About

Dr. Mark Beenhakker is the principal investigator and leader of the Beenhakker Lab at the University of Virginia School of Medicine. He is described as the "lab wrangler," indicating his role in managing and guiding the research team. His lab focuses on understanding the mechanisms underlying seizures and epilepsy, employing a variety of experimental approaches including electrophysiology, imaging, and computational methods. The lab's research involves developing tools to record neuronal activity and resolve brain structures responsible for seizure initiation and propagation. Dr. Beenhakker's work supports a collaborative environment with PhD students, postdoctoral researchers, and research scientists who investigate topics such as metabolism's role in seizures, noradrenergic signaling related to arousal and attention, and genetic models of epilepsy. Outside of his scientific endeavors, Dr. Beenhakker enjoys spending time in nature, wandering forests and floating down rivers.

Research topics

• Medicine
• Internal medicine
• Immunology
• Anesthesia
• Cell biology
• Chemistry
• Biology
• Endocrinology
• Pharmacology

Selected publications

• Combining Residual U-Net and Data Augmentation for Dense Temporal Segmentation of Spike Wave Discharges in Single-Channel EEG

  arXiv (Cornell University) · 2026-01-01

  preprintOpen access

  Manual annotation of spike-wave discharges (SWDs), the electrographic hallmark of absence seizures, is labor-intensive for long-term electroencephalography (EEG) monitoring studies. While machine learning approaches show promise for automated detection, they often struggle with cross-subject generalization due to high inter-individual variability in seizure morphology and signal characteristics. In this study we compare the performance of 15 machine learning classifiers on our own manually annotated dataset of 961 hours of EEG recordings from C3H/HeJ mice, including 22,637 labeled SWDs and find that a 1D U-Net performs the best. We then improve its performance by employing residual connections and data augmentation strategies combining amplitude scaling, Gaussian noise injection, and signal inversion during training to enhance cross-subject generalization. We also compare our method, named AugUNet1D, to a recently published time- and frequency-based algorithmic approach called "Twin Peaks" and show that AugUNet1D performs better on our dataset. AugUNet1D, pretrained on our manually annotated data or untrained, is made public for other users.

  PublisherDOI

• Combining Residual U-Net and Data Augmentation for Dense Temporal Segmentation of Spike Wave Discharges in Single-Channel EEG

  arXiv (Cornell University) · 2026-01-01

  articleOpen access

  Manual annotation of spike-wave discharges (SWDs), the electrographic hallmark of absence seizures, is labor-intensive for long-term electroencephalography (EEG) monitoring studies. While machine learning approaches show promise for automated detection, they often struggle with cross-subject generalization due to high inter-individual variability in seizure morphology and signal characteristics. In this study we compare the performance of 15 machine learning classifiers on our own manually annotated dataset of 961 hours of EEG recordings from C3H/HeJ mice, including 22,637 labeled SWDs and find that a 1D U-Net performs the best. We then improve its performance by employing residual connections and data augmentation strategies combining amplitude scaling, Gaussian noise injection, and signal inversion during training to enhance cross-subject generalization. We also compare our method, named AugUNet1D, to a recently published time- and frequency-based algorithmic approach called "Twin Peaks" and show that AugUNet1D performs better on our dataset. AugUNet1D, pretrained on our manually annotated data or untrained, is made public for other users.

  PublisherOA PDF

• Targeting microRNA-dependent control of X chromosome inactivation improves the Rett Syndrome phenotype

  Nature Communications · 2025-07-04 · 9 citations

  articleOpen access

  X chromosome inactivation (XCI) is induced by Xist long non-coding RNA and protein-coding genes. However, the role of small non-coding RNA function in XCI remains unidentified. Our genome-wide, loss-of-function CRISPR/Cas9 screen in female fibroblasts identified microRNAs (miRNAs) as regulators of XCI. A striking finding is the identification of miR106a among the top candidates from the screen. Loss of miR106a is accompanied by altered Xist interactome, leading to dissociation and destabilization of Xist. XCI interference via miR106a inhibition has therapeutic implications for Rett syndrome (RTT) girls with a defective X-linked MECP2 gene. Here, we discovered that the inhibition of miR106a significantly improves several facets of RTT pathology: it increases the life span, enhances locomotor activity and exploratory behavior, and diminishes breathing variabilities. Our results suggest that miR106a targeting offers a feasible therapeutic strategy for RTT and other monogenic X-linked neurodevelopmental disorders.

