Large scale ECoG recording during social scene watching and social behaviors in marmosets
Authors: *H. Xu, Y. Su, J. Sharma, W. Menegas, C. Trimmer, F. Liang, R. Landman, B. Zhang, M. Sur, R. Saxe, G. Feng, R. Desimone
Abstract: Humans can understand social scenes at a glance. Abilities such as these are thought to rely on a network of brain regions often referred to as the social brain. Although people have found cortical patches selective for social content including faces, bodies and social interactions, it is mostly unknown how these domains work collectively dynamically during perception. People with autism spectrum disorders typically have deficits in some aspects of social cognition, but there is uncertainty over which brain regions are responsible. Here we “map” the regions of cortex important for social perception and cognition in wild type marmosets which is a new world monkey that is highly social. In the traditional paradigm we showed images and videos with marmoset faces and marmosets engaged in several forms of social behavior, together with matched control images and videos. We mapped brain regions responsive to the visual stimuli with micro-electrocorticogram (ECoG) recordings over the posterior and middle temporal cortex and lateral prefrontal cortex in animals seated in a primate chair with head fixed and gaze monitored. Stimulus selectivity was evident in the high gamma band. We localized several face/body/object patches in temporal and frontal lobe, which appear to be in similar locations to areas mapped with fMRI scanning. We also identified patches that appear to be selective for social interactions. Timing data suggests a first posterior to anterior feedforward wave of activation, followed by a second feedback wave. Moreover, in the naturalistic paradigm, we also recorded with the same ECoGs but with untethered, wireless recordings in the same animal, during social behaviors in the home cage. A topdown-view camera was used to record and track the marmosets’ behaviors. We found the face and social patches showed higher activity when the animal with the electrodes was facing towards the cage-mate, compared with when they were facing apart. With these recordings, we hope to compare patterns of activation during the perception of social stimuli with patterns activated during performance of social behaviors.
Mitochondrial dynamics and astrocyte contribution to development of Rett pathophysiology in human ESC derived cerebral organoids
Authors: *D. Tomasello, I. Barrasa, R. Jaenisch
Abstract: Rett syndrome (RTT) is a postnatal neurodevelopmental disorder, largely due to mutation of the Methyl CpG-binding Protein 2 gene (MECP2), and is associated with severe mental disability and autism-like syndromes that manifests during early childhood. An important aspect of neurodevelopmental disorders are the cell non-autonomous functions that are sparsely understood. Astrocytes provide structural and molecular support for neurons that are essential for proper development and maturation. We found human embryonic stem cell (ESC) derived RTT astrocytes have impaired glutamate uptake, a key function that prevents excitotoxicity. We have identified RTT astrocyte mitochondrial function and glycolytic capacity are severely diminished. Interestingly, RTT astrocytes revealed increased mitochondrial mass, a possible mechanism to compensate for limited mitochondrial function. When mitochondrial activity was stimulated, RTT astrocytes maintain a threshold of activity irrespective of glycolytic shift. Metabolomic analysis defined metabolic disruption in energy balance and amino acid supply. RTT cerebral organoids revealed developmental delay in maturation of both neurons and astrocytes. Additionally, we observed mitochondrial transplantation from RTT astrocytes to RTT neurons, indicating dysfunctional mitochondria in astrocytes may amplify disease pathology. These findings highlight cell non-autonomous contribution to RTT, impacting neuronal bioenergetics and function.
Layer-specific deactivation of rhesus prefrontal cortex to test predictive coding feedback onto visual cortex
Authors: *A. Major, A. Bastos, S. Brincat, J. Roy, M. Mahnke, E. Miller
Abstract: Disrupted organization of neocortical layers during development can lead to altered neuronal circuits and conditions such as autism and schizophrenia. There is burgeoning interest in the physiology of superficial vs deep cortical layers and they are now recognized to perform different functions. In the laminar model of Predictive Coding, deep cortical layers of higher-order cortex send feedback signals via alpha/beta rhythms (10-30 Hz) to sensory regions (Bastos et al., PNAS, 2020). Conversely, superficial layers of sensory regions send feedforward projections (via gamma rhythms, > 35 Hz) to update higher-order regions. For example, it is proposed that prefrontal cortex (PFC) sends feedback alpha/beta rhythms to suppress feedforward information in visual cortex. This project will test this layer-specific model of Predictive Coding using layer-specific pharmacological deactivation of PFC. Using custom laminar probes with embedded drug-injection ports, either the superficial or deep layers of rhesus PFC were deactivated with GABAA receptor agonist muscimol.
