2024

The role of altered neuromodulation in motor dysfunction in Rett syndrome. Giselle Fernandes, Hiroki Sugihara, Ruby Lam, Yuma Osako and Mriganka Sur. The Society for Neuroscience, 2024.

Rett Syndrome (RTT) is a devastating neurodevelopmental disorder that affects 1 in 10,000 females. It is caused by mutations in the X-linked gene methyl-CpG binding protein 2 (MECP2), which is present in both neuronal and non-neuronal cells in the brain. Loss of MeCP2 in astrocytes, the most abundant non-neuronal cell type in the brain, leads to morphological impairments in the dendritic processes of neuronal cells and a range of phenotypes, whereas re-expression of the protein in astrocytes substantially rescues phenotypes arising from neuronal loss. These findings suggest that MeCP2 has an important role in astrocytes, which are in turn crucial for regulating neuronal function in RTT. However, the mechanisms by which MeCP2 exerts its function in astrocytes remains unknown. To address this question, we used primary astrocyte-neuron cultures from WT and MeCP2 KO mice and carried out proteomics and transcriptomics analyses of astrocytes as well as a full proteome analysis of the Rett astrocyte-conditioned media (ACM). Synaptogenic proteins that were dysregulated in RTT ACM included SPARC and SPARC-like protein 1. Mutant astrocytes showed a significant suppression of mitochondrial function and cellular respiration related proteins. Importantly, both mutant astrocytes and Rett ACM revealed an upregulation of insulin-like growth factor binding proteins (IGFBPs) that are known to sequester insulin-like growth factor-1 (IGF-1) and affect IGF-1 signaling, with IGFBP2 being the most upregulated. We confirmed the upregulation of IGFBPs by western blots and quantitative real-time PCR. We asked whether the recent FDA-approved drug for treatment of Rett syndrome, a modified N-terminal peptide of IGF-1 [(1-3)IGF-1], targets astrocyte mechanisms. Our proteomics analysis revealed that treatment of cultures with IGF1 peptide led to reversal of expression of a substantial number of proteins that were down- or up-regulated in MeCP2 null astrocytes. Activated pathways included those related to mitochondrial function and matrix metalloproteinases. Thus astrocytes contribute importantly to the downregulation of IGF-1 signaling in Rett Syndrome, and are a major substrate for the action of IGF1 peptide in the treatment of the disorder.

The JPB Foundation – Picower Postdoctoral Fellowship (G.F.), National Institute of Mental Health/NRSA T32MH112510 (R.M.L.), Japan Society for Promotion of Science-Overseas Research Fellowship (Y.O.), Uehara Memorial Foundation – Postdoctoral Fellowship (Y.O.), R01MH085802, R01MH126351, R01NS130361, R01MH133066 and the Simons Foundation Autism Research Initiative (M.S.)

The Role of Norepinephrine in Astrocyte Signaling in Reinforcement Learning. Gabrielle T. Drummond, Arundhati Natesan, Jennifer Shih, Marco Celotto, Yuma Osako, Stefano Panzeri and Mriganka Sur. The Society for Neuroscience, 2024.

The locus coeruleus (LC) is a small brainstem nucleus that serves as the primary source of the neuromodulator norepinephrine (NE) in the brain. While LC neurons release NE throughout the brain to regulate arousal and attention, recent work from our lab has also shown two distinct functions of LC-NE in reinforcement learning. In an instrumentally conditioned go/no-go task with graded auditory stimulus detection, phasic LC-NE activity is crucial for task execution under high uncertainty conditions and for optimizing task performance after surprising outcomes. LC-NE silencing during this task reduces the amount of information about the stimulus and the choice in individual neurons in the prefrontal cortex (PFC) and motor cortex (MC). Large LC-NE release following a false alarm trial affects neuronal population dynamics to improve discrimination between go and no-go stimuli on the next trial. While the effect of a previous trial’s outcome on the following trial requires LC-NE signals to persist for several seconds, phasic LC-NE activity only lasts for tens of milliseconds. LC-NE has been shown to act on astrocytes, which can integrate and alter neuron signals over diverse timescales, suggesting a means by which information from phasic LC-NE signals can be sustained through the subsequent trial. Here, we explored the effects of LC-NE on astrocytes in the PFC and MC of mice performing our reinforcement learning task. Using 2-photon calcium imaging to record astrocyte and neuronal calcium dynamics in the MC, we find that astrocytes show long-lasting increases in calcium activity following a false alarm. Chemogenetic manipulation of astrocyte calcium using astrocyte-specific GqDREADD-CNO blocks the improvement in performance following a false alarm, suggesting that astrocyte signaling influences this behavioral outcome. To determine how astrocyte manipulations affect neuronal responses during the task, we used high-density neuropixels recordings to analyze PFC and MC neuronal activity while chemogenetically manipulating astrocyte calcium. Disrupting astrocyte calcium activity decreases neuronal encoding during the task. Finally, by knocking down alpha-1a-adrenergic receptors in PFC and MC astrocytes while imaging astrocyte and neuronal responses during the task, we are exploring the role of NE specifically on astrocytes during reinforcement learning. Taken together, our findings indicate a critical role for cortical astrocyte signaling in reinforcement learning.

Support: NIH grants R01MH126351, R01NS130361, R01MH133066; MURI Grant W911NF2110328; the Picower Institute Innovation Fund; the Japan Society for the Promotion of Science

Impaired astrocytic function in Rett syndrome and a potential role of insulin-like growth factor signaling pathway. Prachi Ojha, Alexandria Barlowe, Velina Kozareva, Fabian Schulte, Danielle Tomasello, Yuma Osako, Joachim Theilhaber, Ernest Fraenkel, Rudolf Jaenisch and Mriganka Sur. The Society for Neuroscience, 2024.

Rett Syndrome (RTT) is a devastating neurodevelopmental disorder that affects 1 in 10,000 females. It is caused by mutations in the X-linked gene methyl-CpG binding protein 2 (MECP2), which is present in both neuronal and non-neuronal cells in the brain. Loss of MeCP2 in astrocytes, the most abundant non-neuronal cell type in the brain, leads to morphological impairments in the dendritic processes of neuronal cells and a range of phenotypes, whereas re-expression of the protein in astrocytes substantially rescues phenotypes arising from neuronal loss. These findings suggest that MeCP2 has an important role in astrocytes, which are in turn crucial for regulating neuronal function in RTT. However, the mechanisms by which MeCP2 exerts its function in astrocytes remains unknown. To address this question, we used primary astrocyte-neuron cultures from WT and MeCP2 KO mice and carried out proteomics and transcriptomics analyses of astrocytes as well as a full proteome analysis of the Rett astrocyte-conditioned media (ACM). Synaptogenic proteins that were dysregulated in RTT ACM included SPARC and SPARC-like protein 1. Mutant astrocytes showed a significant suppression of mitochondrial function and cellular respiration related proteins. Importantly, both mutant astrocytes and Rett ACM revealed an upregulation of insulin-like growth factor binding proteins (IGFBPs) that are known to sequester insulin-like growth factor-1 (IGF-1) and affect IGF-1 signaling, with IGFBP2 being the most upregulated. We confirmed the upregulation of IGFBPs by western blots and quantitative real-time PCR. We asked whether the recent FDA-approved drug for treatment of Rett syndrome, a modified N-terminal peptide of IGF-1 [(1-3)IGF-1], targets astrocyte mechanisms. Our proteomics analysis revealed that treatment of cultures with IGF1 peptide led to reversal of expression of a substantial number of proteins that were down- or up-regulated in MeCP2 null astrocytes. Activated pathways included those related to mitochondrial function and matrix metalloproteinases. Thus astrocytes contribute importantly to the downregulation of IGF-1 signaling in Rett Syndrome, and are a major substrate for the action of IGF1 peptide in the treatment of the disorder.

Support: NIH grant R01MH085802 and Simons Foundation Autism Research Initiative (SFARI).

CRISPR/Cas9-based tool reveals contribution of astrocyte-specific GABA transporter, GAT-3, to information processing capacity of neurons in the mouse visual cortex. Jiho Park, Grayson Sipe, Xin Tang, Prachi Ojha, Giselle Fernandes, Yi-Ning Leow, Caroline Zhang, Arundhati Natesan, Gabrielle T. Drummond, Rudolf Jaenisch, and Mriganka Sur. The Society for Neuroscience, 2024.

Astrocytes are increasingly recognized as pivotal constituents of neural circuits governing a wide range of brain functions. They express a diverse array of neurotransmitter receptors and transporters, enabling them to monitor and regulate synaptic and neuronal activity. Specifically, astrocytes express a rich repertoire of GABA-related proteins, suggesting they have a role in the regulation of proper inhibitory signaling in brain circuits. GABA transporter 3 (GAT-3) is exclusively expressed in distal astrocytic processes and is the major astrocyte-specific transporter responsible for maintenance of extrasynaptic GABA levels. Here, we have examined the functional significance of GAT-3 in astrocyte-mediated modulation of neuronal activity. First, we developed a multiplexed CRISPR construct applicable for efficient genetic manipulation of one or more genes in a cell-type specific manner with precise temporal and spatial control and tailored this tool to efficiently knock out astrocyte GAT-3 in the visual cortex of adult mice. GAT-3 knockdown was confirmed with immunohistochemistry and whole-cell patch clamp recordings. GAT-3 knockdown increased the frequency of spontaneous inhibitory currents in the visual cortex, indicating an increase in the levels of ambient GABA. We then examined the effects of GAT-3 knockdown in vivo using 2-photon calcium imaging of visual cortex neurons. GAT-3 reduction led to alterations in single neuronal responses to drifting gratings and natural movies, including a decrease in maximum response magnitude to preferred stimuli and an increase in response variability. While orientation selectivity of neurons did not change on average, GAT-3 knockout exerted a pronounced influence on population-level neuronal activity, impairing the capacity of neuronal populations to accurately represent stimulus information, as revealed by decoding analyses, and integrate network dynamics, as revealed by encoding models of population activity. These findings demonstrate that reducing GAT-3 in astrocytes profoundly alters the information processing capacity of neurons and networks within the visual cortex. Furthermore, the CRISPR/Cas9-based approach employed here represents a powerful tool for dissecting astrocyte-neuron interactions mediated by different types of neurotransmitters and neuromodulators.

Support: F32EY022264 (GS), Simons Foundation Autism Research Initiative Bridge to Independence (XT), The JPB Foundation (GF), R01MH126351, R01NS130361, R01MH133066 (MS)

Population geometry for generalization of cognitive operations in frontal and parietal cortex. Yuma Osako, Yi Ning Leow, Greggory Robert Heller, Sofie Ahrlund-Richter, Yoshihito Saito, Gabrielle T. Drummond, Ziyu Wang, Arundhati Natesan, Tatsuya Osaki, Timothy J. Buschman, and Mriganka Sur. The 47th Annual Meeting of the Japan Neuroscience Society, Fukuoka, Japan, 2024.

The brain flexibly employs various cognitive operations depending on the situation. However, the brain has limited resources and therefore cannot have unique resources to address every possible cognitive operation the organism might encounter. To address this constraint, it can be desirable to encode an information subspace that can be shared and reused among somewhat overlapping cognitive operations (cognitive generalization). However, the mechanism by which the brain shares a common neural space remains elusive. In this study, we investigated the shared subspace between working memory (WM) and action planning (AP) in the mouse. We trained mice on a behavioral task involving distinct cognitive operations and recorded neuronal activity simultaneously in the medial prefrontal cortex (mPFC) and posterior parietal cortex (PPC), regions thought to be associated with cognitive generalization. Mice were trained on a delayed match-to-sample (DMS) task, requiring them to discern whether two consecutive stimuli were matched. During the delay period between stimuli, the stimulus information was maintained as WM. Subsequently, a second delay period served as an AP period, allowing mice to make decisions and prepare for the behavior. This task enabled investigation of different cognitive processes (WM and AP) driven by identical stimuli. Our findings revealed that both the mPFC and PPC represented WM and AP using linearly separable neural decoders. Moreover, the neural subspaces corresponding to these representations were partially overlapped in mPFC, while relatively independent in PPC. Notably, this population architecture in mPFC, but not in PPC, robustly correlated with animal behavior such that the amount of overlapping subspace increased as a function of behavioral performance. Additionally, recurrent neural networks trained on the same task exhibited latent representations akin to those observed in the mPFC. In simulation, network performance degraded when neurons represented WM and AP (mixed selective neurons) were inactivated, implying an efficient geometric structure for achieving cognitive generalization by mixed selective neurons. These results suggest that the mPFC represents WM and AP in a more abstract manner within a high-dimensional space to facilitate cognitive generalization, while the PPC achieves flexibility by utilizing distinct neural subspaces. Future experiments will explore potential interactions between these geometric structures in efficiently representing cognitive flexibility.

Support: NIH grants R01NS130361, R01MH133066, R01MH126351, Wenner-Gren Postdoctoral Fellowship WGF2020-0019​

Integration of visually tuned inputs weighted by dendritic organization in the mouse visual cortex. Kyle Jenks, Greggory Heller, Katya Tsimring, Emma Odom, and Mriganka Sur. Cosyne, 2024.

Neurons in the mouse visual cortex (V1) receive broadly tuned inputs at thousands of dendritic spines. Despite this, only a fraction of neurons respond to any given stimulus. It is unclear how variability in neuronal responses is influenced by the structural and functional properties of individual spines. Here, we address this by performing in-vivo two-photon imaging of more than 4,500 dendritic spines on L2/3 pyramidal neurons in the mouse binocular visual cortex during presentation of drifting gratings. With this dataset we compare the orientation tuning of spines to their respective soma. Visually responsive neurons have a higher fraction of responsive spines compared to unresponsive neurons. Additionally, responsive spines on responsive somas have higher orientation selectivity and are more similarly tuned to the soma than responsive spines on unresponsive somas. The results demonstrate that the quantity and quality of visual selectivity of spine inputs correlates with the visual selectivity of the soma. The structure and location of dendritic spines within the dendritic arbor are hypothesized to weight the influence of a given spine on the soma’s response. To test how dendritic organization impacts the spine-input to soma-output relationship we developed a linear integration model which linearly sums recorded inputs from spines and generates an in-silico orientation tuning curve. We compared the model to the in-vivo visual tuning of the soma and found that the model generates a similar tuning curve for responsive somas, but not for unresponsive somas as expected. Next, we differently weight the inputs by biophysical features such as spine size or distance to identify those which are relevant to the neuron’s input-output computation. Results from our study will not only enhance our understanding on how presynaptic inputs are transformed into an informative and reliable single neuron output but will also provide insight into building biologically constrained neural networks.ients and suggests the potential translational application of these findings for drug screening.

Support: NIH grant MH126351, Picower Institute Innovation Fund, F31 EY033649-01 (KT)

Hebbian and heterosynaptic plasticity regulate orientation matching in binocular cortical circuits during the critical period. Katya Tsimring, Claudia Cusseddu, Kyle Jenks, Greggory Heller, Julijana Gjorgjieva, Jacque Ip, and Mriganka Sur. Cosyne, 2024.

Experience-dependent plasticity refines immature sensory circuits during critical periods in development. In the binocular visual cortex (bV1), visual experience aligns information from the ipsilateral and contralateral eye onto bV1 neurons. It remains unclear how this experience-dependent alignment between the two eyes takes place at the synaptic level. We propose that Hebbian and heterosynaptic mechanisms regulate the alignment of inputs onto bV1 neurons during the critical period by modifying synapses based on their correlation to the post-synaptic neuron and their synaptic neighbors, respectively. To investigate the synaptic mechanisms that underlie somatic orientation matching, we used in vivo two photon calcium imaging to chronically track eye-specific and binocular visual responses of neurons and dendritic spines. We observed significant turnover of dendritic spines from textasciitilde postnatal (p)22 (D1) to textasciitilde p32 (D10), as well as shifts in the functional properties of retained spines. To examine Hebbian-like mechanisms of regulating dendritic spine turnover, we compared tuning curves of spines with the soma’s. Early in development, we found that newly added spines were less aligned to the soma’s tuning properties than retained spines. By D10, spines that were misaligned to the soma were lost, suggesting Hebbian refinement of visual circuits. To examine heterosynaptic interactions, we measured trial-to-trial correlations between spine pairs and found a distance-dependent relationship between co-active spines. Furthermore, this correlation between spine neighbors increased from D1 to D10 with the addition of co-tuned spines. Our results suggest that heterosynaptic mechanisms stabilized newly added spines. To quantify contributions of Hebbian and heterosynaptic plasticity in somatic orientation matching, we built a computational model to simulate dendritic spine turnover. Simulations removing heterosynaptic plasticity prevented orientation matching at the soma, indicating that Hebbian mechanisms alone cannot drive this alignment. Overall, our research provides critical insights into how synaptic mechanisms shape sensory circuits and enhance information encoding across development.

Support: NIH grant MH126351, Picower Institute Innovation Fund, F31 EY033649-01 (KT)

Engineering 3D endothelial vascular networks from Rett syndrome patient-derived iPS cells. Tatsuya Osaki, Zhengpeng Wan, Koji Haratani, Marco Campisi, David A. Barbie, Roger Kamm, and Mriganka Sur. MPS World Summit, 2024.

The mutations in the X-linked methyl-CpG binding protein 2 (MECP2) genes, Rett syndrome (RTT)-associated gene mutations, are widely believed to downregulate and upregulate numerous downstream genes, resulting in abnormal neuronal activity. Their symptoms typically appear between 6-12 months along with the loss of purposeful hand skills, speech and social engagement, motor abnormalities, and cognitive impairments. Disease severity can vary significantly among individuals due to the variation of MeCP2 mutations. Although MeCP2 is a non-cell type-specific DNA binding protein and its mutation influences not only neural cells but also non-neural cells in the brain including vascular networks associated with endothelial cells. To date, limited evidence of impaired vascular function has been reported in MeCP2-null mouse models. However, whether altered brain vascular homeostasis and subsequent blood-brain barrier (BBB) breakdown occur in Rett syndrome, and whether they contribute to disease-related cognitive impairment, especially with specific MECP2 SNP mutations, remains unknown. Here, we engineer a 3D endothelial vascular network in microfluidic devices by Rett syndrome patient-derived induced pluripotent stem (iPS) which carries MeCP2[R306C] mutation. Notably, the MeCP2[R306C] mutation does not affect affinity to the CpG binding domain but disrupts the interaction with the NCoR corepressor complex, a key corepressor. To expedite endothelial cell differentiation, doxycycline (DOX)-inducible ETV2 expression vectors were inserted into the AAVS1 locus of Rett syndrome patient-derived iPS cells and its isogenic control by CRISPR/Cas9. This strategy facilitated rapid differentiation into endothelial cells (iEC) within five days in the presence of DOX. Differentiated endothelial cells were subsequently embedded in a fibrin gel and seeded into a microfluidic device with supporting cells. Permeability assays using fluorescent-conjugated dextran solutions revealed that the 3D vascular network formed by Rett syndrome patient-derived iPS cells exhibited higher permeability and slightly smaller vessel diameters compared to isogenic control vascular networks. Immunostaining analysis and flow cytometry-based high-content analysis demonstrated downregulation of gene expression related to tight junctions, such as ZO-1, claudin-5, and occludin, in mutated endothelial cells. To understand how the MeCP2 mutation influenced this impaired vascular integrity, we also performed RNAseq. This result indicated that MeCP2 mutation might caused lower TIMP1 gene expression. This study provides evidence of impaired vascular function associated with TIMP1 and the MECP2 mutation in Rett syndrome patients and suggests the potential translational application of these findings for drug screening.

Support: NIH grant R01MH085802, the Simons Foundation Autism Research Initiative (SFARI) through the Simons Center for the Social Brain (SCSB), U01-CA214381(NCI), and U01-CA214381(NCI)

2023

Behavioral states modulate the relationship between top-down input and inhibition within the mouse visual cortex that differentially shape visual processing. Kyle Jenks, Sofie Ährlund-Richter, Grayson O Sipe, Mriganka Sur. Society for Neuroscience, 2023.

Sensory encoding within the visual cortex can shift dynamically based on an animal’s behavioral state, which may serve to optimize encoding for distinct behavioral needs. Locomotion and pupil diameter (a measure of arousal) are two behavioral state spaces known to alter visual encoding. However, the signals that drive these state-dependent shifts in encoding are unclear. Top-down input from higher cortical areas, such as the anterior cingulate cortex (ACC), can modulate information encoding in the visual cortex and likely contributes to these state-dependent shifts. The activity of a subset of inhibitory interneurons in the visual cortex, primarily VIP and NDNF positive interneurons, correlates with locomotion and pupil diameter; importantly, these neuron classes also receive input from ACC and other higher cortical areas and thus can mediate top-down influences. Indeed, their activity may represent a confluence of top-down modulation and state-dependent signals. However, these roles have largely been studied in isolation using different paradigms, and we have little understanding of how, for example, top-down signals in these interneurons are related to distinct behavioral states or vary across behavioral state transitions. To better understand these relationships, we leveraged dual-color in vivo calcium imaging to simultaneously record combinations of excitatory neurons, VIP interneurons, NDNF interneurons, and ACC axons within the mouse visual cortex as the mice viewed visual stimuli. We also simultaneously recorded locomotion and pupil diameter and delivered unexpected air puffs to evoke behavioral state shifts.

As expected, we found that excitatory visual responses and baseline excitatory population activity vary across behavioral states. VIP and NDNF interneurons both respond to visual stimuli and increase their activity in response to pupil dilation, but their time courses post-dilation diverge, with VIP activity remaining elevated longer than NDNF activity. VIP activity, but not NDNF activity, increases when the mouse begins running. ACC axons are visually responsive and selective, and these responses are modulated by running speed and pupil diameter. The different patterns and time courses of activity we observe in ACC axons and inhibitory interneurons across behavioral state transitions, and their correlations with changes in excitatory visual encoding, suggest that these signals coregulate state-dependent transitions within the visual cortex.

Support: KRJ; F32EY032756. SAR; WGF2020-0019. GS; K99AA028579. MS; R01EY028219, R01MH126351.

Long distance cortical projections from anterior cingulate cortex to visual and retrosplenial cortical areas in Rett syndrome model mice. Jonathan Harpe, Sofie Ährlund-Richter, Mriganka Sur. Society for Neuroscience, 2023.

Neurodevelopmental disorders, including autism spectrum disorders, are commonly characterized as “connection disorders”, with emerging evidence suggesting a role for disrupted local and long-range connections as features underlying their phenotypes. Rett Syndrome is a devastating neurodevelopmental disorder that largely affects girls and is caused by mutations in MECP2. Rett model mice with MeCP2 mutations are known to have disrupted synaptic function and plasticity as well as reduced neuronal maturation, with a variety of motor and cognitive deficits likely related to altered brain connections. Whether or not MECP2 mutations disrupt long-range connections in the brain is unknown. The anterior cingulate cortex (ACA) is a crucial brain region involved in cognitive processes. It plays a significant role in top-down control of attention, decision-making and emotional regulation, via its long-range influences on diverse brain structures. Here, we examined whether top-down long-range projections from the ACA to visual and adjacent areas of neocortex are altered in MeCP2 mutant mice. AAV1-CAG-TdTomato was injected in ACA of 3-month-old MECP2 heterozygous mice to trace anterograde projections in target regions, and compared with projections in age-matched wild-type control mice. Confocal microscopy was used to capture serial sections, and brain regions were aligned and registered with the Allen Brain Atlas for quantifying anterograde axonal densities based on anatomical location. In particular, we aimed to identify whether cortical regions such as primary visual cortex, medial and lateral higher visual cortical areas, and retrosplenial cortex, displayed differences in areal and laminar termination sites and densities. Preliminary analysis of the data reveals that similar to wild-type mice, Rett model mice show dense ACA projections to retrosplenial cortex with sparser projections to lateral visual areas, especially targeting superficial cortical layers including layer 1. Axonal termination densities in the Rett model mice may exhibit a tendency towards a broader distribution across the layers and borders of target cortical areas when compared to wild-type mice. Thus, MECP2 mutations may thus be associated with a more generalized pattern of top-down axonal projections originating from the ACA, potentially leading to altered laminar and areal specificity in target regions. While further analyses are required to confirm these observations, they are consistent with reduced neuronal maturation hypothesized to be a key feature of Rett Syndrome, and suggest that changes in long-range cortical projection systems may be an important feature of the disorder.

