Conceptualization of the experimental strategy. As a representative mutation commonly found in CADASIL patients, we studied the human Arg169Cys mutation in EGFr4; this mutation, located in a mutational hotspot (EGFr1–6), is associated with a severe and highly penetrant phenotype (Figure 1A) (29). We used 2 distinct mouse models — the Notch3R170C and TgNotch3R169C lines — that express a distinct number of copies of WT and mutant Notch3 with this mutation (Figure 1B). The Notch3R170C line is a knockin model harboring a cysteine mutation inserted at codon 170 in the mouse Notch3 gene that corresponds to the human Arg169Cys mutation (30). Mice were bred to homozygosity (Notch3R170C/R170C) to express 2 copies of the endogenous mutant Notch3 gene at normal levels, and no copies of WT Notch3. This homozygous mutation is clinically relevant since several CADASIL patients with a homozygous cysteine mutation in EGFr1–6 have been described, particularly in Finland, where there is a founder effect (31). The phenotype of these patients is within the clinical spectrum of CADASIL, although at its more severe end (32). In contrast, the TgNotch3R169C line is a transgenic model, created using a phage P1–derived artificial chromosome, that overexpresses rat Notch3 containing the Arg169Cys mutation (4 copies) but also overexpresses normal murine Notch3 (2 copies) (33). The Notch3R170C and TgNotch3R169C lines, initially developed on a mixed SV/129-BL/6 background (Notch3R170C) or FVB/N (TgNotch3R169) background, were backcrossed onto the same C57BL/6 background, based on recent work suggesting that the C57BL/6 background may be more permissive for the expressivity of cSVD phenotypes (34).

Schematic overview of the experimental strategy and working model. (A) Mechanisms put forward to explain how the Arg169Cys mutation causes arterial pathology. (B–E) Four main types of experiments used to discriminate between a loss/gain of NOTCH3 activity and a neomorphic effect. (F) Proposed model emerging from this study. NOTCH3 activity is unaffected and arterial maturation and vascular function proceed normally despite a CADASIL mutation, but mutant Notch3ECD aggregates with WT Notch3ECD, possibly via a mechanism involving thiol disulfide exchange reactions. This leads initially to reversible vascular dysfunction owing to an excess of TIMP3, which compromises activity of Kv2.1 voltage-dependent K+ channels (23–25). Upon reaching a threshold, accumulated Notch3ECD aggregates cause arterial SMC loss and irreversible vascular dysfunction, possibly through accumulation and inactivation of other proteins (e.g., HTRA1).

With these 2 mouse models in hand, we were poised to determine which of the 2 proposed mechanisms accounts for arterial SMC loss in CADASIL. Our approach was to compare the burden of Notch3ECD accumulation and the severity of arterial pathology between TgNotch3R169C and Notch3R170C/R170C mice and profile the transcriptome of NOTCH3-regulated genes (Figure 1, B and C). If arterial SMC loss is caused by a reduction in NOTCH3 activity, the prediction was that arterial pathology would be more severe in Notch3R170C/R170C mice, which lack WT NOTCH3. If instead arterial loss is caused by a neomorphic effect or an increase in NOTCH3 activity, arterial pathology should be more severe in TgNotch3R169C mice, which would presumably also show greater aggregation and accumulation of Notch3ECD. In the latter scenario, gene expression profiling should enable discrimination between a neomorphic effect and increased NOTCH3 activity. To perform this phenotypic comparison, we developed dedicated histopathological and multiscale imaging modalities that allowed us to precisely quantify Notch3ECD accumulation in dissected brain arteries and — more importantly — to identify and quantify arterial SMC defects in brain and retinal arteries. To further investigate the contribution of a neomorphic effect, we examined whether WT NOTCH3 participates in the accumulation process and assessed the impact of reducing Notch3ECD accumulation on arterial pathology (Figure 1, D and E).

