Biology subjects

Schrag, M. S.

Publications and source records attributed to Schrag, M. S..

7 recordsLinked to original sources

bioRxiv2026

Serpin A3 is Increased in Cerebral Amyloid Angiopathy and ARIA

Vascular pathology is an important contributor to dementia, and astrocyte dysfunction is increasingly recognized as a key mediator of cerebrovascular injury. Serpin A3 is a well-known marker of reactive astrocytes, yet its role in vascular degeneration in cerebral amyloid angiopathy and Alzheimers disease remains poorly defined. We combined analyses of single-nucleus transcriptomic datasets with western blotting and quantitative two- and three-dimensional microscopy of human brain tissue to characterize Serpin A3 expression, distribution, cellular sources, and neuropathological correlates in the human brain. Serpin A3 protein abundance was increased in cases with cerebral amyloid angiopathy and Alzheimers disease compared to neurological controls, both by western blot and quantitative microscopy. The tissue distribution of Serpin A3 was heterogeneous and was predominantly within astrocytes associated with degenerating arterioles. Serpin A3 was also detected around parenchymal beta-amyloid plaques. In a postmortem case of active and severe lecanemab-associated amyloid-related imaging abnormalities, light sheet microscopy revealed prominent Serpin A3 immunoreactivity around inflamed, dilated vessels and microhemorrhages, with Serpin A3 levels higher than those found in the Cerebral Amyloid Angiopathy cases. These findings identify Serpin A3 as a marker of cerebrovascular injury and astrocyte-driven inflammatory responses in cerebral amyloid angiopathy.

neuroscience↗

bioRxiv2026

A geometric-to-neural cascade for cerebral microbleed detection in susceptibility-weighted MRI

Cerebral microbleeds (CMBs) are established imaging biomarkers of cerebral small vessel disease and are a defining feature of cerebral amyloid angiopathy (CAA), yet their automated detection in susceptibility-weighted imaging (SWI) remains challenging due to high false-positive rates from vessel cross-sections, iron and calcium deposits, and other hypointense mimics. We present a fully automated, three-stage cascade pipeline that combines subject-adaptive unsupervised candidate generation with two successive lightweight 3D ResNet classifiers, trained with only human-in-the-loop quality-assurance (QA) labels (yes/no per candidate) rather than dense voxel-wise segmentation masks. The candidate generation stage is performed by fitting a Gaussian Mixture Model (GMM) to each subjects SWI intensity histogram to define an adaptive low-intensity threshold, followed by anatomical masking to exclude physiologically irrelevant regions (image edges, ventricles/CSF/choroid plexus, and cerebellum), and filters candidates by size and sphericity. The model was trained and evaluated across a nested 3x5-fold cross-validation on N = 30 subjects from a publicly available labeled microbleed dataset and a CAA cohort (11,424 CMB candidate lesions) with data augmentation during training. Stage A classifies all geometric candidates as CMB or non-CMB and Stage B refines the predicted positives to suppress false positives (cascade AUC = 0.9587, sensitivity = 0.712, specificity = 0.975, PPV = 0.676, F1 = 0.693). The cascade reduces Stage A false positives by 76.8% (888/1,157 false positives eliminated) while retaining competitive sensitivity. Inference was performed on 141 SWI scans, detecting a mean 40.3 CMBs per scan and being preferred for use in 85% of high CMB cases, as evaluated by a blinded neurologist. The inference pipeline outputs binary CMB segmentation NIfTI images and radiologist-ready QA visualizations.

bioinformatics↗

bioRxiv2026

Multimodal Molecular Mapping of the Vasculature in Human Cortex Reveals Lipid Markers of Cerebral Amyloid Angiopathy

Cerebral amyloid angiopathy (CAA) commonly co-occurs with Alzheimers disease (AD), yet the molecular changes that accompany vascular {beta}-amyloid deposition in human tissue remain incompletely defined. Herein, we use a novel imaging approach that combines matrix-assisted laser desorption/ionization imaging mass spectrometry (IMS) with immunofluorescence microscopy on the same sections of postmortem human frontal cortex to map the lipid microenvironment of leptomeningeal vasculature in cases with and without CAA. Autofluorescence-guided regions-of-interest were imaged by IMS in both negative and positive ion modes and registered to post-IMS-acquired microscopy images. Immunofluorescence microscopy using collagen IV, -smooth muscle actin (SMA), and thiazine red enabled automated segmentation of total, amyloidpositive, and amyloid-negative vasculature regions. A CAA index, the ratio of amyloid-positive area to total vasculature area in a region imaged by IMS, was used to define vasculature and classify each case into having CAA, or CAA-present and not having CAA, or CAA-absent. An interpretable machine learning approach (XGBoost models with Shapley additive explanations for interpretation) was trained on pixel-level spectra and identified lipid signatures of vascular identity shared across groups as well as class-specific marker candidates that distinguished CAA-present from CAA-absent vasculature. CAA-present vessels were enriched for gangliosides (e.g., GM1), whereas CAA-absent vessels were characterized by higher contributions from phosphatidylserines (e.g., long-chain polyunsaturated PS species). Univariate differences were inconsistent between the two groups, but multivariate models in negative mode yielded stable discriminatory features. These results define spatial lipid correlates of vascular amyloid pathology in the human brain and establish a multimodal framework for mechanistically linking lipid metabolism, vascular integrity, and CAA in AD.

neuroscience↗

bioRxiv2025

Spatial Dependence and Heterogeneity in Molecular Imaging: Moran Quadrant Maps Enable Advanced Spatial-Statistical Analysis

