Faculty Bibliography

Neuronal oscillations at about 10 Hz, called alpha oscillations, are often thought to arise from synchronous activity across the occipital cortex and are usually largest when the cortex is inactive. However, recent studies measuring visual receptive fields have reported that local alpha power increases when cortex is excited by visual stimulation. This contrasts with the expectation that alpha oscillations are associated with cortical inactivity. Here, we used intracranial electrodes in human patients to measure alpha oscillations in response to visual stimuli whose location varied systematically across the visual field. We hypothesized that stimulus-driven local increases in alpha power result from a mixture of two effects: a reduction in alpha oscillatory power and a simultaneous increase in broadband power. To test this, we implemented a model to separate these components. The two components were then independently fit by population receptive field (pRF) models. We find that the alpha pRFs have similar center locations to pRFs estimated from broadband power but are several times larger and exhibit the opposite effect: alpha oscillatory power decreases in response to stimuli within the receptive field, reinforcing the link between alpha oscillations and cortical inactivity, whereas broadband power increases. The results demonstrate that alpha suppression in the human visual cortex can be precisely tuned, but that to measure these effects, it is essential to separate the oscillatory signal from broadband power changes. Finally, we show how the large size and the negative valence of alpha pRFs can explain key features of exogenous visual attention.

Sentence production is the uniquely human ability to transform complex thoughts into strings of words. Despite the importance of this process, language production research has primarily focused on single words. It remains a largely untested assumption that the principles of word production generalize to more naturalistic utterances like sentences. Here, we investigate this using high-resolution neurosurgical recordings (ECoG) and an overt production experiment where ten patients produced six words in isolation (picture naming) and in sentences (scene description). We trained machine learning classifiers to identify the unique brain activity patterns for each word during picture naming, and used these patterns to decode which words patients were processing while they produced sentences. Our findings confirm that words share cortical representations across tasks, but reveal a division of labor within the language network. In sensorimotor cortex, words were consistently activated in the order in which they were said in the sentence. However, in prefrontal cortex, the order in which words were processed depended on the syntactic structure of the sentence. In non-canonical sentences (passives), we further observed a spatial code for syntactic roles, with subjects selectively encoded in inferior frontal gyrus (IFG) and objects selectively encoded in middle frontal gyrus (MFG). We suggest that these complex dynamics of prefrontal cortex may impose a subtle pressure on language evolution, potentially explaining why nearly all the world's languages position subjects before objects.

We introduce an intracranial EEG (iEEG) dataset collected as part of an adversarial collaboration between proponents of two theories of consciousness: Global Neuronal Workspace Theory and Integrated Information Theory. The data were recorded from 38 patients undergoing intracranial monitoring of epileptic seizures across three research centers using the same experimental protocol. Participants were presented with suprathreshold visual stimuli belonging to four different categories (faces, objects, letters, false fonts) in three orientations (front, left, right view), and for three durations (0.5, 1.0, 1.5 s). Participants engaged in a non-speeded Go/No-Go target detection task to identify infrequent targets with some stimuli becoming task-relevant and others task-irrelevant. Participants also engaged in a motor localizer task. The data were checked for its quality and converted to Brain Imaging Data Structure (BIDS). The de-identified dataset contains demographics, clinical information, electrode reconstruction, behavioral performance, and eye-tracking data. We also provide code to preprocess and analyze the data. This dataset holds promise for reuse in consciousness science and vision neuroscience to answer questions related to stimulus processing, target detection, and task-relevance, among many others.

OBJECTIVE:New-onset refractory status epilepticus (NORSE) occurs in people without pre-existing epilepsy or a rapidly identified structural, toxic, metabolic, or other cause. NORSE is a rare disorder with high morbidity and mortality rates and limited evidence for effective therapies. We aimed to assess whether the gut microbiome of NORSE and status epilepticus (SE) differs from that of chronic epilepsy, whether NORSE differs from SE at different disease time points, and to examine the correlations between specific gut microbiota and cytokines in NORSE and SE. METHODS:This longitudinal cohort study observed patients with NORSE (n = 15), SE (n = 17), and chronic epilepsy who were not in SE (n = 12). NORSE patients were recruited through the NORSE Consortium. Patients with NORSE and SE underwent longitudinal serial biospecimen collection. Fecal samples were subjected to whole-community shotgun metagenomics to characterize microbiome features. Cohorts were evaluated for prokaryotic, eukaryotic, and functional diversity. Correlations between blood inflammatory cytokine levels and microbiome features and covariate analysis with critical illness and clinical treatments were examined for NORSE and SE patients during and after SE resolution. RESULTS:During SE, NORSE and SE patients had significantly different prokaryotic, eukaryotic, and functional microbiome levels compared to chronic epilepsy patients without SE. Limited microbiome differences were observed within and between NORSE and SE, although these groups displayed differing correlation patterns between microbial species and cytokines. Patients who later died or were tube-fed harbored significantly different microbiomes than those who survived or were orally fed. SIGNIFICANCE/CONCLUSIONS:NORSE and SE patients present with a more variable and dramatically different fecal microbiome than chronic epilepsy patients, which may indicate gut dysbiosis that may be reciprocally linked to inflammatory responses. Although NORSE and SE patients had similar microbiome structures, fungal and bacterial correlates with inflammatory cytokines differed between NORSE and SE, with confounding factors influencing microbiome structure. Our data suggest a microbiome-specific response to NORSE and SE, with implications for future treatment strategies.

