Attention-Based CNN-BiLSTM for Sleep State Classification of Spatiotemporal Wide-Field Calcium Imaging Data.

BACKGROUND: Wide-field calcium imaging (WFCI) with genetically encoded calcium indicators allows for spatiotemporal recordings of neuronal activity in mice. When applied to the study of sleep, WFCI data are manually scored into the sleep states of wakefulness, non-REM (NREM) and REM by use of adjunct EEG and EMG recordings. However, this process is time-consuming, invasive and often suffers from low inter- and intra-rater reliability. Therefore, an automated sleep state classification method that operates on spatiotemporal WFCI data is desired. NEW METHOD: A hybrid network architecture consisting of a convolutional neural network (CNN) to extract spatial features of image frames and a bidirectional long short-term memory network (BiLSTM) with attention mechanism to identify temporal dependencies among different time points was proposed to classify WFCI data into states of wakefulness, NREM and REM sleep. RESULTS: Sleep states were classified with an accuracy of 84% and Cohen's kappa of 0.64. Gradient-weighted class activation maps revealed that the frontal region of the cortex carries more importance when classifying WFCI data into NREM sleep while posterior area contributes most to the identification of wakefulness. The attention scores indicated that the proposed network focuses on short- and long-range temporal dependency in a state-specific manner. COMPARISON WITH EXISTING METHOD: On a 3-hour WFCI recording, the CNN-BiLSTM achieved a kappa of 0.67, comparable to a kappa of 0.65 corresponding to the human EEG/EMG-based scoring. CONCLUSIONS: The CNN-BiLSTM effectively classifies sleep states from spatiotemporal WFCI data and will enable broader application of WFCI in sleep.

Myeloid cell activation during Zika virus encephalitis predicts recovery of functional cortical connectivity.

Neurologic complications of Zika virus (ZIKV) infection across the lifespan have been described during outbreaks in Southeast Asia, South America, and Central America since 2016. In the adult CNS ZIKV tropism for neurons is tightly linked to its effects, with neuronal loss within the hippocampus during acute infection and protracted synapse loss during recovery, which is associated with cognitive deficits. The effects of ZIKV on cortical networks have not been evaluated. Although animal behavior assays have been used previously to model cognitive impairment, in vivo brain imaging can provide orthogonal information regarding the health of brain networks in real time, providing a tool to translate findings in animal models to humans. In this study, we use widefield optical imaging to measure cortical functional connectivity (FC) in mice during acute infection with, and recovery from, intracranial infection with a mouse-adapted strain of ZIKV. Acute ZIKV infection leads to high levels of myeloid cell activation, with loss of neurons and presynaptic termini in the cerebral cortex and associated loss of FC primarily within the somatosensory cortex. During recovery, neuron numbers, synapses and FC recover to levels near those of healthy mice. However, hippocampal injury and impaired spatial cognition persist. The magnitude of activated myeloid cells during acute infection predicted both recovery of synapses and the degree of FC recovery after recovery from ZIKV infection. These findings suggest that a robust inflammatory response may contribute to the health of functional brain networks after recovery from infection.

Open-source statistical and data processing tools for wide-field optical imaging data in mice.

Significance: Wide-field optical imaging (WOI) can produce concurrent hemodynamic and cell-specific calcium recordings across the entire cerebral cortex in animal models. There have been multiple studies using WOI to image mouse models with various environmental or genetic manipulations to understand various diseases. Despite the utility of pursuing mouse WOI alongside human functional magnetic resonance imaging (fMRI), and the multitude of analysis toolboxes in the fMRI literature, there is not an available open-source, user-friendly data processing and statistical analysis toolbox for WOI data. Aim: To assemble a MATLAB toolbox for processing WOI data, as described and adapted to combine techniques from multiple WOI groups and fMRI. Approach: We outline our MATLAB toolbox on GitHub with multiple data analysis packages and translate a commonly used statistical approach from the fMRI literature to the WOI data. To illustrate the utility of our MATLAB toolbox, we demonstrate the ability of the processing and analysis framework to detect a well-established deficit in a mouse model of stroke and plot activation areas during an electrical paw stimulus experiment. Results: Our processing toolbox and statistical methods isolate a somatosensory-based deficit 3 days following photothrombotic stroke and cleanly localize sensory stimulus activations. Conclusions: The toolbox presented here details an open-source, user-friendly compilation of WOI processing tools with statistical methods to apply to any biological question investigated with WOI techniques.

