Amyotrophic lateral sclerosis and frontotemporal dementia mutation reduces endothelial TDP-43 and causes blood-brain barrier defects

Abstract

Mutations in the TARDBP gene encoding TDP-43 protein are linked to loss of function in neurons and familial frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS). We recently identified reduced nuclear TDP-43 in capillary endothelial cells (ECs) of donors with ALS-FTD. Because blood-brain barrier (BBB) permeability increases in ALS-FTD, we postulated that reduced nuclear TDP-43 in ECs might contribute. Here, we show that nuclear TDP-43 is reduced in ECs of mice with an ALS-FTD-associated mutation in TDP-43 (Tardbp^G348C) and that this leads to cell-autonomous loss of junctional complexes and BBB integrity. Targeted excision of TDP-43 in brain ECs recapitulates BBB defects and loss of junctional complexes and ultimately leads to fibrin deposition, gliosis, phospho-Tau accumulation, and impaired memory and social interaction. Transcriptional changes in TDP-43-deficient ECs resemble diseased brain ECs. These data show that nuclear loss of TDP-43 in brain ECs disrupts the BBB and causes hallmarks of FTD.

Fig. 1.. BBB disruption astrogliosis and microgliosis in Tardbp^G348C/+ mice.

(A) A schematic illustration of the assay for measuring BBB permeability. mo, months; iv, intravenous; 15′ circulation, 15 min of circulation. (B) The quantification process involved homogenizing brain tissue samples from 3-month-old wild-type (n = 3) and Tardbp^G348C/+ mice (n = 3) and 10- to 11-month-old wild-type Tardbp^+/+ mice (n = 8) and their heterozygous littermates, Tardbp^G348C/+ mice (n = 15), followed by measuring fluorescence intensity at 590 nm. n.s., not significant. Representative images of (C) 3-kDa Texas Red–dextran leakage, (E) NHS-biotin, (G) glial fibrillary acidic protein (GFAP) staining of astrocytes, and (I) Iba1 staining of microglia in the mouse brain cortex reveal consistent results across Tardbp^+/+ mice (n = 3) and Tardbp^G348C/+ mice (n = 3). [(B), D, and F] Quantification of data, with each data point representing the fluorescence image intensity in one image. (H and J) Field of view (FoV) is 0.16 mm². Quantification of data with each data point representing the number of activated cells in an image, with multiple images per mouse. Scale bars, 50 μm. Data are presented as means ± SEM. Statistical analysis was conducted using an unpaired Mann-Whitney test, with significance levels indicated as follows: ***P < 0.001; ****P < 0.0001.

Fig. 2.. Defects in permeability and cell junctions in Tardbp^G348C/+ ECs.

(A) Schematic illustration of the isolation and purification of ECs. (B) Immunofluorescence image of cells on transwell filter at endpoint of in vitro permeability experiment. Color of phalloidin is preferentially enhanced in G348C image to show cell coverage. Scale bars, 50 μm. (C) Passage of 10-kDa fluorescein isothiocyanate (FITC)–dextran dye across confluent monolayers of ECs isolated from Tardbp^+/+ (n = 6) and Tardbp^G348C/+ (n = 6), where each replicate indicates cells isolated from a separate mouse with age matched controls. Dye leak is normalized to wild type (WT) at 120 min. KI, knock-in. Statistical analysis was conducted using two-way analysis of variance (ANOVA) with repeated measurements. (D, G, and I) Representative images of mouse brain ECs isolated from 3-month-old wild-type Tardbp^+/+ (n = 3) and their heterozygous littermates, Tardbp^G348C/+ mice (n = 3), immunostained with antibodies to the indicated proteins. (E, F, H, and J) Quantification of data, with each data point representing the fluorescence image intensity in one image, multiple images taken of cells from each mouse. Scale bars, 50 μm. Data are presented as means ± SEM. Statistical analysis was conducted using an unpaired Mann-Whitney test, with significance levels indicated as follows: ****P ≤ 0.0001.

Fig. 3.. Reduced nuclear TDP-43 in ECs of Tardbp^G348C/+ mice.

(A) Representative confocal images of mouse brain frontal cortex sections from 10-month-old wild-type Tardbp^+/+ (n = 6) and their heterozygous littermates, Tardbp^G348C/+ mice (n = 6). Shown are representative low-magnification and high-magnification immunofluorescence images for the indicated markers. Each red arrow indicates vascular endothelial nuclear TDP-43 immunostaining. (B) Quantification of TDP-43; each data point represents one vascular EC nucleus. (C) Representative confocal images of mouse brain ECs isolated from 3-month-old wild-type Tardbp^+/+ (n = 6) and their heterozygous littermates, Tardbp^G348C/+ mice (n = 6), immunostained with antibodies against endogenous TDP-43 (green) and DAPI (blue). (D) Quantification of nuclear TDP-43 levels in isolated ECs, confirming a reduction in TDP-43 in Tardbp^G348C/+ mice; each data point represents one EC nucleus. (E) RNA-sequencing analysis of ECs derived from Tardbp^G348C/+ mice and littermate controls showed no significant differences in mRNA expression. Phased analysis of transcripts derived from the mutated G348C allele and the wild-type allele in the same cells revealed no differences in mRNA transcript levels. Data are presented as means ± SEM. Statistical analysis was conducted using an unpaired Mann-Whitney test, ****P ≤ 0.0001. [(A) and (C)] FPKM, fragments per kilobase of exon per million mapped fragments. Scale bars, 50 μm.

