Ensemble learning and ground-truth validation of synaptic connectivity inferred from spike trains.

Probing the architecture of neuronal circuits and the principles that underlie their functional organization remains an important challenge of modern neurosciences. This holds true, in particular, for the inference of neuronal connectivity from large-scale extracellular recordings. Despite the popularity of this approach and a number of elaborate methods to reconstruct networks, the degree to which synaptic connections can be reconstructed from spike-train recordings alone remains controversial. Here, we provide a framework to probe and compare connectivity inference algorithms, using a combination of synthetic ground-truth and in vitro data sets, where the connectivity labels were obtained from simultaneous high-density microelectrode array (HD-MEA) and patch-clamp recordings. We find that reconstruction performance critically depends on the regularity of the recorded spontaneous activity, i.e., their dynamical regime, the type of connectivity, and the amount of available spike-train data. We therefore introduce an ensemble artificial neural network (eANN) to improve connectivity inference. We train the eANN on the validated outputs of six established inference algorithms and show how it improves network reconstruction accuracy and robustness. Overall, the eANN demonstrated strong performance across different dynamical regimes, worked well on smaller datasets, and improved the detection of synaptic connectivity, especially inhibitory connections. Results indicated that the eANN also improved the topological characterization of neuronal networks. The presented methodology contributes to advancing the performance of inference algorithms and facilitates our understanding of how neuronal activity relates to synaptic connectivity.

DeePhys: A machine learning-assisted platform for electrophysiological phenotyping of human neuronal networks.

Reproducible functional assays to study in vitro neuronal networks represent an important cornerstone in the quest to develop physiologically relevant cellular models of human diseases. Here, we introduce DeePhys, a MATLAB-based analysis tool for data-driven functional phenotyping of in vitro neuronal cultures recorded by high-density microelectrode arrays. DeePhys is a modular workflow that offers a range of techniques to extract features from spike-sorted data, allowing for the examination of functional phenotypes both at the individual cell and network levels, as well as across development. In addition, DeePhys incorporates the capability to integrate novel features and to use machine-learning-assisted approaches, which facilitates a comprehensive evaluation of pharmacological interventions. To illustrate its practical application, we apply DeePhys to human induced pluripotent stem cell-derived dopaminergic neurons obtained from both patients and healthy individuals and showcase how DeePhys enables phenotypic screenings.

Functional imaging of brain organoids using high-density microelectrode arrays.

Abstract: Studies have provided evidence that human cerebral organoids (hCOs) recapitulate fundamental milestones of early brain development, but many important questions regarding their functionality and electrophysiological properties persist. High-density microelectrode arrays (HD-MEAs) represent an attractive analysis platform to perform functional studies of neuronal networks at the cellular and network scale. Here, we use HD-MEAs to derive large-scale electrophysiological recordings from sliced hCOs. We record the activity of hCO slices over several weeks and probe observed neuronal dynamics pharmacologically. Moreover, we present results on how the obtained recordings can be spike-sorted and subsequently studied across scales. For example, we show how to track single neurons across several days on the HD-MEA and how to infer axonal action potential velocities. We also infer putative functional connectivity from hCO recordings. The introduced methodology will contribute to a better understanding of developing neuronal networks in brain organoids and provide new means for their functional characterization. Impact statement: Human cerebral organoids (hCOs) represent an attractive in vitro model system to study key physiological mechanisms underlying early neuronal network formation in tissue with healthy or disease-related genetic backgrounds. Despite remarkable advances in the generation of brain organoids, knowledge on the functionality of their neuronal circuits is still scarce. Here, we used complementary metal-oxide-semiconductor (CMOS)-based high-density microelectrode arrays (HD-MEAs) to perform large-scale recordings from sliced hCOs over several weeks and quantified their activity across scales. Using single-cell and network metrics, we were able to probe aspects of hCO neurophysiology that are more difficult to obtain with other techniques, such as patch clamping (lower yield) and calcium imaging (lower temporal resolution). These metrics included, for example, extracellular action potential (AP) waveform features and axonal AP velocity at the cellular level, as well as functional connectivity at the network level. Analysis was enabled by the large sensing area and the high spatiotemporal resolution provided by HD-MEAs, which allowed recordings from hundreds of neurons and spike sorting of their activity. Our results demonstrate that HD-MEAs provide a multi-purpose platform for the functional characterization of hCOs, which will be key in improving our understanding of this model system and assessing its relevance for translational research. Supplementary Information: The online version contains supplementary material available at 10.1557/s43577-022-00282-w.

