Misdirected yet intact TREX1 exonuclease activity causes human cerebral and systemic small vessel disease

Abstract

Retinal vasculopathy with cerebral leukoencephalopathy and systemic manifestations (RVCL-S) is an incurable microvascular disease caused by C-terminus truncation of the TREX1 exonuclease. There is a pressing need to understand disease mechanisms and identify therapeutic targets. We evaluated TREX1 sequencing data from 469 229 UK Biobank participants together with RVCL-S-related microvascular clinical and imaging outcomes. We show that mono-allelic truncating mutations in TREX1 require intact nuclease activity in order to cause endothelial disease. Differential proteomics identifies loss of interaction with endoplasmic reticulum insertion proteins such as Guided Entry of Tail-Anchored Proteins Factor 3 as a major consequence of pathogenic TREX1 truncation, and this altered trafficking results in the unregulated presence of enzymatically active TREX1 in the nucleus. In endothelial cells with a patient mutation, mislocalized yet enzymatically active TREX1 causes accumulation of a spectrum of DNA damage. These pathological changes can be rescued by inhibiting exonuclease activity. In summary, our data implicate exonuclease-dependent DNA damage in endothelial cells as a key therapeutic target in the pathogenesis of RVCL-S.

Keywords: DNA damage; cell cycle; endothelial; exonuclease; vascular dementia.

Figure 1

TREX1 truncating mutations require intact 3′-5′ exonuclease activity to cause RVCL-S. (A) TREX1 function was assayed by expression of an EGFP-TREX1 fusion protein in Trex1^−/− mouse embryonic fibroblasts (MEFs). Wild-type (WT) and exonuclease dead [Asp18Asn (D18N), a dominant negative AGS variant]^, are shown for comparison. Eight truncating TREX1 variants were introduced into an EGFP-TREX1 expression plasmid in order to express the truncated protein with an N-terminal EGFP fusion. Representative images taken at ×40 from at least three independent experiments are shown. Scale bar = 5 μm. (B) Total protein lysate was extracted from transfected MEFs to assay 3′-5′ exonuclease activity. WT activity over time is shown by the black line. Asp18Asn (D18N) activity over time is shown by the orange line. Background lysate exonuclease activity (untransfected cells) is shown by the grey line. (Further details of the nuclease activity assay are in Supplementary Fig. 3) Data presented is from at least three independent experiments, with non-linear curve fit. (C) Localization of and (D) 3′-5′ exonuclease activity of RVCL-S variants. (E) Localization of and (F) 3′-5′ exonuclease activity of truncating frameshift variants identified in the UK Biobank. (G) Area under the curve was measured with baseline set at time = 0 for UT or D18N, whichever was lowest, and normalized within each experiment to WT = 100%. Data-points show independent experiments, columns show mean. RVCL-S variants in purple, Biobank variants in blue, nuclease dead variant (D18N) in orange. Error bars are standard error of the mean. ****P < 0.0001, One way ANOVA with mixed effects model comparing each data set with WT before data normalization, individual P-values are from Dunnett’s multiple comparison test. (H) Schematic diagram of protein effects of TREX1 variants. Exo = exonuclease region; PPII = polyproline II; ER TM = endoplasmic reticulum transmembrane. Created in BioRender. McGlasson, S. (2025) https://BioRender.com/k45a920. (I) Table summarizing TREX1 variant effects on ER localization, 3′-5′ exonuclease function and clinical disease association. ^aPathogenic consequence refers to heterozygotes. ER = endoplasmic reticulum; AGS = Aicardi–Goutieres Syndrome. *M232 variant has slightly reduced nuclease activity, with a possible low penetrance RVCL-S phenotype (see text) Created in BioRender. McGlasson, S. (2025) https://BioRender.com/k45a920. (J) Forest plot showing a summary of clinical phenotype associations of truncating TREX1 variants in the UK Biobank. The frequency of 30 RVCL-S-associated outcomes, summarized into five categories, were tested using logistic regression. (Full phenotype outcomes are detailed in Supplementary Fig. 5) (K) Forest plot showing neuroradiological associations of truncating TREX1 variants in the UK Biobank. The distribution of three imaging derived phenotypes were tested using linear regression. aOR = adjusted odds ratio; CI = confidence interval; ER = endoplasmic reticulum; IQR = interquartile range; RVCL-S = retinal vasculopathy with cerebral leukoencephalopathy; SD = standard deviation; UT = untreated.

