About

Karen Wilcox, PhD, is a Distinguished Professor and Chair of Pharmacology & Toxicology at the College of Pharmacy, University of Utah. She also holds the Richard L. Stimson Presidential Endowed Chair. Her research focuses on understanding the basic mechanisms underlying epileptogenesis, seizure generation, and therapy-resistance to anticonvulsant drugs. To achieve these goals, her laboratory employs electrophysiological, calcium imaging, pharmacological, behavioral, genetic, immunoblot, and immunohistochemical techniques in various in vitro preparations and animal models of epilepsy. Dr. Wilcox's working hypothesis is that insight into disease-induced changes in neuronal and glial function will provide new avenues for therapeutic interventions for patients at risk of developing epilepsy or those who are refractory to current treatments. She is the Director of the Anticonvulsant Development (ADD) Program, where she evaluates proprietary investigational compounds through a contract with NINDS at the NIH to determine their antiseizure and disease-modifying potential. Her contributions to the field have been recognized through awards such as the 2023 University of Utah Distinguished Professor Award, the 2022 NINDS Landis Outstanding Mentorship Award, and the 2025 University of Utah Distinguished Researcher Award.

Research topics

• Medicine
• Psychiatry
• Internal medicine
• Pharmacology
• Neuroscience
• Political Science
• Anesthesia
• Pathology
• Psychology
• Biology
• Intensive care medicine
• Medical physics
• Pediatrics
• Immunology
• Bioinformatics

Selected publications

• Investigating the Role of Cortical Microglia in a Mouse Model of Viral Infection-Induced Seizures

  eNeuro · 2026-02-01

  articleOpen accessSenior author

  Microglia, resident immune sentinels in the brain, are crucial in responding to tissue damage, infection, damage signals like purines (ATP/ADP), and clearing cellular debris. It is currently unknown how microglial reactivity progresses and contributes to seizure development following Theiler's murine encephalomyelitis virus (TMEV) infection. Previously, it has been demonstrated that purinergic signaling in microglia is disrupted in the hippocampus of TMEV-infected mice. However, whether reactive cortical microglia also exhibit changes in purinergic signaling, cytokine levels, and purinergic receptors is unknown. Thus, we seek to evaluate region-based differences in microglial reactivity in the TMEV model. We employed a custom triple transgenic mouse line expressing tdTomato and GCaMP6f under a CX3CR1 Cre promoter and exogenously applied ATP/ADP to acute brain slice preparations from TMEV-infected mice and controls of either sex. Interestingly and in contrast to what is observed in the hippocampus, we found that despite microglial reactivity in the cortex, microglia can respond to purinergic damage signals and engage calcium signaling pathways, comparable to PBS controls. Using a cytokine panel, we also found that proinflammatory cytokine levels (TNF-α, IL-1α, and IFN-γ) are brain region dependent in mice infected with TMEV. Using RNAscope FISH, we observed increases in expression of purinergic receptors responsible for microglial motility (P2Y 12 R) and inflammation (P2X 7 R) in the cortex. Collectively our results suggest that following TMEV infection, microglial response to novel damage signals, as well as the production of proinflammatory cytokines, varies as a function of the brain region.

  PublisherDOI

• Epilepsy Therapy Screening From 1975 to 2026 and Beyond: Merging Established and New Approaches to Develop Novel Therapies

  Epiliepsy currents/Epilepsy currents · 2026-04-20

  articleOpen accessSenior author

  Antiepileptic Drug Development: II. Anticonvulsant Drug Screening Krall RL, Penry JK, White BG, Kupferberg HJ, Swinyard EA. Epilepsia . 1978 Aug;19(4):409-28. doi:10.1111/j.1528-1157.1978.tb04507.x. PMID: 699894 By means of the maximal electroshock seizure test, the subcutaneous pentylenetetrazol seizure threshold test, and the Rotorod minimal neurotoxicity test, the Anticonvulsant Screening Project has evaluated the activity of 1,495 experimental compounds accessioned in the first 2 years. A three-screen protocol for appraising these compounds has proved reliable, fast, and inexpensive. Preliminary data show that 430 of these compounds have good anticonvulsant activity. Completed evaluations of 352 identified 16 that have anticonvulsant activity at doses less than 75 mg/kg and protective indices greater than 5.0.