Deactivation of deep layer PFC (the feedback layer) is hypothesized to cause a greater reduction in feedback alpha/beta to lower-order cortex. Preliminary results support this. Muscimol deactivation of PFC reduced feedback alpha/beta connectivity with visual area V4 and reduced V4 stimulus information carried by spiking. The effects were more pronounced when the stimulus was predictable (vs unpredictable). This work will perform causal tests on a prevalent model of cortical communication. It can yield insights into conditions with altered cortical communication, such as autism spectrum disorder and schizophrenia. This work was supported by Simons Center for the Social Brain Postdoctoral Fellowship, NIMH R37MH087027, ONR MURI N00014-16-1-2832, and The JPB Foundation.
Molecular imaging of extracellular glutamate in the rat and marmoset brain
Authors: *M. Schwalm, R. Ohlendorf, S. Bricault, J. Sharma, A. Jasanoff
Abstract: Monitoring extracellular glutamate dynamics provides a way to discern the engagement of excitatory activity that conducts the bulk of long-distance signaling in the mammalian brain. We recently introduced a molecular imaging approach that could permit noninvasive imaging of neurochemicals like glutamate using vasoactive sensors called AVATars. By coupling analyte detection to relaxation of vascular smooth muscle cells, these sensors induce artificial hemodynamic signals detectable by functional magnetic resonance imaging (fMRI) or functional ultrasound. Here we describe the design and in vivo validation of Glu-AVATar, a novel protein-based AVATar for selective hemodynamic imaging of extracellular glutamate in the brain. We demonstrate that this probe produces robust hemodynamic responses in the presence of exogenous glutamate, while control experiments performed with vehicle stimulation or control proteins do not result in signal changes. Glu-AVATar functionality is exhibited in both rodent and primate brains, indicating potential translatability of the technology. We further demonstrate that Glu-AVATar reveals correlations between probe-infused regions of the dorsal rat hippocampus and distal regions such as the ventral hippocampus. These measurements constitute a neurochemically specific form of functional connectivity analysis that can be combined with conventional fMRI techniques to reveal correlates of excitatory signaling between different brain regions. Future AVATar-based molecular fMRI approaches will allow for more precise and less invasive methods for imaging extracellular glutamate dynamics over wide fields of view in intact brains.
Huntington’s disease produces multiplexed transcriptional vulnerabilities of striatal D1-D2 and Striosome-Matrix Neurons
Authors: *A. Matsushima, S. S. Pineda, J. R. Crittenden, H. Lee, K. Galani, J. Mantero, M. Kellis, M. Heiman, A. M. Graybiel
Abstract: Striatal cell-type-specific vulnerability in Huntington’s disease (HD) preferentially affects dopamine D2R-expressing projection neurons (SPNs), compatible with manifest motor symptomatology in HD. A second, less fully studied feature of striatal vulnerability involves the compartmental organization of the striatum, with neurochemically specialized labyrinthine ‘striosomes’ thought to be affected especially in relation to premanifest mood symptomatology.To disentangle the cell-type-specific vulnerability in HD, we performed single-nucleus RNA sequencing on striatal samples from two murine models (zQ175 and R6/2) and rare Grade 1 HD patient tissue, and examined striosome and matrix sub-clusters within parent D1 and D2 SPN clusters. In the Grade 1 human HD, striosomal SPNs were the most depleted SPN population. Surprisingly, for both mouse models, transcriptomic distinctiveness was diminished more for striosome-matrix SPNs than for D1-D2 SPNs. Compartmental markers tended to cancel endogenous identities of striosomal and matrix SPNs; striosomal makers were downregulated in striosomal SPNs and upregulated in matrix SPNs, and matrix markers were upregulated in striosomal SPNs and downregulated in matrix SPNs. On the contrary, markers for D1-D2 SPNs exhibited less identity obscuring; they appeared up- and down-regulated in a non-systematic way. The degree of dysregulation (i.e., absolute values of up- or down-regulations) was largest in D2R-expressing SPNs, recapitulating the D2-dominent vulnerability in HD, and reflected in genes upregulated in specific cell types and downregulated in others.These results suggest that striosomes are the first to die in human HD, and that striosome-matrix identities are more vulnerable than those of D1-D2, a pattern that could reflect a differentiation deficiency during development due to loss of function of normal Huntingtin, as proposed previously. Given that D2-dominent transcriptional dysregulation is observed from only about the age of onset, the two axes of striatal organization might be affected differentially in time and in nature, with the striosome-matrix axis affected during development, leading to deficient compartmental identities, and the D1-D2 axis affected later, around the age of onset of motor symptoms.