Figure 2

RVCL-S mutations cause loss of interactions with key components of the post-translational, tail-anchored ER protein insertion pathway. (A) Schematic outline of differential proteomic analysis to identify differential protein interactions between TREX1^WT and TREX1^V235fs. Created in BioRender. McGlasson, S. (2025) https://BioRender.com/b34z593. (B) Volcano plot showing the differential protein interactions between TREX1^WT and TREX1^V235fs. The x-axis displays the log₂-fold-change of protein abundance between the conditions, where positive values indicate increased interaction with TREX1^V235fs versus TREX1^WT, and negative values indicate decreased interactions with TREX1^V235fs versus TREX1^WT. The P-values are presented on −log10 scale on the y-axis. Proteins determined significantly differently bound between TREX1^V235fs versus TREX1^WT at thresholds P < 0.05 are indicated in the uppermost segment. Protein interactions that are more than log₂(2) increased or decreased are indicated by coloured dots (blue = decreased interactions with TREX1^V235fs versus TREX1^WT; red = increased interactions with TREX1^V235fs versus TREX1^WT). (C) Volcano plot of differential protein interactions colour coded by canonical subcellular localization (green = ER; pink = nucleus; grey = other). (D) Quantification of the number of proteins that are altered in each subcellular location, in each condition (TREX1^WT or TREX1^V235fs). Data summarized from two independent data sets^, and Uniprot., (green = ER; pink = nucleus; grey = other). (E) Mapping of Go cell component terms associated with decreased (left) or increased (right) interactions using SubcellulaRVis. Colour intensity indicates FDR. FDR significance threshold set to 0.01. (F) Super resolution (STED) microscopy of EGFP-TREX1^WT or EGFP-TREX1^V235fs, co-stained with calnexin ER membrane protein. Images taken at ×100. Images below quantify the signal intensity of calnexin (magenta) and EGFP-TREX1 (green) across each cell. ImageJ was used to quantify signal intensity across a single cross section. (G) GO Process analysis of significant interactions that are increased with TREX1^V235fs versus TREX1^WT. ER = endoplasmic reticulum; FDR = false discovery rate; RVCL-S = retinal vasculopathy with cerebral leukoencephalopathy; STED = Stimulated Emission Depletion; WT = wild-type.

Figure 3

Endothelial disease is an important component of the microvascular disease observed in RVCL-S. (A) Electron micrograph of microvessel (renal) showing endotheliopathic features such as loss of endothelial cell fenestration (green arrow) and subendothelial lucency (blue arrow). Scale bar = 5 μm. (B) For comparison, electron microscopy of normal endothelial cell with fenestrations (red arrows). (C) PRISMA pipeline for systematic literature review of RVCL-S pathology. Created in BioRender. McGlasson, S. (2025) https://BioRender.com/x56u663. (D) Schematic of TREX1-dsRed transgenic reporter. These mice express TREX1 and dsRed under the control of the endogenous TREX1 promoter.⁸ Created in BioRender. McGlasson, S. (2025) https://BioRender.com/f91o214. (E) Flow cytometry histogram overlaying WT CD31⁺ cells and TREX1-dsRed CD31⁺ cells. (F) Quantification of median fluorescent intensity (MFI) of dsRed from independent samples (n = 4 per condition). Data-points show individual experiments, columns show mean with standard error of the mean (SEM). P = 0.0079, unpaired t-test. (G) Representative flow cytometry plots of the frequency of vascular endothelial cells (VECs, CD31⁺ popoplanin⁻) and lymphatic endothelial cells (LECs, CD31⁺ popoplanin⁺) in the spleen of Tie2-Cre LSL hTREX1^WT or Tie2-Cre LSL hTREX1^V235Gfs mice. (H) Quantification of the frequency and number of VECs from (H). Points indicate data from individual mice (n = 3 per group). Columns show mean. Error bars show SEM. (I) Quantification of the frequency and number of LECs from (H). Points indicate data from individual mice (n = 3 per group). Columns show mean. Error bars show SEM. (J) Representative flow cytometry plots showing the frequency of hTREX1⁺ vascular endothelial cells via staining the HA tag. (K) Quantification of the frequency of hTREX1⁺ VECs from (J). Points indicate data from individual mice (n = 3 or 4 per group). Columns show mean. Error bars show SEM. P = 0.0167 by unpaired t-test. RVCL-S = retinal vasculopathy with cerebral leukoencephalopathy with systemic manifestations; PRISMA = Preferred Reporting Items for Systematic reviews and Meta-Analyses; WT = wild-type.