Supported by: MIT Research Scholars Program (JH), Wenner-Gren Fellowship (SAR), Picower Institute Innovation Fund, NIH grant MH085802

Discrete PFC subregions provide distinct feedback modulation to the visual cortex that differentially shape visual processing. Sofie Ährlund-Richter, Yuma Osako, Kyle Jenks, Mriganka Sur. Society for Neuroscience, 2023.

The prefrontal cortex (PFC) is a source of top-down feedback modulation for the visual cortex- guiding, biasing or modulating activity in the visual cortex to optimize visual processing. The visual cortex of the mouse receives monosynaptic input from two discrete PFC subregions, the anterior cingulate cortex (ACA) and the orbitofrontal cortex (ORB), which are implicated in distinct aspects of cognition. However, it is currently unknown if these discrete PFC subregions contribute to different aspects of visual attention and perception, and how this process relates to the behavioral state of the animal and previously experienced visual contexts. To better understand how discrete PFC subregions shape the activity of the visual cortex, we performed in-vivo calcium imaging of ACA and ORB axonal activity in the visual cortex of mice viewing visual stimuli. In parallel, behavioral variables such as running speed, pupil size and face movements were recorded, and unexpected air puffs were delivered to evoke behavioral state shifts. To be able to parcellate the discrete activity communicated uniquely to the visual cortex versus more global behavioral signals, the ACA and ORB axonal activity was also recorded in the primary motor cortex to allow for comparison. We found that both ACA and ORB axonal activity in the visual cortex incorporate visual and behavioral variables. Using generalized linear models (GLMs) of axonal activity, we found that ACA axons encode visual information more strongly than ORB axons, and that their visual responses scale with contrast. Importantly, the observed PFC axons’ activity was more aligned to its source region, rather than the cortical regions they terminated within (i.e., they conveyed activity globally across cortex). To directly evaluate the impact of distinct PFC inputs on the activity of the visual cortex, we also recorded the activity of visual cortex neurons with or without DREADD-mediated inhibition of ACA or ORB neurons projecting to the visual cortex. We are currently evaluating the effects of feedback modulation from the ACA or the ORB on the visual cortex at both population and single neuron levels. Our work so far suggests that even in a passive viewing context, the ACA and the ORB are actively conveying information in regards to the visual scene and behavioral state of the animal that shape the activity of the visual cortex.

Support: R01MH126351, R01NS130361, R01MH133066, F32EY032756, Picower Institute Innovation Fund (PIIF)

Pulvinar-prefrontal inputs facilitate adaptive updating with sensory history during visual decision-making. Yi-Ning Leow, Alexandria Barlowe, Cindy T Luo, Yuma Osako, Mehrdad Jazayeri, Mriganka Sur. Society for Neuroscience, 2023.

Processing sensory information to generate decisions and action is a central component of learned, goal-directed behavior. Our sensory landscape isconstantly filtered through our prior expectations and ongoing goals. Under greater perceptual uncertainty, perceptual processing becomes moresusceptible to the influence of prior history, which can induce biases that compromise decision-making. However, trial history can also be usedadaptively to guide decisions when perceived similarity across stimuli is used to guide switching or repeating actions. We trained mice on a two-choicerandom dot motion discrimination task, varying sensory certainty with the proportion of coherent target dot directions. We found that mice adopted astrategy of comparing the similarity of stimulus directions across trials to make switch/stay decisions. The Anterior Cingulate Cortex (ACC) is a frontalregion that integrates uncertainty for switch/stay decisions from diverse inputs including the pulvinar/lateral posterior (LP) nucleus (LP, rodenthomolog), a higher-order visual thalamic nucleus. The pulvinar serves as an integrative hub with reciprocal connectivity with the ACC and many othervisual and associative cortices. Pulvinar, specifically its interactions with the frontal cortex, has been implicated in effective visual decision-making andpredictive processing, but the information conveyed by this pathway has been challenging to resolve given the broadly divergent projections ofpulvinar neurons. Leveraging genetic tools available in rodent models, we performed projection-specific two-photon calcium imaging and optogeneticmanipulations of LP-ACC axons in mice performing the visual discrimination task. We show that LP-ACC visual stimulus representations were gated bytask engagement, and encoded multiple task variables scaled by stimulus uncertainty. LP-ACC responses were also highly modulated by previous trialinformation, which was also graded by the uncertainty associated with the previous trial. Critically, we found that estimated similarity in stimulusdirections across trials also modulated LP-ACC activity. Optogenetic activation of LP-ACC axons during stimulus evaluation impaired perceptualdiscrimination and reduced psychometric slopes, due to altered interactions between current and previous trial information. Our findings demonstratethat the LP-ACC inputs support decision-making by providing a read-out of ongoing uncertainty, integrated over time with behavioral history, toadaptively tune neuronal responses and guide goal-directed behavior on a trial-to-trial basis.

Support: A*STAR NSS (BS-PhD) (YNL), R01MH126351, R01NS13036, R01MH133066, ARO MURI Grant W911NF2110328, Picower Institute Innovation Fund (PIIF)

Effects of locus coeruleus norepinephrine signaling on cortical neurons, neuronal populations, and astrocytes during reinforcement learning. Gabi Drummond, Jennifer Shih, Marco Celotto, Yuma Osako, Jiho Park, Vincent Breton-Provencher, Stefano Panzeri, Mriganka Sur. Society for Neuroscience, 2023.

The locus coeruleus (LC) is a small brainstem nucleus and the primary source of the neuromodulator, norepinephrine (NE), in the brain. Through a widely divergent set of projections, LC neurons release NE throughout the brain to regulate arousal and attention. However, our recent work has also indicated two distinct functions for LC-NE in reinforcement learning. In an instrumentally conditioned go/no-go task with graded auditory stimulus detection, phasic LC-NE activity is critical for task execution under high uncertainty conditions, and for optimizing the decision boundary and task performance after surprising outcomes. It is unknown how LC-NE activity alters neuronal responses and population dynamics in target regions to facilitate task execution and to promote reinforcement encoding and influence behavior. Here, we explored the effects of LC-NE on cortical neurons and astrocytes in prefrontal cortex (PFC) and motor cortex (MC) during our reinforcement learning task. We used neuropixels probes to generate high density recordings of single units in PFC and MC in mice performing the task while silencing LC-NE neurons optogenetically on a subset of trials. Using mutual information analyses, we find that LC-NE silencing reduces the amount of information about the stimulus (go or no-go tone) and the choice (press or no press) in individual cortical neurons. Using population analyses, we find that LC-NE release following a false alarm changes population dynamics to improve discrimination between go and no-go tones on the next trial. Phasic LC-NE activity lasts only milliseconds, while the effect of a trial outcome on the next trial requires that the signal persists for several seconds. LC-NE has been shown to act on astrocytes, the major glial cell of the brain, which can integrate and alter neuronal signals over diverse timescales, presenting a means by which the information from phasic LC-NE signals can be sustained through the next trial. Thus, we used 2-photon calcium imaging to record astrocyte and neuron calcium dynamics in the cortex during the task. We find that astrocytes show reliable and long-lasting increases in calcium following a false alarm. When we chemogenetically manipulate astrocyte calcium, we no longer see an improvement in performance following a false alarm, indicating that astrocyte dynamics are important for this behavioral outcome. Finally, by chemogenetically manipulating LC-NE neuronal activity while imaging astrocyte calcium, we are exploring the role of NE action on astrocytes in mediating this sustained increase in astrocyte calcium and its impact on learned behavior. Together, our results indicate that phasic LC-NE during pre-execution and post-reinforcement epochs alter neuronal population dynamics and astrocyte calcium to facilitate task execution and optimization.

Support: F31MH129112 (GD); the JPB Foundation fellowship (JS); JSPS fellowship (YO); NIH grants R01MH126351, R01NS130361 and R01MH133066; MURI Grant W911NF2110328; and the Picower Institute Innovation Fund. VBP: BBRF Young Investigator Award , Future Leaders in Canadian Brain Research Program, FRQS Salary Award and Start up grant (#324009)

Distinct population architecture for cognitive flexibility in frontal and parietal cortex. Yuma Osako, Yoshihito Saito, Yi Ning Leow, Greggory Robert Heller, Sofie Ährlund-Richter, Ziyu Wang, Gabrielle T. Drummond, Tatsuya Osaki, Mriganka Sur. Society for Neuroscience, 2023

Cognitive flexibility, a vital ability for mammals to adapt to diverse situations and environments, is hindered by the curse of dimensionality in neuronal activity. To address this challenge, leveraging shared information among cognitive operations within the system to reduce information dimensionality is advantageous. However, the underlying mechanisms of such cognitive operations remain elusive. In this study, we trained mice on a behavioral task involving distinct cognitive operations and recorded neuronal activity simultaneously in the anterior cingulate cortex (ACC) and posterior parietal cortex (PPC), regions thought to be associated with cognitive flexibility. Mice were trained on a delayed match-to-sample (DMS) task, requiring them to discern whether two consecutive stimuli were matched. During the delay period between stimuli, the stimulus information was maintained as working memory (WM). Subsequently, a second delay period served as an action planning (AP) period, allowing mice to make decisions and prepare for the behavior. This task enabled investigation of different cognitive processes (WM and AP) driven by identical stimuli. Our findings revealed that both the ACC and PPC represented WM and AP using linearly separable neural decoders. Moreover, the neuronal subspaces corresponding to these representations were independent (orthogonal) in PPC, while partially overlapping in ACC. Notably, this population architecture correlated with animal behavior, collapsing during error trials. Additionally, recurrent neural networks trained on the same task exhibited latent representations akin to those observed in the ACC, implying an efficient geometric structure for achieving cognitive flexibility. These results suggest that the ACC represents WM and AP in a more abstract manner within a higher-dimensional space to facilitate cognitive flexibility, while the PPC achieves flexibility by utilizing distinct neural subspaces. Future experiments will explore potential interactions between these geometric structures in efficiently representing cognitive flexibility.

Support: R01NS130361, R01MH133066

Impaired astrocytic signaling contributes to the synaptic deficits seen in Rett Syndrome. Prachi Ojha, Mriganka Sur. Society for Neuroscience, 2023.

Rett syndrome (RTT) is a devastating neurodevelopmental disorder that is caused by mutations in the X-linked gene, methyl-CpG binding protein 2 (MECP2). Astrocytic loss of MeCP2 has been implicated in diminished neurite outgrowth seen in Rett syndrome. Restoration of MeCP2 specifically in astrocytes rescues the behavioral deficits seen in RTT. However, the mechanisms underlying regulation of astrocytic function by MeCP2 are unclear. We hypothesized that knockdown (KD) of MeCP2 leads to abnormal secretion of synaptogenic proteins from astrocytes which leads to impaired synaptogenesis and activity-dependent synaptic plasticity in RTT. To test this hypothesis, primary cortical neurons and astrocytes from P0 mice were co-cultured and the effects of knocking down MeCP2 specifically in astrocytes were examined. We observed that wild-type (WT) neurons cultured with MeCP2-KD astrocytes had reduced synaptogenesis comparable to MeCP2-KD neurons cultured with MeCP2-KD astrocytes. This decrease was rescued when MeCP2-KD neurons were cultured with WT astrocytes. Interestingly, this decrease in synaptogenesis was seen only in excitatory synapses (colocalized PSD-95/ Bassoon punctae) and not in inhibitory synapses (colocalized Gephyrin/Bassoon punctae). To elucidate the mechanism by which loss of MeCP2 in astrocytes causes this decrease, a qPCR screen of all the known astrocyte- secreted synaptogenic molecules in mutant astrocytes was done. We observed that Ephrin A3 (EphA3) was highly upregulated in mutant astrocytes. Interestingly, EphA3 in astrocytes interacts with EphA4 receptors on neurons to regulate the levels of synaptic AMPA receptors and glutamate transporters. It has also been shown to regulate hippocampal dendritic spine morphology. Thus, we predict that overexpressing EphA3 in WT astrocytes will partially recapitulate the synaptic deficits seen in neurons cocultured with MeCP2-KD astrocytes. Some other genes that were dysregulated in mutant astrocytes include EphA2, EphB4, Nlgn-2 and 3, and thrombospondin. Most of these genes are involved in excitatory synapse formation which may underlie the decrease in the number of excitatory synapses observed earlier. The active peptide of insulin-like growth factor 1 [(1-3)IGF-1] which has been FDA-approved for treating RTT has been shown to increase the synaptic marker synapsin in neurons and activate PI3 kinase signaling in astrocytes. Currently, we are investigating whether (1-3)IGF-1 treatment of mutant astrocytes can rescue the levels of the altered synaptogenic proteins in astrocytes, pointing to additional mechanisms by which (1-3)IGF-1 exerts its therapeutic effects on neurons.

Support: Simons Foundation Autism Research Initiative (SFARI), R01MH085802

Determining the synaptic mechanisms behind somatic orientation matching in the binocular visual cortex. Katya Tsimring, Claudia Cusseddu, Kyle R. Jenks, Greggory Heller, Julijana Gjorgjieva, Jacque P.K. Ip, Mriganka Sur. Society for Neuroscience, 2023

During critical periods in postnatal development, cortical circuits undergo significant plasticity and refinement due to sensory experience. This is particularly evident in the rodent binocular visual cortex (bV1), where visual experience from postnatal day (p)22 to p34 is critical for rewiring binocular circuits to align visual information coming from the ipsilateral (ipsi) and contralateral (contra) eye. As a result, the preferred orientation of bV1 neurons becomes matched between ipsi and contra eye inputs. While the maturation of bV1 neurons has been well-characterized at the somatic level, no studies have thus far investigated whether somatic orientation matching is linked to modifications at the synaptic level and which plasticity mechanisms facilitate this alignment. Here, we hypothesized that Hebbian and heterosynaptic plasticity regulate somatic orientation matching during the critical period by strengthening synaptic inputs correlated with the postsynaptic neuron or with synaptic neighbors, respectively, and by weakening those that are uncorrelated. To test our hypothesis, we used in vivo two photon calcium imaging to longitudinally track the eye-specific and binocular visual responses of neurons and their dendritic spines. At the somatic level, we found that responses to the ipsi eye become gradually aligned with the soma’s binocular preference over development, whereas contra eye responses are already matched at the onset of the critical period (p22). These result suggests that inputs from the ipsi eye undergo plasticity to drive somatic orientation matching. At the synaptic level, we found that a majority of dendritic spines exhibit dynamics in their structural and functional properties from p22 to p34. To determine whether Hebbian and heterosynaptic plasticity impact the structural turnover of dendritic spines, we performed a signal correlation analysis between the tuning properties of dendritic spines and the soma, and a trial-to-trial correlation analysis between dendritic spines and their neighbors (within 5um). We found that retained spines are more correlated to somatic tuning preferences than those that have been added or lost, indicating a role for Hebbian-like mechanisms. On the other hand, trial-to-trial correlations to spine neighbors are not significantly different among retained, lost, or added spines, suggesting that hetero-synaptic interactions may not affect spine turnover. To quantify the necessity and sufficiency of Hebbian and heterosynaptic plasticity rules in driving somatic orientation matching, we built a computational model that simulates the turnover of dendritic spines from p22 to p24. On simulated data, we find that Hebbian plasticity alone cannot drive the alignment of orientation matching. We are currently testing whether our model can recapitulate the final state of synapses we observe in vivo using the initial functional and structural properties of dendritic spines at p22 to evaluate whether the plasticity mechanisms in our model are reflective of physiological processes. Overall, our research provides critical insights into how synaptic mechanisms shape the refinement of sensory circuits.

Support: F31EY033649-01(KT), F32EY032756 (KJ); R01EY028219, R01MH126351 (MS), SmartNets 860949 (CC)

Uncovering the synaptic basis for the alignment of orientation preference in the binocular visual cortex. Katya Tsimring, Kyle R. Jenks, Claudia Cusseddu, Julijana Gjorgjieva, Jacque P.K. Ip, Mriganka Sur. Society for Neuroscience, 2023.

Experience-dependent plasticity is important for refining immature sensory circuits during critical periods in development. Through visual experience, neurons in the binocular visual cortex (bV1) become aligned to the same orientation preference from the ipsilateral (ipsi) and contralateral (contra) eye during the critical period for ocular dominance plasticity (~p22-p34 in rodents). However, the synaptic mechanisms underlying the alignment of somatic orientation tuning remains unclear. We propose that Hebbian and heterosynaptic mechanisms regulate the alignment of visual inputs onto bV1 neurons during the critical period by modifying synapses based on their correlation to the postsynaptic neuron and to synaptic neighbors, respectively. To test this hypothesis, we have chronically tracked the eye-specific and binocularly driven visual responses of neurons and their dendritic spines using in vivo two-photon calcium imaging. We find that ipsi eye driven responses become aligned to the soma’s binocular preference over development, whereas contra eye driven responses are already matched at the start of the critical period. At the synaptic level, about 60% of dendritic spines exhibit structural turnover between p22 and p34. However, of the dendritic spines that are retained, about 85-90% exhibit loss, gain or changes in their eye-specific preferences. What can explain the dynamics we observe in the functional and structural properties of synaptic inputs? Intriguingly, we find that dendritic spines driven by both contra and ipsi eye inputs exhibit the least structural turnover among eye-specific inputs, suggesting that of these inputs may be essential for driving binocularity in somatic responses. Furthermore, retained dendritic spines tend to be more aligned to the soma’s binocular orientation preference and are more correlated to their neighboring spines than those that are lost. This indicates that Hebbian and heterosynaptic-like mechanisms are at play in stabilizing these inputs. To quantify the contribution of Hebbian and heterosynaptic mechanisms in development, we are building a computational model that simulates the dynamics of dendritic spines between p22 and p34 by incorporating the structural and functional properties of dendritic spines imaged in vivo. By comparing the final state of synapses in the model with our experimental data at p34, we can assess which plasticity rules are necessary and sufficient to explain our results. Our work will thus provide critical insight into the nature and mechanisms of experience-dependent plasticity during development.               

Support: F31EY033649-01(KT), F32EY032756 (KJ); R01EY028219, R01MH126351 (MS)

The geometry of cognitive flexibility in frontal and parietal cortex. Yuma Osako, Yi Ning Leow, Greggory Robert Heller, Sofie Ährlund-Richter, Yoshihito Saito, Gabrielle T. Drummond, Tatsuya Osaki, Mriganka Sur. The 46th Annual Meeting of the Japan Neuroscience Society, 2023.

Cognitive operations that can adapt flexibly to different situations and environments are a fundamental function for mammals to survive. Various　cognitive operations are plagued by the curse of dimensionality in neural activity. To resolve this, it would be beneficial to reuse information that is shared among different cognitive operations within the system to reduce the dimensionality of information representation. However, it is still unclear how the brain realizes such cognitive operations. In this study, we trained mice on a behavioral task requiring different cognitive operations and simultaneously recorded the neuronal activity of the anterior cingulate cortex (ACC) and the posterior parietal cortex (PPC), areas thought to be involved in cognitive flexibility. Mice were trained in a delayed match-to-sample (DMS) task in which two consecutive stimuli were presented with an intervening delay, and animals were required to make a choice to indicate whether the stimuli were matched. During the delay period between the stimuli, the first stimulus was maintained as working memory (WM) in the brain. After the second stimulus was presented, a second delay served as an action planning (AP) period to give the mice time to make a decision. This task enabled us to investigate different cognitive processes (WM and AP) driven by identical stimuli. We found that both the ACC and PPC represented WM and AP with linearly separable neural decoders. Furthermore, the neuronal subspaces of these two representations were found to be independent (orthogonal) from each other in PPC, but partially overlapped in ACC. Additionally, recurrent neural networks trained on the same task showed latent representations comparable to the ACC, suggesting that the ACC representation may have an efficient geometric structure for achieving cognitive flexibility. These results suggest that the ACC represents WM and AP in a more abstract way in a higher-dimensional space to achieve cognitive flexibility, while the PPC achieves cognitive flexibility by using different neural subspaces. In future experiments, we will investigate whether these different geometric structures interact to represent cognitive flexibility efficiently.

Support: R01NS130361, R01MH133066

Label-free three-photon imaging of intact human cerebral organoids: tracking early events in brain development and deficits in Rett Syndrome. Murat Yildirim, Chloe Delepine, Danielle Feldman, Vincent Pham, Stephanie Chou, Jacque Pak Kan Ip, Alexi Nott, Li-Huei Tsai, Guo-li Ming, Peter T.C. So, Mriganka Sur. Optica OLS Meeting, 2023.

We demonstrate label-free three-photon imaging of intact organoids (~2 mm depth) derived from Rett syndrome patients. Long-term imaging of live organoids shows that mutant neurons have shorter migration distances, slower migration speeds and tortuous trajectories.

Brain-wide, specialized and state-dependent cortical encoding of reward, value and action switching during reversal learning. Murat Yildirim, Nhat Le, Yuma Osako, Yizhi Wang, Abigail Dulski, Alexandria Barlowe, Hiroki Sugihara, Peter So, Mriganka Sur. COSYNE, 2023.

2022

Large scale ECoG recording during social scene watching and social behaviors in marmosets. H. Xu, Y. Su, J. Sharma, W. Menegas, C. Trimmer, F. Liang, R. Landman, B. Zhang, M. Sur, R. Saxe, G. Feng, R. Desimone.

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.

Spatiotemporal dynamics and targeted functions of locus coeruleus norepinephrine in a learned behavior. Gabi Drummond, Vincent Breton-Provencher, Mriganka Sur. COSYNE, 2022

The locus coeruleus (LC) serves as the primary source of the neuromodulator, norepinephrine (NE), in the brain. Through a widely divergent set of projections, LC neurons have been suggested to release NE ubiquitously to regulate arousal and attention. However, LC-NE is implicated in more precise roles such as mediating learning, promoting task execution, and signaling unexpected uncertainty. Whether and how LC-NE activity facilitates these distinct aspects of behavior is unknown. Here, we show that LC-NE activity displays distinct spatiotemporal dynamics to enable two functions during a learned behavior—facilitating task execution under conditions of uncertainty, and encoding reinforcement to improve performance accuracy. To examine these functions, we used a behavioral task with graded auditory stimulus detection and task performance. Optogenetic inactivation of the LC demonstrated that LC-NE activity was causal for both task execution and optimization. Targeted recordings of LC-NE neurons using photo-tagging, two-photon micro-endoscopy, and two-photon output monitoring showed that transient LC-NE activation preceded behavioral execution and followed reinforcement. These two components of phasic activity were heterogeneously represented in LC-NE cortical outputs, such that the behavioral response signal was higher in motor cortex and facilitated task execution, whereas the negative reinforcement signal was widely distributed among cortical targets and improved performance on the subsequent trial. Modular targeting and spatiotemporal integration in target regions thus enable diverse functions, whereby some NE signal are segregated amongst targets while others are broadly distributed. 