Functionally distinct sub-regions of the parahippocampal place area revealed by model-based neural control
Authors: *N. Ratan Murty, F. S. Kamps, A. Abate, J. DiCarlo, N. G. Kanwisher
Abstract: Abundant evidence supports a role for the parahippocampal place area (PPA) in visual scene perception, but fundamental questions remain. Here we ask whether the PPA contains distinct sub-regions that encode different aspects of scenes. To address this question, we used data-driven clustering to identify groups of PPA voxels with similar responses to a large set of images in extensively scanned individual brains (185 images, 20 repetitions per image, N = 4). We found that >95% of the variance of PPA voxel responses was explained by just two clusters, mapped approximately along the anterior-posterior axis, consistent with previous findings (Baldassano et al., 2013; Nasr et al., 2013; Cukur et al., 2016; Steel et al., 2021). But what distinct scene features do these sub-regions encode? Responses profiles of the two subregions were quite correlated, and visual inspection of stimuli eliciting high and low responses in each sub-region did not reveal any obvious functional differences between them. We therefore built artificial neural network-based encoding models of each PPA sub-region, which were highly accurate at predicting responses to held-out stimuli (each R > 0.70, P< 0.00001), and harnessed these models to find new images predicted to maximally dissociate responses of the two sub-regions. These predictions were then tested in a new fMRI experiment, which produced a clear double dissociation between the two sub-regions in all four PPAs tested (two participants x two hemispheres each): The anterior sub-region responded more to images containing relatively bare spatial layouts than images containing object arrays and textures, while the more posterior region showed the opposite pattern. Taken together, this approach revealed distinct sub-regions of the PPA and produced highly accurate computational models of each, which in turn identified stimuli that could differentially activate the two subregions, providing an initial hint about the functional differences between them.
Causal inference using the experiences of self and others
Authors: *S. Radkani, S. M. Yoo, M. Jazayeri
Abstract: In social contexts, humans and animals infer latent causes from the experiences of others. However, the computational and neural mechanisms of social inference remain elusive. Here, we report our initial findings in a project aimed at understanding how humans and monkeys infer the latent states of the environment based on their own experience (self-learning) as well as experiences of others (observation-learning). We designed a novel foraging task with two levels of hierarchy. First, players had to choose between two foraging sites and then had to control an avatar with a joystick to intercept passing-by tokens on a display. Importantly, the site with the reward switched covertly after a random number of trials and reward delivery was stochastic with a probability that increased with the number of intercepted tokens. This causal structure made it necessary for players to (1) track the history of decisions, performances, and outcomes to infer the rewarding site, and (2) try to intercept as many of the tokens as possible. Starting with a single-player version, we verified that humans (N=10) and monkeys (N=2) integrate information across multiple trials to update their belief about the rewarding site. Next, we had the same subjects play the two-player version in conspecific pairs. In this version, both players reported their preferred site but only one randomly selected player proceeded to collect tokens in their chosen site (‘actor’) while the other (‘observer’) watched the actor play. Importantly, the observer had visual access to the entire sequence of actions and outcomes making it possible to compare the efficacy of self-learning to observation-learning. We found that, on average, humans discount the outcome brought about by others compared to themselves. This finding reveals an inherent asymmetry between self- and observation-learning even when the external information is perfectly matched between the two modes of learning. Monkeys were also able to learn from one another. However, compared to humans, their behavior showed more sensitivity to outcomes and less sensitivity to decisions and performance of the other animal. They also did not discount observational outcomes as much as humans. We are currently using modeling and electrophysiology to understand the mechanistic factors that lead to discounting social outcomes, and putative species’ differences in observation-learning.
