Citations

Abstract Microglia are essential for maintaining brain homeostasis, including the sleep-wake cycle, but the underlying mechanisms remain unclear. Here, using in vivo two-photon imaging in both head-fixed and freely moving mice with simultaneous electroencephalogram–electromyogram (EEG–EMG) recording, we investigated microglial process dynamics and their interactions with locus coeruleus (LC) noradrenergic (NE) axons in the frontal cortex across sleep-wake states. During sleep or chemogenetic suppression of LC neurons, microglia enhanced their surveillance and increased interactions with NE axons. Correlative electron microscopy at nanometer resolution identified structural contacts between microglial process endings and NE boutons. Mechanistically, microglia-bouton interaction is primed by NE–β2-adrenergic receptor (β2AR) signaling, leading to enhanced Ca²⁺ activity in NE axons. These findings suggest that microglial interactions with NE axons promote cortical norepinephrine transmission and maintain stable wakefulness.

Abstract Error analysis is a critical step in evaluating and improving clinical concept extraction models, the most common clinical natural language processing (NLP) task. Unlike corpus annotation, which follows standardized protocols for creating gold-standard datasets, error analysis requires nuanced judgment grounded in both clinical expertise and NLP knowledge. This is especially important given the heterogeneity of clinical text, where variations in documentation style, note structure, and terminology can substantially influence model behavior. Despite its importance, there is currently no standardized, user-level framework to support systematic error analysis in clinical concept extraction task. In this study, we developed and validated MedError, a machine-assisted, human-in-the-loop framework designed to standardize and enhance error analysis for clinical concept extraction tasks. We collected and manually curated a corpus of 1,187 unique errors from a total of 4,237 notes across three different distinct hospitals. The error categories were defined using our previously validated error taxonomy and included 480 false negatives and 707 false positives across 25 error types and 48 clinical concept categories. We evaluated the performance of three proprietary and three open-source large language models (LLMs) in automatically classifying these errors into 26 and 15 predefined categories. We further developed a machine-assisted framework, MedError, which integrates best practices in error analysis, LLM-assisted classification and reasoning, and a user-friendly interface to enable more efficient, reproducible, and context-aware error analysis. The framework supports both single-site and federated multisite error analysis, facilitating the effective translation of clinical NLP systems into real-world settings.

Background: Acute disseminated encephalomyelitis (ADEM) is a demyelinating syndrome observed in myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD) but there are limited data on application of this definition to its cerebral attacks or in adults with this disease. Methods: We analyzed patients from the Mayo Clinic MOGAD cohort ( n = 263) with cerebral attacks. We assessed whether children (<18 years-old) and adults (⩾18 years-old) fulfilled the 2013 ADEM definition. Results: We identified 89 patients (children, 34; adults, 55) and 55/89 (62%) were of female sex. Acute magnetic resonance imaging (MRI) abnormalities included: poorly demarcated T2 lesions (white matter, 85 (96%); deep gray matter, 41 (46%); T1-hypointense lesions (42 (47%)). Forty-two patients did not fulfill the ADEM definition due to: monofocal syndrome, 9; or no encephalopathy, 33. The remaining 47 patients (53%: children, 27/34 (79%); adults, 20/55 (36%)) fulfilled the ADEM definition, and 14 had preceding attacks, and 34 had subsequent attacks relative to ADEM episode. The features in children and adults with ADEM were broadly comparable. Conclusion: Applying the ADEM definition to MOGAD cerebral attacks highlighted that an update of the 2013 ADEM definition would be beneficial and could explicitly state its application to both adults and children and allow it to be applied to attacks at any time during the disease course.