Predicting in vitro single-neuron firing rates upon pharmacological perturbation using Graph Neural Networks.

Modern Graph Neural Networks (GNNs) provide opportunities to study the determinants underlying the complex activity patterns of biological neuronal networks. In this study, we applied GNNs to a large-scale electrophysiological dataset of rodent primary neuronal networks obtained by means of high-density microelectrode arrays (HD-MEAs). HD-MEAs allow for long-term recording of extracellular spiking activity of individual neurons and networks and enable the extraction of physiologically relevant features at the single-neuron and population level. We employed established GNNs to generate a combined representation of single-neuron and connectivity features obtained from HD-MEA data, with the ultimate goal of predicting changes in single-neuron firing rate induced by a pharmacological perturbation. The aim of the main prediction task was to assess whether single-neuron and functional connectivity features, inferred under baseline conditions, were informative for predicting changes in neuronal activity in response to a perturbation with Bicuculline, a GABA A receptor antagonist. Our results suggest that the joint representation of node features and functional connectivity, extracted from a baseline recording, was informative for predicting firing rate changes of individual neurons after the perturbation. Specifically, our implementation of a GNN model with inductive learning capability (GraphSAGE) outperformed other prediction models that relied only on single-neuron features. We tested the generalizability of the results on two additional datasets of HD-MEA recordings-a second dataset with cultures perturbed with Bicuculline and a dataset perturbed with the GABA A receptor antagonist Gabazine. GraphSAGE models showed improved prediction accuracy over other prediction models. Our results demonstrate the added value of taking into account the functional connectivity between neurons and the potential of GNNs to study complex interactions between neurons.

Electrophysiological Phenotype Characterization of Human iPSC-Derived Neuronal Cell Lines by Means of High-Density Microelectrode Arrays.

Recent advances in the field of cellular reprogramming have opened a route to studying the fundamental mechanisms underlying common neurological disorders. High-density microelectrode-arrays (HD-MEAs) provide unprecedented means to study neuronal physiology at different scales, ranging from network through single-neuron to subcellular features. In this work, HD-MEAs are used in vitro to characterize and compare human induced-pluripotent-stem-cell-derived dopaminergic and motor neurons, including isogenic neuronal lines modeling Parkinson's disease and amyotrophic lateral sclerosis. Reproducible electrophysiological network, single-cell and subcellular metrics are used for phenotype characterization and drug testing. Metrics, such as burst shape and axonal velocity, enable the distinction of healthy and diseased neurons. The HD-MEA metrics can also be used to detect the effects of dosing the drug retigabine to human motor neurons. Finally, it is shown that the ability to detect drug effects and the observed culture-to-culture variability critically depend on the number of available recording electrodes.

Versatile live-cell activity analysis platform for characterization of neuronal dynamics at single-cell and network level.

Chronic imaging of neuronal networks in vitro has provided fundamental insights into mechanisms underlying neuronal function. Current labeling and optical imaging methods, however, cannot be used for continuous and long-term recordings of the dynamics and evolution of neuronal networks, as fluorescent indicators can cause phototoxicity. Here, we introduce a versatile platform for label-free, comprehensive and detailed electrophysiological live-cell imaging of various neurogenic cells and tissues over extended time scales. We report on a dual-mode high-density microelectrode array, which can simultaneously record in (i) full-frame mode with 19,584 recording sites and (ii) high-signal-to-noise mode with 246 channels. We set out to demonstrate the capabilities of this platform with recordings from primary and iPSC-derived neuronal cultures and tissue preparations over several weeks, providing detailed morpho-electrical phenotypic parameters at subcellular, cellular and network level. Moreover, we develop reliable analysis tools, which drastically increase the throughput to infer axonal morphology and conduction speed.

Micro-connectomics: probing the organization of neuronal networks at the cellular scale.