Mecp2 deletion results in profound alterations of developmental and adult functional connectivity.

As a regressive neurodevelopmental disorder with a well-established genetic cause, Rett syndrome and its Mecp2 loss-of-function mouse model provide an excellent opportunity to define potentially translatable functional signatures of disease progression, as well as offer insight into the role of Mecp2 in functional circuit development. Thus, we applied widefield optical fluorescence imaging to assess mesoscale calcium functional connectivity (FC) in the Mecp2 cortex both at postnatal day (P)35 in development and during the disease-related decline. We found that FC between numerous cortical regions was disrupted in Mecp2 mutant males both in juvenile development and early adulthood. Female Mecp2 mice displayed an increase in homotopic contralateral FC in the motor cortex at P35 but not in adulthood, where instead more posterior parietal regions were implicated. An increase in the amplitude of connection strength, both with more positive correlations and more negative anticorrelations, was observed across the male cortex in numerous functional regions. Widespread rescue of MeCP2 protein in GABAergic neurons rescued none of these functional deficits, nor, surprisingly, the expected male lifespan. Altogether, the female results identify early signs of disease progression, while the results in males indicate MeCP2 protein is required for typical FC in the brain.

SIRT1 mediates hypoxic postconditioning- and resveratrol-induced protection against functional connectivity deficits after subarachnoid hemorrhage.

Functional connectivity (FC) is a sensitive metric that provides a readout of whole cortex coordinate neural activity in a mouse model. We examine the impact of experimental SAH modeled through endovascular perforation, and the effectiveness of subsequent treatment on FC, through three key questions: 1) Does the endovascular perforation model of SAH induce deficits in FC; 2) Does exposure to hypoxic conditioning provide protection against these FC deficits and, if so, is this neurovascular protection SIRT1-mediated; and 3) does treatment with the SIRT1 activator resveratrol alone provide protection against these FC deficits? Cranial windows were adhered on skull-intact mice that were then subjected to either sham or SAH surgery and either left untreated or treated with hypoxic post-conditioning (with or without EX527) or resveratrol for 3 days. Mice were imaged 3 days post-SAH/sham surgery, temporally aligned with the onset of major SAH sequela in mice. Here we show that the endovascular perforation model of SAH induces global and network-specific deficits in FC by day 3, corresponding with the time frame of DCI in mice. Hypoxic conditioning provides SIRT1-mediated protection against these network-specific FC deficits post-SAH, as does treatment with resveratrol. Conditioning-based strategies provide multifaceted neurovascular protection in experimental SAH.

Functional Connectivity of the Developing Mouse Cortex.

Cross-sectional studies have established a variety of structural, synaptic, and cell physiological changes corresponding to critical periods in cortical development. However, the emergence of functional connectivity (FC) in development has not been fully characterized, and hemodynamic-based measures are vulnerable to any neurovascular coupling changes occurring in parallel. We therefore used optical fluorescence imaging to trace longitudinal calcium FC in the awake, resting-state mouse cortex at 5 developmental timepoints beginning at postnatal day 15 (P15) and ending in early adulthood at P60. Calcium FC displayed coherent functional maps as early as P15, and FC significantly varied in connections between many regions across development, with the developmental trajectory's shape specific to the functional region. Evaluating 325 seed-seed connections, we found that there was a significant increase in FC between P15 and P22 over the majority of the cortex as well as bilateral connectivity and node degree differences in frontal, motor, and retrosplenial cortices after P22. A rebalancing of inter- and intrahemispheric FC and local-distal FC dominance was also observed during development. This longitudinal developmental calcium FC study therefore provides a resource dataset to the field and identifies periods of dynamic change which cross-sectional studies may target for examination of disease states.

A Multivariate Functional Connectivity Approach to Mapping Brain Networks and Imputing Neural Activity in Mice.