The ability to connect the form and meaning of a concept, known as word retrieval, is fundamental to human communication. While various input modalities could lead to identical word retrieval, the exact neural dynamics supporting this process relevant to daily auditory discourse remain poorly understood. Here, we recorded neurosurgical electrocorticography (ECoG) data from 48 patients and dissociated two key language networks that highly overlap in time and space, critical for word retrieval. Using unsupervised temporal clustering techniques, we found a semantic processing network located in the middle and inferior frontal gyri. This network was distinct from an articulatory planning network in the inferior frontal and precentral gyri, which was invariant to input modalities. Functionally, we confirmed that the semantic processing network encodes word surprisal during sentence perception. These findings elucidate neurophysiological mechanisms underlying the processing of semantic auditory inputs ranging from passive language comprehension to conversational speech.

Presence of focal to bilateral tonic-clonic seizures (FBTCS) in focal epilepsy is associated with increased morbidity and mortality. Risk factors for FBTCS are poorly understood, and little is known regarding FBTCS recurrence after treatment initiation. This study aimed to investigate factors related to FBTCS in newly diagnosed focal epilepsy and their recurrence after starting antiseizure medications (ASMs) in the Human Epilepsy Project (HEP) cohort. HEP was an international, prospective cohort study that enrolled people with newly diagnosed focal epilepsy within 4 months of treatment initiation and followed them for up to 6 years. Baseline characteristics, treatment choices, and seizure outcomes were collected. Descriptive and inferential statistical analysis was conducted to assess the differences between study participants who had FBTCS and those who never experienced FBTCS. A total of 443 participants were included in this analysis; 77% (n = 342) had FBTCS at some point prior to or within the study period. In participants with FBTCS, regardless of initial seizure type, diagnosis was mostly made after FBTCS (335/342, 98%). After treatment initiation, FBTCS did not recur in 57% (n = 194/342) of cases. A higher number of total pretreatment seizures (median = 16 vs. 11, p = .048, Mann-Whitney U-test), predominantly focal aware seizures (FAS) or focal impaired awareness seizures (FIAS; median = 15 vs. 10, p = .049, Mann Whitney U-test), was associated with no recurrence in FBTCS after treatment initiation. Of 108 participants without FBTCS prior to treatment, only seven (6%) developed FBTCS after treatment initiation. There was no significant difference in choice of initial ASM class (levetiracetam vs. sodium channel blockers) between participants who experienced FBTCS and those who did not. This study highlights the significance of FBTCS among individuals with newly diagnosed focal epilepsy. The majority of participants who experienced FBTCS were diagnosed with epilepsy after experiencing their first FBTCS despite preceding FAS/FIAS. The more frequent FAS/FIAS in participants whose FBTCS resolved may be a characteristic of their epilepsy.

The cerebellum, a subcortical structure, is traditionally linked to sensorimotor integration and coordination, although its role in cognition and affective behavior, as well as epilepsy, is increasingly recognized. Cerebellar dysfunction in patients with epilepsy can result from genetic disorders, antiseizure medications, seizures, and seizure-related trauma. Impaired cerebellar function, regardless of cause, can cause ataxia (imbalance, impaired coordination, unsteady gait), tremor, gaze-evoked nystagmus, impaired slow gaze pursuit and saccade accuracy, as well as speech deficits (slurred, scanning, or staccato). We explore how cerebellar dysfunction can contribute to epilepsy, reviewing data on genetic, infectious, and neuroinflammatory disorders. Evidence of cerebellar dysfunction in epilepsy comes from animal studies as well as human neuropathology and structural magnetic resonance imaging (MRI), functional and diffusion tensor MRI, positron emission and single photon emission computerized tomography, and depth electrode electro-encephalography studies. Cerebellar lesions can infrequently cause epilepsy, with focal motor, autonomic, and focal to bilateral tonic-clonic seizures. Antiseizure medication-resistant epilepsy typically presents in infancy or before age 1 year with hemifacial clonic or tonic seizures ipsilateral to the cerebellar mass. Lesions are typically asymmetric benign or low-grade tumors in the floor of the fourth ventricle involving the cerebellar peduncles and extending to the cerebellar hemisphere. Electrical stimulation of the cerebellum has yielded conflicting results on efficacy, although methodological issues confound interpretation. Epilepsy-related comorbidities including cognitive and affective disorders, falls, and sudden unexpected death in epilepsy may also be impacted by cerebellar dysfunction. We discuss how cerebellar dysfunction may drive seizures and how genetic epilepsies, seizures and seizure therapies may drive cerebellar dysfunction, and how our understanding of epilepsy-related comorbidities through basic neuroscience, animals models and patient studies can advance our understanding and improve patient outcomes.

