Cortical Computations Underlying the Integration of Perceptual Priors and Sensory Processing

Tahereh Toosi, Ph.D.

Postdoctoral Research Scientist, Mortimer B. Zuckerman Mind Brain Behavior Institute, Columbia University

The ability of the visual system to store and use learned information, or perceptual priors, is essential for interpreting complex visual scenes, such as identifying obscured objects or imagining scenes not currently visible. This process relies on the interaction between processing incoming sensory data and existing knowledge stored in the synaptic strengths throughout the brain. Although the importance of top-down and bottom-up integration is recognized, the precise ways in which they enable the brain to piece together information from different sources remain largely unknown.

My research aims to reveal the mechanisms underlying these processes by demonstrating how the brain’s need to function reliably in noisy environments influences the development of these pathways, enabling visual processing abilities like resolving visual occlusion and visual imagination. The phenomenon of illusory contours and shapes, exemplified by the Kanizsa optical illusion and Rubin’s face-vase illusion, serves as an ideal case study for how the brain combines sensory input with past experiences to create a coherent perception. Previous studies have shown that such illusory contours invoke activation in specific layers (L2/3) of the early visual cortex but not in others (L4). I will demonstrate the recapitulation of these findings within a deep convolutional model optimized for object recognition, powered by a theory-grounded, biologically plausible algorithm that processes activations through forward and feedback pathways iteratively. This represents the first instance of a large-scale, image-computable model that, while primarily optimized for recognizing objects, also explains how illusions are perceived in the visual cortex as a result of integrating sensory data with learned information.

Zooming out, the insights from this computational modeling suggest a resolution to the debate over whether the brain functions primarily as a generative or a pattern recognition neural network, and explaining a number of experimental findings regarding specificity of computations in cortical layers.

Join the Carney Institute for the Brain Science for its External Postdoc Seminar Series (BrainExPo), featuring Tristan Geiller, Postdoctoral Researcher at Columbia University.

Abstract: The hippocampus is a multi-stage neural circuit, where local interactions between excitatory principal cells and inhibitory interneurons are thought to contribute distinct computations important for memory formation and retrieval. The overarching goal of my research is to uncover the architecture of the local circuits that provides the scaffolding for such interactions, by developing and using variety of experimental methods in behaving mice. 

Join the Carney Institute for the Brain Science for its External Postdoc Seminar Series (BrainExpo), featuring Feng-Kuei Chiang, Postdoctoral Fellow at Icahn School of Medicine at Mount Sinai.

Abstract: Cognitive strategies, such as processing information in sequences, can improve behavioral performance in working memory tasks, but how this is accomplished at the neural level remains unclear. Here we created a non-human primate model of self-generated search strategies to study prefrontal functions and found that sequencing strategies shift information from single, highly tuned neurons to more distributed population codes in lateral prefrontal cortex.

Join the Carney Institute for the Brain Science for its External Postdoc Seminar Series (BrainExPo), featuring Elif Tunc-Ozcan, Postdoctoral Fellow at Northwestern University.

Abstract: I will talk about how chemogenetically regulating the activity of adult-born hippocampal neurons, without changing their numbers, affects stress-related phenotypes and antidepressant action. Additionally, I will describe how we confirmed bone morphogenetic protein (BMP) signaling as a common downstream pathway that mediates the behavioral effects of different classes of antidepressants.

Join the Carney Institute for the Brain Science for its External Postdoc Seminar Series (BrainExPo), featuring Hanieh Falahati, Ph.D., Postdoctoral Fellow at Yale School of Medicine.

Abstract: The spine apparatus is a morphologically peculiar specialization of the endoplasmic reticulum at dendritic spines, but studying this organelle has remained a longstanding challenge in neuroscience for the past 60 years. I have used an interdisciplinary approach to characterize the morphological features and molecular components of the spine apparatus, which finally allows us to address the longstanding questions of biogenesis and function of this enigmatic organelle.

Join the Carney Institute for its Brain Science External Postdoc Seminar Series, featuring Chen Ran, Ph.D., a postdoctoral fellow in the Department of Cell Biology at Harvard Medical School.

Abstract: In vivo brainstem two-photon calcium imaging analyses of sensory inputs from the internal organs reveal fundamental features of the interoceptive nervous system.

Join the Carney Institute for the Brain Science for its External Postdoc Seminar Series (BrainExPo), featuring Aaron Kuan, Ph.D., postdoctoral fellow in the Neurobiology Department at Harvard Medical School. 

Abstract: One of the grand quests in neuroscience is to build complete maps of the brain, charting all of its cells and the connections between them. In this talk, I describe how innovations in X-ray and electron microscopy that are expanding the scope and detail at which we can image the brain, and enable us to investigate the circuit basis of cognitive tasks such as decision-making, and will soon allow us to tackle the massive scaling challenges involved in comprehensively mapping mammalian brains.

