SUFI: an automated approach to spectral unmixing of fluorescent multiplex images captured in mouse and post-mortem human brain tissues.

BACKGROUND: Multispectral fluorescence imaging coupled with linear unmixing is a form of image data collection and analysis that allows for measuring multiple molecular signals in a single biological sample. Multiple fluorescent dyes, each measuring a unique molecule, are simultaneously measured and subsequently "unmixed" to provide a read-out for each molecular signal. This strategy allows for measuring highly multiplexed signals in a single data capture session, such as multiple proteins or RNAs in tissue slices or cultured cells, but can often result in mixed signals and bleed-through problems across dyes. Existing spectral unmixing algorithms are not optimized for challenging biological specimens such as post-mortem human brain tissue, and often require manual intervention to extract spectral signatures. We therefore developed an intuitive, automated, and flexible package called SUFI: spectral unmixing of fluorescent images. RESULTS: This package unmixes multispectral fluorescence images by automating the extraction of spectral signatures using vertex component analysis, and then performs one of three unmixing algorithms derived from remote sensing. We evaluate these remote sensing algorithms' performances on four unique biological datasets and compare the results to unmixing results obtained using ZEN Black software (Zeiss). We lastly integrate our unmixing pipeline into the computational tool dotdotdot, which is used to quantify individual RNA transcripts at single cell resolution in intact tissues and perform differential expression analysis, and thereby provide an end-to-end solution for multispectral fluorescence image analysis and quantification. CONCLUSIONS: In summary, we provide a robust, automated pipeline to assist biologists with improved spectral unmixing of multispectral fluorescence images.

Single-nucleus transcriptome analysis reveals cell-type-specific molecular signatures across reward circuitry in the human brain.

Single-cell gene expression technologies are powerful tools to study cell types in the human brain, but efforts have largely focused on cortical brain regions. We therefore created a single-nucleus RNA-sequencing resource of 70,615 high-quality nuclei to generate a molecular taxonomy of cell types across five human brain regions that serve as key nodes of the human brain reward circuitry: nucleus accumbens, amygdala, subgenual anterior cingulate cortex, hippocampus, and dorsolateral prefrontal cortex. We first identified novel subpopulations of interneurons and medium spiny neurons (MSNs) in the nucleus accumbens and further characterized robust GABAergic inhibitory cell populations in the amygdala. Joint analyses across the 107 reported cell classes revealed cell-type substructure and unique patterns of transcriptomic dynamics. We identified discrete subpopulations of D1- and D2-expressing MSNs in the nucleus accumbens to which we mapped cell-type-specific enrichment for genetic risk associated with both psychiatric disease and addiction.

KCNH2-3.1 mediates aberrant complement activation and impaired hippocampal-medial prefrontal circuitry associated with working memory deficits.

Increased expression of the 3.1 isoform of the KCNH2 potassium channel has been associated with cognitive dysfunction and with schizophrenia, yet little is known about the underlying pathophysiological mechanisms. Here, by using in vivo wireless local field potential recordings during working memory processing, in vitro brain slice whole-cell patching recordings and in vivo stereotaxic hippocampal injection of AAV-encoded expression, we identified specific and delayed disruption of hippocampal-mPFC synaptic transmission and functional connectivity associated with reductions of SERPING1, CFH, and CD74 in the KCNH2-3.1 overexpression transgenic mice. The differentially expressed genes in mice are enriched in neurons and microglia, and reduced expression of these genes dysregulates the complement cascade, which has been previously linked to synaptic plasticity. We find that knockdown of these genes in primary neuronal-microglial cocultures from KCNH2-3.1 mice impairs synapse formation, and replenishing reduced CFH gene expression rescues KCNH2-3.1-induced impaired synaptogenesis. Translating to humans, we find analogous dysfunctional interactions between hippocampus and prefrontal cortex in coupling of the fMRI blood oxygen level-dependent (BOLD) signal during working memory in healthy subjects carrying alleles associated with increased KCNH2-3.1 expression in brain. Our data uncover a previously unrecognized role of the truncated KCNH2-3.1 potassium channel in mediating complement activation, which may explain its association with altered hippocampal-prefrontal connectivity and synaptic function. These results provide a potential molecular link between increased KCNH2-3.1 expression, synapse alterations, and hippocampal-prefrontal circuit abnormalities implicated in schizophrenia.