Notch3ECD aggregates and accumulates to a greater extent in TgNotch3R169C than in Notch3R170C/R170C mice. We first assessed the burden of Notch3ECD aggregation and accumulation between TgNotch3R169C and Notch3R170C/R170C mice. Brain arteries from both models were dissected at 4 months of age, labeled in toto, and imaged using high-resolution confocal microscopy (Figure 2, A and B). Measurements of Notch3ECD immunostaining in individual SMCs showed that both the area and the mean fluorescence intensity of immunostaining were significantly higher in TgNotch3R169C mice compared with Notch3R170C/R170C mice (Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/JCI175789DS1). Because these 2 metrics do not discriminate between a global increase in Notch3ECD expression and genuine aggregation of Notch3ECD, we re-ran the image analysis protocol using more stringent parameters, adjusted so as to predominantly detect Notch3ECD aggregates. This analysis confirmed the previous analysis (Figure 2, C and D), establishing that Notch3ECD accumulation is more pronounced in TgNotch3R169C mice. In WT mice, a few spots were detected along the internal elastic lamina, likely corresponding to clusters of NOTCH3 receptors expressed on myoendothelial projections (Figure 2B). Using this approach, we were also able to determine the number of Notch3ECD aggregates as well as their size and mean intensity distribution per SMC. The 2 most pronounced differences between TgNotch3R169C and Notch3R170C/R170C mice were the significantly higher number of aggregates and increased proportion of brighter aggregates (i.e., aggregates containing a higher number of Notch3ECD molecules) in TgNotch3R169C mice (Figure 2, E and F). Interestingly, this finding was not associated with a greater proportion of larger aggregates (Figure 2G). Collectively, these results confirm that the Notch3ECD aggregates and accumulates to a greater extent in TgNotch3R169C than in Notch3R170C/R170C mice.

Notch3ECD accumulation is greater in TgNotch3R169C than in Notch3R170C/R170C mice. (A) Schematic overview of the protocol for labeling dissected brain arteries. (B) Representative confocal images of brain arteries from WT, Notch3R170C/R170C (KinR170C), and TgNotch3R169C (TgN3R169C) mice aged 4 months, stained for α-SMA, perlecan, and Notch3ECD. (C–E) Quantification of the mean area (C), mean intensity (D), and number (E) of Notch3ECD aggregates in brain arterial SMCs of WT, KinR170C, and TgN3R169C mice (n = 5 mice per genotype with 20 SMCs per mouse). Shown are violin plots (individual data points represent individual SMCs) and scatter dot plots (individual data points represent individual animals). Data from individual animals were analyzed by 1-way ANOVA and Tukey’s post hoc test. (F and G) Intensity distribution of individual Notch3ECD aggregates (F) and intensity versus size distribution of Notch3ECD aggregates (G) in brain arteries of KinR170C and TgN3R169C mice. Data were analyzed by Kolmogorov-Smirnov test. Scale bars: 9 μm (B).

Detection of arterial defects in the brains of CADASIL mice is enabled by the use of thick brain slices. Next, we tackled the problem of detecting brain arterial pathology in CADASIL mouse models. Pathological analyses of postmortem brain tissues from CADASIL patients have shown prominent loss of SMCs in small brain arteries, a hallmark of the disease (35, 36). Specifically, reconstruction of 1,000 serial paraffin sections from an autopsy subject revealed a diffuse loss of SMCs along the entire length of frontal cerebral medullary arteries from the penetrating site at the cortical surface to the distal end in the white matter (37). However, to date, we and others have failed to convincingly demonstrate the presence of arterial lesions in the brains of mouse models of CADASIL (30, 33, 38). Here, we suggest the possibility that arterial SMC loss is focal and segmental at the early stage of the disease and, assuming that this is recapitulated in mouse models, might go undetected by an analysis of a limited number of 5- to 10-μm-thick sections using conventional histological techniques. Thus, rather than using cross sections, we sought to image long arterial segments by immunostaining thick brain slices.