Multiplexed molecular imaging, such as imaging mass spectrometry, provides spatially-contextualized insights that are transforming biomedical research by advancing our understanding of tissue organization and disease mechanisms. However, the spatial-statistical properties of molecular imaging data are often underutilized, and scalable computational tools to analyze them are lacking. Here, we demonstrate how local and global spatial autocorrelation (SAC) metrics can be used to quantify spatial dependence and spatial heterogeneity, properties that violate the common assumption of independent and identically distributed measurements. We develop mathematically rigorous methods for SAC-based exploratory analysis. Furthermore, we introduce a novel spatial feature extractor, the Moran quadrant map, and develop two advanced workflows based on it: Moran-Felsenszwalb segmentation for tissue domain segmentation and Moran-HOG clustering for colocalization-based image analysis. Finally, our open-source Moran Imaging toolbox (https://github.com/vandeplaslab/Moran_Imaging) provides scalable Python implementations, including a novel parallelized spatial lag algorithm, unlocking SAC-based analyses and biological insights for large-scale imaging.

bioinformatics↗

bioRxiv2025

Amyloid-β-regulated gene circuits for programmable Alzheimer's disease therapy

Alzheimer's disease (AD) is a neurodegenerative disease characterized in part by the accumulation of the protein amyloid-{beta} (A{beta}). Monoclonal antibodies (mAbs) that target A{beta} for clearance from the brain have received FDA approval; however, these therapies are accompanied by serious side effects, and their cognitive benefit for patients remains of tremendous debate. Here, we present a potential engineered cell therapy for AD in which we enlist cells of the central nervous system as programmable agents for sculpting the neurodegenerative niche toward one that mitigates glial reactivity and neuronal loss. We constructed a suite of A{beta}-sensitive synthetic Notch (synNotch) receptors from clinically tested anti-A{beta} mAbs and show that cells expressing these receptors can recognize synthetic A{beta}42 and A{beta}40 with differential sensitivity. We express these receptors in astrocytes, cells native to the brain that are known to become dysfunctional in AD. These synNotch astrocytes, which upregulate selected transgenes upon exposure to synthetic and human brain-derived amyloid, were engineered to express potential therapeutic transgenes in response to A{beta}, including brain-derived neurotrophic factor and antagonists of the cytokines tumor necrosis factor and interleukin-1. SynNotch astrocytes that express such antagonists in response to A{beta} partially attenuate a cytokine-induced reactive astrocyte phenotype and promote barrier properties in brain microvascular endothelial cells. Additionally, engineered A{beta}-synNotch cells potently upregulate transgene expression in response to A{beta} deposited in the 5xFAD mouse brain, demonstrating the capacity to recognize A{beta} in situ. Overall, our work supports A{beta}-synNotch receptors as promising tools to generate a cell-based therapy for AD with targeted functionalities to positively influence the AD niche.

bioengineering↗

bioRxiv2021

CRISPRi screens in human astrocytes elucidate regulators of distinct inflammatory reactive states

In response to central nervous system injury or disease, astrocytes become reactive, adopting context-dependent states and functional outputs. Certain inflammatory insults induce reactive astrocytes that lose homeostatic functions and gain harmful outputs through cellular pathways that are not fully understood. Here, we combined single-cell transcriptomics with CRISPRi screening in human iPSC-derived astrocytes to systematically interrogate inflammatory astrocyte reactivity. We found that autocrine-paracrine IL-6 and interferon signaling downstream of canonical NF-{kappa}B activation drove two distinct inflammatory reactive signatures - one promoted by and the other inhibited by STAT3. These signatures overlapped with those observed in other experimental contexts, including mouse models, and their markers were upregulated in the human brain in Alzheimers disease and hypoxic ischemic encephalopathy. Furthermore, we validated that these signatures were regulated by Stat3 in vivo. These results and the platform we established have the potential to guide the development of therapeutics to selectively modulate different aspects of inflammatory astrocyte reactivity.

neuroscience↗

bioRxiv2019

PLD3 is a neuronal lysosomal phospholipase D associated with β-amyloid plaques and cognitive function in Alzheimer’s disease

Phospholipase D3 (PLD3) is a protein of unclear function that structurally resembles other members of the phospholipase D superfamily. A coding variant in this gene confers increased risk for the development of Alzheimers disease (AD), although the magnitude of this effect has been controversial. Because of the potential significance of this obscure protein, we undertook a study to determine whether PLD3 is relevant to memory and cognition in sporadic AD, to observe its distribution in normal human brain and AD-affected brain, to describe its subcellular localization, and to evaluate its molecular function. PLD3 mRNA levels in the human pre-frontal cortex inversely correlated with {beta}-amyloid pathology severity and rate of cognitive decline in 531 participants enrolled in the Religious Orders Study and Rush Memory and Aging Project. PLD3 levels across genetically diverse BXD mouse strains and strains crossed with 5xFAD mice correlated strongly with learning and memory performance in a fear conditioning task. In human neuropathological samples, PLD3 was primarily found within neurons and colocalized with lysosome markers (LAMP2, progranulin, and cathepsins D and B). This colocalization was also present in AD brain with prominent enrichment on lysosomal accumulations within dystrophic neurites surrounding {beta}-amyloid plaques. This pattern of protein distribution was conserved in mouse brain in wild type and the 5xFAD mouse model of cerebral {beta}-amyloidosis. We discovered PLD3 has phospholipase D activity in lysosomes. A coding variant in PLD3 reported to confer AD risk significantly reduced enzymatic activity compared to wild-type PLD3. In summary, this study identified a new functional mammalian phospholipase D isoform which is lysosomal and closely associated with both {beta}-amyloid pathology and cognition.

neuroscience↗