Li-Huei Tsai

Our laboratory is interested in elucidating the pathogenic mechanisms underlying neurological disorders that impact learning and memory and to identify new therapeutic targets and candidate therapeutics. We study disease pathogenesis both on the cellular and on the systems level, using a multidisciplinary approach that combines molecular biology and electrophysiology with computational biology and bioengineering.

Systems-level analyses

In addition to investigating the molecular and cellular processes underlying Alzheimer’s disease and other neurodegenerative conditions, my lab also seeks to understand disease pathogenesis on the systems level, i.e., on the level of neuronal circuits or the entire brain. Our goals are to understand how neuronal activity is impacted by brain pathology, to reveal how it contributes to progression of pathologies through the brain, and to identify strategies for detecting and preventing this progression.

To this end, we use a wide range of methods in mouse models of disease, including classical electrophysiological techniques such as intra-cranial EEG and patch-clamping as well as real-time monitoring of biosensors through fiber optics or two-photon microscopy; targeted stimulation or silencing of neurons using opto- and chemogenetics; and rabies virus-mediated retrograde tracing of neurons and whole-brain clearing and light-sheet microscopy to visualize neuronal circuits. Using these methods on a mouse model of Alzheimer’s disease, we found that dense early amyloid-beta deposits in the mammillary body (MB), a subcortical node of the medial limbic circuit, correlate with neuronal hyper-excitability in lateral MB neurons. Attenuating this aberrant hyperactivity in lateral MB neurons ameliorated memory deficits in the Alzheimer’s mice, while inducing hyperactivity in lateral MB neurons in wild-type mice impaired their performance on memory tasks.

A recent focus in our lab has been the effects of non-invasive sensory stimulation on Alzheimer’s disease symptoms. Synchronized rhythmic activity of groups of neurons, producing the electrical oscillations detectable by electroencephalography (EEG), is believed to mediate long-range communication within the brain. Oscillations in the gamma frequency range of 30 Hertz (Hz) and above, which are decreased in both Alzheimer’s patients and Alzheimer’s mouse models, are thought to be particularly important for learning and memory. Some years ago, we found that increasing gamma oscillations through external stimuli pulsating at 40 Hz not only reduced pathologies such as amyloid plaques, tau tangles, and neuronal loss in the brain of Alzheimer’s model mice, but also led to their improved performance in tasks related to learning and memory. While we initially applied this gamma stimulation directly to parvalbumin-positive interneurons through use of optogenetics, similar effects could be achieved by non-invasive sensory stimulation with a flickering light and/or a clicking sound. Recording from intra-cranial electrodes showed that gamma oscillations were not only induced in the sensory cortices but could also be detected in the hippocampus and prefrontal cortex. We termed this non-invasive sensory stimulation “GENUS,” for Gamma ENtrainment Using Sensory stimuli.

In recent studies, we found that the benefits of GENUS may not be limited to Alzheimer’s disease but extend to other neurodegenerative disease. We showed that daily GENUS treatment over several weeks reduced neuroinflammation and demyelination observed in mouse models of chemotherapy-related cognitive impairment, known as “chemobrain,” and of cuprizone-induced demyelination, a model of multiple sclerosis.

Given the promising results in mouse models and the completely non-invasive nature of GENUS, clinical trials were quickly initiated. Several small clinical studies conducted by us and others showed GENUS to be safe and tolerable in humans and gave encouraging early treatment results such as less brain loss, improved sleep quality, and better performance on a memory test in patients with mild Alzheimer’s disease. Ongoing clinical trials in our lab and by others are exploring the effects of long-term GENUS treatment on cognitive performance as well as brain-imaging- and blood-based biomarkers.

While the mechanisms underlying the beneficial effects of GENUS have not yet been fully elucidated, we recently showed that amyloid reduction observed in Alzheimer’s mice after GENUS is at least partially mediated through enhancement of glymphatic clearance. Glymphatic clearance describes the “flushing out” of waste products from brain tissue by the directed flow of cerebrospinal fluid (CSF) from the perivascular spaces surrounding arterial brain vessels through the brain parenchyma into perivenous spaces and finally lymphatic vessels. This CSF flow is believed to be driven by arterial pulsation and supported by the water channel aquaporin-4 (AQP4) located on astrocytic endfeet surrounding the brain blood vessels. We showed that GENUS promoted arterial pulsatility, AQP4 polarization along astrocytic endfeet, influx of a CSF tracer into and its efflux from the cortex, and amyloid accumulation in cervical lymph nodes in a mouse model of Alzheimer’s disease. Pharmacological or genetic inhibition of AQP4 reduced CSF tracer influx into the brain and attenuated stimulation-mediated amyloid clearance.