Support: R01EY028219, R01MH126351, R01MH085802, PIIF, Simons Foundation Autism Research Initiative, F31MH129112-01A1​

Effects of locus coeruleus norepinephrine signaling on target regions during reinforcement learning. Gabi Drummond, Jennifer Shih, Yuma Osako, Jiho Park, Vincent Breton-Provencher, Mriganka Sur. Society for Neuroscience, 2022

The locus coeruleus (LC) is a small brainstem nucleus and the primary source of the neuromodulator, norepinephrine (NE), in the brain. Through a widely divergent set of projections, LC neurons release NE throughout the brain to regulate arousal and attention. However, recent work from our lab has also shown two distinct functions for LC-NE in reinforcement learning. In an instrumentally conditioned go/no-go task with graded auditory stimulus detection, phasic LC-NE activity is critical for task execution under high uncertainty conditions, and for optimization after surprising outcomes. These results suggest that spatiotemporal dynamics enable LC-NE circuits to facilitate task execution and to promote reinforcement encoding and performance optimization. Yet, it remains unknown how these signals alter target circuits to change behavior. Phasic LC-NE activity lasts only milliseconds, while the effect of a trial outcome on the next trial requires that the signal persists for several seconds. LC-NE has been shown to act on astrocytes, the major glial cell of the brain. Astrocytes can integrate and alter neuronal signals over diverse timescales, presenting a means by which the information from phasic LC-NE signals can be sustained through the next trial. Alternatively, neurons have been suggested to hold information in memory through recurrent network dynamics. Here, we explored the effects of LC-NE on cortical neurons and astrocytes in prefrontal cortex (PFC) and motor cortex (MC) during our reinforcement learning task. We used neuropixels probes to generate high density recordings of single units in PFC and MC in mice performing the task while silencing LC-NE neurons on a subset of trials. We then used population analyses to determine the effects of LC-NE on cortical population dynamics in each region during select task epochs. Using 2-photon calcium imaging, we recorded astrocyte and neuron calcium dynamics in MC and PFC in mice performing the task, and measured the effects of silencing LC-NE neurons on these cell types. Finally, by manipulating astrocyte calcium while recording MC and PFC neuronal activity, we explored the effects of astrocyte dynamics on behavioral performance and on neuronal population activity in our reinforcement learning task. We find that phasic LC-NE during pre-execution and post-reinforcement epochs alter neuronal population dynamics and astrocyte calcium to facilitate task execution and optimization.

Support: R01MH126351, R01EY028219, W911NF-21-1-0328 , R01-DA049005-01

The synaptic basis for orientation matching in binocular visual cortex circuits. Katya Tsimring, Kyle Jenks, Jacque P.K. Ip, Mriganka Sur. Society for Neuroscience, 2022

Experience-dependent plasticity refines sensory circuits during critical periods in development. In the binocular visual cortex (bV1), visual experience aligns neuronal orientating tuning from the ipsilateral and contralateral eyes during the critical period for ocular dominance plasticity (~p22-p32). However, it remains unclear whether the alignment of somatic orientation tuning is coupled with changes at the synaptic level and through what forms of plasticity these synaptic changes take place. We propose that Hebbian and heterosynaptic plasticity shape visual responses of bV1 neurons during the critical period by strengthening synapses based on their correlation to the postsynaptic neuron and to synaptic neighbors, respectively. To examine how visual properties of bV1 neurons and their inputs change across development, we used in vivo two-photon calcium imaging to chronically track eye-specific and binocular visually driven responses of neurons and their dendritic spines. We find that ipsilateral eye driven responses become aligned to the soma’s binocular preference over development, whereas contralateral eye driven responses are already matched at the start of the critical period. This indicates that Hebbian plasticity is at least partially responsible for the alignment of ipsilateral responses during the critical period. At the synaptic level, there is turnover of over half of the existing dendritic spines between p22 and p32. Of the dendritic spines that are retained, about a third undergo changes in their functional responses. Intriguingly, the responses of these stable spines become more aligned to the soma by the end of the critical period, further suggesting that correlated pre- and post-synaptic neurons strengthen their connectivity through Hebbian-like mechanisms to converge to a similar orientation preference. Heterosynaptic plasticity, on the other hand, may play a role in gating the output of a neuron’s visual response, as we find that neighboring spines (< 5 um apart) are more correlated in responsive than unresponsive neurons. To test the contribution of Hebbian and heterosynaptic plasticity to the development of binocular matching, we are currently building a biophysical model that incorporates the structural and functional properties of dendritic spines imaged in vivo to simulate their dynamics between p22 and p32. By comparing the final state of synapses in the model with our experimental data at p32, we can assess which plasticity rules are necessary and sufficient to explain our observed results. Our work will thus provide critical insight into the nature and mechanisms of experience-dependent plasticity during development.

Support: F32EY032756 (KJ); R01EY028219, R01MH126351 (MS)

Alignment of visual features in binocular cortical circuits through experience-dependent plasticity. Katya Tsimring, Kyle Jenks, Dae Hee Yun, Jose Zepeda, Jacque P.K. Ip, Kwanghun Chung, Mriganka Sur. EMBO Dendrites, 2022

Experience-dependent synaptic plasticity is essential for fine-tuning immature sensory circuits during critical periods in development. In visual circuits, this fine-tuning includes the alignment of information from the ipsilateral and contralateral eye onto neurons in the binocular primary visual cortex (bV1). At the onset of the critical period for binocular vision, bV1 neurons have distinct orientation preferences for each eye, which become matched by the end of the critical period. Disrupting visual experience during this critical period can elicit severe impairments in binocular vision. While we know that synapses on these neurons are highly plastic during the binocular critical period, it remains unclear how this plasticity leads to properly aligned bV1 neurons. Hebbian and heterosynaptic plasticity are mechanisms of experience-dependent plasticity that modify synaptic inputs based on the correlation of pre- and postsynaptic neuronal responses and on the activity of neighboring synapses, respectively. We hypothesize that Hebbian and heterosynaptic mechanisms cooperatively regulate the binocular alignment of bV1 neurons during the critical period by strengthening synapses that are correlated with the postsynaptic neuron or with synaptic neighbors, and by weakening those that are uncorrelated.

To test this hypothesis, we have chronically tracked the eye-specific visual responses of bV1 neurons and of their dendritic spines using in vivo two-photon calcium imaging. Our preliminary findings indicate that while contralateral eye driven somatic responses are already matched to the soma’s binocular orientation preference at the start of the critical period, ipsilateral eye driven responses become aligned to the soma’s binocular preference over development. At the synaptic level, we observe minimal changes in the functional responses of contralateral eye driven dendritic spines (< 5%), whereas over 25% of dendritic spines lose visual responses to the ipsilateral eye. These results suggest that ipsilateral inputs uncorrelated to the somatic binocular preference are lost over the critical period. Furthermore, we observe that the functional responses between neighboring dendritic spines (< 10 um) are more correlated in adult mice than in critical period mice, indicating that heterosynaptic mechanisms may be at play in stabilizing neighboring co-active inputs over development. To determine whether ipsilateral inputs with orientation preferences that match the soma’s binocular preference are stronger than unmatched ipsilateral inputs, we have mapped the synaptic composition of the physiologically characterized dendritic spines post hoc by implementing a novel tissue expansion technology called eMAP (Epitope-preserving magnified analysis of proteome) that allows for super-resolution imaging of intact tissue. Our preliminary results suggest that the size of synaptic protein puncta, such as PSD-95 and GluA2, tends to be larger in dendritic spines with matching preferences to the soma’s. Finally, to probe the contributions of binocular visual experience in the developmental binocular alignment, we are performing monocular deprivation (MD) in mice during the critical period. Together, these experiments will answer fundamental questions on the nature of experience-dependent plasticity in bV1 and provide critical insight into the synaptic basis of amblyopia and neurodevelopmental disorders arising from synaptic dysfunction.

Support: NIH grant EY028219 and Picower Institute Innovation Fund

Assessing the computational power of astrocyte Ca2+ transients in the motor cortex during learning. Jennifer Shih, Gabi Drummond, Chloe Delepine, Nhat Minh Le, Yi Ning Leow, Mriganka Sur. Society for Neuroscience, 2022

Astrocytes are the most abundant glial cell type in the brain. Increasing evidence shows that they respond to and influence neuronal function at the synaptic, cellular, and network levels, and that this relationship may be reflected in transient astrocytic calcium (Ca2+) activity. These Ca2+ transients are spatiotemporally diverse, which may indicate complex bidirectional astrocyte-neuron interactions that drive learning and behavior. We have previously shown that correlated neuronal activity in the motor cortex is associated with learning and hypothesize that astrocyte Ca2+ signaling is critical for the temporal specificity of neuronal activity during behavioral epochs. To test this, we measured astrocytic Ca2+ during motor skill learning and reinforcement learning paradigms, and manipulated Ca2+ activity using DREADDs. Our preliminary data show that astrocyte Ca2+ signals are increased during motor execution and that the slower temporal scale of astrocyte activity relative to neuronal activity allows for task optimization. Disruption of astrocyte Ca2+ signals alters neuronal ensemble formation and learning. Using support vector classifiers and generalized linear models, we show that astrocyte Ca2+ signals can be used to decode behavioral outcomes and encode task parameters, suggesting that astrocyte Ca2+ is a critical part of cortical information processing during learned behaviors.

Support: F32NS110481 (JS), F31MH129112 (GD), R01EY028219 (MS), R01DA049005 (MS)

Rodent Lateral Posterior (LP) thalamus input to Anterior Cingulate Cortex (ACC) tracks ongoing and post-factum perceptual uncertainty. Yi Ning Leow, Yuma Osako, Alexandria Barlowe, Cindy Luo, Mriganka Sur. Society for Neuroscience, 2022

Ongoing uncertainty in our sensory environment is constantly being estimated in order to effectively guide attention and determine perceptual confidence. The pulvinar, or homologous rodent lateral posterior (LP) nucleus, is a higher order visual thalamic structure critical for regulating selective attention for behaviorally-relevant visual stimuli. Pulvinar interactions with the prefrontal cortices have been particularly implicated in regulating attentional processes. Here, we hypothesized that the rodent medial LP, by broadly integrating input from multiple visual areas as well as the frontoparietal cortices, is able to provide an ongoing estimate of the visual evidence, while the Anterior Cingulate Cortex (ACC) can utilize this information to guide sensorimotor decisions. We trained mice on a two-choice random dot (RDK) motion discrimination task, varying sensory certainty with the proportion of coherent target dot directions, while examining LP→ACC axonal calcium activity with two-photon microscopy. We found LP→ACC axons responded robustly during the task, not just during the stimulus period but also strongly upon reinforcement. We performed population analyses to determine the selectivity subspaces occupied by LP→ACC activity during the task. In the stimulus direction coding subspace, we found that a subset of LP→ACC axons represented the sensory uncertainty with reference to the dominant movement direction, while others represent absolute sensory uncertainty regardless of direction. Intriguingly, even after stimulus offset, the coherence of the stimulus appears to be represented even after reinforcement. This information persists in the LP→ACC axons across to the next trial which may potentially guide top down allocation of attentional resources. Our results suggest LP involvement in decision-making beyond ongoing visual processing and perceptual evidence accumulation to include a potential involvement in conveying information on post-factum perceptual uncertainty.

Supported by ASTAR (YL), PIIF, and NIH grant R01MH126351

 

Novel computational approaches for characterizing correlated astrocyte and neuron activity. Jennifer Shih, Grayson Sipe, Yi Ning Leow, Nhat Minh Le, Mriganka Sur. CHSL Glia, 2022

Astrocytes represent a significant population of non-neuronal cells with roles in ionic homeostasis, neurotransmitter transport, and metabolic coupling. How astrocytes dynamically influence neuronal circuits remains unclear. Two-photon microscopy has been increasingly used to characterize astrocyte physiology in the awake animal primarily through the imaging and analysis of dynamic calcium activity in both astrocytes and neurons simultaneously. Diverse spatiotemporal astrocyte calcium signatures are thought to underlie distinct functions in physiological and pathophysiological conditions. However, the relationship between astrocyte calcium events at different spatiotemporal scales and neuronal activity remains poorly understood. Here, we present novel computational methods for the analysis of simultaneously collected, dual-color astrocyte/neuron activity. We hypothesize that astrocyte-neuron interactions can be defined by the encoding capabilities of each signal, and more specifically, that calcium activity in astrocyte microdomains encodes behavior and neuronal activity. Using generalized linear models (GLMs) to model astrocyte events in the visual cortex during passive visual stimuli, we show that neuronal activity, particularly in the dendrites, is a strong predictor for astrocyte activity, which suggests that calcium activity in microdomains preferentially encodes dendritic activity. We also use hidden Markov models to identify latent neuronal network states and map large scale astrocyte calcium activity onto those state transitions. These approaches provide new insight into the dynamic relationship between neuronal and astrocytic physiology.

Supported by F32NS110481 (JS), K99AA028579 (GS), ASTAR (YL), the Newton Fund (NML), NIH grants EY028219, DA049005 and MURI W911NF-21-1-0328 (MS)

State-dependent Reward Encoding in Cortical Activity During Dynamic Foraging

Nhat Minh Le, Murat Yildirim, Hiroki Sugihara, Yizhi Wang, Mriganka Sur
COSYNE, 2022

Funding: NIH grants 1R01MH126351, EY028219, K99EB027706, Paul and Lilah Newton Brain Science Research Award

2021

Shifts in learning strategies underlie rodent behavior during dynamic foraging
Nhat Minh Le, Murat Yildirim, Hiroki Sugihara, Yizhi Wang, Mriganka Sur
Society for Neuroscience, 2021

In uncertain foraging environments, a critical problem agents face is how to adapt their learning in accordance with the uncertainty of the environment and knowledge of the hidden structure of the world. In this context, previous studies distinguished between two modes of learning: model-free learning updates action values while balancing between exploration and exploitation, while inference-based behavior leverages knowledge of world’s dynamics to infer the current state of the world. However, it is unclear how these behavioral strategies can be cleanly distinguished based on behavioral data and how animals transition between these modes during training. Here, we tackled these questions by examining rodent behavior in a dynamic foraging task. We trained head-fixed mice to select between two choices, left and right wheel turns. Only one direction was rewarded in each trial, and the target direction changed in blocks of 15-25 trials. To determine whether rodent behavior was better described by the model-free or inference-based strategies during learning, we simulated choice sequences of model-free and inference-based agents to gain insights into their behavioral signatures. Our model-free agent implements a Q-learning algorithm, while the inference-based agent performs Bayesian inference assuming a 2-state model of the world. We identified four criteria to characterize the behavior of these agents and used them to fit rodent behavioral data: the slope and offset of their switching decisions, the degree of exploration, and the relationship between switching and the performance in the previous block. Across all animals, we observed a consistent decrease in the times to switch the wheel turn direction, and a decrease in the degree of exploration with training. By performing a model comparison using log-likelihoods, these observations were better explained by a transition from model-free to inference-based mode of learning, as opposed to an increase in learning rate in a model-free agent. A 3-state Hidden-Markov model was fitted to the rodent behavioral data, revealing transitions between left-exploit, right-exploit and exploratory states. The duration of exploratory states decreased with training, consistent with the transition to the inference-based strategy. These analyses and characterizations form the basis of understanding shifts in behavioral strategies during dynamic foraging. Together with widefield calcium imaging across the whole cortex during the task, they will be important for identifying unique neural signatures of global brain states that are associated with distinct modes of learning.

Funding: NIH grants 1R01MH126351, EY028219, K99EB027706, Paul and Lilah Newton Brain Science Research Award

Mixture of learning strategies underlies rodent behavior in dynamic foraging
Nhat Minh Le, Murat Yildirim, Hiroki Sugihara, Yizhi Wang, Mriganka Sur
Neuromatch, 2021

In uncertain foraging environments, agents need to adapt their learning in accordance with the uncertainty of the environment and knowledge of the hidden structure of the world. Previous studies distinguished between two modes of learning: model-free learning updates action values using trial-by-trial feedback signals, while inference-based behavior uses past experience to infer the current state of the world. However, it is unclear how well these strategies can be distinguished based on behavioral data, and how animals may acquire appropriate strategies during training. Here, by combining computational simulations and analysis of rodent behavioral data on a dynamic foraging task, we showed that mice adopt a remarkable mixture of strategies that include both model-free and inference-based behavior within single training sessions. We identified signatures of model-based and model-free behavior by simulating choice sequences of a large ensemble of parameterized agents, and characterized the choice switches of these agents around the block transition points. Transition functions of these ensembles clustered into five classes corresponding to distinct regimes of model-free and inference-based strategies, which can be robustly decoded from observed data. We examined how mice trained on a head-fixed dynamic foraging task adapted their behavioral strategies with training. Surprisingly, despite an increase in learning rate, the session-average behavior of expert animals remained model-free. Closer inspection of the behavioral data showed that this average behavior masked the use of multiple strategies that involved both slow and fast switch dynamics within single sessions. We characterized these dynamic shifts using a novel state-space method, block Hidden Markov Model, to infer switching modes for individual animals within single blocks. The model revealed a decrease in the random blocks and an increase in the fast-switching, inference-based blocks with training. Yet, a mixture of strategies, model-free, model-based and random strategies, co-exist even in the expert animals. These analyses and characterizations form the basis of understanding shifts in behavioral strategies during dynamic foraging. Together with widefield calcium imaging across the whole cortex during the task, they are important analytical tools for identifying unique neural signatures of global brain states associated with distinct modes of learning.

Developmental changes in eye-specific inputs to dendritic spines of neurons in mouse visual cortex
Katya Tsimring, Pak Kan IP, Jose C. Zepeda, Kyle Jenks, Dae Hee Yun, Taeyun Ku, Kwanghun Chung, Mriganka Sur
Society for Neuroscience, 2021

Neuronal circuits in the primary visual cortex undergo significant reorganization during the critical period. In mice, this reorganization corresponds with an experience-dependent convergence of ipsilateral and contralateral eye inputs onto neurons in the binocular region of the primary visual cortex (bV1). How synaptic inputs onto bV1 neurons, reflected in the responses of individual spines, change across development remains unknown. We compared the functional properties of synapses on L2/3 bV1 neurons between critical period and adulthood, using two-photon calcium imaging of dendritic spines in awake mice during monocular and binocular presentations of visual stimuli. We find that adult neurons have a significantly higher fraction of spines with a stronger bias to either the ipsilateral or contralateral eye than during the critical period, when spines are more binocular. Some spines in adult neurons that exhibit a preference towards one eye during monocular stimulation show reduced responses to binocular stimulation, indicating that these spines are either directly or indirectly suppressed by the non-preferred eye. To determine how dendritic spine responses evolve over development, we imaged the same spines and dendritic segments at different time points. Our preliminary findings show some visually responsive spines that shift their relative responses to each eye and develop binocular suppression, indicating dynamic changes in eye-specific drive onto individual spines. These changes can arise from presynaptic inputs to spines or be constructed at individual spines, potentially reflecting changes in excitatory inputs from one eye and development of inhibitory inputs from the other eye. To further probe mechanisms of binocular suppression in spines, our ongoing work employs a novel tissue expansion technology, Magnified Analysis of the Proteome (MAP), that allows super-resolution measurement of the synaptic composition of dendritic spines from L2/3 bV1 neurons. We hypothesize that the enrichment and distribution of inhibitory and excitatory synapses onto spines could provide a mechanistic explanation for the emergence of inter-ocular suppression across development.

Support: NIH grants R01 EY028219, R01 EY007023

Temporal Expectation in Marmosets: Global influences of task structure and local modulation by trial history
Tudor Dragoi, Hiroki Sugihara, Nhat Le Minh, Elie Adam, Jitendra Sharma, Guoping Feng, Robert Desimone, Mriganka Sur
Society for Neuroscience, 2021

It is generally believed that in absence of overt cues, humans and macaques implicitly use prior information to predict upcoming events and reduce uncertainty. Temporal expectation has been modeled using the hazard rate, which posits the likelihood of an event to occur in the future provided it has not occurred already. However, previous studies have not addressed how internal models of temporal expectation are acquired as a consequence of learning. To answer this question, we implemented a simple timing task in which freely behaving marmosets were required to make a timed response prompted by a visual stimulus change with a uniform distribution of stimulus durations. Our results demonstrate that, similar to previous findings in humans and macaques, marmoset reaction times follow the hazard rate model of expectation, consistent with the global task structure. Further, we examined how this model emerges from learning and found that with repeated task exposure, trial history and hence local task structure begins to influence reaction time. The combined effects of global and local task structure are well described by a multiple regression model, and computationally by Bayesian updating of the hazard function. Parallel experiments in human subjects similarly demonstrate global and local influences on reaction times and temporal expectation, which are also well captured by multiple regression and Bayesian models. These results demonstrate the evolution of dynamic internal models of temporal expectation based on multiple cues, along with the richness of temporal cognition in marmosets comparable to that in humans. An open question is whether the effects of local and global influences are distinct parallel processes or rather if they synergistically interact to influence temporal expectation. To reveal the neural underpinnings of temporal expectation, future studies in marmosets will measure large scale population activity across visual, parietal and dorsomedial prefrontal cortex.

Support: Simons Foundation Autism Research Initiative, Hock E. Tan and K. Lisa Yang Center for Autism Research, James and Patricia Poitras Center for Psychiatric Disorders Research, Stanley Center for Psychiatric Research

Behavioral state-dependent modulation of medial Lateral Posterior (LP) thalamic visual input to the Anterior Cingulate Cortex (ACC)
Yi Ning Leow, Blake Zhou, Mriganka Sur
Society for Neuroscience, 2021

Selective attention is a critical cognitive process to adaptively highlight task-relevant stimuli while gating indiscriminate stimulus responsivity. Functional connectivity between the reciprocally connected pulvinar and frontal areas like the anterior cingulate cortex (ACC) is associated with impaired sensory filtering and over-responsiveness to stimuli. We sought to examine how the lateral posterior (LP) thalamic nucleus, the rodent homologue of the pulvinar, may exert behavior state-dependent gating of its visual-associated input to ACC. Using two-photon calcium imaging of LP axons in ACC of head-fixed mice presented with visual stimuli, we demonstrate that visual responsiveness and feature tuning to the same stimuli are highly correlated with spontaneous fluctuations in behavioral state as well as direction of self-motion. We also find dissociable contributions of pupil-linked arousal and locomotion. As ACC simultaneously receives visual input from primary visual cortex and other higher visual areas, state-dependent gating by LP could in turn influence the visual information integrated at ACC or bias ACC-dependent top-down modulation of visual inputs. To identify neural sources that may convey such behavioral states, we performed monosynaptic rabies tracing of brain-wide inputs to LP-ACC projections and its top-down complement, ACC-LP projections. This revealed a convergence of multi-sensory and motor input upon LP-ACC neurons, in contrast to ACC-LP neurons, which mainly receive visual and frontal inputs. In addition, the most prominent subcortical input source to LP-ACC neurons is the superior colliculus (SC) — particularly from the intermediate and deep layers which do not receive direct retinal input. With the shared involvement of ACC, LP and SC in visuomotor functions, orienting and coordinating attentional priority, LP can thus serve as one of the pathways for feedback from SC to ACC.