  PublisherDOI

• <scp>Electroencephalographic</scp> signal dimension provides the necessary stability of measurements for the unbiased evaluation of <scp>antiseizure medications</scp> , unlike seizure frequency: Overcoming drawbacks of seizure occurrence variability

  Epilepsia · 2025-10-29

  letterOpen access

  The assessment of antiseizure medication (ASM) efficacy still predominantly relies on changes in seizure frequency (SF). However, in people with epilepsy (PWE), intrinsic variability in seizure timing and patterns significantly influences this assessment, often leading to potential false-negative or false-positive results. A low seizure rate particularly increases false-positive risk, as average seizure-free periods can be statistically comparable to therapy duration. This impacts clinical practice, necessitating extended monitoring and delaying appropriate treatments. For new drug development, PWE with low SFs are typically excluded from clinical trials, hindering patient enrollment and resulting in cohorts that may not adequately represent the general PWE population. To address the well-known limitations of SF as a therapeutic marker, in our recent study published in Epilepsia (doi: https://doi.org/10.1111/epi.18397), we utilized a widely used mouse model of acquired epilepsy that exhibits similar SF variability to human patients. Using recurrence quantification analysis (RQA), a robust method for analyzing short, noisy, and nonstationary time series like the electroencephalogram (EEG), we demonstrated that electroencephalographic signal dimension serves as a reliable, SF-independent measure of both seizure susceptibility and therapeutic response to ASMs. The electroencephalographic signal dimension (DIM), computed from RQA, is an inverse index of the average degree of correlation among brain cells. As we demonstrated in our paper (doi: https://doi.org/10.1111/epi.18397), DIM shows peculiar features to overcome the abovementioned limitations. DIM leverages the network disease characteristic of epilepsy, which arises from excessive correlation among firing neurons, thus acting as an (inverse) proxy of epileptic tissue internal correlation directly underpinning seizure susceptibility and modulating the EEG activity. In our study, DIM provided a stable and consistent biomarker for the risk of developing a seizure (i.e., seizure susceptibility), regardless of whether a seizure actually occurs. This makes DIM a superior measure to SF, which is affected by the inherent stochastic nature of seizure occurrence, which is modulated by the system's susceptibility but is not a direct and continuous measure of it. This makes SF often inconsistent, especially in subjects with low seizure rates. These findings highlighted that DIM offers immediate benefit in preclinical studies and may hold promise for clinical application. We now present new, compelling evidence that further strengthens and extends our initial findings, consolidating the relevance and reliability of electroencephalographic signal dimension as a biomarker for seizure susceptibility compared to seizure rate. We used the same mice and methodology as published, basing our reasoning on the fact that no evidence supports any effect on seizure susceptibility of drug vehicles (NaCl .9% or methylcellulose .5%, administered intraperitoneally). Therefore, for each mouse, both SF and DIM measured during baseline should be statistically equivalent to those measured during vehicle administration. This implies both metrics must satisfy two related statistical relationships: (1) baseline and vehicle measurements should be significantly correlated; and (2) they can be modeled by a regression line with a slope statistically equal to 1, with minor fluctuations around unity. Accordingly, we first established the existence of statistically significant correlations between baseline and vehicle measurements. The DIM measurements were always significantly correlated across all groups of mice, whereas only a minority of SF values showed significant correlation, revealing inconsistencies among mice treated with the same vehicle (Figure 1, top left table). Such differences in correlation statistics are rooted in the respective frequency distributions. Whereas DIM showed a Gaussian-like distribution consistent with a sensible measure of individual system's susceptibility, SF gave rise to exponential frequency rate distribution reminiscent of a Poisson/rare-event-like stochastic phenomenon (Figure 1, top right histograms), due to the inherent stochastic nature of seizure occurrence. To confirm the ability of both DIM and SF to meet the baseline–vehicle equivalence constraints, we adopted an ordinary linear regression approach to evaluate how close slope values were to unity, in the presence of possible systematic biases (e.g., circadian rhythms, manipulation or environmental stress, equipment noise; Figure 1A–F). We found that the slopes of DIM regression lines were all statistically significant and distributed, on average, around unity (Figure 1A–E; mean ± SEM = .92 ± .19). In contrast, SF regression statistics were poor (Figure 1A–F, right graphs) and even inconsistent with their respective correlation analysis, as in the case of the phenytoin–vehicle group, which showed no significant regression line despite a significant correlation. Finally, to exclude that the slopes close to unity in the DIM regressions were spurious, because they were induced by the systematic biases rather than driven by the actual equivalence between baseline and vehicle measurements, we ran a second regression analysis setting the intercepts to zero. The closeness of DIM slope values to unity was consistently confirmed and strengthened for each group of mice (Figure 2A–F, left graphs). In contrast, SF measurements continued to show poor regression statistics, confirming that their intrinsic high variability prevents a stable relationship among measurements (Figure 2A–F, right graphs). The sequence of our analyses provides clear biologically and statistically sound evidence of DIM robustness and stability compared to SF. Such a stable relationship between baseline and vehicle is a necessary requirement for an unbiased evaluation of any ASM efficacy. This critical requirement is poorly satisfied by SF but is perfectly met by DIM, highlighting the importance of shifting the focus from seizure rate to the susceptibility to undergo a seizure. Therefore, our findings further confirm the limitations of SF as a biomarker of seizure susceptibility and support DIM as a valid and robust alternative to SF in preclinical testing of experimental therapeutics with translational potential in clinical settings. If validated for human use, DIM's ability to measure seizure risk could pave the way for a paradigm shift from continuous to on-demand or intermittent therapy to rapidly decrease the risk of a seizure. This targeted approach could also reduce the side effects associated with chronic medication. The successful implementation of DIM's use will rely upon the development of an algorithm for real-time DIM calculation, integrated into advanced electroencephalographic closed loop systems. The authors wish to thank Kyle Thomson for his helpful technical support. The work presented has been funded in whole or in part with federal funds from the National Institute of Neurological Disorders and Stroke, National Institutes of Health, Department of Health and Human Services, under contract No. HHS 75N95022C00007. We acknowledge the CINECA award under the ISCRA initiative for the availability of high-performance computing resources and support. Open access funding provided by BIBLIOSAN. None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. The reference data are available in Rizzi M, West PJ, Vezzani A, Wilcox KS, Giuliani A. EEG signal dimension is an index of seizure propensity and antiseizure medication effects in a mouse model of acquired epilepsy. Epilepsia. 2025;66:3035–3047. https://doi.org/10.1111/epi.18397.