  PublisherOA PDFDOI

• Thalamic hubs as early sources of global neuronal synchrony in absence epilepsy

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-12-22

  articleOpen accessSenior authorCorresponding

  Generalized seizures reflect a pathological state of sudden, global neuronal hypersynchrony. The mechanisms that support the initiation and maintenance of such synchrony remain unknown. Using simultaneous single-unit and electrocorticographic recordings in two mouse models of absence epilepsy, we evaluated the activity of approximately 2,000 individual neurons across 26 brain structures. By doing so, we resolved the temporal progression of single neuron activity prior to and during generalized seizures. Surprisingly, we observed that rhythmic, synchronized activity emerges early and gradually in both thalamus and cortex. Moreover, while we observe that individual neurons across most structures fire rhythmically and synchronously during seizures, a small subset of thalamic nuclei stand out as hubs, displaying activity that is strongly correlated across multiple regions. Thus, our findings collectively highlight the thalamus as an early driver of global and pathological neuronal hypersynchrony.

  PublisherOA PDFDOI

• Osmolarity Controls Oscillatory Calcium Signaling to Reduce Autonomous Aldosterone Production in Zona Glomerulosa Cells

  Endocrinology · 2025-10-14 · 2 citations

  articleOpen access

  Primary aldosteronism (PA) is characterized by autonomous aldosterone (Aldo) production, resulting in blood volume/electrolyte imbalance and hypertension. Intracellular calcium (Ca2+) is the principal signal driving Aldo synthesis in adrenal zona glomerulosa (zG) cells, and mutations in ion transport genes that regulate Ca2+ are frequently mediators of PA. When organized in intact rosette structures, zG cells are voltage oscillators; stimulation by angiotensin II (AngII) or loss of TWIK-related acid-sensitive potassium (TASK) channel function evokes stereotypic Ca2+ oscillations with bursting activity proportional to increased steroidogenesis. Here, we delineate the role of the osmolar-volume regulatory axis in the control of Ca2+ and Aldo production in adrenal slices. Strikingly, in both pharmacological and genetic models of PA, extracellular osmolarity (OSMEC) potently and reversibly regulated Aldo secretion and Ca2+ signaling. Elevated OSMEC progressively suppressed Aldo production from AngII-stimulated adrenal slices and strongly inhibited autonomous production in both zG-specific TASK knockout slices and wild-type slices incubated with TASK inhibitors (TIs). To determine if the effects of OSMEC on Ca2+ dynamics were causative, we imaged adrenal slices expressing zG-specific GCaMP6f incubated in variable osmotic media with TIs or AngII. Consistent with Aldo suppression, increasing osmolarity proportionally reduced the number of active cells and the Ca2+ activity of bursting cells evoked by TASK loss of function or AngII stimulation. Collectively, our findings identify OSMEC as a broad regulator of zG excitability and adrenal steroidogenesis, and suggest that targeting volume-regulatory mechanisms such as the Na+-K+-2Cl- cotransporter may offer a novel strategy to suppress Aldo autonomy in PA.