Support: R01 EY028219, PIIF, A*STAR National Science Scholarship

Early Molecular and Cellular Deficits in 3D Cerebral Organoid models of Rett Syndrome
Chloe Delepine, Vincent Pham, Hayley WS Tsang, Murat Yildirim, Nader Morshed, Xian Adiconis, Sean Simmons, Paola Arlotta, Joshua Levin, Forest White, Li-Huei Tsai, Mriganka Sur
Society for Neuroscience, 2021

Rett Syndrome (RTT, OMIM 312750) is a progressive and pervasive, X-linked neurodevelopmental disorder that predominantly affects girls, who exhibit symptoms by early childhood. The vast majority of typical RTT cases is triggered by a sporadic mutation in the methyl CpG-binding protein 2 (MeCP2) gene. Although most research has focused on postnatal mice and humans displaying clinical symptoms, recent studies have shown that MeCP2 deficiency triggers molecular and cellular defects at very early stage of embryonic brain development, prior to clinical manifestation. We used isogenic RTT patient-derived induced pluripotent stem cells to generate 3D human cerebral organoids to recapitulate these early developmental events and to better understand the molecular and cellular events that underlies RTT pathology. Previously, we showed that MeCP2 deficiency was linked to dysregulation of the AKT/ERK pathway, which led to an increase in neural progenitor proliferation and concomitant decrease in neurogenesis and neuronal migration and maturation (Mellios et al., Molecular Psychiatry 2018). In this current study, we further investigated the molecular underpinnings of the deficits in neuronal migration observed in RTT organoids. Using genetic expression of fluorescent markers, and immunostaining, combined with confocal and multi-photon microscopy, we found that, although the morphology and polarity of radial glial cells were mostly preserved, neuronal migration patterns (speed, trajectory, and distance) were disrupted in RTT MeCP2 mutant organoids compared to isogenic controls. We then used transcriptomic, proteomic and phospho-proteomic techniques to screen for dysregulated adhesion molecules downstream of RTT MeCP2 mutations. We found an increase in the phosphorylation of the Reelin signaling adaptor protein, DAB1, at the tyrosine (Y) 232 residue. By introducing a dominant-negative phospho-null Y232F plasmid construct, and an DAB1 overexpression construct in the ventricles of RTT organoids and isogenic controls, here we probe the contribution and mechanisms of DAB1 in migration deficits observed in RTT organoids, through the regulation of cell adhesion. By using small molecules to modulate the AKT signaling pathway, we interrogate the links between DAB1 phosphorylation, adhesion, neuronal migration and signaling. These findings will further clarify the role and mechanisms of early deficits during cortical development in RTT.

Support: NIH grant MH085802, Picower Neurological Disorder Research Fund, James W. and Patricia T. Poitras Fund, Simons Center for the Social Brain Seed Fund

Astrocytes contribute to motor learning, neuronal correlations and movement encoding by motor cortex neurons
Chloe Delepine, Keji Li, Jennifer Shih, Pierre Gaudeaux, Mriganka Sur
Society for Neuroscience, 2021

While motor cortex is crucial for learning precise and reliable movements, whether and how astrocytes contribute to its plasticity and function during motor learning is unknown. Here we report that primary motor cortex (M1) astrocytes in mice show in vivo plasticity during learning of a lever push task, as revealed by transcriptomic and functional modifications, particularly changes in expression of glutamate transporter genes and increased coincidence of microdomain calcium events. Astrocyte-specific manipulations of M1 are sufficient to alter motor learning and execution, and neuronal population coding, in the same task. Specifically, mice expressing decreased levels of the astrocyte glutamate transporter GLT1 show impaired and variable movement trajectories. Mice with increased astrocyte Gq signaling show decreased performance rates, delayed response times and impaired trajectories, along with abnormally high levels of GLT1. In both groups of mice, M1 neurons have altered inter-neuronal correlations and impaired population representations of task parameters, including response time and movement trajectories. Thus, astrocytes have a specific role in coordinating M1 neuronal activity during motor learning, and control learned movement execution and dexterity through mechanisms that include fine regulation of glutamate transport.

Support: NIH grants EY007023, EY028219

2020

Temporal dynamics of locus coeruleus noradrenaline during learned behavior
Vincent Breton-Provencher, Gabi Drummond, Mriganka Sur
Society for Neuroscience, 2020

Neurons in the locus coeruleus (LC) are the main source of the neuromodulator, noradrenaline (NA), in the forebrain. Despite the brain-wide effects of LC-NA release, the conditions under which LC-NA neurons are activated and the modes of NA action during learned behavior are poorly understood. Previous reports on LC-NA have sought to explain its role in gain control or network reset, yet how these two functions get integrated at the level of LC dynamics remain unknown. Here, we recorded and manipulated genetically identified LC-NA neurons in mice trained on an instrumental learning behavior incorporating stimulus-reward uncertainty. In this auditory detection task, head-fixed mice have to press a lever to a go tone to obtain a water reward, and refrain from pressing the lever to a no-go tone to avoid an air puff punishment. We modulated task difficulty by using tones of four intensities within the two frequencies. Targeted recordings of LC-NA neurons with opto-tagging, or with calcium imaging via two-photon microendoscopy, show that the majority of LC-NA
neurons responds before execution – the lever press – and after punishment, and a smaller fraction responds to reward. The amplitude of the execution and reward response of LC-NA neurons varies with tone intensity: the execution signal increases with tone intensity, while the reward signal decreases with tone intensity, or lower uncertainty of reward. LC-NA
transient activity peaks following a punishment regardless of tone intensity to signal a state of greater uncertainty or surprise. The effect of photo-inactivation of LC-NA activity during selected trials is twofold. First, inactivating LC-NA produces a decrease in lever presses during go trials with low tone intensity, suggesting a role for LC-NA activity during
behavioral response in high uncertainty conditions. Second, inactivating LC-NA impairs active reinforcement learning, or the ability of the mice to optimize its task strategy following a punishment. Together, our results suggest that distinct temporal dynamics enable the simultaneous regulation of execution and active learning by LC-NA during learned behavior.

Support: NIH grant R01EY028219, Picower Innovation Fund

Motor cortex astrocytes contribute to motor learning and neuronal coding in vivo
Chloe Delepine, Keji Li, Mriganka Sur
Glia in Health and Disease, Cold Spring Harbor Laboratory, 2020

Motor cortex, including M1, is crucial for the production of precise and reliable movements, yet its functions and plasticity during motor learning are not fully understood. Furthermore, the contribution of M1 astrocytes and the physiological role of M1 astrocyte-neuron interactions during motor learning is unknown. Here we report that astrocyte-specific manipulation of M1 in mice, targeting glutamate clearance and Gq signaling, regulate motor learning and performance in a lever push task by modulating neuronal activity and encoding. Mice expressing decreased levels of the astrocyte glutamate transporter GLT1 in M1 showed normal success rate and improved response timing but impaired learning of a stereotyped movement; M1 neuronal populations are more active but encode less task information. Mice with increased astrocyte Gq signal activation in M1 show decreased success rate, delayed response time and decreased learning of the stereotyped movement; these features accompany high levels of neuronal signal correlation, low encoding of movement parameters and altered encoding of correct trials. Analysis of gene expression from purified M1 astrocytes during the task suggest subtle gene expression modulation associated with motor learning. This included regulation of genes involved in glutamate transport and GPCR signaling, which may underlie the influence of astrocytes on neuronal circuits and population encoding that we demonstrate.

Support: NIH EY007023, NIH EY028219

Morphological and molecular changes at dendritic spines orchestrate neuronal shifts in functional identity during monocular deprivation
Jose Zepeda, Jacque Ip, Katya Tsimring, Dae Hee Yun, Taeyun Ku, Kwanghun Chung, Mriganka Sur
American Society for Biochemistry and Molecular Biology Annual Meeting, 2020

Experience‐dependent plasticity refers to the brain’s ability to sculpt and remodel its circuits in an experience‐dependent manner. In the binocular visual cortex, where inputs from both eyes converge onto single neurons, the relative responsiveness of a neuron to input from one eye can shift when that eye is deprived of input. Ocular dominance plasticity (ODP) is a well‐established model for experience‐dependent plasticity, and it is triggered by monocular deprivation (MD) by suturing an eyelid. This shift is classically measured by calculating the ocular dominance index (ODI), a measure of responsiveness to either eye. The cellular and molecular mechanisms for this reorganization during ODP are not well understood. We hypothesize that morphological and molecular changes at dendritic spines (synapses) allow for functional reprogramming of neurons. To test this, we introduced a plasmid into binocular neurons via single‐cell electroporation that encodes for the calcium sensor GCaMP6s, together with a structural marker mRuby2. Using two‐photon microscopy, spines were imaged in vivo in awake mice to determine their ODIs. We then imaged at later time points during MD to measure how spine size and their responsiveness was affected. Finally, the same neurons were visualized using magnified analysis of proteome (MAP), which evenly expanded tissue by 4X to closely examine scaffolding and transmission‐related proteins at the postsynaptic density. Preliminary data suggests that a combination of extensive spine loss and potentiation of few specific spines allows for shifts in responsiveness. In addition to synaptic plasticity being important for learning and memory in adulthood, many autism spectrum disorders (ASDs) have been associated with mutations in synaptic proteins. Investigation of the molecular mechanisms behind changes in synaptic morphology are therefore important to better understand the mechanisms of learning and memory, as well as these disorders

Support: IMSD R25GM076321 , EY007023 and EY028219

 

2019

Early Molecular and Cellular Deficits in 3D Cerebral Organoid models of Rett Syndrome
Vincent Pham, Chloe Delepine, Hayley WS Tsang, Nader Morshed, Forest
White, Mriganka Sur
Society for Neuroscience, 2019

Rett Syndrome (RTT, OMIM 312750) is a pervasive, X-linked neurodevelopmental disorder that predominantly affects girls. With an incidence of 1:10 000–15 000, it is the second most common cause of severe intellectual disability in females after Down Syndrome. The vast majority of typical RTT cases is triggered by a sporadic mutation in the gene encoding methyl CpG-binding protein 2 (MeCP2). RTT is a progressive disorder, characterized by deficits that are prominently manifested by early childhood. Although most research has focused on symptomatic mice and humans, recent studies have shown that MeCP2 deficiency triggers molecular and cellular defects at very early stage of embryonic brain development. We used isogenic RTT patient-derived induced pluripotent stem cells to generate 3D human cerebral organoids to recapitulate these early developmental events and to better understand the molecular and cellular complexity that underlies RTT pathology. Previously, we showed that MeCP2 deficiency was associated with an increase in neural progenitor proliferation and concomitant decrease in neurogenesis and neuronal migration and maturation (Mellios et al., Molecular Psychiatry 2018). These data suggested that dysregulation of the AKT/ERK pathway due to MeCP2 deficiency were causal for the observed phenotype. In the current study, we have further investigated the molecular underpinnings of the deficits in neuronal migration observed in RTT organoids. Using transcriptomic, proteomic and phospho-proteomic techniques, we have interrogated the consequences of RTT MeCP2 mutations on molecular signaling pathways and found an associated increase in GSK3β-β-catenin signaling, downstream of AKT. Using fluorescent markers and immunostaining, combined with multi-photon microscopy and confocal microscopy, we find that, although the morphology and polarity of radial glial cells are mostly preserved, adhesion molecules are dysregulated and neuronal migration patterns are disturbed in RTT MeCP2 mutant organoids compared to isogenic controls. These findings confirm and further clarify the role and mechanisms of early deficits during cortical development in RTT.

Support: NIH grant MH085802

Involvement of the Locus Coeruleus Network in an Attention-Demanding Sensory Discrimination Task
Vincent Breton-Provencher, Mriganka Sur
Society for Neuroscience, 2019

Speed and accuracy on tasks involving high levels of uncertainty depend on arousal and attention. For example, driving during a downpour requires heightened attention compared to driving during a sunny day. Recent human and theoretical studies have described a link between uncertainty, arousal, and locus coeruleus (LC) activity. However, due to limitations of causal interrogation of brain function in these studies, how behavioral uncertainty may recruit the LC neuronal network remains unclear. Furthermore, while this recruitment implies regulation of noradrenaline (NA), the neuronal substrates of such a process are unknown. Here we used an attention-demanding task in head-fixed awake mice to assess the role of LC circuits in decisions during high uncertainty. We have used a go/no-go auditory discrimination task in which the mouse must quickly decide to enact or withhold a lever press after hearing a tone to obtain a reward or avoid a punishment. To introduce uncertainty and increase attentional demand, we vary the intensity and the timing of the auditory cues. Direct recordings of LC-NA neurons using phototagging display a high level of modulation of LC-NA spiking activity to specific task epochs. NA activity increases rapidly following the presentation of the tone during correct responses on ‘go’ trials, or after the delivery of reinforcements for both ‘go’ and ‘no-go’ trials. The level of uncertainty, with respect to the identity of the auditory stimulus, scales this response of NA neurons. We are now investigating the involvement of inhibitory neurons of the LC network, the LC-GABA neurons, in the behavior using phototagging. We are also silencing the activity of LC-NA or LC-GABA neurons using archaerhodopsin targeted at each population, to evaluate the causal contribution of the different components of the LC network for task execution. Collectively, our results indicate that decision-making during high levels of uncertainty recruits NA activity.

Support: NIH grants EY007023, EY028219

Label-Free Characterization of Attenuation Lengths of Cortical Regions via Three-Photon Microscopy in Awake Mice
Murat Yildirim, Ming Hu, Peter T. C. So, Mriganka Sur
Society for Neuroscience, 2019

Our understanding of large population activity of cortical neurons has been greatly advanced by two-photon microscopy. However, due to light scattering and absorption, it is still a challenge to be able to imaging neurons at deep layers. Moreover, the attenuation length of the cortex may change from area to area depending on its structure. Thus, it is critical to characterize the attenuation length of the area of interest in order to precisely image/modulate cortical and subcortical regions. Current state of
art for determining the attenuation length of any cortical region is to label blood vessels with high absorption cross-section dyes injected retro-orbitally in anesthetized mice. However, these injections do not last long enough to characterize multiple brain regions at once so multiple retro-orbital injections are required. Attenuation lengths of multiple cortical regions in awake mice has never been reported. Here, we developed two label-free methods to characterize attenuation lengths in multiple brain regions in awake mice via our custom-made three photon microscope. The first method is based on imaging blood vessels in the cortex and axonal tracts in the white matter via third-harmonic generation (THG) microscopy. The second method is to ablate four different depths in the cortex with varying pulse energy. THG imaging of blood vessels is advantageous since it is practical for multi-site imaging whereas it is disadvantageous since it relies on both excitation and emission wavelengths. Ablation is a more precise method since it relies on only the excitation wavelength whereas it is disadvantageous since it creates small lesions in multiple depths. We performed these imaging and ablation experiments in several brain regions including primary visual cortex (V1), and somatosensory cortex (S1) in four awake mice. Comparing the attenuation length between these regions obtained with THG imaging and ablation, we concluded that ablation experiments consistently generate 10% longer attenuation lengths compared to those obtained with THG imaging. Also, we found that there were significant differences in attenuation lengths between cortical regions. For example, V1 has an attenuation length of 254.0 ±8.1 µm whereas S1 has an attenuation length of 312.0 ±10.2 µm. Our work shows that the attenuation lengths of cortical regions in awake mice can be reliably estimated by performing label-free THG imaging of blood vessels at 1300 nm wavelength. These measurements place fundamental constraints on two-photon imaging depths and three-photon imaging power for analysis of cortical activity.

Support: NIH Grants NS090473, EY007023, EB022726

A New Interpretation of Visual Cortical Areas in Mice
Ming Hu, R. V. Rikhye, M. Gourd, M. G. M. Kumar, H. Murthy, Mriganka Sur
Society for Neuroscience, 2019

Much like other mammals, mouse visual cortex is composed of multiple areas, each containing a separate representation of the visual field. Early studies (Caviness, 1975; Drager, 1975; Wagor, Mangini and Pearlman, 1980) found that the region lateral to V1 contains at least two complete representations, labeled V2 and V3, whereas the region medial to V1 contains two partial representations, labeled Vm-r (medial rostral region) and Vm-c (medial caudal region) respectively. Subsequent work (Kalatsky and Stryker, 2003; Wang and Burkhalter, 2007; Marshel et al., 2011; Garrett et al., 2014; Zhuang et al., 2017; Waters et al., 2019) reported a number of extrastriate areas lateral and medial to V1. The sizes of these extrastriate areas are much smaller and more variable than V1. In an attempt to reconcile these studies, we used wide-field and single-cell calcium imaging from awake, head-fixed mice, which transgenically expressed GCaMP6f, to functionally segment the entire visual cortex. We further characterized the responses of neurons within each segmented area to drifting gratings and natural scene movies. Our key observations are two-fold. First, the lateral regions align closely with consistent features of extrastriate visual cortex in other species. Specifically, RL and LM together constitute a single and complete representation, consistent with its potential function as a second tier representation along the extrastriate pathway (V2), whereas AL functions as a third tier representation (V3). The regions within AL, RL/LM and V1 that represent the central visual field are organized closely next to each other in a narrow band which resembles the ‘foveal confluence’ (Zeki 1969; Newsome et al., 1986; Maunsell and van Essen, 1987; Gattass et al., 1988, 2005; Tootell et al., 1998; Schira et al., 2009, 2010) described in other species. These interpretations are strongly supported by the shared vertical meridian border between V1 and RL+LM, and the split representation of the horizontal meridian at the lateral border of RL+LM, including with AL. Second, the two medial areas (AM, PM) align closely with distal visual cortex (prostriata) recently described in marmosets (Yu et al., 2012) and humans (Mikellidou et al., 2017; Tamietto & Leopold, 2018), and they similarly share a peripheral visual field border with V1. We propose a framework of mouse visual cortex organization that follows the same global principles of visual cortex maps in other species, with variations in local layout that possibly reflect the ecological adaption of vision in these species.

Temporal Expectation in Marmosets
Tudor Dragoi, Hiroki Sugihara, Ming Hu, Jitendra Sharma, Mriganka Sur
Society for Neuroscience, 2019

In natural environments, animals continuously use stimuli in specific contexts which help them predict and expect forthcoming events. Expectation is an internally generated state that utilizes prior knowledge of the environment to optimize behavioral responses to changing stimulus contexts. Previous work by our lab and others has shown that humans and macaques utilize prior information such as the probability of an event based on past occurrences, which helps reduce uncertainty and improve performance (Sharma et al. 2003, Summerfield et al. 2006). Expectation in a temporal context has been shown to follow hazard rate measurements of an event’s rate of occurrence over time. (Ghose and Maunsell 2003, Janssen and Shadlen 2005). Higher brain areas likely use prior information and feed it back to update motor plans for behavioral adjustments to changing task variables. To uncover if marmosets utilize similar strategies to optimize behavior, we developed an expectation based task in marmosets using an in-home cage training system. Two marmosets were trained to hold off touching a tablet until a visual cue was provided, at which point a touch must be registered. To understand if they used hazard rate to guide expectation in this simple task, we measured reaction times to uniformly varying lengths of time of cue disappearance. Although tested in a less controlled environment, our early results show that marmoset reaction time shortens with longer stimulus duration. The distribution of touches following the disappearance of a cue sharpened over time, with mean latency (0.6s) and full width half maximum (0.021s) decreasing across 3 months of exposure to the task. An uncertainty-based hazard rate model more closely correlated with reaction time when compared to standard hazard rate models indicating that reaction time is sensitive to the uncertainty of upcoming stimulus events. We believe that elapsed time is differentially used to adjust reaction times in response to changing patterns of visual input. Further work will examine the neural underpinnings of stimulus expectation. Specifically, we will use optogenetic and pharmacological manipulations to change balance between cortical excitation and inhibition in local and long range cortical circuits to investigate their role in temporal prediction and expectation. Atypical temporal prediction has been implicated in autism, thus our research has the potential to uncover elements of neural processes in such disorders such as how integration of bottom-up and top down feedback in cortical networks provides contextual processing.

Support: Simons Center for the Social Brain

Modulation of Neuronal Coding by Astrocytes During Motor Learning
Chloe Delepine, Keji Li, Vaibhavi Shah, Taylor Johns, Mriganka Sur
Society for Neuroscience, 2019

Astrocytes, long thought to operate only as a support network for neurons, are now emerging as key players in the modulation of brain information processing. Astrocytes influence synaptic transmission via glutamate transporters, and respond to, as well as modulate, neuronal activity with Ca2+ signaling. However the contribution of astrocytes in vivo to complex behaviors and cognition remains unresolved. During motor learning, the primary motor cortex (M1) is functionally and structurally reorganized, manifest in changes of neuronal activity and dendritic spine turnover. We hypothesize that astrocytes modulate learning-associated neuronal network reorganization by influencing synaptic strength through glutamate clearance and Ca2+ signaling. Here we investigated the role of cortical astrocytes in a motor learning task in vivo, where mice were rewarded for pushing a lever following an auditory cue. Using the engineered human muscarinic G protein-coupled receptor DREADD-hM3Dq activated by low doses of clozapine-N-oxide (CNO), we found that modulation of astrocyte Ca2+ activity perturbed task performance, causing decreased responses rate and increased delay. Using a transgenic mouse line in which the expression of the glutamate transporter GLT1 was inhibited locally in M1, we found that decreasing astrocyte glutamate clearance prevented learning of a stereotypical motor trajectory while increasing response rate. Using genetically encoded Ca2+ indicators and high-resolution two-photon imaging, we imaged neuron activity during execution of the motor task following training with astrocyte manipulation. Perturbing Ca2+ activity and glutamate clearance both disrupted M1 neurons’ capability to encode the motor command, as in both cases neuron populations showed drastically weakened performance when decoding for lever trajectory. Perturbation of astrocyte Ca2+ activity was found to create a higher level of correlated noise in neural activity and to hamper neurons’ ability to encode for lever trajectory. In contrast, decreased glutamate clearance by astrocytes increased neuronal activity and prevented the emergence of highly motion relevant neurons in the circuit, potentially by decreasing their dynamic range. These findings demonstrate specific contributions by astrocytes to the creation of neuronal ensembles during motor learning, and their representation and encoding of learned trajectories.

Support: NIH grants EY007023, EY028219

Prefrontal Cortex-Superior Colliculus Interactions in Visually Guided Decision-Making
Rafiq Huda, Karen Cruz, Austin Sullins, Mriganka Sur
Society for Neuroscience, 2019

Perceptual decision-making requires mechanisms for promoting and inhibiting specific choices in accordance with task goals. The prefrontal cortex has long been implicated in guiding goal-oriented choices by generating control signals and selectively modulating the activity of downstream target structures. A similarly large body of work also ascribes an important role for midbrain circuits in choice selection and execution. Yet, how prefrontal and midbrain circuits interact to facilitate goal-oriented choices remains unresolved. To address this question, we developed a two-alternative forced choice visual decision making task in which head-fixed mice report the spatial location of a target stimulus while suppressing responses to simultaneously presented distractor cues. Mice learn this task well and perform hundreds of trials per session, allowing us to construct high-quality psychometric functions. Our previous work, in combination with other recent studies, suggests that a subdivision of the mouse prefrontal cortex, the anterior cingulate cortex (ACC), is a crucial anatomical node for coordinating visually-guided behavior. Our anatomical studies show that the ACC provides direct top-down inputs to the superior colliculus (SC). We are defining the function of ACC-superior colliculus interactions using a combination of areal and projection-specific optogenetic manipulations during behavior. Results to date suggest that the ACC and SC play opposing, but complementary functional roles in the task. Targeted projection-specific inactivations suggest that the ACC exerts top-down inhibitory control over the SC. We are currently using two-photon calcium imaging to assess the physiological responses of ACC neurons that target the SC and determine their specific role in the decision-making process. Overall, our work suggests that the prefrontal cortex exerts top-down inhibitory control over midbrain circuits to facilitate goal-oriented choices.