Studying the interactions in a mammalian nerve fiber: a functional modeling approach.

Modern therapeutic interventions are increasingly favoring electrical stimulation to treat neurophysiological dis-orders. These therapies are associated with suboptimal efficacy since most neurostimulation devices operate in an open-loop manner $(i.e.$, stimulation settings like frequency, amplitude are preprogrammed). A closed-loop system can dynamically adjust stimulation parameters and may provide efficient therapies. Computational models used to design these systems vary in complexity which can adversely affect their real-time performance. In this study, we compare two models of varying degrees of complexity. We constructed two computational models of a myelinated nerve fiber (functional versus mechanistic) each receiving two inputs: the underlying physiological activity at one end of the fiber, and the external stimulus applied to the middle of the fiber. We then defined relay reliability as the percentage of physiological action potentials that make it to the other end of the nerve fiber. We applied the two inputs to the fiber at various frequencies and analyze reliability. We found that the functional model and the mechanistic model have similar reliability properties, but the functional model significantly decreases the computational complexity and simulation run time. This modeling effort is the first step towards understanding and designing closed loop, real-time neurostimulation devices.

Selective relay of afferent sensory-induced action potentials from peripheral nerve to brain and the effects of electrical stimulation.

Electrical stimulation of peripheral nerve fibers and dorsal column fibers is used to treat acute and chronic pain. Recent studies have shown that sensitized A-fibers maybe involved in the relay of pain information. These nerve fibers also carry sensory-induced action potentials (APs), such as proprioception, mechanoreception, etc. Electrical stimulation of these nerve fibers can result in interactions between sensory-induced APs and stimulation-induced APs. For example, the sensory-induced APs can collide with stimulus APs, and thus may never be relayed to the brain. In this study, we aimed to quantify the effects of stimulation frequency on these interactions. Specifically, we focused on the goal of stimulation to simultaneously (i) block noxious sensory signals while (ii) relaying innocuous sensory signals from the periphery to the brain via a myelinated nerve fiber. We defined a performance metric called the "selective relay $(SR)$ " measure. Specifically, we constructed a tractable model of a nerve fiber that receives two inputs: the underlying sensory activity at the bottom of the fiber (noxious or innocuous), and the external stimulus applied to the middle of the fiber. We then defined relay reliability, $R$, as the percentage of sensory APs that make it to the top of the fiber. $SR$ is then a product of relaying innocuous sensory information while blocking noxious pain stimuli, i.e., $SR=R_{\mathrm {s}\mathrm {e}\mathrm {n}}(1-R_{\mathrm {p}\mathrm {a}\mathrm {i}\mathrm {n}})$. We applied the two inputs to the fiber at various frequencies and analyzed relay reliability and then we studied selective relay assuming noxious and innocuous stimuli produce APs with distinct frequencies. We found that frequency stimulation between 50-100Hz effectively blocks relay of low-frequency pain signals, allowing mid-to-high frequency sensory signals to transmit to the brain.

Modeling the interactions between stimulation and physiologically induced APs in a mammalian nerve fiber: dependence on frequency and fiber diameter.

Electrical stimulation of nerve fibers is used as a therapeutic tool to treat neurophysiological disorders. Despite efforts to model the effects of stimulation, its underlying mechanisms remain unclear. Current mechanistic models quantify the effects that the electrical field produces near the fiber but do not capture interactions between action potentials (APs) initiated by stimulus and APs initiated by underlying physiological activity. In this study, we aim to quantify the effects of stimulation frequency and fiber diameter on AP interactions involving collisions and loss of excitability. We constructed a mechanistic model of a myelinated nerve fiber receiving two inputs: the underlying physiological activity at the terminal end of the fiber, and an external stimulus applied to the middle of the fiber. We define conduction reliability as the percentage of physiological APs that make it to the somatic end of the nerve fiber. At low input frequencies, conduction reliability is greater than 95% and decreases with increasing frequency due to an increase in AP interactions. Conduction reliability is less sensitive to fiber diameter and only decreases slightly with increasing fiber diameter. Finally, both the number and type of AP interactions significantly vary with both input frequencies and fiber diameter. Modeling the interactions between APs initiated by stimulus and APs initiated by underlying physiological activity in a nerve fiber opens opportunities towards understanding mechanisms of electrical stimulation therapies.