OBJECTIVE:We analyzed the long-term safety and effectiveness of fenfluramine (FFA) in patients with Dravet syndrome (DS) in an open-label extension (OLE) study after participating in randomized controlled trials (RCTs) or commencing FFA de novo as adults. METHODS:Patients with DS who participated in one of three RCTs or were 19 to 35 years of age and started FFA de novo were included. Key endpoints were: incidence of treatment-emergent adverse events (TEAEs) in the safety population, and median percentage change in monthly convulsive seizure frequency (MCSF) from the RCT baseline to end of study (EOS) in the modified intent-to-treat (mITT) population. Post hoc analyses compared effectiveness in patients on concomitant stiripentol (STP) vs those not taking STP, and assessed safety (TEAEs) and effectiveness (Clinical Global Impression-Improvement [CGI-I] scale ratings) in patients enrolled as adults. RESULTS:A total of 374 patients, including 45 adults, received ≥1 FFA dose. Median FFA exposure was 824 days (range, 7-1280). TEAEs occurring in ≥10% of patients were pyrexia, nasopharyngitis, decreased appetite, seizure, decreased blood glucose, diarrhea, abnormal echocardiography (only physiologic regurgitation), upper respiratory tract infection, influenza, vomiting, and ear infection; no valvular heart disease or pulmonary arterial hypertension was observed over the OLE. In the mITT population (n = 324), median percentage change in MCSF from baseline to EOS was -66.8% (p < .001). The post hoc analyses of MCSF change from baseline to EOS in patients on concomitant STP (n = 75) was -36.2% vs -71.6% in those not on concomitant STP (n = 234) (p < .0001). In adult patients, 29 of 41 (70.7%) and 29 of 42 patients (69.1%) demonstrated clinically meaningful improvement on CGI-I at last visit as rated by caregivers and investigators, respectively. SIGNIFICANCE/CONCLUSIONS:Our OLE study of FFA in patients with DS confirmed previous positive findings and extended the exposure up to 3.5 years. No new or unexpected safety signals were observed and FFA demonstrated sustained and clinically meaningful reduction in MCSF.

Altered protein conformation can cause incurable neurodegenerative disorders. Mutations in SERPINI1, the gene encoding neuroserpin, can alter protein conformation resulting in cytotoxic aggregation leading to neuronal death. Familial encephalopathy with neuroserpin inclusion bodies (FENIB) is a rare autosomal dominant progressive myoclonic epilepsy that progresses to dementia and premature death. We developed HEK293T and induced pluripotent stem cell (iPSC) models of FENIB, harboring a patient-specific pathogenic SERPINI1 variant or stably overexpressing mutant neuroserpin fused to GFP (MUT NS-GFP). Here, we utilized a personalized adenine base editor (ABE)-mediated approach to correct the pathogenic variant efficiently and precisely to restore neuronal dendritic morphology. ABE-treated MUT NS-GFP cells demonstrated reduced inclusion size and number. Using an inducible MUT NS-GFP neuron system, we identified early prevention of toxic protein expression allowed aggregate clearance, while late prevention halted further aggregation. To address several challenges for clinical applications of gene correction, we developed a neuron-specific engineered virus-like particle to optimize neuronal ABE delivery, resulting in higher correction efficiency. Our findings provide a targeted strategy that may treat FENIB and potentially other neurodegenerative diseases due to altered protein conformation such as Alzheimer's and Huntington's diseases.

This study introduces a unified computational framework connecting acoustic, speech and word-level linguistic structures to study the neural basis of everyday conversations in the human brain. We used electrocorticography to record neural signals across 100 h of speech production and comprehension as participants engaged in open-ended real-life conversations. We extracted low-level acoustic, mid-level speech and contextual word embeddings from a multimodal speech-to-text model (Whisper). We developed encoding models that linearly map these embeddings onto brain activity during speech production and comprehension. Remarkably, this model accurately predicts neural activity at each level of the language processing hierarchy across hours of new conversations not used in training the model. The internal processing hierarchy in the model is aligned with the cortical hierarchy for speech and language processing, where sensory and motor regions better align with the model's speech embeddings, and higher-level language areas better align with the model's language embeddings. The Whisper model captures the temporal sequence of language-to-speech encoding before word articulation (speech production) and speech-to-language encoding post articulation (speech comprehension). The embeddings learned by this model outperform symbolic models in capturing neural activity supporting natural speech and language. These findings support a paradigm shift towards unified computational models that capture the entire processing hierarchy for speech comprehension and production in real-world conversations.