  PublisherOA PDFDOI

• Seizure Circuit Activity in the Theiler’s Murine Encephalomyelitis Virus Model of Infection-induced Epilepsy Using Transient Recombination in Active Populations

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-05-21

  preprintOpen accessSenior authorCorresponding

  Epilepsy affects one in twenty-six individuals. A major cause of epilepsy worldwide is viral encephalitis. Central nervous system infections can provoke seizures in the short term and increase the risk of spontaneous, recurrent seizures post-infection. However, the neural mechanisms underlying seizures during acute infection are unknown. These neuronal changes can be studied in C57BL6/J mice infected with Theiler's murine encephalomyelitis virus (TMEV). TMEV-infected mice experience seizures 3-8 days post-injection (DPI), clear the virus by DPI 14, and may develop chronic, acquired temporal lobe epilepsy. TMEV may incite seizures during the acute infection period through inflammation, reactive gliosis, and cell death in hippocampal area CA1. Here, we explore the neuronal circuits underlying acute seizures in TMEV-injected mice using c-Fos driven TRAP (targeted recombination in active populations). TRAP mice (c-Fos-CreERT2 x CAG-tdTomato) were injected with PBS or TMEV and gently handled on DPI 5 to induce seizures. 4-OHT was administered to mice either 1.5 or 3 hr after seizures to tag the active cells expressing c-Fos with tdTomato. After 1 week, the mice were sacrificed and whole mouse brains were sectioned and immunostained for tdTomato expression. Percent area of fluorescence was quantified, and comparisons were made between TMEV-injected mice and PBS controls, sites ipsilateral vs contralateral to TMEV injection site, and between sexes. TdTomato expression was elevated in the TMEV-injected mice in the ipsilateral and contralateral hippocampus, thalamus, lateral septal nucleus, basal ganglia, triangular septal nucleus, fornix, and corpus callosum. Critically, the expression pattern suggests that seizures induced on DPI 5 arise from the hilus, dentate gyrus, and CA3 hippocampal subregions. Generalized seizures during acute TMEV infection may have propagated to the contralateral hemisphere via CA3 and the hippocampal commissure. TRAP has not been previously utilized in the TMEV mouse model and these experiments address crucial questions regarding seizure spread during TMEV infection.