Temporal correlation analysis of spontaneous brain activity (e.g., Pearson "functional connectivity," FC) has provided insights into the functional organization of the human brain. However, bivariate analysis techniques such as this are often susceptible to confounding physiological processes (e.g., sleep, Mayer-waves, breathing, motion), which makes it difficult to accurately map connectivity in health and disease as these physiological processes affect FC. In contrast, a multivariate approach to imputing individual neural networks from spontaneous neuroimaging data could be influential to our conceptual understanding of FC and provide performance advantages. Therefore, we analyzed neural calcium imaging data from Thy1-GCaMP6f mice while either awake, asleep, anesthetized, during low and high bouts of motion, or before and after photothrombotic stroke. A linear support vector regression approach was used to determine the optimal weights for integrating the signals from the remaining pixels to accurately predict neural activity in a region of interest (ROI). The resultant weight maps for each ROI were interpreted as multivariate functional connectivity (MFC), resembled anatomical connectivity, and demonstrated a sparser set of strong focused positive connections than traditional FC. While global variations in data have large effects on standard correlation FC analysis, the MFC mapping methods were mostly impervious. Lastly, MFC analysis provided a more powerful connectivity deficit detection following stroke compared to traditional FC.

Automated sleep state classification of wide-field calcium imaging data via multiplex visibility graphs and deep learning.

BACKGROUND: Wide-field calcium imaging (WFCI) allows for monitoring of cortex-wide neural dynamics in mice. When applied to the study of sleep, WFCI data are manually scored into the sleep states of wakefulness, non-REM (NREM) and REM by use of adjunct EEG and EMG recordings. However, this process is time-consuming and often suffers from low inter- and intra-rater reliability and invasiveness. Therefore, an automated sleep state classification method that operates on WFCI data alone is needed. NEW METHOD: A hybrid, two-step method is proposed. In the first step, spatial-temporal WFCI data is mapped to multiplex visibility graphs (MVGs). Subsequently, a two-dimensional convolutional neural network (2D CNN) is employed on the MVGs to be classified as wakefulness, NREM and REM. RESULTS: Sleep states were classified with an accuracy of 84% and Cohen's κ of 0.67. The method was also effectively applied on a binary classification of wakefulness/sleep (accuracy=0.82, κ = 0.62) and a four-class wakefulness/sleep/anesthesia/movement classification (accuracy=0.74, κ = 0.66). Gradient-weighted class activation maps revealed that the CNN focused on short- and long-term temporal connections of MVGs in a sleep state-specific manner. Sleep state classification performance when using individual brain regions was highest for the posterior area of the cortex and when cortex-wide activity was considered. COMPARISON WITH EXISTING METHOD: On a 3-hour WFCI recording, the MVG-CNN achieved a κ of 0.65, comparable to a κ of 0.60 corresponding to the human EEG/EMG-based scoring. CONCLUSIONS: The hybrid MVG-CNN method accurately classifies sleep states from WFCI data and will enable future sleep-focused studies with WFCI.

Astrocyte deletion of α2-Na/K ATPase triggers episodic motor paralysis in mice via a metabolic pathway.

Familial hemiplegic migraine is an episodic neurological disorder characterized by transient sensory and motor symptoms and signs. Mutations of the ion pump α2-Na/K ATPase cause familial hemiplegic migraine, but the mechanisms by which α2-Na/K ATPase mutations lead to the migraine phenotype remain incompletely understood. Here, we show that mice in which α2-Na/K ATPase is conditionally deleted in astrocytes display episodic paralysis. Functional neuroimaging reveals that conditional α2-Na/K ATPase knockout triggers spontaneous cortical spreading depression events that are associated with EEG low voltage activity events, which correlate with transient motor impairment in these mice. Transcriptomic and metabolomic analyses show that α2-Na/K ATPase loss alters metabolic gene expression with consequent serine and glycine elevation in the brain. A serine- and glycine-free diet rescues the transient motor impairment in conditional α2-Na/K ATPase knockout mice. Together, our findings define a metabolic mechanism regulated by astrocytic α2-Na/K ATPase that triggers episodic motor paralysis in mice.

