A gift from the Henry L. Hillman Foundation has allowed the Brain Institute to launch a wide variety of multidisciplinary projects to advance research on normal and abnormal brain function. The five new projects include:
- development of new statistical approaches for studying large-scale recordings of neurons;
- a study of how lack of sleep may make teens more vulnerable to substance abuse;
- applying artificial intelligence and machine learning to improve neuroprosthetics for brain-controlled use of robotic arms and hands;
- an examination of obsessive compulsive disorder, including whether OCD behaviors reduce anxiety to the point that they become addictive; and
- creating a new paradigm for studying movement disorders like Parkinson's disease that includes studies of network connectivity.
“Collectively, these projects have the potential to significantly advance our understanding of brain function and shed light on disorders that affect millions of people. New therapies are urgently needed, and the road to these cures begins with this kind of fundamental research,” says Peter L. Strick, PhD, Scientific Director of the Brain Institute and Thomas Detre Professor and Chair of the Department of Neurobiology. “The funded projects will help build our capacity in key areas, enabling us to better frame the national research agenda and to compete more effectively for large-scale federal funding opportunities.”
The principal investigators briefly describe their studies:
Brent Doiron, PhD, Associate Professor of Mathematics and Matthew Smith, PhD, Assistant Professor of Ophthalmology
"Towards a circuit based theory of cortical variability and its impact on neuronal processing"
One of the key limitations to understanding of the brain is the ability to measure, analyze and comprehend the activity of large populations of neurons. Historically, statistical analysis of the spike train responses of single neurons, or small groups, have provided important constraints to computational models of cortical networks. However, modern electrophysiological approaches now provide access to neuronal populations numbering in the tens to hundreds, and can be expected to routinely approach the thousands over the next decade. As the scale of recorded neural populations and the complexity of network simulations increases, though, it is no longer possible to compare data and models in a straightforward way. Advancing understanding of the brain demands statistical approaches to distill the key features of neural interactions from large-scale recordings and models.
Dimensionality reduction techniques have been applied to neuronal population activity with some success, and excel in identifying and studying the activity patterns in a population without requiring advance knowledge of the expected pattern structure. However, such tools give little insight into the physiological mechanisms of observed patterns without explicit grounding in the structure of the data. We propose to develop the computational infrastructure to model and analyze large-scale neural populations by coordinated effort to advance realistic network simulations and matched statistical analysis tools. The key innovation of this proposal is to leverage spiking network models, where we can sample from as many neurons and for as much time as desired and we know the ground truth connectivity among the neurons. This allows us to understand how the outputs of dimensionality reduction relate to the underlying network structure. Simultaneously, the spiking network models are constrained by the population-level statistics of actual neural recordings. This creates a virtuous cycle between data, models and analytic tools that forms the necessary computational infrastructure for investigations into perceptual and cognitive phenomena.
Colleen A. McClung, PhD, Associate Professor of Psychiatry
"Impact of sleep and circadian rhythms on the vulnerability for substance abuse in adolescents"
Adolescence is a critical developmental period marked by major changes in sleep and circadian physiology, changes in function of brain reward circuitry, and heightened vulnerability to substance abuse. We propose that sleep and circadian changes interact with changes in reward function to increase substance abuse risk in vulnerable adolescents. During puberty there is a natural shift in circadian rhythms (particularly among boys) toward delayed clock timing, leading to a preference for later bedtimes and wake times. However, school start times move earlier during adolescence, creating circadian misalignment or “social jet lag” each week. Moreover, the demands of school often lead to sleep restriction, despite a developmental need for longer sleep.
We hypothesize that short sleep duration combined with late sleep timing constitutes a critical risk phenotype among adolescents. However, the extent to which this sleep/circadian phenotype specifically influences addiction vulnerability in this population is still unclear. While we know from work in the McClung lab and others that changes in circadian rhythms and sleep can alter reward circuitry in the brain, the specific circuit functions that are altered in adolescents who have particular sleep/circadian phenotypes is largely unknown. We will combine powerful study designs for humans and rodents to begin to answer these questions.
In one part of the study, we will conduct sleep and circadian manipulations aimed at either restoring these measures in individuals with the high risk phenotype (short/late sleep), or acutely disrupting them in those with the low risk phenotype (long/early sleep). Adolescents will be placed on strict sleep schedules at the WPIC Sleep Center and at home to alter the amount/timing of sleep, and then undergo behavioral tasks to assess reward-circuit function and measures of circadian rhythms to assess phase changes. We seek to demonstrate that our experimental sleep/circadian manipulations can alter rhythms and behavior. In another part of the study, we will examine our manipulations of sleep and circadian rhythm in humans and rats, with electrophysiology and fMRI. We also will look at molecular effects (circadian gene expression and RNA sequencing).