  PublisherOA PDFDOI

• Discovery of novel hybrid pyrrolidin-2-one derivatives exhibiting potent antiseizure and antinociceptive effects in preclinical models

  Biomedicine & Pharmacotherapy · 2025-11-01

  articleOpen access

  In the present study, a series of novel derivatives based on the pyrrolidin-2-one scaffold were designed using a framework combination approach, synthesized, and comprehensively evaluated through in vitro and in vivo assays conducted on Swiss albino male mice. The obtained hybrid molecules demonstrated potent and broad-spectrum antiseizure activity in key preclinical seizure models. Following intraperitoneal (i.p.) administration, compound (R)-9, identified as the lead molecule, provided robust protection across all tested seizure paradigms, with ED₅₀ values of 64.3 mg/kg (maximal electroshock test), 26.3 mg/kg (6 Hz, 32 mA), and 37.8 mg/kg (6 Hz, 44 mA). In addition, (R)-9 was active in the pentylenetetrazole (PTZ)-induced kindling model as well as in spontaneous electrographic bursting, an in vitro model of pharmacoresistant seizure-like activity at a concentration of 120 μM. Moreover, it significantly increased seizure threshold in the ivPTZ test. Importantly, (R)-9 also exhibited strong antinociceptive properties. No adverse effects on motor coordination and grip strength were observed at effective doses. Pharmacokinetic profiling and in vitro ADME-Tox evaluation showed satisfying drug-like characteristics of (R)-9, including metabolic stability in human liver microsomes, interaction with cytochrome P450 enzymes and hepatotoxicity (HepG2 cell line). In vitro binding and functional assays suggest a multimodal mechanism of action. Besides its TRPV1 antagonistic activity, (R)-9 effectively inhibited voltage-gated sodium channels at a concentration of 50 μM in electrophysiological studies. Collectively, these findings support the further preclinical development of (R)-9 as a promising candidate for the treatment of epilepsy and pain-related disorders.

  PublisherDOI

• Development of a preclinical testing platform for clinically relevant therapy for Dravet syndrome

  Epilepsia · 2025-06-30 · 3 citations

  articleOpen accessSenior authorCorresponding

  Abstract Objective Patients with drug‐resistant epilepsy, including Dravet syndrome, are frequently prescribed multiple antiseizure medications. Nevertheless, people with Dravet syndrome often have inadequate seizure control, and there is an ongoing unmet clinical need to identify novel therapeutics. As a proof‐of‐principle study to further validate and characterize the Scn1a A1783V/WT mouse model and identify a drug‐testing paradigm with face, construct, and predictive validity, we assessed the efficacy of subchronic administration of stiripentol add‐on to clobazam and valproic acid at clinically relevant doses using the Scn1a A1783V/WT mouse model. Methods Following a 14‐day treatment, we evaluated the efficacy of stiripentol add‐on to clobazam and valproic acid using hyperthermia‐induced ( n = 6) and video‐electroencephalography (EEG) monitoring of spontaneous seizure tests ( n = 13). Valproic acid was delivered via osmotic minipump, whereas stiripentol and clobazam were administered via food pellets delivered through automatic feeders. Bioanalytical assays were performed to evaluate drug concentrations in plasma and brain using liquid chromatography–tandem mass spectrometry. Results Stiripentol, clobazam, N ‐desmethylclobazam, and valproic acid all yielded plasma concentrations within the therapeutic plasma concentration range for humans. Stiripentol added to clobazam and valproic acid significantly elevated the seizing temperatures in the hyperthermia‐induced seizure assay (** p = 0.0018; Log‐rank test). Clobazam, valproic acid, and stiripentol co‐administration significantly reduced spontaneous seizure frequency compared to clobazam and valproic acid combined (*** p = 0.0003, Mann–Whitney test). Significance This research lays the groundwork for exploring effective add‐on compounds to clobazam and valproic acid in treating Dravet syndrome. The study further highlights the utility of the Scn1a A1783V/WT mice in discovering therapies for Dravet syndrome–associated pharmacoresistant seizures.