Starting with coronal brain sections, we selected a thickness of 200 μm, a depth that allows imaging by different modalities without optical preclearing and over a wide range of pixel size resolutions (Figure 3A). Brain sections were labeled with antibodies against smooth muscle α-actin (α-SMA), a contractile protein expressed by SMCs; perlecan, a major component of vascular basement membranes; and elastin, a marker of the internal elastic lamina, which separates endothelial cells from SMCs in arteries/arterioles. To improve antibody penetration and immunoreactivity, we optimized every step, from fixation to immunostaining, as described in detail in Methods. Three-dimensional (3D) confocal imaging confirmed that antibodies did indeed penetrate throughout the entire thickness of the section (Figure 3B). Imaging at medium resolution (194 × 194 × 750 nm) showed elaborate vascular (perlecan-positive) networks that could be divided into 3 distinct microvascular segments according to their positivity or negativity for α-SMA and elastin labeling (Figure 3, C and D). Following the convention of recent studies (39–41), microvascular segments positive for α-SMA and elastin are considered arteries/arterioles; post-arteriole microvascular segments positive for α-SMA but negative for elastin are termed “transition zone” segments; and distal microvascular segments negative for both α-SMA and elastin are referred to simply as capillaries (Figure 3D). We then confirmed that the processing of brain samples and sections did not affect the integrity of mural cells. Imaging at higher resolution (78 × 78 × 160 nm) showed that SMCs of arteries/arterioles were ring-shaped, whereas mural cells in the post-arteriole regions had more irregular ensheathing processes with weaker α-SMA staining and protruding nuclei (Figure 3, E–J). Because acquisition of a single 581 × 581 × 185 μm field of view by confocal microscopy at medium resolution required approximately 1.5 hours and yielded an image file size of approximately 12.4 gigabytes, this imaging modality was deemed unsuitable for large-scale screening of SMC defects in multiple brain slices from dozens of animals. Seeking a compromise among acquisition, resolution, vessel-sampling, and data-handling criteria, we found that imaging with an epifluorescence microscope equipped with a motorized xyz stage and a ×5 (0.16 numerical aperture) objective provided a cellular resolution (1.3 μm pixel size) sufficient to assess the integrity of arterial SMCs in a coronal section of a whole hemisphere in only 3 minutes with a relatively low image size (200–300 megabytes) (Supplemental Figure 2). Using this approach, we found that the length of all arteries analyzed in 4 coronal sections of a single hemisphere totaled 127.5 ± 13.1 mm.

Thick brain slices enable the analysis of long arterial segments. (A) Schematic overview of the protocol starting with a frozen hemisphere thawed for 1 hour in 4% paraformaldehyde (PFA) and sliced at 200 μm thickness for immunostaining, then postfixed in PFA 4% for 15 minutes before imaging. (B–J) Confocal images of a 200-μm-thick brain section from a WT mouse stained for α-SMA (green), elastin (red), and perlecan (white), acquired at medium (194 × 194 × 750 nm) (B–D) or high (78 × 78 × 160 nm) (E–J) resolution. (B and C) 3D rendering views (Imaris software, Bitplane) showing staining of vessels throughout the entire thickness of a slice (B) and long α-SMA+ vascular segments within a single slice (C). (D) Segmentation of the arterial (elastin+ and α-SMA+) and transition zone (elastin– and α-SMA+) compartments (Imaris software). (E–J) High-magnification images of an arteriole (Ar) (E–G) and transition zone (Tz) segment (H–J) delineated in panel D (yellow outlines), illustrating the distinct shapes of arterial SMCs and transition zone pericytes. Scale bars: 30 μm (B–D) and 10 μm (E–J).

We next assessed the integrity of brain arteries from TgNotch3R169C and Notch3R170C/R170C mice, examining 4 coronal sections per animal, segmented into pial, cortical, and subcortical regions. We focused our analyses on arterial/arteriolar (α-SMA–positive and elastin-positive) segments. Arterial SMC integrity was comparable between Notch3R170C/R170C and age-matched WT mice up to 22 months of age (Supplemental Figure 3). In striking contrast, a comparison of TgNotch3R169C mice and WT mice revealed discrete regions where α-SMA staining was looser or discontinuous, hereafter called “defects,” in 6-month-old TgNotch3R169C mice that were not found in WT mice (Figure 4A and Supplemental Figure 4). Quantitative analyses of the lengths of these defects showed that pial, cortical, and subcortical arteries were similarly affected in mutant mice (Figure 4B) and that these defects worsened with age (Figure 4C). Specifically, 12-month-old TgNotch3R169C mice harbored multiple gaps in α-SMA coverage — defined as arterial areas negative for α-SMA but positive for perlecan over the entire circumference — whereas fewer such gaps were detected in 6-month-old TgNotch3R169C mice (Figure 4D).

Brain arteries of TgNotch3R169C exhibit arterial SMC defects. (A) Confocal images of 200-μm-thick brain sections stained for α-SMA, elastin, and perlecan. Shown are representative subcortical brain arteries from WT and TgNotch3R169C (TgN3R169C) mice. Arteries from TgN3R169C mice aged 6 months and older exhibit SMC defects, characterized by discontinuous or notched α-SMA staining (yellow dotted lines) as well as gaps in α-SMA staining (brackets). (B) Quantification of SMC defects in WT and TgN3R169C mice aged 6 months in pial, cortical, and subcortical arteries (n = 5 mice per genotype). Data were analyzed by 2-way ANOVA and Šidák’s post hoc test. (C and D)Quantification of SMC defects (C) and SMC gap numbers (D) in brain arteries of WT and TgN3R169C mice at 6 and 12 months (n = 5–6 mice per genotype). Data were analyzed by 2-way ANOVA and Tukey’s post hoc test. Scale bars: 20 μm (A).