Maternal Fluoxetine Exposure Alters Cortical Hemodynamic and Calcium Response of Offspring to Somatosensory Stimuli.

Epidemiological studies have found an increased incidence of neurodevelopmental disorders in populations prenatally exposed to selective serotonin reuptake inhibitors (SSRIs). Optical imaging provides a minimally invasive way to determine if perinatal SSRI exposure has long-term effects on cortical function. Herein we probed the functional neuroimaging effects of perinatal SSRI exposure in a fluoxetine (FLX)-exposed mouse model. While resting-state homotopic contralateral functional connectivity was unperturbed, the evoked cortical response to forepaw stimulation was altered in FLX mice. The stimulated cortex showed decreased activity for FLX versus controls, by both hemodynamic responses [oxyhemoglobin (HbO2)] and neuronal calcium responses (Thy1-GCaMP6f fluorescence). Significant alterations in both cortical HbO2 and calcium response amplitude were seen in the cortex ipsilateral to the stimulated paw in FLX as compared to controls. The cortical regions of largest difference in activation between FLX and controls also were consistent between HbO2 and calcium contrasts at the end of stimulation. Taken together, these results suggest a global loss of response signal amplitude in FLX versus controls. These findings indicate that perinatal SSRI exposure has long-term consequences on somatosensory cortical responses.

Separability of calcium slow waves and functional connectivity during wake, sleep, and anesthesia.

Modulation of brain state, e.g., by anesthesia, alters the correlation structure of spontaneous activity, especially in the delta band. This effect has largely been attributed to the ∼ 1 Hz slow oscillation that is characteristic of anesthesia and nonrapid eye movement (NREM) sleep. However, the effect of the slow oscillation on correlation structures and the spectral content of spontaneous activity across brain states (including NREM) has not been comprehensively examined. Further, discrepancies between activity dynamics observed with hemoglobin versus calcium (GCaMP6) imaging have not been reconciled. Lastly, whether the slow oscillation replaces functional connectivity (FC) patterns typical of the alert state, or superimposes on them, remains unclear. Here, we use wide-field calcium imaging to study spontaneous cortical activity in awake, anesthetized, and naturally sleeping mice. We find modest brain state-dependent changes in infraslow correlations but larger changes in GCaMP6 delta correlations. Principal component analysis of GCaMP6 sleep/anesthesia data in the delta band revealed that the slow oscillation is largely confined to the first three components. Removal of these components revealed a correlation structure strikingly similar to that observed during wake. These results indicate that, during NREM sleep/anesthesia, the slow oscillation superimposes onto a canonical FC architecture.

ESM-CT: a precise method for localization of DBS electrodes in CT images.

BACKGROUND: Deep brain stimulation (DBS) of the subthalamic nucleus produces variable effects in Parkinson disease. Variation may result from different electrode positions relative to target. Thus, precise electrode localization is crucial when investigating DBS effects. NEW METHOD: We developed a semi-automated method, Electrode Shaft Modeling in CT images (ESM-CT) to reconstruct DBS lead trajectories and contact locations. We evaluated methodological sensitivity to operator-dependent steps, robustness to image resampling, and test-retest replicability. ESM-CT was applied in 56 patients to study electrode position change (and relation to time between scans, postoperative subdural air volume, and head tilt during acquisition) between images acquired immediately post-implantation (DBS-CT) and months later (DEL-CT). RESULTS: Electrode tip localization was robust to image resampling and replicable to within ∼ 0.2 mm on test-retest comparisons. Systematic electrode displacement occurred rostral-ventral-lateral between DBS-CT and DEL-CT scans. Head angle was a major explanatory factor (p < 0.001,Pearson's r = 0.46, both sides) and volume of subdural air weakly predicted electrode displacement (p = 0.02,r = 0.29:p = 0.1,r = 0.25 for left:right). Modeled shaft curvature was slightly greater in DEL-CT. Magnitude of displacement and degree of curvature were independent of elapsed time between scans. COMPARISON WITH EXISTING METHODS: Comparison of ESM-CT against two existing methods revealed systematic differences in one coordinate (1 ± 0.3 mm,p < 0.001) for one method and in three coordinates for another method (x:0.1 ± 0.1 mm, y:0.4 ± 0.2 mm, z:0.4 ± 0.2 mm, p < 10-10). Within-method coordinate variability across participants is similar. CONCLUSION: We describe a robust and precise method for CT DBS contact localization. Application revealed that acquisition head angle significantly impacts electrode position. DBS localization schemes should account for head angle.