Defining the organizational principles of neuronal networks at the cellular scale, or micro-connectomics, is a key challenge of modern neuroscience. In this Review, we focus on graph theoretical parameters of micro-connectome topology, often informed by economical principles that conceptually originated with Ramón y Cajal's conservation laws. First, we summarize results from studies in intact small organisms and in samples from larger nervous systems. We then evaluate the evidence for an economical trade-off between biological cost and functional value in the organization of neuronal networks. Various results suggest that many aspects of neuronal network organization are indeed the outcome of competition between these two fundamental selection pressures.

Functional brain network changes associated with clinical and biochemical measures of the severity of hepatic encephalopathy.

Functional properties of the brain may be associated with changes in complex brain networks. However, little is known about how properties of large-scale functional brain networks may be altered stepwise in patients with disturbance of consciousness, e.g., an encephalopathy. We used resting-state fMRI data on patients suffering from various degrees of hepatic encephalopathy (HE) to explore how topological and spatial network properties of functional brain networks changed at different cognitive and consciousness states. Severity of HE was measured clinically and by neuropsychological tests. Fifty-eight non-alcoholic liver cirrhosis patients and 62 normal controls were studied. Patients were subdivided into liver cirrhosis with no outstanding HE (NoHE, n=23), minimal HE with cognitive impairment only detectable by neuropsychological tests (MHE, n=28), and clinically overt HE (OHE, n=7). From the earliest stage, the NoHE, functional brain networks were progressively more random, less clustered, and less modular. Since the intermediate stage (MHE), increased ammonia level was accompanied by concomitant exponential decay of mean connectivity strength, especially in the primary cortical areas and midline brain structures. Finally, at the OHE stage, there were radical reorganization of the topological centrality-i.e., the relative importance-of the hubs and reorientation of functional connections between nodes. In summary, this study illustrated progressively greater abnormalities in functional brain network organization in patients with clinical and biochemical evidence of more severe hepatic encephalopathy. The early-than-expected brain network dysfunction in cirrhotic patients suggests that brain functional connectivity and network analysis may provide useful and complementary biomarkers for more aggressive and earlier intervention of hepatic encephalopathy. Moreover, the stepwise deterioration of functional brain networks in HE patients may suggest that hierarchical network properties are necessary for normal brain function.

Simultaneous electroencephalographic and functional magnetic resonance imaging indicate impaired cortical top-down processing in association with anesthetic-induced unconsciousness.

BACKGROUND: In imaging functional connectivity (FC) analyses of the resting brain, alterations of FC during unconsciousness have been reported. These results are in accordance with recent electroencephalographic studies observing impaired top-down processing during anesthesia. In this study, simultaneous records of functional magnetic resonance imaging (fMRI) and electroencephalogram were performed to investigate the causality of neural mechanisms during propofol-induced loss of consciousness by correlating FC in fMRI and directional connectivity (DC) in electroencephalogram. METHODS: Resting-state 63-channel electroencephalogram and blood oxygen level-dependent 3-Tesla fMRI of 15 healthy subjects were simultaneously registered during consciousness and propofol-induced loss of consciousness. To indicate DC, electroencephalographic symbolic transfer entropy was applied as a nonlinear measure of mutual interdependencies between underlying physiological processes. The relationship between FC of resting-state networks of the brain (z values) and DC was analyzed by a partial correlation. RESULTS: Independent component analyses of resting-state fMRI showed decreased FC in frontoparietal default networks during unconsciousness, whereas FC in primary sensory networks increased. DC indicated a decline in frontal-parietal (area under the receiver characteristic curve, 0.92; 95% CI, 0.68-1.00) and frontooccipital (0.82; 0.53-1.00) feedback DC (P<0.05 corrected). The changes of FC in the anterior default network correlated with the changes of DC in frontal-parietal (rpartial=+0.62; P=0.030) and frontal-occipital (+0.63; 0.048) electroencephalographic electrodes (P<0.05 corrected). CONCLUSION: The simultaneous propofol-induced suppression of frontal feedback connectivity in the electroencephalogram and of frontoparietal FC in the fMRI indicates a fundamental role of top-down processing for consciousness.