Both TgNotch3R169C and Notch3R170C/R170C mice develop age-dependent arterial SMC defects in the retina, but defects are more severe in TgNotch3R169C mice. The 3D morphology of small blood vessels in the brain makes it difficult to assess pathological changes at the cellular level and detect minimal lesions, as might occur in Notch3R170C/R170C mice. Moreover, a quantitative analysis of arterial defects in the brain is especially burdensome. To overcome these issues, we exploited the stereotyped and planar angioarchitecture of the retina, a developmental extension of the brain that offers unparalleled accessibility for direct analysis and robust quantification of the vascular network at cellular resolution (39, 42). Analyses of the retina are especially relevant given that CADASIL-specific pathologies, including GOM deposits and loss of SMCs, were identified by early studies in retinal vessels of patients with CADASIL (43). Preliminary analyses using flat-mounted retina preparations immunostained for Notch3ECD, α-SMA, and perlecan confirmed that retinal arteries of both TgNotch3R169C and Notch3R170C/R170C mice displayed Notch3ECD aggregates (Figure 5A and Figure 6A).

Retinal arteries of TgNotch3R169C retinas exhibit progressive degeneration and loss of arterial SMCs. (A) Representative confocal images of retinal arteries from WT and TgNotch3R169C (TgN3R169C) mice, aged 4 months, stained for α-SMA, Notch3ECD, and perlecan. (B) Representative pictures of retinal arteries from WT and TgN3R169C mice stained for α-SMA, perlecan, and MEF2C. Arteries in mutant mice exhibit a discontinuous SMC coating at 6 months of age onward (dashed yellow lines) that worsens with age, ultimately causing focal SMC gaps (brackets). (C and D) Quantification of SMC defects (C) and SMC numbers (MEF2C nuclei) (D) in retinal arteries of WT and TgN3R169C mice at 6 and 12 months of age (n = 8–10 mice per group). Data were analyzed by 2-way ANOVA and Tukey’s post hoc test. (E) Representative pictures of retinal arteries from WT and TgN3R169C mice, stained for α-SMA, cleaved caspase-3 (ClC3), and perlecan. (F) Quantification of cleaved caspase-3 in retinal arteries of WT and TgN3R169C mice at 6 and 12 months of age (n = 8–10 mice per group). Data were analyzed by 2-way ANOVA and Tukey’s post hoc test. (G) Number of TgN3R169C mice without or with arterial SMC gaps at 6 and 12 months. Data were analyzed by Fisher’s exact test. (H) Quantification of SMC gaps in retinal arteries of WT and TgN3R169C mice at 6 and 12 months of age (n = 8–10 mice per group). Data were analyzed by 2-way ANOVA and Tukey’s post hoc test. Scale bars: 5 μm (A) and 10 μm (B and E).

Arterial pathology is milder and appears later in Notch3R170C/R170C mice. (A) Representative confocal images of retinal arteries from Notch3R170C/R170C (KinR170C) mice aged 4 months, stained for α-SMA, Notch3ECD, and perlecan. (B) Representative pictures of retinal arteries (distal segment) from Notch3+/+ (WT) and KinR170C mice, stained for α-SMA and perlecan, showing SMC defects at 12 months. (C and D) Quantification of SMC defects (C) and SMC numbers (D) in the distal segment (900–1,800 μm) of retinal arteries at 6 and 12 months (n = 11 mice per genotype per age). Data were analyzed by 2-way ANOVA and Tukey’s post hoc test. (E) Number of KinR170C mice without or with SMC gaps at 6 and 12 months of age. Data were analyzed by Fisher’s exact test. Scale bars: 5 μm (A) and 10 μm (B).