Join the Carney Institute for the Brain Science for its External Postdoc Seminar Series (BrainExPo), featuring Marino Pagan, Ph.D., postdoctoral researcher at Princeton University. 

Abstract: Using a high-throughput procedure, I trained many rats to perform a task requiring context-dependent selection and accumulation of evidence towards a decision. Detailed neural and behavioral analyses revealed remarkable heterogeneity across rats, despite uniformly good task performance. This approach opens the door to the study of individual variability in neural computations underlying higher cognition. Finally, I will present preliminary data leveraging this behavioral paradigm to study the neural mechanisms underlying cognitive deficits in rat models of autism.

Join the Carney Institute for its Brain Science External Postdoc Seminar Series (BrainExPo), featuring Micaela Chan, Ph.D., a postdoctoral fellow at the University of Texas at Dallas.

Resting-state functional magnetic resonance imaging (fMRI) allows us to examine patterns of large-scale brain network organization. Over the course of healthy adult aging, the functional brain network desegregates (i.e., fewer connections within a system, more connections between systems), which in turn, is predictive of poorer behavioral performance. Brain network segregation is also clinically relevant, providing prognostic value for dementia. We have found that this aging trajectory of brain network desegregation varies across individuals that are embedded in distinct environments. Measures of environment are typically coarse, based on individual-level variables (e.g., education, socioeconomic status). The next step in my research is focused on linking brain measures with a more comprehensive description of an individual by (1) collecting information on an individual’s life history and daily activities (e.g., activity tracking, ecological momentary assessment); and (2) linking neighborhood-level data such as Census data or other geographically anchored data back to the individual. Together, this will better capture how an individual’s environment impacts their brain network organization over the course of aging and disease such as dementia.

Join the Carney Institute for its Brain Science External Postdoc Seminar Series (BrainExPo), featuring Julieta Lischinsky, Ph.D., a postdoctoral fellow at New York University.

Julieta Lischinsky’s research focuses on understanding the neuronal substrates and circuitry for the generation of innate social behaviors in the limbic system.

Abstract: Innate social behaviors are crucial for survival, thus shared across animal species. In humans, psychiatric disorders with deficits in social interactions, e.g. autism spectrum disorders, can be observed during child development and have been associated with amygdala dysfunction. There is still a lack of understanding of the circuitry and developmental mechanisms for the generation of social behaviors. We have focused on the murine medial amygdala (MeA) as it receives conspecific pheromone inputs and projects to hypothalamic regions. The MeA GABAergic cells have been shown to be sufficient for the production of social behaviors including aggression and mating. Given that these diverse social behaviors differ in their sensory trigger and behavioral outcomes, can the neuronal substrates for these behaviors be distinct? Taking a developmental approach, we have previously characterized two MeA GABAergic neuronal subpopulations, marked by the expression of the transcription factors Foxp2 and Dbx1 which originate from the same embryonic region. The Foxp2+ and Dbx1-derived subpopulations are spatially, molecularly and physiologically distinct. Interestingly, I have now observed that these two subpopulations receive distinct inputs and differ in their processing of social conspecific information. Furthermore, I uncovered that these subpopulations differ in their functional roles during social behaviors. In addition, as the Foxp2+ cells respond to conspecific cues even with no/minimal social experience, I aimed to determine the extent to which these neuronal responses are hard-wired by investigating the social tuning of Foxp2+ cells across development. In conclusion, developmentally distinct MeA neuronal subpopulations differ in their anatomical circuitry, are differentially relevant for processing conspecific sensory cues and mediating social behaviors.

Join the Carney Institute for its Brain Science External Postdoc Seminar Series (BrainExPo), featuring Danique Jeurissen, a postdoctoral fellow at Columbia University.

Abstract

Neural substrates of higher cognitive functions like decision-making are distributed across multiple brain areas. The flexibility afforded by such architecture renders some cognitive functions resilient to focal lesions. We used pharmacological and chemogenetic approaches to disrupt activity in the parietal cortex of monkeys performing two perceptual decision-making tasks. Inactivation initially disrupted decision-making in all four monkeys. This was followed by behavioral compensation occurring at two time scales: within experimental sessions and across sessions. Our results suggest that compensatory mechanisms can account for the disparate effects of causal manipulations on higher cognitive functions.

Join the Carney Institute for its Brain Science External Postdoc Seminar Series (BrainExPo), featuring Megha Sehgal, a postdoctoral fellow at the University of California, Los Angeles.