Spatiotemporal reconfiguration of large-scale brain functional networks during propofol-induced loss of consciousness.

Applying graph theoretical analysis of spontaneous BOLD fluctuations in functional magnetic resonance imaging (fMRI), we investigated whole-brain functional connectivity of 11 healthy volunteers during wakefulness and propofol-induced loss of consciousness (PI-LOC). After extraction of regional fMRI time series from 110 cortical and subcortical regions, we applied a maximum overlap discrete wavelet transformation and investigated changes in the brain's intrinsic spatiotemporal organization. During PI-LOC, we observed a breakdown of subcortico-cortical and corticocortical connectivity. Decrease of connectivity was pronounced in thalamocortical connections, whereas no changes were found for connectivity within primary sensory cortices. Graph theoretical analyses revealed significant changes in the degree distribution and local organization metrics of brain functional networks during PI-LOC: compared with a random network, normalized clustering was significantly increased, as was small-worldness. Furthermore we observed a profound decline in long-range connections and a reduction in whole-brain spatiotemporal integration, supporting a topological reconfiguration during PI-LOC. Our findings shed light on the functional significance of intrinsic brain activity as measured by spontaneous BOLD signal fluctuations and help to understand propofol-induced loss of consciousness.

Effects of rapid eye movement sleep deprivation on fear extinction recall and prediction error signaling.

In a temporal difference learning approach of classical conditioning, a theoretical error signal shifts from outcome deliverance to the onset of the conditioned stimulus. Omission of an expected outcome results in a negative prediction error signal, which is the initial step towards successful extinction and may therefore be relevant for fear extinction recall. As studies in rodents have observed a bidirectional relationship between fear extinction and rapid eye movement (REM) sleep, we aimed to test the hypothesis that REM sleep deprivation impairs recall of fear extinction through prediction error signaling in humans. In a three-day design with polysomnographically controlled REM sleep deprivation, 18 young, healthy subjects performed a fear conditioning, extinction and recall of extinction task with visual stimuli, and mild electrical shocks during combined functional magnetic resonance imaging (fMRI) and skin conductance response (SCR) measurements. Compared to the control group, the REM sleep deprivation group had increased SCR scores to a previously extinguished stimulus at early recall of extinction trials, which was associated with an altered fMRI time-course in the left middle temporal gyrus. Post-hoc contrasts corrected for measures of NREM sleep variability also revealed between-group differences primarily in the temporal lobe. Our results demonstrate altered prediction error signaling during recall of fear extinction after REM sleep deprivation, which may further our understanding of anxiety disorders in which disturbed sleep and impaired fear extinction learning coincide. Moreover, our findings are indicative of REM sleep related plasticity in regions that also show an increase in activity during REM sleep.

Development of a large-scale functional brain network during human non-rapid eye movement sleep.

Graph theoretical analysis of functional magnetic resonance imaging (fMRI) time series has revealed a small-world organization of slow-frequency blood oxygen level-dependent (BOLD) signal fluctuations during wakeful resting. In this study, we used graph theoretical measures to explore how physiological changes during sleep are reflected in functional connectivity and small-world network properties of a large-scale, low-frequency functional brain network. Twenty-five young and healthy participants fell asleep during a 26.7 min fMRI scan with simultaneous polysomnography. A maximum overlap discrete wavelet transformation was applied to fMRI time series extracted from 90 cortical and subcortical regions in normalized space after residualization of the raw signal against unspecific sources of signal fluctuations; functional connectivity analysis focused on the slow-frequency BOLD signal fluctuations between 0.03 and 0.06 Hz. We observed that in the transition from wakefulness to light sleep, thalamocortical connectivity was sharply reduced, whereas corticocortical connectivity increased; corticocortical connectivity subsequently broke down in slow-wave sleep. Local clustering values were closest to random values in light sleep, whereas slow-wave sleep was characterized by the highest clustering ratio (gamma). Our findings support the hypothesis that changes in consciousness in the descent to sleep are subserved by reduced thalamocortical connectivity at sleep onset and a breakdown of general connectivity in slow-wave sleep, with both processes limiting the capacity of the brain to integrate information across functional modules.