Gene-level Studies

Epidemiological studies and genome-wide association studies (GWAS) of DNA variation with disease have identified numerous genetic risk variants for Alzheimer’s disease. These genetic risk factors provide powerful model systems for unraveling some of the molecular and cellular mechanisms underlying Alzheimer’s disease and for identifying therapeutic strategies tailored to a patient’s genetic makeup. After employing RNA sequencing to reveal cellular programs impacted by the risk factor, my lab uses sophisticated and highly tractable in vitro systems to dissect the underlying mechanisms.

iPSC-derived in vitro systems

Many of our in vitro systems are built from induced pluripotent stem cells (iPSC) derived from patients harboring a specific risk variant. These iPSC can be differentiated into different brain cell types, and isogenic control cells that share the exact genetic background and differ only at the risk position can be generated by precise DNA editing using CRISPR/Cas9 technology. Our in vitro models range from monocultures of individual cell types to complex co-culture systems modeling the brain vasculature, myelinated neuronal axons, or even a complete vascularized brain tissue. The modular assembly of these co-culture systems allows “mixing and matching” of brain cell types with and without a risk factor, so that the effect of risk variants in a specific cell type can be isolated. In addition, our iPSC-derived in vitro systems enable live imaging using fluorescent biosensors for real-time monitoring of molecular processes or efficient drug screening.

Cell-type specific effects of APOE4

The ε4 allele in the APOE gene is the strongest known genetic risk factor for non-familial Alzheimer’s disease. APOE codes for apolipoprotein E (ApoE), the major protein component of high-density lipoprotein particles in the brain that transport cholesteryl ester and fatty acids between cells. Our transcriptomic studies in postmortem samples from Alzheimer’s patients with and without the APOE4 genotype indicated that presence of APOE4 leads to expression differences for genes associated with lipid metabolism and storage. In follow-up studies using our iPSC-based in vitro systems, we found that presence of the APOE4 allele causes glial accumulation of lipid droplets, small intracellular organelles that store fatty acids in the form of triacylglycerides and cholesteryl ester. Interestingly, this lipid-burdened phenotype in glial cells had been described by Alzheimer himself in his first description of the disease named after him.

Candidate therapeutics for APOE4 effects

Using our iPSC-based in vitro systems, we also identified several different agents that could resolve this APOE4-associated lipid droplet accumulation and related effects in glial cells. In APOE4 astrocytes, supplementation with CDP-choline, a widely available nutritional supplement and the rate limiting intermediate in the synthesis of the membrane lipid phosphatidylcholine, resolved the lipid droplet accumulation. In APOE4 oligodendrocytes, treatment with cyclodextrin abolished both lipid droplet accumulation and myelination defects associated with a redistribution of cholesterol from myelin sheaths around neuronal axons to lipid droplets in the cell bodies. Cyclodextrin, a cyclic polysaccharide that is commonly used as a solubilizer in the food and drug industry, presumably normalizes myelination by APOE4 oligodendrocytes by promoting the redistribution of cholesterol within the cell. Treatment with cyclodextrin also improved learning and executive function in transgenic mice expressing human ApoE with the APOE4 allele, further supporting its therapeutic potential. In APOE4 microglia, lipid droplet accumulation is accompanied by an activated phenotype and a weakened response to neuronal cues mediated by the purinergic P2RY12 receptor. Inhibiting fatty acid synthesis with Triacsin C resolved this lipid droplet accumulation and restored microglial responsiveness to neuronal cues.

Cell-type specific effects of loss-of-function variants in ABCA7

Loss-of-function variants in ABCA7 are also associated with a strong genetic risk for non-familial Alzheimer’s disease. ABCA7 is a membrane protein that effluxes lipids to ApoE in lipoprotein particles and flips phospholipids from the inner to the outer leaflet of the plasma and other membranes. Our transcriptomic studies in postmortem samples from Alzheimer’s patients revealed a metabolic shift away from phosphatidylcholine synthesis towards triacylglyceride accumulation in neurons with ABCA7 loss of function. iPSC-derived neurons with ABCA7 loss of function indeed showed accumulation of lipid droplets and also displayed dysregulation of the mitochondrial membrane potential, indicating a decreased ability to metabolize lipids. Promoting phosphatidylcholine synthesis by supplementation with CDP-choline resolved lipid droplet accumulation and normalized the mitochondrial membrane potential. These findings indicate that lipid dysregulation is not limited to glial cells harboring the APOE4 allele but may play a wider role in Alzheimer’s pathogenesis. In addition, the findings further support the potential of choline supplementation as a therapeutic for Alzheimer’s disease.