Abstract

Events occurring close in time are often linked in memory, providing an episodic timeline and a framework for those memories. Recent studies suggest that memories acquired close in time are encoded by overlapping neuronal ensembles, but whether dendritic plasticity plays a role in linking memories is unknown. Using activity-dependent labeling and manipulation, as well as longitudinal one- and two-photon imaging of RSC somatic and dendritic compartments, we show that memory linking is not only dependent on ensemble overlap in the retrosplenial cortex, but also on branch-specific dendritic allocation mechanisms. These results demonstrate a causal role for dendritic mechanisms in memory integration and reveal a novel set of rules that govern how linked, and independent memories are allocated to dendritic compartments.

Join the Carney Institute for its Brain Science External Postdoc Seminar Series (BrainExPo), featuring Carl Schoonover and Andrew Fink, postdoctoral fellows at Columbia University.

The primary olfactory cortex has traditionally been hypothesized to establish the identity of odorants. Schoonover and Fink will discuss how their research has shown that after just a few weeks odor responses bear little resemblance to their original form, raising basic questions about the role of this brain region in olfactory perception.

Abstract

We have discovered that in the rodent primary olfactory cortex (piriform) the pattern of neural activity evoked by a smell changes with the passage of time. These changes, which unfold absent a task or learning paradigm, accumulate to such an extent that after just a few weeks odor responses bear little resemblance to their original form. The piriform has been traditionally hypothesized to establish the identity of odorants. Our observations have forced us to radically reconsider the role of this vast brain region in olfactory perception. We propose that the piriform operates instead as a flexible learning system, a ‘scratch pad’ that continually learns and continually overwrites itself. This poses the problem of how transient memory traces can subsequently be stored over long timescales.

These results also raise the question of what the piriform learns. We have designed a behavioral assay that provides a sensitive readout of whether mice expect a given sensory event. Using this assay, we have demonstrated that mice learn the identity, order and precise timing of elements in a sequence of neutral odorants, A–>B, without reward or punishment. Simultaneous recordings in naïve primary olfactory cortex (piriform) show strong and distinct responses to both A and B. These diminish with experience in a manner that tracks these expectations: predictable cues, such as B in the A–>B sequence, evoke hardly any response in experienced animals. This does not reflect simple adaptation. When B is presented alone, it elicits robust activation. When B is omitted, and A is presented alone, piriform exhibits vigorous activity at the precise moment when the animal, expecting odor B, encounters nothing. Thus, when the external world conforms to expectation, piriform is relatively quiescent, but any departure from the expected results in vigorous activation. The biological learning mechanisms that generate this predictive activity, a feature more commonly encountered in higher order cortices, can be readily studied and probed in a circuit only two synapses from the sensory periphery.

The next Carney Institute External Postdoc Seminar Series (BrainExPo) will feature Ashley Ingiosi, Ph.D., Ruth L. Kirschstein Postdoctoral Fellow at Washington State University.

Abstract

Sleep is characterized by dynamic changes in neuronal activity, and waking neuronal activity is thought to increase sleep need. Changes in non-neuronal cells (e.g. glia) across the sleep-wake cycle and their role in sleep regulation are comparatively unexplored. I investigated if glial cells called astrocytes also change dynamically across sleep and wake states and if astroglial signaling mechanisms are important for sleep regulation. Astrocytes are not electrically excitable but use calcium (Ca2+) to mediate signaling. Therefore, Ca2+ imaging is optimal for studying astroglial activity. I quantified astroglial Ca2+ activity during sleep-wake behavior in mice expressing the Ca2+ indicator GCaMP6f selectively in cortical astrocytes. Ca2+ activity was captured in vivo with an epifluorescent miniscope in freely-behaving mice and with two-photon microscopy in unanesthetized, head-restrained mice for more detailed exploration of astroglial processes. I recorded astroglial Ca2+ dynamics simultaneously with sleep-wake behavior under baseline conditions, in response to sleep deprivation, and following genetic depletion of astroglial intracellular Ca2+ stores. I found that 1) astroglial Ca2+ signals change dynamically with sleep, wake, and sleep loss, 2) astroglial Ca2+ activity is more pronounced in processes compared to somata, 3) astroglial Ca2+ is responsive to changes in sleep need, 4) synchrony of astroglial Ca2+ signals changes with vigilance state and sleep loss but does not mirror neuronal electrical activity, and 5) reduced astroglial Ca2+ via knockout of stromal interaction molecule 1 reduces sleep drive after sleep loss. Overall, this research revealed a new level of brain organization in a non-neuronal cell type that changes dynamically with vigilance state and plays a role in sleep regulation. These findings could trigger a fundamental shift in our understanding of sleep and sleep regulation.

Join the Carney Institute for the Brain Science for its External Postdoc Seminar Series (BrainExPo), featuring Seungwon (Sebastian) Choi, Ph.D., postdoctoral fellow at Harvard Medical School.