Fig. 4.. BBB disruption in Tardbp BrEC-KO mice.

(A) Schematic representation of the assay for measuring BBB permeability. (B) Measurement of 3-kDa Texas Red–dextran in homogenized brain tissue 1 week after tamoxifen (Tam) treatment of BrEC-KO mice (n = 11) and littermate controls (n = 12). **P < 0.0034. (C and E) Representative images of (C) 3-kDa Texas Red–dextran leakage and (E) NHS-biotin in the cortex of 3- to 7-month-old mice (n = 3 BrEC-KO and n = 3 littermate controls). (D and F) Quantification signal, with each data point representing the fluorescence image intensity from one image, multiple images per mouse. A.U., arbitrary units. (G) Whole-mount immunostaining for VE-cadherin and α–smooth muscle actin (n = 12 arteries in 3 BrECKO mice and n = 14 arteries in 4 controls), with (H) quantitation of signal intensity in the artery, across the yellow line in (G). (I) Quantitation by animal of threshold VE-cadherin staining in arteries. [(C), (E), and (G)] Scale bars, 50 μm. Data are presented as means ± SEM. Statistical analysis was conducted using an unpaired Mann-Whitney test, with significance levels indicated as follows: **P < 0.01; ***P < 0.001; ****P ≤ 0.0001.

Fig. 5.. Core BBB pathways affected by loss of TDP-43.

Analysis of differential gene expression in acutely isolated ECs from BrEC-KO mice and littermate controls. (A) Volcano plot showing differential gene expression by DESeq2. Colors indicate that gene expression changes are consistent in direction between datasets. Inset shows Enrichr analysis of up-regulated genes. Large font points are shown for each dataset in (B), along with changes in expression in human primary brain ECs with and without siTDP-43 in the indicated culture conditions [static culture, static culture with TNF-α (30 ng/ml), or rotational culture and high shear or low shear, from inner or outer region of the well]. IL-6/JAK/STAT3, interleukin-6/Janus kinase/signal transducer and activator of transcription 3; ECM, extracellular matrix; FACS, fluorescence-activated cell sorting. (C) Pathway analysis by GSEApy, of the most consistently regulated transcripts in acutely isolated brain ECs from BrECKO and KI mice, cultured ECs from the same mice, or primary human brain ECs with suppression of TDP-43 by small interfering RNA (siRNA). (D and E) Example gene set enrichment analysis (GSEA) plots for top scoring pathways. Heatmap and clustering (UPGMA, unweighted pair group method with arithmetic mean algorithm) shows the most consistently affected Kyoto Encyclopedia of Genes and Genomes (KEGG) and Hallmark pathways (human to mouse liftOver by BioMart) and custom pathways. Details on the gene sets used for GSEA and the RNA-sequencing samples are contained in table S1. False discovery rate (FDR) value derived from GSEA analysis is shown. (C) *P < 0.05; **P < 0.01; ***P < 0.001.

Fig. 6.. TDP-43 loss results in mis-splicing of key endothelial transcripts.

(A) Graph shows individual splicing events detected by LeafCutter analysis and significantly different between BrECKO and littermate controls, and correlation with the regulation of the same splicing events in a comparison between Tardbp^G348C/+ mice and littermate controls. (B) LeafCutter (LeafViz) analysis of example splicing events differentially regulated in BrEC-KO cells in vivo, some of which are also seen in KI cells in vivo. Line plots to the right show read density in each of the replicates across the alternatively spliced region. (C) Venn diagram showing correlation between the transcripts affected by differential expression or splicing in the comparison between BrECKO and littermate controls. (D) Examples of the top pathways enriched among transcripts affected by either expression or splicing in BrECKO mice, by String analysis in Cytoscape. (E) Pathways most affected by significantly regulated splicing changes that were consistently regulated (same direction) in BrECKO and Tardbp^G348C/+ mice, relative to their littermate controls. UV, ultraviolet.

Fig. 7.. Pathological and behavioral consequences of chronic endothelial TDP-43 loss.

(A) Representative immunofluorescence images of fibrin deposition in mouse brain frontal cortex sections from 8- to 11-month-old mice (n = 3 BrEC-KO and n = 3 littermate controls) are shown. (B) Quantification of data, with each data point representing the fluorescence image intensity in one image. (C and E) Iba1 staining of microglia in the mouse brain cortex reveals consistent results across n = 3 BrEC-KO and n = 3 littermate controls mice (n = 3), as well as n = 3 Grn^R493X/+ and n = 3 littermate controls mice (n = 3), and (G and I) GFAP staining of astrocytes reveals a substantial increase in astrocyte numbers, resembling astrogliosis observed in FTD. (D, F, H, and J) Quantification of data with each data point representing the number of activated cells in an image, with multiple images per mouse. (K) Y-maze and tube dominance test results for BrEC-KO mice. For tube dominance test, Grn^R493X/+ mice are included as a positive control, showing a high loss percentage in both models. [(B), (D), (F), (H), and (J)] FoV is 0.16 mm². Data are presented as means ± SEM. Statistical analysis was conducted using an unpaired Mann-Whitney test, with significance levels indicated as follows: ****P < 0.0001. [(A), (C), (E), (G), and (I)] and *P = 0.044 (K). Scale bars, 50 μm.