Objective To assess integration of a single-use robotic-assisted cochlear implant (RA-CI) electrode array insertion system into surgical workflows, identify potential challenges, and inform best practices. Study Design Survey. Setting Survey data from 16 surgeons across 10 US hospitals. Patients CI candidates 12 years and older with radiographically normal cochleae. Intervention CI surgery with single-use RA-CI insertion system during electrode array insertion. Main Outcome Measures Number of RA-CI to achieve proficiency, estimated additional time for RA-CI, and best practices to successfully incorporate the RA insertion system into CI surgery. Results A total of 121 RA-CI cases were performed. All surgeons completed the survey. Most surgeons (62.5%) reported they would feel comfortable using the device within five cases. Eight (50%) reported use of the RA system added 5 to 10 minutes to the case, and seven (44%) reported an added 10–15 minutes. Providing adequate incision size to secure the unit base superior to the temporal line and maximizing exposure of the facial recess were the most recommended best practices. Additional recommendations included pulling the receiver/stimulator forward within the pocket to facilitate a full electrode array insertion and considering drive head placement and mastoidectomy size to maximize overall visibility during electrode array insertion. Conclusions Proficiency with an RA electrode array insertion system during CI surgery can be achieved with minor adjustments to the surgical approach, within relatively few use cases (generally <5) and with minimal time addition (between 5 and 15 min). Training to specific best practices is important before initial clinical use.

Abstract Intrinsic resistance to targeted therapeutics in PTEN-deficient glioblastoma (GBM) is mediated by redundant signaling networks that sustain critical metabolic functions. Here, we demonstrate that coordinated inhibition of the ribosomal protein S6 kinase 1 (S6K1) and the receptor tyrosine kinase AXL using LY-2584702 and BMS-777607 can overcome network redundancy to reduce GBM tumor growth. This combination of S6K1 and AXL inhibition suppressed glucose flux to pyrimidine biosynthesis. Genetic inactivation studies to map the signaling network indicated that both S6K1 and S6K2 transmit growth signals in PTEN-deficient GBM. Kinome-wide ATP binding analysis in inhibitor-treated cells revealed that LY-2584702 directly inhibited S6K1, and substrate phosphorylation studies showed that BMS-777607 inactivation of upstream AXL collaborated to reduce S6K2-mediated signal transduction. Thus, combination targeting of S6K1 and AXL provides a kinase-directed therapeutic approach that circumvents signal transduction redundancy to interrupt metabolic function and reduce growth of PTEN-deficient GBM. Significance: Therapy for glioblastoma would be advanced by incorporating molecularly targeted kinase-directed agents, similar to standard of care strategies in other tumor types. Here, we identify a kinase targeting approach to inhibit the metabolism and growth of glioblastoma.

Gene fusions involving tumor protein p63 gene (TP63) occur in multiple T and B cell lymphomas and portend a dismal prognosis for patients. The function and mechanisms of TP63 fusions remain unclear, and there is no target therapy for patients with lymphoma harboring TP63 fusions. Here, we show that TP63 fusions act as bona fide oncogenes and are essential for fusion-positive lymphomas. Transgenic mice expressing TBL1XR1::TP63, the most common TP63 fusion, develop diverse lymphomas that recapitulate multiple human T and B cell lymphomas. Here, we identify that TP63 fusions coordinate the recruitment of two epigenetic modifying complexes, the nuclear receptor corepressor (NCoR)—histone deacetylase 3 (HDAC3) by the N-terminal TP63 fusion partner and the lysine methyltransferase 2D (KMT2D) by the C-terminal TP63 component, which are both required for fusion-dependent survival. TBL1XR1::TP63 localization at enhancers drives a unique cell state that involves up-regulation of MYC and the polycomb repressor complex 2 (PRC2) components EED and EZH2. Inhibiting EZH2 with the therapeutic agent valemetostat is highly effective at treating transgenic lymphoma murine models, xenografts, and patient-derived xenografts harboring TP63 fusions. One patient with TP63 -rearranged lymphoma showed a rapid response to valemetostat treatment. In summary, TP63 fusions link partner components that, together, coordinate multiple epigenetic complexes, resulting in therapeutic vulnerability to EZH2 inhibition.