Further analyses of 6-month-old TgNotch3R169C mice revealed that retinal arteries of these mice exhibited numerous defects in SMC coverage similar to those observed in brain arteries, whereas such regions were rare in arteries of WT control mice (Figure 5, B and C). To determine whether these defects resulted from a reduction in the number of SMCs, we counted cells with the aid of an antibody against myocyte-specific enhancer factor 2C (MEF2C), a transcription factor that is predominantly expressed in mural cell nuclei (Supplemental Figure 5). The number of MEF2C-positive nuclei was lower in regions with SMC defects compared with regions with normal SMC coverage, and a quantitative analysis confirmed that the number of arterial SMCs was significantly reduced in TgNotch3R169C mice (Figure 5, B–D). Moreover, we detected cleaved caspase-3 staining within regions with sparse SMC coating, confirming that the SMCs therein were undergoing apoptosis (Figure 5, E and F). As was the case in the brain, arterial SMC defects were segmental and focal; that is, the burden of defects was unevenly distributed across retinal arteries and within a given artery. Moreover, a distal preference was apparent in TgNotch3R169C mice, with more extensive SMC defects observed in the distal region (900–1,800 μm) of arteries compared with the proximal (0–900 μm) region (Supplemental Figure 6). Arterial SMC defects in TgNotch3R169C mice worsened with age, becoming significantly more severe at 12 months compared with those at 6 months (Figure 5, B and C). Gaps, which were occasionally detected in TgNotch3R169C mice at 6 months, were present in all of these mutant mice at 12 months (Figure 5, G and H).

In Notch3R170C/R170C mice, SMC coverage of retinal arteries at 6 months of age appeared comparable to that in WT mice, despite a significant increase in the number of arterial SMCs in these mutant mice. At 12 months, SMC defects were detected in the distal region of retinal arteries of Notch3R170C/R170C mice in association with a return to normal WT numbers of arterial SMCs (Figure 6, B–D). SMC gaps were occasionally observed in 12-month-old Notch3R170C/R170C mice (Figure 6E).

Collectively, these results indicate that TgNotch3R169C and Notch3R170C/R170C mice exhibit focal and segmental SMC defects and losses in cerebro-retinal arteries, effects that progress with age. They further show that arterial pathology is more severe in TgNotch3R169C mice, a feature that is associated with more robust accumulation and aggregation of Notch3ECD. A LOF mechanism predicts that, because Notch3R170C/R170C mice lack normal copies of Notch3, arterial defects should be more severe in these mice; thus, our data showing greater effects in TgNotch3R169C mice suggest that arterial defects do not result from a reduction in NOTCH3 activity.

Expression of genes regulated by Notch3 is not increased in brain arteries of TgNotch3R169C mice. While our findings above argue against a LOF mechanism, they do not discriminate between increased NOTCH3 activity and a neomorphic effect. To explore the possibility that arterial SMC loss in TgNotch3R169C mice might be caused by an increase in NOTCH3 activity, we profiled the transcriptome of brain arteries dissected from TgNotch3R169C mice, Notch3–/– mice, and their WT age-matched littermates using RNA sequencing. We first focused on 11 Notch3 target genes, which, as expected, were significantly downregulated in brain arteries of Notch3–/– mice (Figure 7A) (44, 45). None of these 11 genes showed a significant change in TgNotch3R169C mice (Figure 7B). Likewise, expression of these 11 genes was unchanged in Notch3R170/R170C mice (Supplemental Figure 7). Broadening our analysis to the whole genome to ensure that we did not miss a change in the expression of additional Notch3-regulated genes in TgNotch3R169C mice, we identified 76 differentially expressed genes (DEGs) in TgNotch3R169C mice and 261 DEGs in Notch3–/– mice (Figure 7, C and D). Among these DEGs, only 4 (excluding Notch3) were shared by TgNotch3R169C and Notch3–/–, and none of the genes downregulated in Notch3–/– mice were upregulated in TgNotch3R169C or vice versa (Figure 7E). We then focused on significantly downregulated genes in Notch3–/– mice that have been identified as forming the molecular signature of NOTCH3 activity, assessing the relationship between the log–fold change for each of these genes in Notch3–/– and their respective log–fold change in TgNotch3R169C mice. Again, we found no positive correlation that would be indicative of reduced activity. Importantly, we also found no negative correlation indicative of increased activity (Figure 7F). Taken together, these results suggest that arterial SMC defects in TgNotch3R169C mice are not associated with a substantial change in NOTCH3 activity, thus arguing against aberrant NOTCH3 activity and instead supporting a neomorphic effect.