  PublisherOA PDFDOI

• Acute Single-Unit Multi-Electrode Recordings from the Brainstem of Head-Fixed Mice

  Journal of Visualized Experiments · 2024-10-11 · 1 citations

  article

  Silicon multielectrode-based recordings are increasingly popular for studying neuronal activity at the temporal resolution of action potentials in many brain regions. However, recording neuronal activity from deep caudal structures like the brainstem using multi-channel probes remains challenging. A significant concern is finding a trajectory for probe insertion that avoids large blood vessels, such as the superior sagittal venous sinus and the transverse venous sinus. Injuring these large veins can cause extensive bleeding, damage to the underlying brain tissue, and potentially death. This approach describes targeting brainstem structures by coupling anterior coordinates with an angled approach, allowing the recording probe to penetrate the brain below high-risk vascular structures. Compared to a strictly vertical approach, the angled approach maximizes the number of brain regions that can be targeted. Using this strategy, the ventrolateral periaqueductal gray (vlPAG), a brainstem region associated with REM sleep, can be reproducibly and reliably accessed to obtain single-unit, multi-electrode recordings in head-fixed mice before and during sevoflurane anesthesia. The ability to record neuronal activity in the vlPAG and surrounding nuclei with high temporal resolution is a step forward in advancing the understanding of the relationship between REM sleep and anesthesia.

  PublisherDOI

• Suppress Globally or Seize Locally: Cortical Network Activity Explains Seizure Diversity Among <i>Kcnt1</i> Mutants

  Epiliepsy currents/Epilepsy currents · 2023-11-30

  letterOpen accessSenior author

  [Box: see text]

  PublisherDOI

• Sleep Fragmentation, Electroencephalographic Slowing, and Circadian Disarray in a Mouse Model for Intensive Care Unit Delirium

  Anesthesia & Analgesia · 2023-05-16 · 8 citations

  article

  BACKGROUND: We aimed to further validate our previously published animal model for delirium by testing the hypothesis that in aged mice, Anesthesia, Surgery and simulated ICU conditions (ASI) induce sleep fragmentation, electroencephalographic (EEG) slowing, and circadian disarray consistent with intensive care unit (ICU) patients with delirium. METHODS: A total of 41 mice were used. Mice were implanted with EEG electrodes and randomized to ASI or control groups. ASI mice received laparotomy, anesthesia, and simulated ICU conditions. Controls did not receive ASI. Sleep was recorded at the end of ICU conditions, and hippocampal tissue was collected on EEG recording. Arousals, EEG dynamics, and circadian gene expression were compared with t tests. Two-way repeated measures analysis of variance (RM ANOVA) was used to assess sleep according to light. RESULTS: ASI mice experienced frequent arousals (36.6 ± 3.2 vs 26.5 ± 3.4; P = .044; 95% confidence interval [CI], 0.29-19.79; difference in mean ± SEM, 10.04 ± 4.62) and EEG slowing (frontal theta ratio, 0.223 ± 0.010 vs 0.272 ± 0.019; P = .026; 95% CI, -0.091 to -0.007; difference in mean ± SEM, -0.05 ± 0.02) relative to controls. In ASI mice with low theta ratio, EEG slowing was associated with a higher percentage of quiet wakefulness (38.2 ± 3.6 vs 13.4 ± 3.8; P = .0002; 95% CI, -35.87 to -13.84; difference in mean ± SEM, -24.86 ± 5.19). ASI mice slept longer during the dark phases of the circadian cycle (nonrapid eye movement [NREM], dark phase 1 [D1]: 138.9 ± 8.1 minutes vs 79.6 ± 9.6 minutes, P = .0003, 95% CI, -95.87 to -22.69, predicted mean difference ± SE: -59.28 ± 13.89; NREM, dark phase 2 (D2): 159.3 ± 7.3 minutes vs 112.6 ± 15.5 minutes, P = .006, 95% CI, -83.25 to -10.07, mean difference ± SE, -46.66 ± 13.89; rapid eye movement (REM), D1: 20.5 ± 2.1 minutes vs 5.8 ± 0.8 minutes, P = .001, 95% CI, -24.60 to -4.71, mean difference ± SE, -14. 65 ± 3.77; REM, D2: 21.0 ± 2.2 minutes vs 10.3 ± 1.4 minutes, P = .029, 95% CI, -20.64 to -0.76, mean difference ± SE, -10.70 ± 3.77). The expression of essential circadian genes was also lower in ASI mice (basic helix-loop-helix ARNT like [BMAL1] : -1.3 fold change; circadian locomotor output cycles protein kaput [CLOCK] : -1.2). CONCLUSIONS: ASI mice experienced EEG and circadian changes mimicking those of delirious ICU patients. These findings support further exploration of this mouse approach to characterize the neurobiology of delirium.