Support: NIH grants EY007023, EY028219; K99 MH112855

Dynamic Control of Visually-Guided Locomotion Through Cortico-Subthalamic Projections
Elie M. Adam, Taylor Johns, Mriganka Sur
Society for Neuroscience, 2019

Voluntary movement is fundamental as our means to act upon the world. We are interested in how the brain controls actions aligned with an organism’s behavioral goals. We developed a behavioral task where mice run on a treadmill through a virtual runway and stop at visual landmarks to collect reward. The task enables a voluntary on-off visually guided locomotion pattern, constructed of sudden initiations and sudden halts. Where do control signals enabling such a pattern emanate from and how are they processed to control locomotion? Extracellular recordings in the mesencephalic locomotor region (MLR) reveal neural activity tuned to that locomotion pattern. The visually guided nature of the task further implicates cortical control, hypothesized to be over MLR. Such a control can however only be indirect, as there are no significant projections from cortex to MLR. We find that prefrontal cortex (PFC) projections to subthalamic regions (specifically to the subthalamic nucleus and zona incerta) mediate stop commands, thereby posing the region as a direct controller onto MLR. Through a combination of simultaneous in-vivo extracellular recordings of PFC and MLR, calcium imaging of PFC projection neurons, axonal optogenetic manipulations in subthalamic regions, and targeted recordings of identified subthalamic neurons, all in behaving animals, we are reconstructing the dynamical nature of the control mechanism and the neural circuitry by which it is implemented.

Support: NIH grant EY007023; Picower Fellowship

Identification of KCC2 Expression Enhancer Compounds as a Basis for Treatment of Rett Syndrome
Xin Tang, Jesse Drotar, Keji Li, Cullen Clairmont, A. Sophie Brumm, Austin Sullins, Hao Wu, X. Shawn Liu, Jinhua Wang, Nathanael Gray, Mriganka Sur, Rudolf Jaenisch
Society for Neuroscience, 2019

The delicate balance between excitatory and inhibitory signaling in neural circuits (E/I balance) is critical for brain function. Disruption in E/I balance and hyperexcitability at the synapse, neural circuit, and behavioral levels have emerged as core mechanisms underlying a variety of brain disorders. The neuron-specific K+/Cl- cotransporter-2 (KCC2) is a ‘keystone’ molecule that is critical for the maturation of both GABAergic neurotransmission and excitatory synapse function, and has emerged as a promising therapeutic target to restore E/I balance in brain disorders including epilepsy, schizophrenia, spinal cord injury, and Rett syndrome, a severe neurodevelopmental disorder. Due to the lack of robust high-throughput screening (HTS) assay, it has been challenging to discover chemical compounds that enhance the expression of the KCC2 gene. In this study, we report the development of a novel human neuron-based high-throughput drug screening platform that allows for the rapid assessment of KCC2 gene expression in genome-edited reporter neurons. We have identified a group of compounds from an unbiased screen of over 900 small molecule chemicals that enhance KCC2 expression termed KCC2 expression-enhancer compounds (KEECs). The identified KEECs include FDA-approved drugs that are inhibitors of the FLT3 or GSK3β kinase pathways, and activators of the SIRT1 or TRPV1 pathways. We demonstrate that treatment with these hit compounds robustly increases KCC2 expression in human WT and isogenic Methyl CpG binding Protein 2 (MECP2) mutant RTT neurons, and rescues the deficits in GABA reversal potential, excitatory synaptic transmission, and morphological development of RTT neurons to levels equivalent to WT neurons. Moreover, injection of KEECs KW-2449 or Piperine into a Mecp2 mutant mouse model of RTT ameliorates disease-associated respiratory and locomotion abnormalities. The small molecule compounds described in our study could potentially benefit various brain diseases through a novel mechanism of enhancing KCC2 expression.

3D imaging and High-Content Analysis of Intact Human Cerebral Organoids
Alexandre Albanese, Justin Swaney, Dae Hee Yun, Nick Evans, Jenna Antonucci-Johnson, Chang Ho Sohn, Vincent Pham, Chloe Delepine, Mriganka Sur, Lee Gehrke, Kwanghun Chung
Cell Symposium, 2019

Patient-derived organoids present the opportunity to study complex cellular ecologies. However, dissociation or sectioning are required to characterize cell populations and their cytoarchitecture. The loss of spatial information in these heterogeneous tissues limits characterization by removing the spatial context required to decipher the properties and organization of cell populations. Here, we present a high-throughput pipeline for high-dimensional phenotyping of cerebral organoids at single-cell resolution. Our platform rapidly preserves, clears, labels, and images a large number of intact organoids. Fully automated algorithms extract hundreds of parameters characterizing cell populations, ventricles, and cytoarchitecture from high-resolution volumetric images in a quantitative and unbiased manner. Our pipeline quantified multi-scale changes in cerebral organoid models of Zika virus and Rett syndrome.

Non-Parametric Hyperdimensional Analysis of Multiscale Phenotypic Factors in Intact Human Cerebral Organoids
Justin Swaney, Alexandre Albanese, Dae Hee Yun, Nick Evans, Lee Kamentsky, Minyoung Kim, Chang Ho Sohn, Jenna Antonucci-Johnson, Vincent Pham, Chloe Delepine, Mriganka Sur, Lee Gehrke, Kwanghun Chung
Society for Neuroscience, 2019

Cerebral organoids are the most complex in vitro model of the developing human brain to date. At the single-cell level, cerebral organoids contain diverse cell populations with distinct expression patterns, and at the tissue level, cytoarchitectures observed during human cortical development begin to emerge. However, current techniques used to characterize cerebral organoids do not scale across these different length scales. Here, we present a tissue processing pipeline for whole-organoid antibody labeling and imaging as well as a computational framework for extracting and analyzing multiscale phenotypic factors from these images. SHIELD tissue processing preserves protein epitopes, mRNA, and endogenous fluorescence in intact cerebral organoids and produces optically transparent samples that can be imaged using light-sheet microscopy. The computational framework segments all nuclei using a curvature-based seeded watershed algorithm and classifies cell types based on the mean fluorescence intensity of each cell type marker. The spatial proximity to progenitor cells (SOX2) and post-mitotic neurons (TBR1) revealed four distinct cellular niches within cerebral organoids cultured for 35 days. A clustering analysis of radial cell profiles identified five types of cortical cytoarchitectures corresponding to different layering patterns of progenitor cells and post-mitotic neurons. By concatenating descriptors of single-cells, cortical cytoarchitectures, and whole-organoid morphology, a hyperdimensional feature vector of multiscale phenotypic factors is constructed for each organoid. Correlation analysis between multiscale phenotypic factors quantified the interscale relationships between cerebral organoids cultured for 35 and 60 days. Permutation testing revealed differences in phenotypic factors corresponding to deficits in cortical development in cerebral organoid models of Zika virus and Rett syndrome. These results demonstrate how phenotypic factors observed from the single-cell level to the tissue level can be integrated into a single, unbiased analysis of intact cerebral organoids.

Critical contributions of astrocytes to motor learning in vivo
Chloe Delepine, Keji Li, Vaibhavi Shah, Taylor Johns, Mriganka Sur
Gordon Research Conference, 2019

Astrocytes, long thought to operate only as a support network for neurons, are now emerging as key players in the modulation of brain information processing. Astrocytes influence synaptic transmission via glutamate transporters, and respond to, as well as modulate, neuronal activity with calcium signaling. However, many questions remain in understanding the contribution of astrocytes in vivo to complex behaviors and cognition. During motor learning, primary motor cortex (M1) is functionally and structurally reorganized. The learning of a new movement is associated with changes in neuronal activity and dendritic spine turnover. We hypothesize that astrocytes are modulators of learning-associated neuronal network reorganization by influencing synaptic strength through glutamate clearance and calcium signaling. Here we investigate the role of cortical astrocytes in a motor learning, lever-push task in vivo. Using the engineered human muscarinic G protein-coupled receptor DREADD-hM3Dq activated by low doses of clozapine-N-oxide (CNO), we find that modulation of astrocyte calcium activity perturbs performance of the animal in the lever-push task (causing decreased lever push responses to a cue sound). Moreover, we use a transgenic mouse line in which the expression of the glutamate transporter GLT1 can be inhibited locally in M1 and show that decreasing astrocyte glutamate clearance prevents learning of stereotyped motor trajectory. Using genetically encoded calcium indicators and high-resolution two-photon imaging, we then show that perturbation of astrocyte calcium activity and GLT1 knockout modulate the correlation structure of neuronal population activity and the movement trajectory encoding. This project utilizes cutting-edge imaging techniques to unravel astrocyte function during a physiologically relevant task involving motor cortex plasticity.

Support: NIH EY007023, NIH EY028219

2018

Bidirectional Control of Orienting Behavior by Distinct Prefrontal Circuits
Rafiq Huda, Grayson O. Sipe, Elie Adam, Vincent Breton-Provencher, Gerald Pho, Liadan Gunter, Ian R. Wickersham, Mriganka Sur
Society for Neuroscience, 2018

Animals respond to their environments using complex and diverse motor movements, but are limited by being able to enact only single actions at a time. Hence, voluntary control over behavior requires context-dependent mechanisms that select appropriate actions and suppress complementary but inappropriate ones. Such duality of behavioral control is readily apparent in sets of commonly displayed opposing behaviors, such as freeze/flight, approach/avoidance, and exploration/exploitation. The prefrontal cortex (PFC) has been widely implicated in dynamically coordinating behavior by biasing the flow of activity in downstream cortical and subcortical structures, but a fundamental outstanding question is how the anatomical organization of inputs to and outputs from the PFC enables its proposed role. Here we use multiple approaches to analyze a two-alternative task and deconstruct the circuit logic of the anterior cingulate cortex (ACC), a subdivision of the mouse PFC. We trained mice to perform a leftward-rightward forepaw orienting movement in response to bilaterally presented visual cues. Using a combination of virus-mediated anatomical tracing, projection-specific optogenetic manipulation and multiphoton imaging, we show that the ACC integrates and routes discrete sensory inputs to anatomically segregated populations of projection neurons in order to promote and inhibit goal-directed visual orienting responses. ACC outputs to the superior colliculus principally inhibit incorrect orienting movements. A projection-based activity model predicts that feedback from the ACC to the visual cortex via a non-overlapping set of neurons is critical for correct orienting, which we confirm. Our results suggest that integrating anatomically distinct but functionally complementary projections for bidirectional control may be a general organizing principle for PFC circuits.

Support: National Eye Institute grant F32 EY024857; National Institute of Mental Health grants K99 MH112855, F32 EY028028, R01 EY007023, U01MH106018, U01MH109129; National Institute of Neurological Disease and Stroke grant U01 NS090473; NSERC; National Science Foundation grant EF1451125; Simons Foundation Autism Research Initiative

Patterns of Astrocytic Microdomain Activity in the Motor Cortex During Motor Learning
Jennifer Shih, Chloe Delepine, Mriganka Sur
Society for Neuroscience, 2018

Excitatory neurons of the primary motor cortex (M1) function as part of neuronal ensembles, which are populations of neurons whose activity becomes correlated as an animal learns a stereotyped movement pattern. Astrocytes, the main type of glial cell in the cortex, have been shown to respond to and influence neuronal activity by transiently increasing their intracellular Ca2+ levels. Astrocytic Ca2+ transients have diverse spatiotemporal characteristics and can spread through the entire astrocyte cell body or be limited to primary branches or fine processes. Emerging evidence suggests that different types of astrocytic Ca2+ activity are associated with different types of neuronal activity and signaling by different neurotransmitter systems. However, definitive correlation between astrocyte Ca2+ activity and defined patterns of neuronal activity has not been established. Here, we focus on Ca2+ microdomain activity in astrocyte fine processes, which are closely associated with synapses and are well-positioned to respond to and influence neuronal activity. We trained mice on an auditory tone-cued lever push task, a behavioral paradigm that incorporates associative learning aspect as well as the acquisition of a stereotyped motor movement. Manipulation of astrocyte mechanisms suggest that altered astrocytic Ca2+ signaling influences neuronal activity during motor learning, leading to decreases in task performance. We hypothesize that a stable pattern of M1 astrocyte Ca2+ activity in microdomains develops during acquisition of a stereotyped motor movement, is correlated with neuronal ensemble formation and motor learning, and is potentially causal for learning. We use chronic two-photon imaging of astrocyte Ca2+ microdomain activity in vivo to test this hypothesis and show that distinctive patterns of astrocyte microdomain activity emerge as an animal learns the lever push task. Our results so far suggest that Ca2+ activity in microdomains may encode an astrocyte “signature” that is associated with motor learning, and thus identifies a novel neuron-astrocyte interaction that occurs during motor behavior.

Support: NIH grants EY007023, EY028219

Critical Contribution of Astrocytes to Motor Learning in vivo
Chloe Delepine, Vaibhavi Shah, Taylor Johns, Elie Adam, Mriganka Sur
Society for Neuroscience, 2018

Astrocytes, long thought to operate only as a support network for neurons, are now emerging as key players in the modulation of brain information processing. Astrocytes influence synaptic transmission via glutamate transporters, and respond to, as well as modulate, neuronal activity with calcium signaling. However, many questions remain in understanding the contribution of astrocytes in vivo to complex behaviors and cognition. During motor learning, primary motor cortex (M1) is functionally and structurally reorganized. The learning of a new movement is associated with changes in neuronal activity and dendritic spine turnover. We hypothesize that astrocytes are modulators of learning-associated neuronal network reorganization by influencing synaptic strength through glutamate clearance and calcium signaling. Here we investigate the role and plasticity of cortical astrocytes in a motor learning, lever-push task in vivo. Using the engineered human muscarinic G protein-coupled receptor DREADD-hM3Dq activated by low doses of clozapine-N-oxide (CNO), we find that modulation of astrocyte calcium activity perturbs performance of the animal in the lever-push task (causing decreased lever push responses to a cue sound). Moreover, we use a transgenic mouse line in which the expression of the glutamate transporter GLT1 can be inhibited locally in M1 and show that decreasing astrocyte glutamate clearance prevents learning of smooth motor trajectory. Using genetically encoded calcium indicators and high-resolution two-photon imaging, we then show that perturbation of astrocyte calcium activity modulates the correlation structure of neuronal population activity. In contrast, GLT1 knockout increases neuronal activity and prevents the formation of correlated neuronal ensembles normally associated with the motor learning. This ongoing project utilizes cutting-edge imaging techniques, and novel technologies for manipulating astrocyte activity, to unravel astrocyte function during a physiologically relevant task involving motor cortex plasticity.

Support: NIH grants EY007023, EY028219

Diverse Heterogeneity of Astrocyte Calcium Transients Underlies Distinct Neuronal Activity Profiles
Grayson Sipe, Vincent Breton-Provencher, Xin Tang, Rafiq Huda, Rodrigo Garcia, Rajeev Rikhye, Jeremy Petravicz, Mriganka Sur
Society for Neuroscience, 2018

Astrocytes represent a significant population of non-neuronal cells with roles in ionic homeostasis, neurotransmitter transport, and metabolic coupling. How astrocytes dynamically influence neuronal circuits remains unclear. Previous work has determined that astrocytes spontaneously display both wide-scale global calcium events as well as localized microdomain transients. Though norepinephrine has been linked to the generation of large astrocyte calcium events in vivo, concurrent neuronal activity at the synaptic and network levels during these events have not been well characterized. We have used two-photon microscopy and novel genetic tools to investigate the correlation of neuronal activity with diverse calcium profiles in astrocytes. We find that norepinephrine, which is known to be crucial for global brain state changes in arousal, novelty response, and attentional task switching, is sufficient to drive global calcium events. Additionally, we find that glutamate uptake via GLT-1 mediates a subset of astrocyte microdomain events and is driven by excitatory activity during visual processing. This work highlights the fact that the diverse gradient of calcium activity in astrocytes reflects diverse states of neuronal networks and likely corresponds to different functional roles.

Support: NIH grant 1F32EY028028-01

Active Control of Arousal by a Locus Coeruleus GABAergic Circuit
Vincent Breton-Provencher, Mriganka Sur
Society for Neuroscience, 2018

Arousal and novelty responses linked to locus coeruleus noradrenergic (LC-NA) activity affect cognitive performance. However, the mechanisms that control modes of LC-NA activity remain unknown. Here, we reveal a local population of GABAergic neurons (LC-GABA) capable of modulating LC-NA activity and arousal. Monosynaptic retrograde virus tracing shows that inputs to LC-GABA and LC-NA neurons arise from similar regions, though a few regions provide differential inputs to one subtype over the other. Extracellular targeted recordings to LC-NA and LC-GABA populations demonstrate two modes of LC-GABA responses whereby spiking is either correlated or broadly anti-correlated with LC-NA responses, reflecting anatomically similar and functionally coincident inputs, or differential and non-coincident inputs, to LC-NA and LC-GABA neurons. Optogenetic modulation of the two populations shows that coincident inputs control the gain of phasic LC-NA mediated novelty responses, while non-coincident inputs, such as from the prefrontal cortex to LC, alter overall levels of LC-NA responses without affecting response gain. These findings demonstrate distinct modes by which an inhibitory LC circuit regulates the gain and tone of arousal in the brain.

Support: NIH grants EY007023, NS090473

Enrichment of Plasticity-Related Synaptic Proteins at Functionally Identified V1 Synapses During Ocular Dominance Plasticity in vivo
Jacque P K Ip, Taeyun Ku, Sami El-Boustani, Kwanghun Chung, Mriganka Sur
Society for Neuroscience, 2018

In mice, the primary visual cortex (V1) in one hemisphere receives dominant input from the contralateral eye and relatively weaker input from the ipsilateral eye. Monocular deprivation (MD) induced by brief eyelid closure of one eye results in unbalanced input to V1 neurons. MD applied during the critical period results in ocular dominance plasticity, due to weakening of neuronal responses to the deprived (closed) eye followed by strengthening of responses to the non-deprived (open) eye. Previous studies suggested that persistent spine remodeling can be induced by MD, resulting in a robust effect on functional plasticity of responses in V1 neurons. However, the mechanisms by which MD remodels dendritic spines are unknown. For example, the eye-specific and other functional properties of spines eliminated during short term MD are unclear, as are the molecular mechanisms that lead to dendritic spine elimination and homeostatic spine strengthening or formation. We have previously described a prominent re-distribution of the immediate early gene Arc and surface AMPARs in individual identified neurons following MD and eye re-opening using fluorescently tagged probes together with two-photon imaging of dendritic spines. However, visualization of fluorescent probes under two-photon microscopy is still restricted by the optical diffraction limit and the availability of individual probes. Here, we employed the magnified analysis of the proteome (MAP) to linearly expand the cortex by fourfold to examine, at super-resolution scale, the enrichment of plasticity proteins at the postsynaptic density to correlate to structural changes of the same dendritic spines observed with two-photon microscopy. To achieve this, plasmids expressing GCaMP6s and the structural filler mRuby2 were delivered into individual mouse V1 neurons during the critical period by two-photon guided electroporation. Electroporated neurons were imaged before and after MD to assess how the decrease in synaptic drive remodels spines. By measuring eye-specific drive in neurons and spines in which we expressed GCaMP6s, we found that spines with input from the closed eye rapidly reduced in size, whereas increasing numbers of spines responded to the open eye. Following functional imaging in vivo, the cortices were subjected to MAP processing for identifying the enrichment of plasticity-related synaptic proteins. Various antibodies targeting pre- and post-synaptic terminals were used to verify the enrichment of plasticity-related proteins. Ongoing studies suggest that active redistribution of synaptic proteins underlies functional experience-dependent plasticity of V1 neurons.

Support: NIH grants EY007023, NS090473; HFSP Long-term fellowship

A Feedback Circuit Shaping Spatial and Reward Expectation During Visually Guided Locomotion
Elie M. Adam, Mriganka Sur
Society for Neuroscience, 2018

The ability to build up expectations, and hence predict and anticipate, is at the heart of cognition. This ability is enabled by assimilating prior experience, and, in light of it, completing acquired partial information, which is reincorporated as experience to build future expectations. We are interested in how this interaction is biologically implemented and how it is leveraged through feedback mechanisms to drive expectation and anticipation. We have developed a behavioral task where head-fixed mice run on a treadmill through a virtual linear track, then stop and wait at fixed visual landmarks to collect specified rewards. Mice learn to perform the task reliably, by moving quickly through the track and recognizing landmarks. By altering reward contingencies, we dissociate spatial information from reward information, thereby setting up rule-specific expectations. We hypothesize that the task biologically recruits an interplay of two pathways: one relaying spatial information, mediated through retrosplenial cortex (RSC), and one relaying reward information, mediated through the prefrontal cortex (PFC). Retrograde tracing experiments reveal reciprocal connections between caudal anterior cingulate cortex (ACC), a subdivision of medial PFC, and both rostral and caudal RSC, as an anatomical basis for this interplay. Preliminary two-photon calcium imaging in caudal RSC and caudal ACC reveals neural population activity that is locked to locomotion activity during task performance. The low-dimensional dynamics of the population activity contain significant information on the animal’s speed. Using projection-specific in-vivo two-photon imaging, coupled with causal optogenetic manipulations, we are further investigating the information content and functional role of these feedforward and feedback pathways, with the goal of understanding their specific contributions to experience and expectation.

Support: NIH grants EY007023, NS090473; Picower Fellowship

Emergence of Neuron Clusters in Mouse Motor Cortex During Learning
Keji Li, Jacque P.K. Ip, Mriganka Sur
Society for Neuroscience, 2018

The superficial layers of the mouse motor cortex is connected with both sensory areas and motor thalamus. It is essential for mouse motor learning, and both synaptic strength and temporal activity patterns have been found to change in the motor cortex following motor learning. However, it is yet unclear what computation is carried out in the motor cortex to facilitate learning, and what principle underlies the changes in motor cortex connectivity following learning. To examine the neuronal correlates of motor learning, we performed two photon Ca+ imaging of gCaMP6s- or jR-GECO-expressing neurons in the forelimb representation in the motor cortex of wildtype mice while training them on a simple lever push task with water restriction (Peters, Chan and Komiyama, 2014). The mice were trained to hold a lever still, then push it forward on an audio signal for water, or get punished with white noise and extra wait time.
During training, the mice first learned to start the push at the signal, then gradually refined the push trajectory to make a smooth “expert” push. When clustering the trials by dynamic time warping linkage of their trajectories, the “expert” push cluster started to appear around day 5 of training, and grew with training to include most trials towards day 14. The other clusters representing different push trajectories, while initially seen in similar numbers of trials as the “expert” push, gradually shrank in number and started to disappear after day 11. Wildtype mice were trained to perform the “expert” push reliably within two weeks, and we could track the same 20-30 layer 2/3 neurons between regular imaging sessions during training. We first examined the functional connectivity among tracked neurons as measured with “noise” correlations while the animal was not moving. The neurons initially increased the strength of both positive and negative noise correlations indiscriminately during training. From day 10 to day 13, the same time-period when the mice transitioned to use the “expert” push, two clusters formed among the recorded neurons, where correlations within clusters grew more positive and noise correlations grew more negative between the clusters. When classifying the neurons by their correlation with the “expert” push at the end of training, almost all neurons belonging to one cluster were push-correlated, while the other cluster had all neurons with negative or no correlation. These results suggest that the reorganization of the motor cortex during learning involves the formation of neurons clusters with strong interconnections, which are also activated during movement.