Comparison of gene expression profiles between brain arteries of TgNotch3R169C and Notch3–/– mice. (A and B) Relative mRNA levels for genes regulated by Notch3 in brain arteries from Notch3–/– versus Notch3+/+ mice (A) and brain arteries from TgNotch3R169C versus WT mice (B), measured by RNA sequencing analysis (n = 8 mice per genotype). Data are reported as a fold change relative to control mice. Data were analyzed by DESeq2 (Wald test and Benjamini and Hochberg method to correct for multiple testing). (C and D) Volcano plot representation of RNA sequencing data comparing brain arteries from Notch3–/– (N3–/–) versus Notch3+/+ (N3+/+) mice aged 3 months (C) and brain arteries from TgNotch3R169C (TgN3R169C) versus WT mice aged 12 months (D) (n = 8 mice per genotype). Red symbols indicate DEGs. Data were analyzed by DESeq2 (adjusted P value < 0.05). (E) Venn diagram showing the overlap of upregulated (red) and downregulated (green) DEGs between TgN3R169C mice (left) and N3–/– mice (right). (F) Correlation analysis of fold change (log2) values of downregulated genes expressed in N3–/– mice with their fold change (log2) values in TgN3R169C mice. Data were analyzed with Spearman’s correlation test.

WT Notch3ECD is incorporated within Notch3ECD aggregates. We and others previously showed that overexpression of WT NOTCH3 is not sufficient to cause an abnormal accumulation of Notch3ECD (33, 38, 46). Importantly, prior work has established that, in many neurodegenerative proteinopathies, misfolded proteins that abnormally accumulate act as seeds for aggregation by templating the pathological conversion of their native isoforms, leading to their accumulation (47). In light of this and the findings noted above, we explored the hypothesis that WT NOTCH3 coaggregates with mutant NOTCH3 to participate in the neomorphic effect. To our knowledge, there are no antibodies that specifically recognize the WT or mutant version of NOTCH3. To circumvent this issue, we tested this hypothesis using a transgenic line (TghumNotch3R90C) that expresses WT murine NOTCH3 and a human mutant NOTCH3 harboring the CADASIL Arg90Cys mutation (46), employing murine- and human-specific antibodies against Notch3ECD to distinguish between WT (murine) and mutant (human) forms (Figure 8A). As expected, the murine-specific polyclonal antibody (designated J53), raised against rodent NOTCH3, recognized murine Notch3ECD aggregates in Notch3R170C/R170C mice but did not detect human Notch3ECD aggregates in postmortem brains from CADASIL patients. Conversely, the human-specific monoclonal anti-Notch3ECD antibody (designated 1E4), raised against human NOTCH3, labeled human Notch3ECD aggregates in patients but did not recognize murine Notch3ECD aggregates in Notch3R170C/R170C mice (Figure 8B). Immunostaining of brain arteries from TghumNotch3R90C and WT mice with the human-specific (1E4) antibody showed that human mutant Notch3ECD-containing aggregates were present in TghumNotch3R90C mice, as expected, and absent in WT mice (Figure 8C). Notably, these aggregates were also labeled by the murine-specific (J53) antibody (Figure 8, C and D), consistent with the idea that WT Notch3ECD is also incorporated into the pathological aggregates. Quantification confirmed that the area, mean fluorescence intensity, and number of both murine-positive aggregates and human Notch3ECD-containing aggregates were significantly increased in TghumNotch3R90C mice compared with WT mice (Figure 8, E–H).

WT Notch3ECD coaggregates with mutant Notch3ECD. (A) Schematic depiction of anti-Notch3ECD antibodies and the TgHumNotch3R90C mouse model. (B) Representative confocal images of brain tissue from a Notch3R170C/R170C (KinR170C) mouse and a CADASIL patient stained with the 1E4 anti–human Notch3ECD (yellow), J53 anti–mouse Notch3ECD (red), and anti–α-SMA (white) antibodies. (C) Representative confocal images of brain arteries from WT and TgHumNotch3R90C (n = 6 per genotype) mice, stained with the 1E4 monoclonal anti–human Notch3ECD, J53 anti–mouse Notch3ECD, and anti-perlecan (cyan) antibodies. (D) Quantification of the colocalization between 1E4- and J53-positive aggregates. Data were analyzed with SODA. (E–H) Quantification of the number of aggregates containing human mutant Notch3ECD (E) or murine WT Notch3ECD (G) and the mean intensity of aggregates containing human mutant Notch3ECD (F) or murine WT Notch3ECD (H) (n = 6 per genotype with 20 SMCs per mouse). Shown are violin plots (individual data points represent individual SMCs) and scatter dot plots (individual data points represent individual animals). Data were analyzed with a linear mixed-effects model. Scale bars: 40 μm (B) and 5 μm (C).