Functional connectivity structure of cortical calcium dynamics in anesthetized and awake mice.

The interplay between hemodynamic-based markers of cortical activity (e.g. fMRI and optical intrinsic signal imaging), which are an indirect and relatively slow report of neural activity, and underlying synaptic electrical and metabolic activity through neurovascular coupling is a topic of ongoing research and debate. As application of resting state functional connectivity measures is extended further into topics such as brain development, aging and disease, the importance of understanding the fundamental physiological basis for functional connectivity will grow. Here we extend functional connectivity analysis from hemodynamic- to calcium-based imaging. Transgenic mice (n = 7) expressing a fluorescent calcium indicator (GCaMP6) driven by the Thy1 promoter in glutamatergic neurons were imaged transcranially in both anesthetized (using ketamine/xylazine) and awake states. Sequential LED illumination (λ = 454, 523, 595, 640nm) enabled concurrent imaging of both GCaMP6 fluorescence emission (corrected for hemoglobin absorption) and hemodynamics. Functional connectivity network maps were constructed for infraslow (0.009-0.08Hz), intermediate (0.08-0.4Hz), and high (0.4-4.0Hz) frequency bands. At infraslow and intermediate frequencies, commonly used in BOLD fMRI and fcOIS studies of functional connectivity and implicated in neurovascular coupling mechanisms, GCaMP6 and HbO2 functional connectivity structures were in high agreement, both qualitatively and also quantitatively through a measure of spatial similarity. The spontaneous dynamics of both contrasts had the highest correlation when the GCaMP6 signal was delayed with a ~0.6-1.5s temporal offset. Within the higher-frequency delta band, sensitive to slow wave sleep oscillations in non-REM sleep and anesthesia, we evaluate the speed with which the connectivity analysis stabilized and found that the functional connectivity maps captured putative network structure within time window lengths as short as 30 seconds. Homotopic GCaMP6 functional connectivity maps at 0.4-4.0Hz in the anesthetized states show a striking correlated and anti-correlated structure along the anterior to posterior axis. This structure is potentially explained in part by observed propagation of delta-band activity from frontal somatomotor regions to visuoparietal areas. During awake imaging, this spatio-temporal quality is altered, and a more complex and detailed functional connectivity structure is observed. The combined calcium/hemoglobin imaging technique described here will enable the dissociation of changes in ionic and hemodynamic functional structure and neurovascular coupling and provide a framework for subsequent studies of neurological disease such as stroke.

Biophysical characteristics of cholera toxin and Escherichia coli heat-labile enterotoxin structure and chemistry lead to differential toxicity.

The biophysical chemistry of macromolecular complexes confer their functional characteristics. We investigate the mechanisms that make the AB5 holotoxin of Vibrio cholerae (CT) a significantly more pathogenic molecule than the enterotoxin of Escherichia coli (LT) with which it shares 88% similarity and whose structure is homologous with a backbone RMSD of 0.84 Å and imposes its deleterious effects though the same process to constitutively ADP-ribosylate adenylate cyclase. We present computational data that characterizes the impact of amino acid variations in the A2 tail, which helps to explain experimental data that demonstrate CT's higher toxicity. A hydrophobic patch on the B pentamer interface and its interactions with the A subdomain are partially disrupted by the substitution of an aspartic acid (LT) for glycine in CT. CT's holotoxin has less solvent accessible surface area (94 Å(2) vs 54 Å(2)) and higher contact area (280 Å(2) vs 241 Å(2)) with S228, which is a gatekeeper, partially controlling the diffusion of water into the pore. CT excludes water from the top of the central pore whereas LT allows much more water to interact. These biophysical properties of the toxins lead to their differential toxicity and resulting impact to human health.