Support: NIH grants EY007023, EY028219, MH085802

Imaging neuronal responses through all cortical layers and subplate of visual cortex in awake mice with optimized three-photon microscopy
Murat Yildirim, Hiroki Sugihara, Peter T.C. So, Mriganka Sur
Society for Neuroscience, 2018

In sensory cortices, information arrives from primary nuclei of the thalamus into multiple cortical layers, but most densely into layers 4 and 6, and is subsequently transformed by inter-laminar circuits to enable cortical information processing. Two-photon microscopy has been used to measure neuronal activity mainly in superficial cortical layers, but has severe limitations for imaging deeper layers. Thus, response features of identified neurons in deeper cortical layers have remained unclear. Here, we demonstrate the optical design of a custom-made three-photon microscope to image a vertical column of the cerebral cortex >1 mm in depth in awake mice with low (<20 mW) average laser power. In order to optimize microscope design for minimal tissue damage, we developed two label-free methods to determine optical properties of cortical tissue in awake mouse brain, based on third harmonic generation imaging of blood vessels and tissue ablation respectively. We demonstrate functionality of the microscope by imaging the cross-laminar dendritic structure of layer 5 neurons in GFP-M mice, and functional visual responses of neurons expressing GCaMP6s across all layers of the primary visual cortex (V1) as well as in the subplate in awake mice. These recordings of identified deep layer neurons reveal that layer 5 neurons are more broadly tuned to visual stimuli whereas layer 6 neurons are more sharply tuned compared to neurons in other layers. Subplate neurons, located in the white matter below cortical layer 6 and also characterized here for the first time, are less visually responsive, and their orientation selectivity is broader compared to those of neurons in the cortical layers. These results demonstrate the design and utility of a custom-made three-photon microscope in revealing fundamental differences between neurons in different cortical layers, and between cortical and subplate neurons – which have been implicated in the pathogenesis of developmental brain disorders including autism and schizophrenia. Furthermore, the principles we describe can guide the development of damage-free three-photon microscopy for live functional imaging with subcellular resolution in other complex tissues.

Support: NIH grants EY007023, NS090473; NSF grant EF1451125

Diverse heterogeneity of astrocyte calcium transients underlies distinct neuronal activity profiles
Grayson Sipe, Vincent Breton-Provencher, Xin Tang, Rafiq Huda, Rodrigo Garcia, Rajeev Rikhye, Jeremy Petravicz, Mriganka Sur
Glia in Health and Disease, Cold Spring Harbor Laboratory, 2018

Astrocytes represent a significant population of non-neuronal cells with roles in ionic homeostasis, neurotransmitter transport, and metabolic coupling. How astrocytes dynamically influence neuronal circuits remains unclear. Previous work has determined that astrocytes spontaneously display both wide-scale global calcium events as well as localized microdomain transients. Though norepinephrine has been linked to the generation of large astrocyte calcium events in vivo, concurrent neuronal activity at the synaptic and network levels during these events have not been well characterized. We have used two-photon microscopy and novel genetic tools to investigate the correlation of neuronal activity with diverse calcium profiles in astrocytes. We find that norepinephrine, which is known to be crucial for global brain state changes in arousal, novelty response, and attentional task switching, is sufficient to drive global calcium events. Additionally, we find that glutamate uptake via GLT-1 mediates a subset of astrocyte microdomain events and is driven by excitatory activity during visual processing. This work highlights the fact that the diverse gradient of calcium activity in astrocytes reflects diverse states of neuronal networks and likely corresponds to different functional roles.

Support: NIH grants 1F32EY028028-01; EY028219

2017

Information Flow Across Long-Range Cortical and Subcortical Circuits Coordinates Sensorimotor Behavior
Rafiq Huda, Gerald Pho, Liadan Gunter, Grayson Sipe, Mriganka Sur
Society for Neuroscience, 2017

Converting sensory inputs into goal-oriented motor actions is a fundamental task of the nervous system. While previous studies have shown that even simple sensorimotor behaviors require coordinated activity across a diverse array of brain regions, how flow of information between areas contributes to specific aspects of decision-making remains unclear. Consistent with anatomical criteria of a sensorimotor area, we found that the anterior cingulate cortex subdivision of the prefrontal cortex (PFC) receives strong sensory inputs from the visual cortex (VC) and sends outputs to the motor layer of the superior colliculus (SC), suggesting that it bridges sensation with action. To test this hypothesis, we developed a two-alternative forced choice task in which head-fixed mice report the spatial location of a visual stimulus by rotating a ball. We found that optogenetic inactivation of VC, PFC, or SC compromised performance on the task. Importantly, projection-specific optogenetic experiments revealed that VC inputs to PFC and PFC inputs to SC are necessary, suggesting that information flow across these long-range circuits plays a critical role during task performance. Next, we used two-photon imaging to record the activity of PFC neurons during the task. Decoding analysis revealed that PFC neurons encode information about the location of the sensory stimulus and the choices made by the animals, suggesting that PFC neurons participate in the sensorimotor transformation process. We are currently performing recordings from PFC neurons that project to the SC to determine if they encode specific aspects of the task. Together, our experiments suggest that visual information flows from VC to PFC, where it is converted into a motor plan and subsequently routed to the SC for motor execution, highlighting how communication across cortical and subcortical circuits contributes to sensorimotor behaviors.

Support: NIH grants EY007023, NS090473; NSF grant EF1451125

Using Machine Learning for Automated Animal Call Detection and Classification (ACDC)
Sharma, R. Landman, K. Srinivasan, R.T. Cheung, J. Sharma, M. Sur, G. Feng, R. Desimone
Society for Neuroscience, 2017

In order to use behavioral observations of freely moving animals for neuroscientific research, objective measurement and quantification is essential. In the context of vocalization studies involving the common marmoset monkey (Callithrix Jacchus), data processing tends to take the form of analyzing hundreds of hours of audio recordings. Processing this data manually has several drawbacks, such as being slow, labor-intensive, imprecise, and subjective. We present a software framework for automated Animal Call Detection and Classification (ACDC), designed for researchers to be able to easily train and utilize models to turn hours of recordings into structured data that specifies the type and timestamp of each animal vocalization. There are two main tasks that this software performs: detection, and classification. The detection task involves determining which segments of the audio include relevant animal vocalizations, while rejecting environmental noise, audio artifacts, human voices etc. The classification task involves taking these extracted audio segments and categorizing them into the types of vocalizations the software has been trained to detect. In order to train ACDC, a “call dictionary” of audio samples of each type of vocalization, as well as samples of noise, is provided. The detection approach then utilizes traditional audio feature extraction techniques in combination with a neural-network-based model to learn from these samples and provide a set of timestamps specifying where vocalizations were detected. The classification approach, in turn, mimics how a human might classify these distinct sounds. Since different types of vocalizations tend to have a unique, easily recognizable shape to their spectrogram, we treat this as an optical character recognition problem. We converted spectrograms into images and trained a convolutional neural network to perform classification on the resulting shapes. We used ACDC to analyze audio from small groups of marmosets, training the models to detect the five or so most common types of vocalizations. Our initial results are promising, reaching > 80% accuracy on the detection task and better than 90% accuracy on the classification task. Our priority for further work is to continue to improve accuracy, while keeping the code clean, modular, and adaptable for use by other researchers, with a view to open sourcing in the future.

Three-Photon Imaging of Intact Human Cerebral Organoids to Assess Key Components of Early Neurogenesis in Rett Syndrome
Murat Yildirim, Chloe Delephine, Danielle Feldman, Tianyu Wang, Dimitre Ouzounov, Stephanie Chou, Justin Swaney, Kwanghun Chung, Chris Xu, Peter So, Mriganka Sur
Society for Neuroscience, 2017

Rett Syndrome (RTT) is a pervasive, X-linked neurodevelopmental disorder that predominantly affects girls. In the vast majority of cases, it is caused by a sporadic mutation in the gene encoding methyl CpG-binding protein 2 (MeCP2). In this study, we have used RTT patient-derived induced pluripotent stem cells to generate 3D human cerebral organoids that can serve as a model for human neurogenesis in vitro. To gain a complete understanding of the structural and functional complexity that underlies human neurogenesis, we developed a three-photon microscope and performed Third Harmonic Generation (THG) imaging as a label-free, nondestructive 3D tissue visualization method in conjunction with multiphoton fluorescence microscopy (MPM) to complement our findings with specific neuronal markers in both fixed and live organoids.
As a proof of concept, we have performed THG imaging in healthy and mutant intact human cerebral organoids. We acquired an intrinsic THG signal and extrinsic MPM signal with the following laser configurations: 800 kHz repetition rate, 65 fs pulse width laser at 1300 nm wavelength. In these THG images, nuclei are clearly delineated and cross sections demonstrate the depth penetration capacity (> 1mm) that extends throughout the organoid. Imaging control and MeCP2-deficient human cerebral organoids in 3D reveals structural and protein expression-based alterations via both THG and MPM microscopy that are not possible to observe with 2D sections.

Support: NIH grant MH085802; NSF grant EF1451125

The Role of Astrocytic GABA Transport During Sensory Processing in Visual Cortex
Grayson Sipe, Rodrigo Garcia, Rajeev Rikhye, Jeremy Petravicz, Mriganka Sur
Society for Neuroscience, 2017

Increasing evidence indicates that astrocytes actively modulate neuronal network activity through dynamic neurotransmitter clearance at the synaptic cleft. Although neurotransmitter clearance maintains synaptic homeostasis, it can also affect synaptic efficacy and subsequent information processing. Most work on astrocytic neurotransmitter clearance focuses on glutamate uptake via the transporters GLT-1 and GLAST, however astrocytes also express the GABA transporters GAT-1 & GAT-3. In particular, multiple studies have indicated that GAT-3 is selectively expressed in astrocyte processes near synaptic clefts and is thought to mediate tonic inhibitory states. Though GAT-3 blockade has been shown to delay the onset of seizure events in epilepsy models, it is not known how GAT-3 activity influences cortical information processing. To explore this question, we investigated the function of GAT-3 in primary visual cortex of the adult mouse. We characterized GAT-3 function in the cortex using immunohistochemistry and confirmed previous reports that GAT-3 expression heavily co-localizes with astrocytic processes and displays layer-specific heterogeneity. We then characterized GAT-3 function in the visual cortex using a combination of 2-photon microscopy, optogenetic activation, and slice electrophysiology. Our data suggests that neuronal reliability and tuning properties are disrupted with GAT-3 blockade. In conjunction with previous work in the lab investigating the role of astrocytic GLT-1 in visual cortex, these data indicate that astrocytes contribute to the crucial balance of excitation and inhibition necessary for cortical information processing.

Altered Intracellular Chloride Level Leads to Reduced Inhibition and Cortical Network Deficits in Rett Syndrome
K. Li, R. V. Rikhye, C. Li, Z. Fu, M. Sur
Society for Neuroscience, 2017

Rett Syndrome (RTT) is caused by genetic mutations in MEPC2 and leads to autism-like symptoms. Disruption of excitation/inhibition (E/I) balance in the brain has been observed in mouse RTT models, and is considered a potential mechanism for RTT symptoms. Previous in vitro studies had showed that Mecp2 mutation in cell culture causes decreased expression of a K ⁺/Cl ^–cotransporter KCC2 that pumps chloride into neurons, and increased intracellular concentration of chloride, which leads to higher GABA _AR reversal potential and reduced inhibition. We explored this specific mechanism in the brain and the impact of its pharmacological reversal on cortical neuron response properties as well as animal physiology. Bumetanide blocks a Na ⁺/K ⁺/Cl ^– cotransporter, NKCC1, that pumps chloride out of neurons, and was used in this study to reverse the impact of KCC2 suppression in Mecp2 homozygous male (KO) or heterozygous female (HET) mice. First we performed perforated patch recordings in brain slices, and found that KO mouse neurons had higher GABA _AR reversal potential than wildtype (WT). Bumetanide treatment restored this heightened reversal potential to WT levels (Banerjee et al., PNAS 2016). Super-clomeleon is a Cl- sensitive FRET fluorophore that has been used for in vitro ratiometric measurement of Cl- concentration. To examine the intracellular Cl ^– change in vivo, we developed two photon imaging of super-clomeleon in cortical neurons of awake mice. Preliminary data showed increased Cl- in HET mice compared to WT, which was reduced after bumetanide treatment.
To probe the direct effect of altered inhibition in the cortex, we carried out two photon calcium imaging in vivo of primary visual cortex (V1) pyramidal neurons, and tested visual response properties of the same neurons in HET mice before and after bumetanide treatment. Contrast gain control, a property of feedforward inhibition in cortical circuits, was reduced in HET compared to WT mice, and was partially rescued by bumetanide treatment. Apnea is a prominent symptom of RTT: we performed full body plethysmograph on KO mice to examine the physiological effect of reversing the altered chloride level. While sham injected KO mice showed increased frequency of breath pauses between P28 and P42, bumetanide treated KO mice showed no respiration decline, similar to WT mice. These results suggest that a specific mechanism of Cl ^– imbalance contributes to multiple stages of RTT pathology in a mouse model, and demonstrate that reversing the imbalance can alleviate some symptoms at these stages.

Plasticity of Identified, Functionally Heterogeneous Synapses Shapes Cell-Wide Plasticity of V1 Neurons in vivo
Sami El-Boustani*, Jacque P K Ip*, Vincent Breton-Provencher, Hiroyuki Okuno, Haruhiko Bito, Mriganka Sur
(* These authors contributed equally to this work)
Society for Neuroscience, 2017

Neuronal circuits in the developing and mature brain are subject to dramatic changes driven by sensory inputs or motor learning, causing individual cells to modify their responses to individual inputs while maintaining a relatively stable level of overall activity. Cell-wide homeostatic plasticity was initially reported as a global mechanism for stabilizing the output firing rate of a cell by uniformly scaling up or down the effective strength of all its synapses. More recent experimental evidence in vitro has suggested the existence of local homeostatic mechanisms acting at the level of dendritic stretches or even at single synapses that would potentially confer rich functional compartmentalization within the dendritic tree. However, the existence and nature of local homeostatic plasticity in vivo and its implications for the coherent reorganization of single cell responses remains unexplored. Here we have used visual-optogenetic pairing to demonstrate that induction of receptive field plasticity in single visual cortex neurons of awake mice alters identified synapses on neuronal dendrites. Such plasticity potentiates specific synapses and depresses others within short stretches of the same dendrite, consistent with functionally heterogeneous local sets of synapses that convey diverse receptive field inputs to a neuron. Crucially, depressed spines lie in close proximity to potentiated spines, indicating coordinated Hebbian and homeostatic plasticity in vivo that involves neighboring synapses. AMPA receptors are trafficked into potentiated spines and removed from depressed spines via targeted expression of the immediate early gene product Arc in the latter spines. The spatially local distribution of depressed spines around potentiated spines, in conjunction with functionally intermixed synaptic inputs to dendrites, highlight a possible mechanism that organizes cell-wide plasticity with local dendritic interactions.

Support: NIH grants EY007023, NS090473

Redistribution of Synaptic Proteins Between Identified Synapses of V1 Neurons During Experience-Dependent Plasticity in vivo
Jacque P K Ip*, Sami El-Boustani*, Vincent Breton-Provencher, Hiroyuki Okuno, Haruhiko Bito, Mriganka Sur
(* These authors contributed equally to this work)
Society for Neuroscience, 2017

Cortical circuits are remodeled in response to changes driven by sensory inputs. Regulation of AMPA receptor (AMPAR) membrane expression and targeting of immediate early genes such as Arc is critical for synaptic plasticity. However, our understanding of the mechanisms of AMPAR and Arc targeting in vivo is limited, in part by a lack of time-lapse imaging strategies at the synaptic level. Furthermore, because conventional transfection methods such as viral vectors require weeks to express, they are less suitable for the expression of molecular probes to study synaptic dynamics during development. To overcome these issues, we delivered plasmids into individual neurons of mouse primary visual cortex (V1) by two-photon guided electroporation. Expression of fluorescently tagged probes in spines and dendrites enabled us to visualize surface AMPAR subunits and Arc in awake mice and to study their synaptic targeting in single, identified layer II-III neurons. We observed prominent redistribution of AMPAR and Arc in spines and along dendritic branches under different conditions in vivo. Ocular dominance (OD) plasticity in V1 during a critical period is a well-established model for studying experience-dependent cortical changes. V1 neurons show reduced responses from a deprived eye following monocular deprivation (MD), followed by recovery of responses when the deprived eye is re-opened. We imaged V1 neurons following MD, before and a few hours after eye re-opening during the critical period for OD plasticity, to assess how an increase in synaptic drive from the re-opened eye remodeled spines. By measuring eye-specific drive in neurons and spines in which we expressed GCaMP6s, we found that spines with input from the re-opened eye rapidly enlarged and showed reduced Arc expression, whereas nearby spines were reduced in size and showed increased Arc expression. These studies suggest that active redistribution of synaptic proteins underlies functional experience-dependent plasticity of V1 neurons.

Support: NIH grants EY007023, NS090473

Activity of Local Inhibitory Neurons Gates Locus Coeruleus Noradrenergic Activity
Vincent Breton-Provencher, Mriganka Sur
Society for Neuroscience, 2017

Locus coeruleus noradrenergic (LC-NE) neurons send broad neuromodulatory projections to the central nervous system. While these neurons are thought to be involved in a variety of functions, including arousal, attention, and cognitive shifts, the behavioral contexts and the underlying mechanisms that drive LC-NE neurons remain largely unknown. Here, we examined the microcircuit mechanism of LC-NE activity by investigating a population of local GABAergic neurons located in the ventromedial region of the LC (LC-GABA). We monitored pupil size as a behavioral marker of LC-NE activity, and performed cell type-specific optogenetic manipulations and electrophysiological recordings within the LC of awake head-fixed mice. We found that activating LC-GABA neurons resulted in constriction of the pupil, similar to direct inactivation of LC-NE activity, suggesting a functional coupling between these cell types. In support of this hypothesis, simultaneous recordings using multi-channel extracellular probes revealed an ongoing interaction between putative LC-GABA and LC-NE activity, in which GABAergic activity followed the increase in NE associated with pupil dilation. Importantly, LC-GABA neurons may critically gate LC-NE function, as increasing their activity impaired responses of LC-NE neurons to novel sensory stimuli. We are currently investigating brain-wide inputs to LC-NE and LC-GABA neurons using monosynaptic rabies virus tracing to identify upstream brain regions that regulate this gating mechanism. Overall, these results demonstrate the existence of a local population of interneurons in the LC capable of rapidly controlling overall NE activity by modulating LC responses to unexpected sensory stimuli.

Support: NIH grants EY007023, NS090473

Large-Scale Imaging and Functional Parcellation of Mouse Visual Cortex
Ming Hu, Rajeev V. Rikhye, Aadhirai R, Mari Ganesh Kumar M, Hardik Suthar, Michael J. Goard, Hema A. Murthy, Mriganka Sur
Society for Neuroscience, 2017

Although it is known that mice have roughly ten distinct visual areas, the specific function of each area and interactions between areas remain poorly understood. In this study, we used wide-field imaging from awake, head-fixed mice, which transgenically expressed GCaMP6f, to segment the entire visual cortex into functionally different areas and investigate each areas’ responses to a battery of visual stimuli, including simple drifting gratings (varying orientation, spatial frequency and temporal frequency), random moving dots and natural movies. Despite the well-known “salt-and-pepper” distribution of orientation preference at a microscopic scale in mouse V1, we found a retinotopically dependent mesoscale bias of averaged orientation preference and spatial and temporal frequency preferences, in response to drifting gratings. In addition, we found higher orientation selectivity in the binocular region. Representation of the amplitude and phase of natural movies were biased towards the binocular zone and monocular zone respectively. In addition to these conventional input-output characterizations, we also applied various machine learning techniques on this dataset to explore its intrinsic structure. Specifically, we used supervised Gaussian Mixture Models (GMM) and unsupervised agglomerative clustering of Maximum a Posteriori (MAP) model to estimate area borders. The generated models in both cases predicted area borders consistent with that of the physiologically identified retinotopic map, suggesting that each area had consistent and correlated responses to visual stimuli. In another experiment, we observed that spatial temporal change points in the visual stimuli correlated well with the change points in the cortical responses. Altogether, by integrating tools developed from mouse genetics, large-scale imaging and modern statistical analysis, we obtained novel insights into the functional organization of mouse visual cortex. These hypotheses are being verified with further experiments.

Support: NIH grant EY007023; NSF grant EF1451125

The Role of Astrocytic GABA Transport in Visual Cortex
Grayson Sipe, Rodrigo Garcia, Rajeev Rikhye, Jeremy Petravicz, Mriganka Sur
Gordon Research Conference, 2017

Increasing evidence indicates that astrocytes actively modulate neuronal network activity through dynamic neurotransmitter clearance at the synaptic cleft. Although neurotransmitter clearance maintains synaptic homeostasis, it can also affect synaptic efficacy and subsequent information processing. Most work on astrocytic neurotransmitter clearance focuses on glutamate uptake via the transporters GLT-1 and GLAST, however astrocytes also express the GABA transporters GAT-1 & GAT-3. In particular, multiple studies have indicated that GAT-3 is selectively expressed in astrocyte processes near synaptic clefts and is thought to mediate tonic inhibitory states. Though GAT-3 blockade has been shown to delay the onset of seizure events in epilepsy models, it is not known how GAT-3 activity influences cortical information processing. To explore this question, we investigated the function of GAT-3 in primary visual cortex of the adult mouse. We characterized GAT-3 function in the cortex using immunohistochemistry, slice electrophysiology and in vivo­ 2-photon microscopy. Immunohistochemical staining confirmed previous reports that GAT-3 expression heavily co-localizes with astrocytic processes and displays layer-specific heterogeneity. Preliminary data suggests that astrocytic GABA uptake via GAT-3 maintains inhibitory network states critical for sensory processing. In conjunction with previous work in the lab investigating the role of astrocytic GLT-1 in visual cortex, these data indicate that astrocytes contribute to the crucial balance of excitation and inhibition necessary for cortical information processing.

Role of Astrocyte Glutamate Transporters in Function and Plasticity of Visual Cortex Circuits
J. Petravicz, R. Rikhye, G. Sipe, R. Garcia, N. Mellios, M. Sur
Gordon Research Conference, 2017

Early in life, neuronal circuits in primary visual cortex (V1) are shaped by activity-dependent critical periods that require balances of excitation and inhibition. At excitatory synapses, astrocytes are responsible for nearly 90% of glutamate uptake in the cortex, the majority of which is mediated by glutamate transporter 1 (GLT-1). This uptake is critical for the maintenance of synaptic transmission not only at excitatory to excitatory synapses, but at excitatory to inhibitory synapses as well. Pharmacological blockade of GLT-1 is known to induce changes in synaptic transmission in situ, but little is known about its role in cortical transmission and sensory processing in vivo. To examine visual cortex plasticity during development, we employed a monocular deprivation (MD) paradigm. During the critical period, depriving V1 of inputs from one eye induces ocular dominance (OD) plasticity by reducing responses from the closed eye and increasing responses from the open eye. The mechanisms underlying this scaling of responses in OD plasticity involve synaptic potentiation and depression, both of which have been shown to be influenced by glutamate transporter activity. Using a combination of in situ and in vivo approaches, we report that reduced GLT-1 availability modulates normal visual cortex plasticity and function. Using mice that carry a heterozygous deletion of GLT-1 (GLT-1-/+), we find an altered ocular dominance index (ODI) with no monocular deprivation due to reduced contralateral eye response amplitude. Furthermore, while both control and GLT-1-/+ mice display significant ODI shifts with short-term (4 days) MD, this shift is absent after long-term (7 days) MD in GLT-1-/+ mice. Electrophysiological recordings of layer 2/3 pyramidal neuron mEPSCs reveal alterations in homeostatic scaling normally associated with ocular dominance plasticity in GLT-1-/+ mice. Using in vivo recordings of single eye responses of excitatory neurons in the binocular region of V1, we find alterations to eye specific drive that underlie the reduced pre-MD ODI in these mice. These findings indicate that a critical threshold of glutamate reuptake exists for the establishment of contralateral bias in the binocular region of the visual cortex and for ocular dominance plasticity. In mature circuits, we find that orientation tuning is sharpened in the excitatory pyramidal neurons with reduced GLT-1. Taken together, our results provide new insights into how astrocyte glutamate uptake influences the development of synapses, cortical plasticity, and response properties of intact visual cortex circuits.