Modeling electrical stimulation of mammalian nerve fibers: A mechanistic versus probabilistic approach.

Electrical neurostimulation is increasingly used over neuropharmacology to treat various diseases. Despite efforts to model the effects of electrical stimulation, its underlying mechanisms remain unclear. This is because current mechanistic models just quantify the effects that the electrical field produces near the fiber and do not capture interactions between stimulus-initiated action potentials (APs) and underlying physiological activity initiated APs. In this study, we aim to quantify and compare these interactions. We construct two computational models of a nerve fiber of varying degrees of complexity (probabilistic versus mechanistic) each receiving two inputs: the underlying physiological activity at one end of the fiber, and the external stimulus applied to the middle of the fiber. We then define reliability, R, as the percentage of physiological APs that make it to the other end of the nerve fiber. We apply the two inputs to the fiber at various frequencies and analyze reliability. We find that the probabilistic model captures relay properties for low input frequencies (<; 10 Hz) but then differs from the mechanistic model if either input has a larger frequency. This is because the probabilistic model only accounts for only (i) inter signal loss of excitability and (ii) collisions between stimulus-initiated action potentials (APs) and underlying physiological activity initiated APs. This first step towards modeling the interactions in a nerve fiber opens up opportunities towards understanding mechanisms of electrical stimulation therapies.

Voltage-sensitive dye imaging of mouse neocortex during a whisker detection task.

Sensorimotor processing occurs in a highly distributed manner in the mammalian neocortex. The spatiotemporal dynamics of electrical activity in the dorsal mouse neocortex can be imaged using voltage-sensitive dyes (VSDs) with near-millisecond temporal resolution and [Formula: see text] spatial resolution. Here, we trained mice to lick a water reward spout after a 1-ms deflection of the C2 whisker, and we imaged cortical dynamics during task execution with VSD RH1691. Responses to whisker deflection were highly dynamic and spatially highly distributed, exhibiting high variability from trial to trial in amplitude and spatiotemporal dynamics. We differentiated trials based on licking and whisking behavior. Hit trials, in which the mouse licked after the whisker stimulus, were accompanied by overall greater depolarization compared to miss trials, with the strongest hit versus miss differences being found in frontal cortex. Prestimulus whisking decreased behavioral performance by increasing the fraction of miss trials, and these miss trials had attenuated cortical sensorimotor responses. Our data suggest that the spatiotemporal dynamics of depolarization in mouse sensorimotor cortex evoked by a single brief whisker deflection are subject to important behavioral modulation during the execution of a simple, learned, goal-directed sensorimotor transformation.

Using demographic and time series physiological features to classify sepsis in the intensive care unit.

Sepsis, a systemic inflammatory response to infection, is a major health care problem that affects millions of patients every year in the intensive care units (ICUs) worldwide. Despite the fact that ICU patients are heavily instrumented with physiological sensors, early sepsis detection remains challenging, perhaps because clinicians identify sepsis by (i) using static scores derived from bed-side measurements individually, and (ii) deriving these scores at a much slower rate than the rate for which patient data is collected. In this study, we construct a generalized linear model (GLM) for the probability that an ICU patient has sepsis as a function of demographics and bedside measurements. Specifically, models were trained on 29 patient recordings from the MIMIC II database and evaluated on a different test set including 8 patient recordings. A classification accuracy of 62.5% was achieved using demographic measures as features. Adding physiological time series features to the model increased the classification accuracy to 75%. Although very preliminary, these results suggest that using generalized linear models incorporating real time physiological signals may be useful for an early detection of sepsis, thereby improving the chances of a successful treatment.