  PublisherDOI

• The Beginning of Everything: Finding the Seizure Onset

  Epiliepsy currents/Epilepsy currents · 2023-01-23

  letterOpen accessSenior author

  [Box: see text]

  PublisherDOI

• Anatomical Substrates of Rapid Eye Movement Sleep Rebound in a Rodent Model of Post-sevoflurane Sleep Disruption

  Anesthesiology · 2023-12-29 · 1 citations

  articleOpen access

  BACKGROUND: Previous research suggests that sevoflurane anesthesia may prevent the brain from accessing rapid eye movement (REM) sleep. If true, then patterns of neural activity observed in REM-on and REM-off neuronal populations during recovery from sevoflurane should resemble those seen after REM sleep deprivation. In this study, the authors hypothesized that, relative to controls, animals exposed to sevoflurane present with a distinct expression pattern of c-Fos, a marker of neuronal activation, in a cluster of nuclei classically associated with REM sleep, and that such expression in sevoflurane-exposed and REM sleep-deprived animals is largely similar. METHODS: Adult rats and Targeted Recombination in Active Populations mice were implanted with electroencephalographic electrodes for sleep-wake recording and randomized to sevoflurane, REM deprivation, or control conditions. Conventional c-Fos immunohistochemistry and genetically tagged c-Fos labeling were used to quantify activated neurons in a group of REM-associated nuclei in the midbrain and basal forebrain. RESULTS: REM sleep duration increased during recovery from sevoflurane anesthesia relative to controls (157.0 ± 24.8 min vs. 124.2 ± 27.8 min; P = 0.003) and temporally correlated with increased c-Fos expression in the sublaterodorsal nucleus, a region active during REM sleep (176.0 ± 36.6 cells vs. 58.8 ± 8.7; P = 0.014), and decreased c-Fos expression in the ventrolateral periaqueductal gray, a region that is inactive during REM sleep (34.8 ± 5.3 cells vs. 136.2 ± 19.6; P = 0.001). Fos changes similar to those seen in sevoflurane-exposed mice were observed in REM-deprived animals relative to controls (sublaterodorsal nucleus: 85.0 ± 15.5 cells vs. 23.0 ± 1.2, P = 0.004; ventrolateral periaqueductal gray: 652.8 ± 71.7 cells vs. 889.3 ± 66.8, P = 0.042). CONCLUSIONS: In rodents recovering from sevoflurane, REM-on and REM-off neuronal activity maps closely resemble those of REM sleep-deprived animals. These findings provide new evidence in support of the idea that sevoflurane does not substitute for endogenous REM sleep.

  PublisherOA PDFDOI

Recent grants

• Energy and Neural Circuit Excitability

  NIH · $1.4M · 2016–2021

• NIH Grant F31NS041894

  NIH · $93k · 2004

• NIH Grant K99NS065130

  NIH · $180k · 2011

Frequent coauthors

• Michael P. Nusbaum

  University of Pennsylvania

  17 shared

• Kathryn A. Salvati

  Neurological Surgery

  12 shared

• Adam C Lu

  University of Virginia

  12 shared

• John R. Huguenard

  Stanford University

  10 shared

• Peter Klein

  Stanford University

  7 shared

• Dag K.J.E. Von Lubitz

  7 shared

• Nadia Lunardi

  University of Virginia

  6 shared

• David T. Breault

  Harvard University

  5 shared

Labs

• Beenhakker LabPI