Support: NIH grant EY-018648; NRSA grant EY-022284

In vivo Exploration of Astrocyte Contribution to Motor Learning
Chloe Delepine, Jacque Ip, Keji Li, Jeremy Petravicz, Mriganka Sur
Gordon Research Conference, 2017

Astrocytes, long thought to operate only as a support network for neurons, are now emerging as key players in the modulation of brain information processing. Astrocytes influence synaptic transmission via glutamate transporters, and respond to, as well as modulate, neuronal activity with increases in calcium signaling. However, many questions remain in understanding the contribution of astrocytes in vivo to complex behaviors and cognition. During motor learning, primary motor cortex (M1) is functionally and structurally reorganized. The learning of a new movement is associated with changes in neuronal activity and dendritic spine turnover. We hypothesize that astrocytes are modulators of the learning-associated neuronal network reorganization by influencing synaptic strength through glutamate clearance and calcium signaling. Here we investigate for the first time the role and plasticity of cortical astrocytes in a motor learning task in vivo. Using tools such as the engineered human muscarinic G protein-coupled receptors DREADD-hM3Dq activated by the pharmacologically inert chemical clozapine-N-oxide (CNO) and transgenic mouse lines in which the expression of the glutamate transporter GLT1 can be inhibited, we find that modulation of both astrocyte calcium activity and of astrocyte glutamate clearance perturbs motor learning performance. Using genetically encoded calcium indicators and high-resolution two-photon imaging, we then show that perturbation of astrocyte calcium activity modulates the neuronal activity pattern associated with the learning. This ongoing project utilizes cutting-edge imaging techniques, and novel technologies for manipulating astrocyte activity, to unravel astrocyte function during a physiologically relevant task involving motor cortex plasticity.

2016

Astrocyte Glutamate Transporters in Visual Cortex Circuit Function and Plasticity
Jeremy Petravicz, Rajeev Rikhye, Ming Hu, Nikolaos Mellios, Mriganka Sur
CSHL, 2016

Early in life, neuronal circuits in primary visual cortex (V1) are shaped by activity-dependent critical periods that require balances of excitation and inhibition. At excitatory synapses, astrocytes are responsible for nearly 90% of glutamate uptake in the cortex, the majority of which is mediated by glutamate transporter 1 (GLT-1). This uptake is critical for the maintenance of synaptic transmission not only at excitatory to excitatory synapses, but at excitatory to inhibitory synapses as well. Pharmacological blockade of GLT-1 is known to induce changes in synaptic transmission in situ, but little is known about its role in cortical transmission and sensory processing in vivo. To examine visual cortex plasticity during development, we employed a monocular deprivation (MD) paradigm. During the critical period, depriving V1 of inputs from one eye induces ocular dominance (OD) plasticity by reducing responses from the closed eye and increasing responses from the open eye. The mechanisms underlying this scaling of responses in OD plasticity involve synaptic potentiation and depression, both of which have been shown to be influenced by glutamate transporter activity. Using a combination of in situ and in vivo approaches, we report that reduced GLT-1 availability modulates normal visual cortex plasticity and function. Using mice that carry a heterozygous deletion of GLT-1 (GLT-1 ^-/+), we find an altered ocular dominance index (ODI) with no monocular deprivation due to reduced contralateral eye response amplitude. Furthermore, while both control and GLT-1 ^-/+ mice display significant ODI shifts with short-term (4 days) MD, this shift is absent after long-term (7 days) MD in GLT-1 ^-/+ mice. Electrophysiological recordings of layer 2/3 pyramidal neuron mEPSCs reveal alterations in homeostatic scaling normally associated with ocular dominance plasticity in GLT-1 ^-/+ mice. Using in vivo recordings of single eye responses of excitatory neurons in the binocular region of V1, we find alterations to eye specific drive that underlie the reduced pre-MD ODI in these mice. These findings indicate that a critical threshold of glutamate reuptake exists for the establishment of contralateral bias in the binocular region of the visual cortex and for ocular dominance plasticity. In mature circuits, we find that orientation tuning is sharpened in the excitatory pyramidal neurons with reduced GLT-1. Taken together, our results provide new insights into how astrocyte glutamate uptake influences the development of synapses, cortical plasticity, and response properties of intact visual cortex circuits.

Cortical Circuit Mechanisms Underlying Sensorimotor Behavior
Gerald Pho*, Rafiq Huda*, Liadan Gunter, Ian Wickersham, Wasim Malik, Emery Brown, Mriganka Sur
2016 BRAIN Initiative Investigators Meeting

A fundamental building block of voluntary behavior is the ability to respond to information from the environment with appropriate motor actions. While seemingly simple, this process of transforming sensory information into motor commands consists of at least two distinct steps – detecting sensory stimuli relevant to behavioral goals in a noisy environment, and selecting appropriate motor actions based on those stimuli. While neural substrates of sensorimotor mapping and attention have been widely studied in non-human primates, many questions remain due to lack of tools that can selectively probe and manipulate neural activity, especially in a projection-specific manner.
In this project, we developed visual sensorimotor tasks for head-fixed mice to probe the contribution of specific cortical regions and projections in attention and sensorimotor mapping. Using optogenetics and large-scale calcium imaging during a go/no-go lick-based task, we previously identified the posterior parietal cortex (PPC) as an important locus for sensorimotor transformation (Goard et al 2016). By varying task parameters such as attentional engagement, stimulus contrast, and reward contingency, we discovered that PPC encodes multiple task variables across heterogeneous cell types in an engagement-specific manner. We are now developing statistical models to describe how sensory and motor variables are combined with network activity to generate PPC responses.
We have additionally developed a two-alternative sensorimotor task that requires sustained attention. Using retrograde tracing, we identified a prefrontal region, the anterior cingulate cortex (ACC), that is anatomically poised to contribute to both attentional processing, owing to its top-down projections to visual cortex, and sensorimotor transformation, due to its feed-forward projections to superior colliculus. Optogenetic inactivation of ACC compromised task performance, and projection-specific inactivation revealed dissociable contributions of corticocortical versus corticotectal neurons to attention and sensorimotor transformation, respectively. We are now combining retrograde tracers with calcium imaging to investigate the physiological signatures of these projection neurons.

Massive-Scale Multi-Area Single Neuron Recordings to Reveal Circuits Underlying Short-Term Memory
Murat Yildirim, Michael J Goard, Christopher J. Rowlands, Gerald Pho, Rafiq Huda, Peter So, Mriganka Sur
2016 BRAIN Initiative Investigators Meeting

Short-term memory is a fundamental cognitive process underlying an array of complex abilities, but its neural mechanism is not fully understood. Many brain regions are implicated in memory-guided decisions, including visual, association, and motor cortices as well as subcortical structures. However, it is not mechanistically understood what regions are involved when, what neuronal subsets are recruited within these regions, or how they interact to represent information relevant to behavior. Conventional two-photon imaging systems offer high resolution imaging of the required brain regions, but the field of view is typically smaller than a cortical area in a mouse (e.g., primary visual cortex), which in turn inhibits the simultaneous measurement of neural activity in different brain regions. In this project, we aim to elucidate the role of specific cortical regions and neuronal subsets in a visually-cued memory-guided discrimination task, in order to understand the neural mechanism of short-term memory. For this purpose, we will perform very large scale (>10 mm2) calcium imaging through two-photon microscopy in behaving mice to measure activity of thousands of neurons simultaneously across multiple brain regions. To obtain larger scale (>10 mm2) calcium imaging through two-photon microscopy, we have developed a novel system which includes non-descanned multifocal microscopy with a single photon counting camera, custom-made intermediate optics, and a custom made objective and tube lens (5x/0.7NA) to preserve the signal to noise ratio (SNR) and high resolution across the entire field of view. In parallel, we have designed and implemented three-photon microscopy to perform structural and functional imaging for discovering how task-relevant information is coded in deeper cortical layers and brain structures. We acquired an intrinsic third harmonic generation (THG) signal and three photon fluorescence (3PEF) images. In these three-photon images, we can excite either green fluorescent protein (GFP) or green genetically encoded calcium indicator (GCamp6s) in the whole visual cortex, external capsule, and hippocampal region. In addition, we have performed THG imaging in healthy intact human cerebral organoids cleared with SWITCH. In these THG images, nuclei are clearly delineated and cross sections demonstrate the depth penetration capacity (> 1mm) that extends throughout the organoid. Imaging control and MeCP2-deficient human cerebral organoids in 2D sections reveals structural and protein expression-based alterations that we expect will be clearly elucidated via both THG and three-photon fluorescence microscopy.

MeCP2 Deficient Astrocytes Have Altered Signaling Pathway Activation and Reduced Visually-Evoked Microdomain Sizes
R. Garcia, R. V. Rikhye, J. Petravicz, C. Delépine, M. Sur
Society for Neuroscience, 2016

Loss of function mutations in the X-linked gene encoding for MeCP2 are the underlying genetic cause for Rett Syndrome (RTT), a devastating neurodevelopmental disorder that primarily affects girls. Loss of function mutations in this ubiquitously expressed transcriptional regulator leads to imbalances in excitation and inhibition and disruption to neuronal circuit function. While the function of this transcriptional regulator remains elusive and complex, recent focus has turned to downstream signaling pathways as putative targets for novel therapeutics. The complexity of MeCP2 function is compounded by the heterogeneity of cell types in the brain, with recent evidence implicating glia cells in RTT pathophysiology. MeCP2 expression has been detected in astrocytes, and selective deletion or re-introduction of MeCP2 in astrocytes alone has been sufficient to induce or ameliorate pathological symptoms, respectively. Previously, we identified signaling pathways upstream of synaptic function that are impaired in MeCP2 mouse models, yet the downstream molecular and signaling effects resulting from a loss of MeCP2 function in astrocytes remains unknown. Here we measure signaling and astrocyte-specific proteins in a heterogeneous MeCP2-expressing population. We find that activated mTOR and AKT are reduced in astrocytes lacking MeCP2, while levels of cortical glutamate transporter 1 (GLT-1) are upregulated. We have recently shown that astrocytes in layer 2/3 of rodent visual cortex can respond to visual stimuli with robust and reliable microdomain Ca2+ elevation and that this effect is influenced by the availability and function of GLT-1. In MeCP2-/+ astrocytes, we find that the microdomain areas evoked during visual stimulation are reduced, in line with reduced circuit function. These data identify novel, cell-specific effects in astrocytes lacking MeCP2 and offer insight on their signaling and circuit interactions.

Circuit Mechanisms of Prefrontal Contribution to Visual Behavior
R. Huda, G. Pho, L. Gunter, I. Wickersham, M. Sur
Society for Neuroscience, 2016

A fundamental building block of voluntary behavior is the ability to respond to information from the environment with appropriate motor actions. While seemingly simple, this process of transforming sensory information into motor commands consists of at least two distinct steps – detecting sensory information relevant to current behavioral goals from a noisy environment and selecting appropriate motor actions based on those stimuli. While neural substrates of sensorimotor mapping and attention have been widely studied in non-human primates, many questions remain due to lack of tools that can selectively probe and manipulate neural activity, especially in a projection-specific manner. The availability of optogenetic and viral-based gene expression tools in mice enables the interrogation of cell type-specific contributions to circuit-level mechanisms underlying these cognitive phenomenon. Here, we devised a visual sustained-attention, sensorimotor task for head-fixed mice to probe the selective contribution of specific neuronal subsets. First, we used rabies viruses to identify a caudal midline prefrontal region (anterior cingulate cortex) that is anatomically poised to contribute to both attentional processing of visual stimuli, owing to its top-down projections to visual cortex, and sensorimotor transformation, due to its feed-forward projections to superior colliculus. Mice were trained to report the spatial location of a visual stimulus by rotating a ball. We made the task attention-demanding by including a variable foreperiod that introduces uncertainty in temporal expectancy for visual stimulus onset. In the first set of experiments, we found that optogenetic inactivation of this prefrontal area compromises performance on the task. Subsequently, two-photon microscopy and projection-specific optogenetic inactivation has allowed us to probe the functional contribution of visual cortex- and superior colliculus-projecting prefrontal neurons to sustained attention and sensorimotor transformation, respectively.

Third Harmonic Generation Imaging of Intact Human Cerebral Organoids to Assess Key Components of Early Neurogenesis in Rett Syndrome
M. Yildirim, D. Feldman, T. Wang, D. Ouzounov, S. Chou, J. M. Swaney, K. Chung, C. Xu, P. So, M. Sur
Society for Neuroscience, 2016

Rett Syndrome (RTT) is a pervasive, X-linked neurodevelopmental disorder that predominantly affects girls. It is mostly caused by a sporadic mutation in the gene encoding methyl CpG-binding protein 2 (MeCP2).The clinical features of RTT are most commonly reported to emerge between the ages of 6-18 months and implicating RTT as a disorder of postnatal development. However, a variety of recent evidence from our lab and others demonstrates that RTT phenotypes are present at the earliest stages of brain development including neurogenesis, migration, and patterning in addition to stages of synaptic and circuit development and plasticity. We have used RTT patient-derived induced pluripotent stem cells to generate 3D human cerebral organoids that can serve as a model for human neurogenesis in vitro. We aim to expand on our existing findings in order to determine aberrancies at individual stages of neurogenesis by performing structural and immunocytochemical staining in isogenic control and MeCP2-deficient organoids. In addition, we aim to use Third Harmonic Generation (THG) microscopy as a label-free, nondestructive 3D tissue visualization method in order to gain a complete understanding of the structural complexity that underlies human neurogenesis. As a proof of concept, we have performed THG imaging in healthy intact human cerebral organoids cleared with SWITCH. We acquired an intrinsic THG signal with the following laser configurations: 400 kHz repetition rate, 65 fs pulse width laser at 1350 nm wavelength. In these THG images, nuclei are clearly delineated and cross sections demonstrate the depth penetration capacity (> 1mm) that extends throughout the organoid. Imaging control and MeCP2-deficient human cerebral organoids in 2D sections reveals structural and protein expression-based alterations that we expect will be clearly elucidated via both THG and three-photon fluorescence microscopy.

Growth, Differentiation and Connectivity of Implanted Human Neuronal Precursor Cells in the Mouse Visual Cortex
J. Benoit, H. Wu, V. Breton-Provencher, J. P. K. Ip, D. Feldman, S. Chou, R. Jaenisch, M. Sur
Society for Neuroscience, 2016

The use of induced pluripotent stem cells (iPSCs) to recapitulate the effect of human genetic diseases in an experimentally tractable system requires both a rich genomic as well as cellular context. Current models for neurological diseases generally consist of human induced-neuronal (iN) cells engineered with the specific mutations of interest, or derived from patient cells with those same mutations, and grown in vitro in a 2D culture. In order to create a more natural and 3D environment in which to grow and assess human iNs, we differentiated “wild-type” AAVS1-CAG-tdTomato human neuronal precursor cells (NPCs) which were then transplanted into the mouse cortex to form a “humanized” functional network in a model system. We injected approximately 10,000 NPCs into the primary visual cortex (V1) of SCID immunodeficient mice at P21 and then using a craniotomy, examined their morphological development in vivo from 6 weeks post-injection onwards. Of the injected NPCs, several hundred survived and were localized mainly to the injection tract although some cells in deeper layers (~300 um from pia) were well-intercalated between endogenous mouse cells within several hundred um of the injection. NPCs sent wide-ranging projections, some of which reached to adjacent cortical areas and extended beyond the craniotomy (1.5 mm lateral distance) in some cases. We found evidence of filopodia and potential immature spines on human dendrites which suggests that the NPCs have differentiated into a neuronal phenotype and may be forming synapses with endogenous mouse neurons. We are currently examining calcium responses in these cells to determine if they possess functional contacts with the mouse cortical circuit and are therefore responsive to visual input. We posit that this system represents a more realistic environment with superior experimental validity in which the development of normal and patient-derived human neurons can be studied.

A Computational Model of Astrocyte Induced Modulation of Synaptic Plasticity and Normalization
V. Sreerag ¹, Ryan Phillips ¹, Srinivasa Chakravarthy ¹, Mriganka Sur ² 
Society for Neuroscience, 2016

¹Department of Biotechnology, Bhupat and Jyoti Mehta School of Biosciences, IIT Madras, India; ²Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA, USA.

It has been suggested that perisynaptic astrocytic processes have a role in synaptic plasticity. Here we propose a simple model for astrocytic modulation of synaptic plasticity. The model consists of presynaptic, postsynaptic and astrocyte compartments of the tripartite synapse. Presynaptic activation initiates neurotransmitter release into the synaptic cleft. This neurotransmitter binds with NMDA receptors leading to calcium influx into the postsynaptic compartment. However, NMDAR opening is also contingent upon postsynaptic potential-dependent magnesium blockage. The postsynaptic compartment is influenced by both the presynaptic compartment and astrocytic processes. Astrocytes release gliotransmitters including D-serine and glutamate, which in turn regulate the synaptic neurotransmitter concentration and influence the postsynaptic calcium concentration. Following the ‘calcium control hypothesis,’ we assume that the postsynaptic calcium controls the “synaptic strength”, w, which corresponds to the strength of response of AMPA receptors to synaptic glutamate. This synaptic strength is modelled by a cubic nonlinearity exhibiting bistable dynamics. The proposed model is comprised of a presynaptic variable (firing rate) and postsynaptic voltage (controlled by an external current), which are independently varied to display LTD and LTP. Interestingly, the model exhibits BCM-like dynamics wherein the plasticity switches from LTD to LTP at a threshold value of postsynaptic voltage. A key feature of the model is that this threshold voltage is modulated by gliotransmitter released by the astrocyte. We next simulate weight dynamics at two synapses supplied by the same astrocyte such that that the total gliotransmitter released to the two synapses is conserved. Simulations showed a mechanism of normalization of the two synaptic weights in that growth in one synapses is accompanied by attenuation in the other. In summary, the model demonstrates: (1) astrocyte induced modulation in LTD to LTP threshold, and (2) weight normalization of multiple synapses controlled via astrocytic gliotransmitter release.

Development of an Image-Based High-Content Screening Assay for Tau Clearing Drugs in a Human iPSC-Derived Neuronal Cell Model of Frontotemporal Dementia
Chialin Cheng, Surya A. Reis, Emily T. Adams, M. Catarina Silva, Krista M. Hennig, Daniel M. Fass, Danielle A. Feldman, Mriganka Sur, Bradford C. Dickerson, Kenneth S. Kosik, Tau Consortium, Stephen J. Haggarty
ISSCR, 2016

Autosomal dominant mutations in the microtubule-associated protein gene (MAPT) encoding the protein tau cause frontotemporal dementia spectrum disorders (FTD-s). These MAPT mutations are associated with pathologically abnormal tau phosphorylation levels and intracellular accumulation of aggregated protein predominantly in neurons (“tauopathy”). Recently, a rare variant of tau, Tau-A152T, located N-terminal of the microtubule-binding domain has been described. This variant of tau has decreased affinity for binding microtubules in vitro, and has been shown to increase the risk for FTD-s, Alzheimer’s disease, and synucleinopathies. Here we used human induced pluripotent stem cells (iPSC) from a FTD-s subject diagnosed with progressive supranuclear palsy carrying this Tau-A152T variant as a genetically accurate cell model of tauopathy. To utilize these cells to create a rapid and robust biological cellular assay system capable of supporting the discovery of novel therapeutics for tauopathies, we have adapted strategies for the inducible expression of the pro-neural transcription factor Neurogenin 2 in stably transduced iPSC-derived neural progenitor cells (iNgn2-NPCs). We demonstrate the ability to efficiently and reproducibly generate nearly limitless numbers of excitatory, glutamatergic-like neurons from these iNgn2-NPCs in a 96-well plate format with abundant expression of tau with enhanced polarized distribution to axonal processes. In order to monitor potential mutation-induced aberrant subcellular tau distribution, as well as drug-induced tau clearance, we developed a high-content image-based screen utilizing automated confocal microscopy and an advanced image-processing pipeline optimized for analysis of morphologically complex neuronal cultures. We summarize the results of a pilot screen for tau clearing compounds targeting autophagy and protein homeostasis pathways with an emphasis on clinically used FDA-approved drugs with potential for repurposing. This strategy will aid in expediting the translational research in elucidating novel targets for therapeutic intervention for neurological diseases involving tauopathy such as FTD-s.

Major Vault Protein, a Candidate Gene in 16p11.2 Microdeletion Syndrome, is Required for Homeostatic Regulation of Cortical Plasticity
Jacque Pak Kan Ip, Ikue Nagakura, Jeremy Petravicz, Jamie Benoit, Erik A.C. Wiemer, Mriganka Sur
Society for Neuroscience, 2016

Microdeletion of a region in the chromosome 16p11.2 increases susceptibility to autism and accounts for up to 1% of this population. Although this region contains 29 genes, disrupting only a small piece of this region, which spans 5 genes, is sufficient to cause autistic traits. One candidate gene in this region is the major vault protein (MVP), which has been implicated in the regulation of several cellular processes including transport mechanisms and multidrug resistance. We found that MVP expression levels in MVP Het mice closely phenocopy those of 16p11.2 mice, suggesting MVP Het mice may serve as a model of MVP function in 16p11.2 microdeletion. However, the function of MVP in the central nervous system, in particular its role in brain function and plasticity, has not been investigated. To determine the role of MVP in experience-dependent synaptic and circuit plasticity, we first measured ocular dominance plasticity (ODP) in primary visual cortex (V1). We found that MVP Het mice show impairment in strengthening of open eye responses in V1 after 7 days monocular deprivation (MD), resulting in reduced overall plasticity. Furthermore, electrophysiology experiments showed that the frequency of mEPSCs was decreased in MVP Het mice after 7 days MD, suggesting a decrease in the number of functional synapses, which may underlie the reduced plasticity in MVP Het mice. By a biotin-labeling assay, we found impaired homeostatic upregulation of surface GluA1 in MVP Het mice after longer term MD. To investigate the underlying molecular mechanism, we measured intracellular signaling in MVP WT and MVP Het mice and found ERK activation was significantly increased in MVP MVP Het mice, while activation of other signals such as Akt and JAK were normal. STAT1 is a downstream molecule of JAK signaling and reported to be inhibited by MVP. We found a tendency towards increased expression of STAT1 in MVP Het mice, suggesting the possibility of MVP inhibition on STAT1. We have previously examined ODP in STAT1 KO mice and shown that they have an accelerated increase in open eye responses and enhanced plasticity after 4 days of MD. These results suggest that MVP may interact with STAT1 to regulate plasticity, and one function of MVP may be to negatively regulate STAT1. Collectively, we find a highly specific role for MVP as a critical molecule in the homeostatic or response-restoring component of activity-dependent synaptic plasticity. Thus, this study helps reveal a new mechanism for an autism-related gene in brain function, and suggests a broader role for neuro-immune interactions in circuit level plasticity.