Figure 4

Human endothelial cells with RVCL-S patient mutations show increased DNA damage, cell cycle defects and increased chromosomal abnormalities. (A) Human brain endothelial cells were transformed using lentivirus to express TREX1 with a patient mutation. The MRI brain scan from a patient with this mutation is shown. Created in BioRender. McGlasson, S. (2025) https://BioRender.com/l89e736. (B) Lentiviral expression of EGFP-TREX1 in human brain endothelial cells (hBEC-5i), co-stained with 4′,6-diamidino-2-phenylindole and 53BP1. Illustrative image taken by confocal microscope at ×20 magnification, images shown with digital zoom. (C) Quantification of 53BP1 foci in response to lentiviral expression of EGFP-TREX1^WT, EGFP-TREX1^V235fs or an empty EGFP vector. Data-points show three independent experiments and columns show mean. Error bars are standard error of the mean (SEM), P = 0.0315, unpaired t-test. (D) Representative image of chromatin bridge seen in human brain endothelial cells expressing EGFP-TREX1^V235fs. Confocal images taken at ×63 magnification. (E) Quantification of percentage of cells with chromatin bridges. Data-points show three independent experiments and columns show mean. Error bars are SEM, P < 0.0001, unpaired t-test. (F) Quantification of cells in G2/M in response to lentiviral expression of EGFP-TREX1^V235fs, EGFP-TREX1^WT or an empty EGFP vector. Data-points show three independent experiments and columns show mean. Error bars are SEM, P = 0.0248, unpaired t-test. (G) Clonal expression of EGFP-TREX1^V235 in hBECs. Top row: ×20 magnification; bottom row: ×63 magnification. Line drawn to indicate GFP⁺ versus GFP⁻ cell clones. (H) Quantification of the correlation (Pearson's r) between corrected total cell fluorescence (CTCF) of EGFP and 53BP1 foci. n = 2 per group, with at least 30 cells counted per experimental group (for full breakdown see Supplementary Fig. 8F and G). Columns show mean. Error bars show standard deviation. P = 0.0117, unpaired t-test. (I) Quantification of the correlation (Pearson's r) between nuclear EGFP-TREX1 and 53BP1 foci. n = 3 WT, n = 4 V235. Points show independent experiments with mean (for full breakdown see Supplementary Fig. 8D and E). Error bars show SEM. P = 0.0158, unpaired t-test. RVCL-S = retinal vasculopathy with cerebral leukoencephalopathy with systemic manifestations; WT = wild-type.

Figure 5

DNA damage and cell cycle abnormalities in cells expressing TRE­X1^V235fs, are rescued by inhibition of nuclease activity. (A) Schematic overview of TREX1 dimeric structure generated using Pymol. Created in BioRender. McGlasson, S. (2025) https://BioRender.com/k20e227. The structure of the dimer was extracted from Pymol using RCSB PDB 7TQN. The position of the D18N and V235 mutations were annotated in Pymol. The insoluble C-terminal domains were added manually. Representative images of 53BP1 foci in cells stably expressing EGFP-TREX1 variants, from three independent 6-week experiments. Images were taken on a confocal microscope at ×63. Scale bar = 5 μm. Experiments performed in HeLa cell lines. (B) Overview of flow cytometry strategy. EGFP-TREX1 expressing cells were selected via an EGFP gate. EGFP⁺ cells were sorted into cell cycle phases by quantification of propidium iodide (PI) intensity. (C) Quantification of proportions of tetracycline-induced cells (EGFP⁺) in G2/M, normalized to uninduced cells (EGFP⁻) at each time point over 6 weeks. Data presented is baseline corrected to uninduced cells from the same cell line at each time point. P-values are from a two-way ANOVA with mixed effects, with Tukey’s test for multiple comparisons. *P < 0.05, **P < 0.01; purple stars = EGFP-TREX1^V235fs versus EGFP-TREX1^WT; green stars = EGFP-TREX1^D18N/V235fs versus EGFP-TREX1^V235fs. (D) Quantification of 53BP1 foci after 4 weeks of exposure to the EGFP-TREX1 transgene. Data-points are from at least four independent experiments, columns show mean. Error bars are standard error of the mean. P < 0.0001 EGFP-TREX1^WT versus EGFP-TREX1^V235fs, P < 0.0001 EGFP-TREX1^V235fs versus EGFP-TREX1^D18N/V235fs, one-way ANOVA with Tukey’s correction for multiple comparisons. WT = wild-type.