What experiments would you like to do next?

RB: We would like to understand variations in brain activity and metabolism across young adults, people as young as their twenties, and ask if those variations are risk factors for AD. We are pushing back the exploration of Alzheimer disease from the beginning points of the disease pathology, to understanding the variation across individuals in the antecedent conditions, the stage-setting conditions that may be the precursors to the early stages of the disease.

ARF: Are you able to look at genetic effects on that, like FAD mutations?

RB: That is what we're doing next. In the first phase we will look at the known risk factors of AD and ask to what degree this new hypothesis explains any of them. Do variations in ApoE relate to variation in default activity in young adults? More importantly, we plan to explore whether variations in default activity, and the metabolism associated with it, might help us find genetic variations that are risk factors for AD. For example, what if we found that variation in a new gene associated with metabolism had variations that associated with default activity and metabolism in young adults? That genetic variation would be a good candidate to explore as a risk factor for AD and, if identified as such, would be a place to then explore more to understand underlying disease mechanisms.

ARF: How does that differ from other studies looking for antecedents of AD?

RB: People are imaging antecedent markers in a wide range of studies. These are important studies and the most appropriate next steps. Typically, they are looking at brain atrophy and the early phases of amyloid deposition before clinical symptoms. They are pushing back the exploration of AD 10 years or so prior to the clinical symptoms. When I'm speaking of antecedent conditions here, we are taking the approach to its extreme; we are talking about metabolic effects beginning in childhood that set the stage for what happens when you're 80. I admit, it may be a little bit out there, but the data are pointing us in this direction and we are going to follow the data. There may be intervention opportunities such as drug therapies that prevent AD by affecting lifelong brain metabolism.

ARF: Those are tough experiments.

RB: They are actually whole research programs. We are looking for convergence across many different views of the same phenomenon. It might be an unwise path to pursue career-wise, but I think there's enough evidence and it's sufficiently important that it's worth exploring.

We are also just trying to understand these phenomena. Why is there a preferential activity mode (default mode) in the human brain? The data start to suggest that this default activity is what we do most of the time. What is its purpose? It is a difficult question because it is not directly linked to immediate behavior.

ARF: Do you see these same patterns in animals?

RB: There is an important body of work emerging from Matt Wilson's laboratory at MIT with David Foster looking at the stop periods when rats pause between their explorations. He shows that in the hippocampus, the animals spontaneously replay recent events. Perhaps this is a related phenomenon—spontaneous replay of activity when people aren't engaged in the environment. It may well be functional for learning and plasticity and other reasons, and it may go on a lot of the time. Certainly the human data point to the fact that this is what is going on most of the time we are not focused on a specific task. I feel fortunate to have stumbled into such an interesting research arena where so many questions remain unanswered. My intuition is that to understand brain disease we will need to understand these spontaneous brain activity patterns that are so common in the brain.

ARF: We thank you for this conversation.

RB: It's been my pleasure.

 

Comments

Make a Comment

To make a comment you must login or register.

Comments on this content

No Available Comments

References

Other Citations

  1. ARF related news story

Further Reading

No Available Further Reading