  PublisherOA PDFDOI

• DynamoSort: Using machine learning approaches for the automatic classification of seizure dynamotypes

  Neurocomputing · 2025-07-15

  article
  PublisherDOI

• <scp>EEG</scp> signal dimension is an index of seizure propensity and antiseizure medication effects in a mouse model of acquired epilepsy

  Epilepsia · 2025-04-09

  articleOpen access

  OBJECTIVE: Variability in the frequency, timing, and pattern of seizures may influence the assessment of the effect of antiseizure medications (ASMs) when measuring seizure frequency, especially in patients with infrequent seizures. A low seizure rate is an exclusion criterion for enrollment of patients with epilepsy in clinical trials and requires prolonged periods of seizure monitoring, thus delaying appropriate treatment interventions. We investigated whether an electroencephalogram (EEG)-based complexity measure of seizure susceptibility of epileptic mice provides a reliable alternative to seizure frequency for evaluating the efficacy of ASMs. METHODS: We used a mouse model of acquired epilepsy characterized by variability in seizure frequency over time and among mice, as observed in humans. We analyzed EEG recordings from chronic epileptic mice (n = 106) at baseline and during treatment with phenobarbital, valproate, carbamazepine, or phenytoin. We used recurrence quantification analysis to detect increased autocorrelation and critical slowing-down, two signatures of criticality that together contribute to estimate the dimension of phase-space of the EEG signals. The measurements of dimension and seizure frequency were compared as proxies for seizure susceptibility by correlation tests and evaluation of ASM efficacy. RESULTS: Dimension provided a statistically robust (inverse) estimate of seizure susceptibility of mice, including mice with low seizure frequency or no seizures during the observation periods. In contrast, seizure frequency provided a reliable measure only in mice with a high seizure rate. Consistently, evaluation of ASM efficacy using dimension variations accurately reproduced seizure responsiveness patterns in this mouse model. SIGNIFICANCE: EEG-based dimension provides a reliable measure of mouse propensity to experience seizures as well as ASM efficacy, regardless of seizure rates. Measuring dimension variation should facilitate the inclusion of subjects with low seizure rate in preclinical and clinical trials while also shortening the periods of monitoring. This could accelerate both the development of new treatments and therapeutic decisions in the medical field.

  PublisherOA PDFDOI

• DynamoSort: Using machine learning approaches for the automatic classification of seizure dynamotypes

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-02-17

  preprintOpen access

  Abstract Objective Epilepsy is characterised by unprovoked and recurring seizures, which can be electrically measured using electroencephalograms (EEG). To better understand the underlying mechanisms of seizures, researchers are exploring their temporal dynamics through the lens of dynamical systems modelling. Seizure initiation and termination patterns of spiking amplitude and frequency can be sorted into “dynamotypes”, which may be able to serve as biomarkers for intervention. However, manual classification of these dynamotypes requires trained raters and is prone to variability. To address this, we developed DynamoSort, a machine-learning algorithm for automatic seizure onset and offset classification. Methods We used approximately 2100 seizures from an intra-amygdala kainic acid (IAK) mouse model of mesial temporal lobe epilepsy, categorized by five trained raters. MATLAB’s classification learner application was used to create an ensemble model to score and label dynamotypes of individual seizures based on spiking and frequency features. Results Dynamotype classification of real EEG data lacks a definitive ground truth, with mean inter-rater agreement at 73.4% for onset and 64.2% for offset types. Despite this, DynamoSort achieved a mean area under the curve (AUC) of 0.81 for onset and a mean AUC of 0.75 for offset types. Machine-human agreement was not significantly different from human-to-human agreement. To address the lack of ground truth in ratings, DynamoSort assigns probabilistic scores (-20 to 20), to indicate similarity to each seizure dynamotype based on spiking features, allowing for a characterization of seizure dynamics on a spectrum rather than the traditional qualitative taxonomy. Significance Automating the classification of dynamotypes is a critical step for their inclusion as a biomarker in clinical and research applications. DynamoSort is a straightforward, open-access tool that uses automatic labelling and probabilistic scoring to quantify subtle changes in seizure onset and offset dynamics. Key Points Dynamotypes are a promising seizure categorization system, but is prone to interrater variability and lacks a ground truth. Machine learning can be used to automatically classify seizure onsets and offsets into appropriate dynamotypes based on spike features. Agreement between DynamoSort and human raters rivals typical agreement rates in trained human raters. DynamoSort uses probabilistic scoring to quantify subtle changes in seizure onset and offset, allowing for a quantitative characterisation.