Abstract

Each day we experience myriad somatosensory stimuli: hugs from loved ones, warm showers, a mosquito bite, and sore muscles after a workout. These tactile, thermal, itch, and nociceptive signals are detected by peripheral sensory neuron terminals distributed throughout our body, propagated into the spinal cord, and then transmitted to the brain through ascending spinal pathways. Primary sensory neurons that detect a wide range of somatosensory stimuli have been identified and characterized. In contrast, very little is known about how peripheral signals are integrated and processed within the spinal cord and conveyed to the brain to generate somatosensory perception and behavioral responses. We tackled this question by developing new mouse genetic tools to define projection neuron (PN) subsets of the anterolateral pathway, a major ascending spinal cord pathway, and combining these new tools with advanced anatomical, physiological, and behavioral approaches. We found that Gpr83+ PNs, a newly identified subset of spinal cord output neurons, and Tacr1+ PNs are largely non-overlapping populations that innervate distinct sets of subnuclei within the lateral parabrachial nucleus (PBNL) of the pons in a zonally segregated manner. In addition, Gpr83+ PNs are highly sensitive to cutaneous mechanical stimuli, receive strong synaptic inputs from primary mechanosensory neurons, and convey tactile information bilaterally to the PBNL in a non-topographically organized manner. Remarkably, Gpr83+ mechanosensory limb of the anterolateral pathway controls behaviors associated with different hedonic values (appetitive or aversive) in a scalable manner. This is the first study to identify a dedicated spinal cord output pathway that conveys affective touch signals to the brain and to define parallel ascending circuit modules that cooperate to convey tactile, thermal and noxious cutaneous signals from the spinal cord to the brain. This study has also revealed exciting new therapeutic opportunities for developing treatments for neurological disorders associated with pain and affective touch.

Join the Carney Institute for the Brain Science for its External Postdoc Seminar Series (BrainExPo), featuring Lisa Scheunemann, Ph.D., an independent research fellow at Freie Universität in Berlin.

Abstract

A key function of the brain is to decide which information is relevant enough to be stored as a stable memory. Pathological perturbation of this filtering process can have catastrophic consequences for later decision making. The molecular and circuit mechanisms that gate memory formation by inhibiting the storage of irrelevant information remain yet largely elusive. I have recently identified a memory suppressor mechanism in the Drosophila (fruit fly) brain within a serotonergic circuit (specifically the SPN “Serotonergic Projection Neurons”) upstream of the fly’s memory center. This “memory checkpoint” sustains a default inhibition of memory consolidation for aversive associations controlled by phosphodiesterase (PDE)-mediated suppression of neuronal activity in the SPN. Strikingly, my studies revealed that the dedicated memory checkpoint is modulated by the mating state of female flies: memory suppression by PDE is constantly inhibiting aversive memory consolidation in virgin females and is only released after mating. This mating-dependent switch is mediated by the sex peptide, a sperm-bound peptide transferred to females during copulation. Such a mechanism could promote foraging behavior in virgin females by suppression of risk-related behavior while promoting it after mating to protect the offspring. Thus, I propose that this type of memory suppression represents an important intersection between behavioral- and memory-dependent plasticity to guarantee consolidation of relevant information and context-appropriate decision making.

Join the Carney Institute for the Brain Science for its External Postdoc Seminar Series (BrainExPo), featuring Allan-Hermann Pool, Ph.D., postdoctoral researcher at the California Institute of Technology. 

Abstract: Animal behavior is governed by innate and hardwired biological drives. Allan-Hermann Pool will describe a single cell RNA-sequence based stimulus-to-cell-type mapping approach for scalable mapping of these drive states to the underlying neural circuitry. He will also outline how this information can be used to functionally reprogram circuits governing innate motivations.

Join the Carney Institute for its first Brain Science External Postdoc Seminar Series (BrainExPo), featuring Sergey Stavisky, postdoctoral research fellow in the Neurosurgery Department of Stanford University. 

Stavisky will discuss “Intracortical brain-computer interfaces: from fundamental science and engineering to restoring speech, reach and grasp.” 

Abstract: Brain-computer interfaces (BCIs) are poised to profoundly transform human neuroscience and health by treating devastating – and currently incurable – nervous system injuries and diseases with precise, circuit-level measurements and interventions. BCIs can potentially restore the ability to speak, move, remember, and more. However, going from proof-of-concept studies in animal models to repairing or replacing patients’ damaged abilities requires a platform for understanding human-specific neural functions and designing, testing, and refining therapies in people. My strategy for accomplishing this is to develop advanced intracortical BCIs to restore reach & grasp movement and speech for people with paralysis. Motor BCI clinical trials can help individuals with severe speech and motor impairment in the near-term, and in doing so, validate the safety of new human-use devices capable of reading from and writing to thousands of neurons. These clinical trials also provide direct access to human neural circuits for gaining a deeper neuroscientific understanding of how the brain generates movements, which I believe will ultimately lead to better BCI therapies.