Observational learning in marmosets during a temporal expectation task
Authors: H. Sugihara, S. Mahajan, J. Sharma, G. Feng, R. Desimone, M. Sur
Abstract: Observational learning is the ability to acquire a new skill or optimize one’s behavior through observation, and is a crucial component of social behavior. Multiple species are reported to have this ability, including common marmosets (Callithrix jacchus). Common marmosets are coming into wide use as a model animal for neurological disorders. Here we investigated how observation affects marmosets’ learning strategy using a temporal expectation task. Two marmosets were trained with a home-cage training setup for a visual change detection task. After an initial cue touch to a tablet screen to initiate a trial, an image was presented for a variable duration. The animals were required to report the offset of the image by a second touch to earn a reward. The stimulus duration was randomly chosen from a uniform distribution ranging from 500 to 2500 milliseconds. After prolonged training, the reaction time was modulated by stimulus duration: reaction times were shorter with stimuli of longer duration. The data were well fit by a modified hazard rate model, consistent with an internal model of temporal expectation based on the probability distribution of stimulus durations. Next, we used a well-trained animal as a demonstrator and trained two naive animals on the same task, while the animals were able to observe the demonstrator performing the task. Animals’ behavior was recorded video graphically and the observational behavior was analyzed with DeepLabCut. We split the original temporal prediction task to five simpler steps and measured the acquisition speed of each step. The observer animals mastered the steps faster than animals who were trained alone. The animals learned the task mechanics, such as touching the screen and the association between touch and reward, quicker than the cognitive component of the task, i.e., developing a model of temporal expectation. The relationship between stimulus duration and reaction time of the observer animals, however, was inversely correlated, indicating that observer animals also started developing an internal model of the task. These data demonstrate that observational learning can be accomplished by marmosets, and suggest layered effects of observation onto learning of complex behaviors.
Support: Simons Center for the Social Brain
Temporal expectation in marmosets: changes in event related responses with training
Authors: S. Mahajan, H. Sugihara, T. Dragoi, J. Sharma, G. Feng, R. Desimone, M. Sur
Abstract: Temporal expectation is the ability of animals to use information from past events to predict their occurrence in the future, which allows their brain to allocate appropriate sensory resources in preparation for the event. The brain achieves this by creating an internal model of the event. In the case of temporal expectation, this model can be represented by the hazard rate function which posits the likelihood of an event to occur in the future provided it has not occurred already. To study how this model is acquired by the brain, we devised a simple timing task for freely behaving marmosets and implemented it on a touchscreen tablet. After initiating a trial by touching a cue image, animals were presented with a stimulus image for a duration of time randomly selected from a uniform distribution between 0.5s to 2s. Once the stimulus image disappeared, animals were trained to touch the screen as quickly as possible for a juice reward. As the animals gained expertise at the task, their reaction times were faster for longer stimulus durations, consistent with the hazard rate model. We recorded intracranial electroencephalogram (EEG) signals from four brain regions- the occipital, frontal and right and left parietal cortices, as well as behavioral task data. A trial by trial analysis of event related potentials in the EEG data showed a significant increment in EEG potentials correlated with the anticipation of known events occurring. The most prominent such increment was seen in anticipation of reward delivery after a correct trial. As training proceeded from the naive to expert level, the amplitude of the reward related potentials became inversely proportional to the stimulus duration. Stimulus onset responses increased in the signal from all four brain regions over the course of training, while the modulation of reward related potentiation was significant mainly in the occipital and frontal cortices. Thus, the acquisition and representation of temporal expectation engages widespread cortical regions, with specific loci representing specific epochs of the task as a consequence of training.
Support: Simons Center for the Social Brain