Depolarizing GABA Receptor Causes Cortical Network Deficits in Rett Syndrome
Keji Li, Rajeev V. Rikhye, Abhishek Banerjee, Mriganka Sur
Society for Neuroscience, 2016

Rett Syndrome (RTT), a form of autism spectrum disorder, is mainly caused by mutations of a single gene, Methyl CpG binding protein 2 (MECP2). Symptoms of RTT include developmental regression of acquired motor and language skills, and severe cognitive impairment. The mechanism behind these symptoms are not yet clear, but in mouse RTT models the disruption of excitation/inhibition (E/I) balance in the brain has been observed. The E/I balance is crucial for the normal functioning of cortical networks. Even subtle E/I imbalance is reflected in changes in spike rate and timing, measured by signal-to-noise-ratio (SNR), and coding reliability and sparseness. Indeed, all these properties are decreased in RTT model mice. This E/I imbalance is moved in the direction of hyper-excitation in RTT, yet RTT also seems to feature hyper-connection of inhibitory neurons and reduced excitation. Reduced efficacy of inhibitory GABAergic transmission following Mecp2 mutation is a likely mechanism for the above observations in RTT. At the synaptic level, Tang et al. (2015) found that K⁺/Cl⁻ cotransporter 2 (KCC2) was a critical downstream target of Mecp2, and KCC2 reduction following MECP2 mutation lowered intracellular Cl⁻. Low intracellular Cl⁻ concentration caused depolarizing postsynatpic responses of GABA _AR, and weakened inhibitory transmission. However, it is not clear whether this synaptic mechanism can cause the observed E/I balance shift in cortical circuits in vivo. To specifically rescue the abnormal intracellular Cl⁻ levels and observe its effect on cortical network behavior, we used bumetanide, a Na⁺K⁺Cl⁻ cotransporter 1 (NKCC1) antagonist that raises intracellular Cl⁻ concentration, and measured the network responses of V1 superficial layer neurons to a range of visual stimuli. In particular, we evaluated SNR, sparseness and reliability, in response to natural movies. After daily injections of bumetanide for a week, RTT model mice showed increased SNR and sparseness compared to sham-injected RTT model control, showing clear rescue of both properties. These results indicate that the altered intracellular Cl⁻ level, which leads to depolarizing GABAergic transmission, underlies a major portion of deficits observed in cortical network responses in RTT.

Dissecting Functional Organization of Mouse Visual Cortex
M. Hu, R. V. Rikhye, M. J. Goard, M. Sur
Society for Neuroscience, 2016

Although it is known that mice have roughly ten distinct visual areas, the specific function of each area and interactions between areas remains poorly understood. In this study, we used both wide-field and single-cell calcium imaging from awake, head-fixed mice, which transgenically expressed GCaMP6f, to functionally segment the entire visual cortex. First, to identify the gross organization of the visual areas, we performed retinotopic mapping using a custom-built wide-field epi-fluorescence microscope. Next, we characterized the responses of neurons within each segmented area to drifting gratings with various directions and spatial temporal frequencies. Within V1, we found a gradient of bias of averaged peak orientation (from vertical to horizontal orientations) along binocular to monocular axis. We also found systematic variations in averaged peak spatial and temporal frequencies along the same axis. The averaged peak temporal frequency in area LM exhibit a similar distribution (a mirror of V1) in visual space, indicating its role as the second visual area along the feedforward visual pathway. These results suggest that the different visual areas are sensitive to different visual features. To further test this hypothesis, we perturbed either the phase or the amplitude content of natural movies. The phase spectrum of natural movies contains information about salient image features, such as edges; whereas the amplitude spectrum contains low-level information, such as luminance. Interestingly, we found that medial regions (such as medial posterior part of V1 and PM) was more selective to phase, while lateral regions (such as lateral anterior part of V1 and AL) was more selective to the amplitude spectrum. Together, our results reveal that mouse visual cortex has organized representation of different visual features, which is globally extended through an entire set of areas. This distribution may already present in V1 and further elaborated in different higher visual areas.

Interaction Between Parasympathetic and Sympathetic Pathways on Prediction of Noradrenergic Activity by Pupil Size
V. Breton-Provencher, M. Sur
Society for Neuroscience, 2016

Pupil diameter has been used as a predictor of brain arousal. However, little is known about the neuromodulators responsible for pupil-mediated brain state, and to what extent pupil size can predict these neuromodulatory tones. Here we recorded the activity of noradrenergic (NE) neurons by using both functional imaging of NE axons in the cortex, and by single unit recordings in the locus coeruleus (LC) of awake mice. We show that pupil dilation predicts an increase of correlated NE activity both at a single cell and population level. The increase in LC-NE firing rate is linearly correlated with the amplitude of dilation events. This coherence between NE and pupil signals peaks in the low frequency range (10-2 to 10-1 Hz). Direct activation of LC-NE neurons by optogenetics further demonstrates the causal relationship between NE and pupil dilation. We are currently investigating the interaction between the pathways governing light mediated pupil constriction and internal state driven dilation. Altogether, our results show that pupil diameter can be used as a tool to track noradrenergic tone in the brain.

Asymmetry in Vocal Communication in Marmosets – Influence of Social Context and Gender Differences
J. Sharma, R. Landman, J. HYMAN, L. Brattain, K. Johnson, T. Quatieri, K. Srinivasan, A. Wisler, G. Feng, M. Sur, R. Desimone
Society for Neuroscience, 2016

Common marmosets are highly gregarious animals and use a rich vocal repertoire while communicating with conspecifics during social interactions. Recent studies indicate that they take turns in uttering calls (anti­phonal calling). There is also evidence that these exchanges are modulated by social context and specific calls are used to locate a group member when out of visual contact, when under threat or to show anxiety. To study these vocal/social interactions in a naturalistic home-­cage environment, we developed a wireless lightweight flexible neck collar, equipped with a micro­electro­mechanical system (MEMS) acoustic microphone, a non­-acoustic contact microphone for detecting caller vocal ­fold vibration, and a Bluetooth module for wireless data transmission. The initial testing was done on marmoset dyads and spectral analysis of their calls was performed to identify individual caller. Approximately 80% of calls could be attributed to an individual based on relative sound pressure alone. The remaining 20% were attributed with addition of data from the contact microphone. Cross­-correlation between audio channel and vibrational signal allows identification of most likely caller. We examined vocal interactions between dyads within the home-­cage environment. We find that antiphonal calling occurs not only for ‘phee’­calls, but also for other calls such as ‘trill’, and even while in visual contact. We selectively removed one member of a dyadic pair (male or a female) from the home cage for short periods, while within or without visual contact. Analysis shows that there is directionality in these interactions ­ when the female is out of the home cage but within view, there is an increase in temporal coherence, where one animal calls at a lag of about 0.5 sec after his/her partner. An asymmetry between dyads is also found when one animal is taken out of the room, but within audible range. When the female is taken out, dyads ‘phee’­call back and forth, but when the male is taken out, there is no ‘phee’­calling. The first ‘phee’­call is typically uttered by the animal that is outside. Directionality in vocal interactions may be thus associated with the sex of the animal, social context, dominance and relatedness. Further research is underway on multiple dyads to confirm and to explore neural underpinnings and behavioral consequences of this asymmetry.

2015

Distinct Roles of Mouse Visual and Parietal Cortex During Perceptual Decisions
G. N. Pho, M. J. Goard, B. Crawford, M. Sur
Society for Neuroscience, 2015

The posterior parietal cortex (PPC) has been implicated in perceptual decisions, but its specific role at the interface between sensation and action remains unresolved. Here, we provide evidence that mouse PPC, in functional analogy to primates, is neither a pure sensory area, nor directly involved in control of motor output, but rather is important for the mapping of sensory inputs to motor commands. Mice were trained on a visual discrimination task with distinct stimulus and motor epochs. We first tested the necessity of both PPC and the primary visual cortex (V1) during the different task epochs using VGAT-ChR2 transgenic mice, which express ChR2 in inhibitory neurons. Optogenetic inactivation revealed that both V1 and PPC were necessary during the stimulus period, but not for execution of the motor response. We then used two-photon calcium imaging to measure population activity in V1 and PPC, both during engagement in the task and during passive viewing of the same stimuli. Whereas V1 responses were driven by visual stimuli alone and only mildly modulated by task engagement, PPC responses were strongly gated by engagement and signaled the impending response. PPC responses exhibited both signatures of classical decision neurons: they reflected both the animal’s choice on error trials, as well as the degree of sensory evidence, which was manipulated using stimuli of varying contrasts. Lastly, to test whether PPC primarily encoded information about the stimulus or the choice, we re-trained mice with a reversed stimulus-reward contingency, and imaged the same neurons before and after the switch. We found that stimulus selectivity in PPC, but not V1, was dramatically reversed after retraining on the new contingency. Our results are consistent with a role of the mouse posterior parietal cortex in transforming sensory information to motor commands during perceptual decisions.

Human and Mouse Models of Rett Syndrome Exhibit Altered Prenatal Cortical Development Due to Alterations in Neurogenesis
P. Ip, N. Mellios, D. Feldman, S. D. Sheridan, S. Kwok, B. Rosen, B. Crawford, Y. Li, R. Jaenisch, S. J. Haggarty, M. Sur
Society for Neuroscience, 2015

Rett Syndrome (RTT) is a neurodevelopmental disorder that, in the vast majority of cases, arises from mutations in the X-linked gene MECP2. MeCP2 is an epigenetic modulator of gene expression that has recently been shown to interact with miRNA machinery. In addition, MeCP2 itself has been implicated in several neurodevelopmental disorders. Multiple lines of evidence point to the importance of miRNA-mediated pathways downstream of MeCP2 in different stages of brain development and plasticity. We hypothesized that the pleiotropic effects of MeCP2 in prenatal development are mediated via a set of early regulated miRNAs. Towards that end, we used induced pluripotent stem cell (iPSC) RTT lines generated from patients, virally-mediated knockdown of MeCP2 in human embryonic stem cells (ESCs), TALEN-derived isogenic ESC RTT lines, and an Mecp2 mutant mouse model as complementary approaches to identify novel MeCP2-regulated miRNAs and examine their respective influence on neurogenesis and neuronal differentiation.. Via BrdU pulse labelling, we found that the proliferation rate of patient-derived and MeCP2-deficient neuronal progenitor cells was significantly altered relative to control cells; this was accompanied by reductions in the expression of early neuronal markers and immature dendritic morphology. Our findings to date implicate aberrant regulation of prenatal neurogenesis as a result of MeCP2 deficiency. Taken together, our data support a novel miRNA-mediated pathway downstream of MeCP2 capable of influencing neurogenesis via interactions with central molecular hubs linked to autism spectrum disorders. Ongoing experiments are focused on elucidating the mechanisms of disease-related impairments in neurogenesis in both mouse and human organoid models of RTT, and translating these findings to scalable assays for novel therapeutic discovery.

Acetylcholine Drives Cortical Microcircuit and Modulates Temporal Dynamics in V1
H. Sugihara, N. Chen, M. Sur
Society for Neuroscience, 2015

Acetylcholine (ACh) modulates cortical functions including information processing and plasticity. To understand the physiological basis of these functions, it is critical to identify the cortical circuit elements involved. We have previously shown that cholinergic activation of astrocytes and their facilitatory influences on pyramidal neurons (PYR) are crucial to induce plasticity (Chen N, Sugihara H et al., PNAS 2012). In this work, we aim to dissect the neural circuit involved in cholinergic modulation of sensory processing. Specifically, we focus on the temporal dynamics of cortical activity: decorrelation of neuronal responses and desynchronization of local field potential (LFP) using L2/3 mouse primary visual cortex (V1) as a model. Recent studies suggest that inhibitory neurons are important for mediating temporal changes in neural activity. Candidate neurons include regular-spiking inhibitory neurons: somatostatin-expressing (SOM), vasoactive intestinal peptide-expressing (VIP) and layer 1 (L1) neurons. We recorded the cholinergic responses of these inhibitory neuronal subtypes in slice preparations. ACh induced concentration-specific responses in these neurons: SOM neurons were activated by a range of ACh concentrations while VIP/L1 neurons were activated only at high concentration. We further show that this is likely due to the active shaping of inhibitory neuronal responses through defined inhibitory connections between them: SOM neurons inhibit VIP/L1 neurons and this counters the ACh-induced facilitatory responses in the VIP/L1 neurons. In addition, we show that ACh-activated SOM (but not VIP/L1) induced inhibitory currents in parvalbumin-expressing (PV) and PYR neurons. This suggests the presence of an ACh-activated neural circuit comprising direct SOM-PYR and indirect SOM-PV-PYR connections. We next tested the causal relationship between this SOM-driven circuit and decorrelation/desynchronization through hyperpolarizing Arch-expressing SOM neurons in vivo. Indeed, hyperpolarization of SOM neurons blocked the cholinergic-mediated desynchronization/decorrelation. Hyperpolarization of VIP neurons did not affect the LFP desynchronization. Finally, we stimulated SOM neurons directly by expressing ChR2 in these neurons. Photostimulation of SOM neurons, in the absence of cholinergic stimulation, induced LFP desynchronization. This suggests that direct activation of SOM-driven circuit is sufficient to change temporal dynamics of V1. Collectively, these findings reveal the powerful role of SOM neurons in dynamically shaping the temporal pattern of cholinergic-mediated responses.

Robust and Reliable Ca2+ Response in Microdomains of Astrocytes
R. Garcia, R. Rikhye, M. Sur
Society for Neuroscience, 2015

Astrocytic intracellular Ca2+ signaling has come to light as a prominent feature of neuronal-glial interactions. The majority of astrocyte Ca2+ signaling studies are performed in either culture or in situ brain slices, both of which rely on electrical stimulation or pharmacological methods to examine the spatial and temporal coding of astrocyte Ca2+ signals. We have investigated visually evoked Ca2+ responses in astrocytes of the visual cortex of awake, head-fixed mice using two-photon microscopy. Initially, our studies involved the use of viral-mediated delivery of genetically encoded calcium indicators. However, in order maintain an intact and unperturbed cortex, we have chosen to use a recently developed line of conditional transgenic animals that express GCaMP5G in astrocytes throughout the mouse brain. We have found that Ca2+ transients in distal processes of cortical astrocytes are more frequent than those observed in anesthetized preparations, exhibiting a variable relationship to somatic responses. Furthermore, we found discrete structural regions of distal processes of single astrocytes that responded to sinusoidal drifting gratings and were tuned to specific orientations. Natural movies (NM) are known to evoke sparse, but reliable, responses from V1 pyramidal neurons. Surprisingly, we found discrete processes of astrocytes also respond reliably to natural movies. Responses to sinusoidal gratings were also less reliable than to natural movies. We hypothesize that these reliable astrocytic microdomain Ca2+ transients are due to the synchronous activation of neighboring ensembles of synapses. Together our results suggest that astrocytes could play an important role in modulating information processing in V1, potentially by modulating response reliability at pyramidal cell synapses.

Two Photon Imaging With Genetically-Encoded Calcium Indicators in New World Primates
J. Sharma, R. Landman, F. Yoshida, H. Sugihara, M. Sur
Society for Neuroscience, 2015

Two photon imaging using calcium sensors has provided important insights into neural circuit mechanisms with unprecedented detail, particularly in mice. With the advent of highly sensitive and targettable genetically encoded-calcium indicators (such as variants of GCaMP) riding on viral backbones of various vectors and aided by cell -specific promoters, the same population of neurons can be repeatedly imaged over several weeks or months to investigate neural circuit function (and dysfunction), and study neural mechanisms and plasticity underlying perception, cognition, learning and memory. However, achieving a similar level of sophistication in imaging higher mammals, particularly primates with more complex and dense cerebral architecture, poses new challenges. From the published literature, it is already clear that a new set of tools needs to be developed for this purpose. These include hardware for animal stabilization, movement correction of the imaging data, the development of chronic optical windows for accessing population of neurons expressing florescent proteins over several weeks, and the optimization of viral vectors and promoters while keeping brain tissue physiologically viable and in good health. Here we present our experience in developing these tools for 2 photon calcium imaging with GCaMP6 variants for chronic imaging in anesthetized new world monkeys, from several striate and extrastriate areas of the visual cortex. We modified a 2 photon imaging system (Sutter Instruments) to couple with a custom-designed movable objective that affords maximal rotational degrees of freedom along X and Y axes. We also developed a flexible imaging platform that provides precise alignment with the imaging plane of the objective, and can be tailored to suit primates of various sizes. For head stabilization we designed a dual, low-profile head post system that provides excellent rigidity while allowing flexibility to orient the head in any position to target multiple areas on the cortical surface. To minimize body movement due to respiration, we designed a simple trampoline suspension. We have also developed and tested several versions of chronic imaging windows, including sealed, removable and replaceable optical windows that are easy to maintain, while minimizing risk of infection and with flexibility to re-inject viral vectors or remove interfering pial-tissue growth. Large field ex-vivo confocal imaging was used to confirm GCaMP expression in several layers of the cortex with hSynapsin and CaMKII promoters. Two photon imaging in new world primates provides exciting possibilities as the next generation of transgenic primates come online.

Circuit Mechanisms Underlying Visual Responses of the Anterior Cingulate Cortex
R. Huda, G. Pho, I. R. Wickersham, M. Sur
Society for Neuroscience, 2015

Neural dynamics in sensory cortices are shaped by bottom-up inputs relaying the physical features of sensory stimuli and by top-down projections that modulate their encoding. The anterior cingulate division of the prefrontal cortex is known to provide top-down input to the visual cortex. Here, we use multiple approaches to delineate the functional role of visual inputs to the anterior cingulate cortex (ACC) and of the feedback from ACC to V1. Using rabies virus-mediated anatomical tracing to identify sources of inputs to the ACC, we found that V1 as well as other cortical and subcortical brain regions project to the ACC. Using rabies viruses that express the genetically encodable calcium indicator GCaMP6f and two-photon microscopy, we characterized the functional properties of ACC-projecting visual cortex neurons in passively viewing, awake head-fixed mice. We found that many of these neurons are tuned to the orientation and direction of drifting gratings. Next, we expressed GCaMP6s in the ACC and imaged the calcium activity of ACC axons found in layer 1 of V1. A subset of ACC axons were visually driven and displayed sharply tuned responses to the orientation and direction of drifting gratings. To assess the contribution of V1 to this property, we used a chemogenetic approach. We expressed the inhibitory hM4Di DREADD (designer receptors exclusively activated by designer drugs) in V1, GCaMP6s in the ACC, and monitored the calcium responses of ACC axons to oriented drifting gratings before and after systemic application of the DREADD agonist clozapine-N-oxide (CNO). While CNO application in control animals had no effect on the visual responses of ACC axons, it reduced responses in DREADD expressing animals. Together, these findings show that a projection from the visual cortex contributes to the visual responsiveness of the ACC. Since the ACC has been proposed to play a crucial role in cognition, and in particular reward processing, we propose that the ACC processes visual information in the context of its behavioral significance and relays a saliency signal back to the visual cortex to modulate the encoding of relevant visual stimuli.

Dissecting the Inhibitory Mechanisms of Reliable Coding in Mouse Primary Visual Cortex
R. V. Rikhye, M. Sur
Society for Neuroscience, 2015

Neurons in the primary visual cortex (V1) respond to full-field natural scenes with spike trains that are highly reliable between trials. While it has been argued that local inhibitory interneurons are responsible for modulating reliable coding, no study has yet systematically detailed the role of different interneuron subtypes. Our goal was to show how Parvalbumin (PV), Somatostatin (SST) and Vasoactive Intestinal Peptide (VIP) expressing interneurons modulate reliable coding in mouse V1. Specifically, we aimed to: (1) show how subnetworks of these interneurons process natural scenes and (2) determine how they contribute to reliable coding. To address these questions, we performed in vivo two-photon calcium imaging in awake, head-fixed mice by conditionally expressing GCaMP6f in PV, SST or VIP neurons. This allowed us to minimize the effect of contamination from nearby excitatory neurons and permitted us to study population coding within these interneuron subnetworks. SST neurons also preferred lower spatial frequencies than PV neurons, consistent with their role in integrating information from a larger visual area. Not surprisingly, VIP neurons responded poorly to gratings. PV neurons responded strongly, but unreliably, to full-field natural scenes. In contrast, SST neurons were more selective and were highly reliable between trials. SST cell reliability was comparable to excitatory neurons. This suggests that SST neurons are selectively driven by specific features in natural scenes and, consequently, provide reliable dendritic inhibition on their target cells. We also found that VIP neurons responded more strongly to natural scenes than gratings, suggesting that these interneurons are driven more by “salient” stimuli. Next, we investigated how these interneurons modulated pyramidal cell reliability. To do so, we conditionally expressed ChR2 in both PV or SST neurons and GCaMP6f in pyramidal neurons. We reasoned that reliability arose due precisely timed excitatory (E) and inhibitory (I) synaptic currents. Thus, we used a stimulation protocol to decorrelate these E- and I-currents in pyramidal cells. Specifically, we pulsed a blue LED for 100ms at random times during a natural movie. This allowed us to activate PV/SST neurons during periods when pyramidal cells were most reliable. We discovered that activating SST neurons during epochs of reliability increased reliability. In contrast, stimulating PV neurons reduced reliability. Taken together, our work demonstrates that SST neurons play an important role in shaping the reliability of pyramidal cell responses to natural scenes in mouse visual cortex.

Robustly Dysregulated miRNAs Downstream of MeCP2 Control Human Prenatal Brain Development Through Differential Effects on Autism-Related Signaling Pathways
N. Mellios, D. Feldman, S. D. Shericdan, P. K. Ip, S. Kwok, B. Rosen, B. Crawford, Y. LI, R. Jaenisch, S. J. Haggarty, M. Sur
Society for Neuroscience, 2015

Rett Syndrome (RTT) is a neurodevelopmental disorder primarily caused by mutations in methyl-CpG-binding protein 2 (MECP2), a potent epigenetic regulator whose role in prenatal brain development is poorly understood. Given the known effects of MeCP2 on miRNA biogenesis, we hypothesized that neurogenesis may be impacted in RTT via MeCP2-regulated miRNAs that are enriched in early brain development; and hence modulate critical molecular components of neuronal progenitor proliferation and differentiation. Focusing on the most dysregulated miRNAs we found two prenatal brain-enriched miRNAs – miR-199 and miR-214 – to be robustly increased in human patient-derived culture, cerebral organoid, and mouse models of MeCP2 deficiency. Increases in miR-199 and miR-214 in MeCP2 mutant or deficient neuronal progenitors were a consequence of altered miRNA biogenesis and were associated with reduced expression of their targets PAK4 and PTEN, which in turn resulted in differential changes in Erk and Akt phosphorylation. Inhibiting miR-199 or miR-214 expression in induced pluripotent stem cell-derived neuronal progenitors deficient in MeCP2 restored Akt and Erk activation, respectively, and ameliorated the observed alterations in neuronal differentiation. Collectively, our data suggest that MeCP2-mediated dysregulation of miR-199 and miR-214 expression influences early neurogenesis through the differential regulation of molecular pathways with known links to autism spectrum disorders.

                                                                                                                           Conference Abstracts 2010 – 2014

                                                                                                                          Conference Abstracts 2005 – 2009

                                                                                                                           Conference Abstracts 2000 – 2004