Andrew Schwartz, PhD, Distinguished Professor of Neurobiology
"Enhanced Neural Prosthetics Using Shared-Mode Control"
Unlike many injuries, damage to the central nervous system is usually permanent, so that the vast majority of people who become paralyzed will never move again. The University of Pittsburgh has the world’s most advanced program in brain-controlled robotic arms and hands for people who are paralyzed. Several notable programs exist elsewhere in this research area (e.g., at Brown, Duke, Stanford, UCSF). To date, Pitt’s advances are unsurpassed, with achievements that include: 1) enabling two people with quadriplegia to control movements of a prosthetic shoulder, elbow, wrist, and fingers all at the same time; and 2) providing shared-mode control (in other words, improving brain-controlled performance with a boost from vision-guided, autonomous robots that can predict what a person wants to grasp). This research proposal seeks to improve the technology both for decoding the user’s intentions from his or her brain signals, and for shared-mode control using artificial intelligence. As a result of this research, we expect to enable users to be able to manipulate objects with much greater dexterity than is possible now. For instance, they will learn to operate a doorknob, open a jar, and handle soft fruit without damaging it. In addition, this research project will help us move brain-controlled prosthetics out of the laboratory and into the everyday world, so that users can become more independent.
Pitt’s brain-controlled prosthetics program includes dozens of scientists, bioengineers and clinicians who have built on decades of basic research on the neural mechanisms involved in movement. The artificial intelligence system we propose is similar to that being used for driverless cars. Members of the same group that has provided such systems to Uber will be partners in this project. In the future we envision a wheelchair-mounted prosthetic arm with integrated, miniature cameras; the software will run on a laptop attached to the wheelchair. We believe this system will dramatically increase independence among those who are paralyzed, by not only restoring their ability to grasp, but by restoring their dexterity.
Susanne E. Ahmari, MD, PhD, Assistant Professor of Psychiatry
"Identifying the circuits underlying abnormal anxiety and reward processing in OCD patients and mouse models"
Obsessive compulsive disorder (OCD) is characterized by intrusive distressing thoughts and/or repetitive mental or behavioral acts. Since a majority of patients display illness onset during childhood or adolescence, and experience symptoms throughout their lives, this chronic disorder leads to significant morbidity. However, the underlying pathophysiologic mechanisms are still unclear, because OCD is disproportionately understudied compared to other mental illnesses. This limited understanding of neural mechanisms limits the development of effective treatments, with approximately half of all patients seeing minimal or partial benefit with available therapy, and up to 10% of patients with treatment-resistant symptoms so severe that they are potential surgical candidates. We will examine the role of cortico-limbic regions in negative
reinforcement paradigms in humans and rodents, to determine how the rewarding properties of anxiety relief may lead to compulsive behaviors. This is the first time the potential link between compulsions in OCD patients and reinforcement processes seen in addiction will be formally investigated, which may open up new avenues for OCD treatment.
Efforts to find improved treatments for people with Parkinson’s disease, dystonia and other basal ganglia-related disorders have been hindered by our poor understanding of why the neuronal dysfunctions seen in those disorders leads to the specific symptoms observed. This project is the next step in an ongoing collaborative effort at Pitt to build a new model based on three key concepts: Basal ganglia-related disorders are “circuit disorders” in which dysfunction is transmitted across whole neural networks; the critical dysfunction is one of circuit dynamics (e.g., of aberrant time-varying neuronal responses to inputs); and the root causes are pathology-induced alterations in network connectivity. We will use mutually-reinforcing approaches in humans, non-human primates and computational models. These approaches will measure disease-related abnormalities in circuit connectivity and dynamics by: (1) estimating the directed flow of information between recording locations in the cortex and basal ganglia; and (2) measuring the relative strength and segregation of communication along cortico-basal ganglia pathways by stimulating electrically at discrete sites in cortex and studying the responses evoked in basal ganglia output neurons.
We will apply these approaches in parallel in awake human subjects undergoing implantation surgery for deep brain stimulation for Parkinson’s or for dystonia and in normal and parkinsonian non-human primates. We will then use computational models to generate mechanistic predictions about how the observed changes in information flow and in circuit connectivity result in the observed abnormalities in circuit dynamics.