Genome-level Analyses

My lab has a long-standing interest in understanding regulation of gene expression in the context of cognitive function. In earlier work, we identified the chromatin-modifier histone deacetylase 2 (HDAC2) as a master regulator of synaptic gene expression and showed that HDAC inhibitors enhance and restore learning and memory, demonstrating their therapeutic potential.

Some years ago, we turned to single-nucleus RNA sequencing (snRNAseq) of postmortem brain tissue to assemble a comprehensive atlas of cell-type specific gene expression changes associated with Alzheimer’s disease. Samples from both Alzheimer’s disease patients and age- and sex-matched cognitively normal controls were obtained from the longitudinal ROSMAP studies, which also collect rich metadata on participants. We performed snRNAseq on the postmortem prefrontal cortex from 392 individuals. For 48 individuals, we sequenced five additional brain regions known to be affected in early (entorhinal cortex), intermediate (hippocampus and anterior thalamus), or late disease stages (angular gyrus, midtemporal cortex, and prefrontal cortex). In addition, we also performed single-nucleus ATAC-seq on postmortem prefrontal cortex samples from 92 individuals. By performing clustering analysis of the RNAseq data and assigning clusters to brain cell types based on known cell-type specific marker genes, we were able to identify many distinct subtypes of inhibitory and excitatory neurons and delineate different microglial cell states. The resulting genome-level gene expression atlas has recently been published, providing a comprehensive resource for the medical, scientific, and pharmaceutical communities.

Identifying determinants of vulnerability and resilience to Alzheimer’s disease

Based on our comprehensive snRNAseq data, we were able to identify several neuronal subtypes that were depleted in samples with Alzheimer’s disease, indicating increased vulnerability. These vulnerable neuronal subclasses included three subtypes of somatostatin (SST) inhibitory neurons as well as Reelin-expressing subtypes of excitatory neurons in the entorhinal cortex. Notably, a gain-of-function variant in Reelin has previously been associated with resilience to familial Alzheimer’s disease.

The strongest association of gene expression with cognitive resilience to Alzheimer’s pathology was detected in astrocytes. Remarkably, several of the resilience-associated genes coded for proteins involved in regulating intracellular choline availability. This finding strongly supports our independent results from in vitro studies that indicate the potential of choline supplementation as a therapeutic strategy for AD.

Role of DNA damage in Alzheimer’s disease

Double-strand breaks (DSBs) resolve topological constraints on the DNA and are necessary for the rapid expression of early-response genes in neurons. However, while such physiological DSBs are typically repaired quickly, DSB accumulation is an early pathological hallmark of neurodegeneration. Based on snRNAseq on postmortem samples from human AD patients and the CK-p25 mouse model of neurodegeneration, we found that DSB accumulation leads to activation of antiviral inflammatory pathways and cellular senescence in neurons. Neurons with senescence gene signatures displayed increased occurrence of gene-fusions that disproportionately affected DSB hotspots such as highly expressed or long genes. snATACseq data from human postmortem samples further indicated a weakening of epigenomic cell identity – known as epigenomic erosion – and a break-down of epigenomic delineations between active and inactive genomic regions. The predicted disruption of the normal 3D genome organization was validated using the Hi-C technique for detecting long-range chromosomal interactions. Together, these finding strongly suggest a role for DNA damage accumulation and subsequent breakdown of chromosomal organization in Alzheimer’s pathogenesis.

Human microglial state dynamics in Alzheimer’s disease progression

Microglia are the resident immune cells of the brain. When activated, they secrete pro-inflammatory cytokines and become phagocytic. Since this normally protective response becomes damaging upon chronic activation in Alzheimer’s disease, chronically activated “disease associated microglia” (DAM) represent an important therapeutic target. While microglia are often poorly represented in snRNAseq data, our large dataset allowed us to identify 12 different microglial states including homeostatic, neuronal surveillance, lipid processing, phagocytic, stress related, glycolytic, and antiviral states as well as three different inflammatory states. Surprisingly, the lipid processing state showed the most significant correlation with Alzheimer’s pathology and cognitive decline, and only one of the inflammatory states was significantly increased in Alzheimer’s samples. We also identified a series of transcription factors whose expression promoted the homeostatic state or whose inhibition weakened the inflammatory state, providing a roadmap for the development of therapies aimed at curbing neuroinflammation with disease-stage specificity.