  PublisherOA PDFDOI

• Ontology accelerates few-shot learning capability of large language model: A study in extraction of drug efficacy in a rare pediatric epilepsy

  International Journal of Medical Informatics · 2025-04-21 · 2 citations

  articleOpen access

  OBJECTIVE: Dravet Syndrome (DS) is a developmental and epileptic encephalopathy that is characterized by severe, prolonged motor seizures and high resistance to multiple antiseizure medications (ASMs) with multiple comorbidities. Evaluating the efficacy of new drugs in DS preclinical models and mapping them to human phenotypes of DS through analysis of published literature is an important goal for improving outcomes in this rare pediatric epilepsy. MATERIALS AND METHODS: Large language models (LLM) have demonstrated great promise in parsing published literature; however, the performance of LLMs falls short in medical applications. In this study, we investigate the effectiveness of domain ontology developed by human experts to optimize LLMs for medical text processing in a rare disease. Utilizing a benchmark dataset that describes the efficacy of 17 ASMs tested in preclinical models and DS patients, we define a new ontology-augmented phased in-context learning (PCL) approach to process 4935 full-text DS articles. We expand this analysis to 7 new drugs that demonstrate efficacy in reducing seizures to identify gaps in current knowledge for designing new experimental studies for drug discovery in DS. RESULTS: Few-shot or in-context learning is a foundational capability of LLMs and the few-shot learning capability of the Gemini 1.0 Pro version LLM dramatically increases when we augment prompts with the DS epilepsy ontology. The DS epilepsy ontology is the largest epilepsy and seizure ontology in clinical use that was developed by DS basic scientists and clinical neurologists. The ontology-augmented PCL prompt achieves 100% accuracy in reproducing the benchmark drug efficacy dataset for 17 ASMs with only two examples for in-context learning. CONCLUSION: The new ontology-augmented PCL approach significantly accelerates the few-shot learning capabilities of the Gemini LLM, thereby reducing the number of required examples and time needed to optimize LLMs for medical applications.

  PublisherDOI

Recent grants

• NIH Grant RC1NS069033

  NIH · $923k · 2012

• NIH Grant R01NS065434

  NIH · $887k · 2015

• NIH Grant R01NS044210

  NIH · $1.3M · 2009

• NIH Grant R01NS078331

  NIH · $2.0M · 2018

• NIH Grant R21NS062419

  NIH · $362k · 2011

Frequent coauthors

• H. Steve White

  University of Washington

  51 shared

• Cameron S. Metcalf

  University of Utah

  33 shared

• Peter J. West

  University of Utah

  23 shared

• Misty D. Smith

  University of Utah

  17 shared

• Marc A. Dichter

  17 shared

• Robert S. Fujinami

  15 shared

• John D. Rolston

  Brigham and Women's Hospital

  15 shared

• Kristina Johnson

  Harvard University

  11 shared

Labs

• Karen Wilcox LabPI

Education

• Ph.D., Pharmacology & Toxicology

  University of Utah

Awards & honors

• 2025 University of Utah Distinguished Researcher Award
• 2023 University of Utah Distinguished Professor Award
• 2022 NINDS Landis Outstanding Mentorship Award