Elimination of 1 copy of WT murine Notch3 in TgNotch3R169C mice is sufficient to mitigate Notch3ECD accumulation and arterial pathology. We next sought to directly establish the centrality of Notch3ECD accumulation in the neomorphic effect and consequent arterial SMC loss by examining how reducing Notch3ECD accumulation impacts arterial pathology. Our demonstration that WT Notch3ECD molecules are present in pathological aggregates predicts that eliminating 1 copy of WT Notch3 in TgNotch3R169C mice would lead to a reduction in Notch3ECD aggregates. To test this, we crossed TgNotch3R169C mice with Notch3-heterozygous mice (Notch3+/–), generating TgNotch3R169C mice lacking 1 allele of WT Notch3 (TgNotch3R169C Notch3+/– mice). Quantitative immunofluorescence analyses showed a trend toward a reduction in the number of aggregates in arterial SMCs of TgNotch3R169C Notch3+/– mice compared with TgNotch3R169C Notch3+/+ mice (Figure 9A and Supplemental Figure 8). Although this difference did not reach statistical significance using a linear mixed model, employed to assess the interdependence of measurements of distinct SMCs from the same animal (Figure 9, B–D), we found a significant leftward shift in the mean intensity distribution of aggregates in TgNotch3R169C Notch3+/– mice (Figure 9E). These results indicate that the proportion of bright aggregates was lower, and thus the number of Notch3ECD molecules in aggregates was fewer, in mice lacking 1 copy of WT Notch3 (TgNotch3R169C Notch3+/– mice) than in those expressing both copies (TgNotch3R169C Notch3+/+ mice).

Elimination of one copy of WT Notch3 attenuates Notch3ECD accumulation in TgNotch3R169C mice. (A) Representative confocal images of brain arteries from TgNotch3R169C Notch3+/+ (TgN3R169C;N3+/+) and TgNotch3R169C Notch3+/– (TgN3R169C;N3+/–) mice stained for α-SMA, perlecan, and Notch3ECD. (B–D) Quantification of the area (B), mean intensity (C), and number (D) of Notch3ECD aggregates in SMCs of TgN3R169C;N3+/+ and TgN3R169C;N3+/– mice (n = 6–7 mice per genotype). Shown are violin plots (individual data points represent individual SMCs) and scatter dot plots (individual data points represent individual animals). Data were analyzed with a linear mixed-effects model. (E) Intensity distribution of individual Notch3ECD aggregates in TgN3R169C;N3+/+ and TgN3R169C;N3+/– mice. Data were analyzed by Kolmogorov-Smirnov test. Scale bars: 5 μm (A).

We then compared the severity of arterial lesions between TgNotch3R169C Notch3+/+ and TgNotch3R169C Notch3+/– mice and their littermate controls at 6 months of age using the retina model. Elimination of 1 copy of Notch3 in WT mice (Notch3+/–) at 6 months of age produced arterial SMC defects that were associated with a reduction in the number of arterial SMCs, a finding consistent with the concept that the function of NOTCH receptors is sensitive to gene dosage (2). Strikingly, arterial pathology was attenuated in TgNotch3R169C Notch3+/– mice, which showed significantly fewer arterial SMC defects and a greater number of arterial SMCs compared with TgNotch3R169C Notch3+/+ mice (Figure 10, A–C). Moreover, unlike TgNotch3R169C Notch3+/+ mice, no TgNotch3R169C Notch3+/– mice exhibited SMC gaps (Figure 10D). To ensure that the amelioration of arterial pathology in TgNotch3R169C Notch3+/– mice did not result from a change in NOTCH3 activity, we quantified brain expression levels of Grip2 and Nrip2 — the 2 most sensitive Notch3 target genes (44) — in addition to murine Notch3, by quantitative reverse transcriptase PCR. We found that mRNA levels of Grip2 and Nrip2 were indistinguishable between TgNotch3R169C Notch3+/+ and TgNotch3R169C Notch3+/– mice, whereas levels of Notch3 mRNA were decreased by half in TgNotch3R169C Notch3+/– mice, as expected (Figure 10E). Collectively, our results provide compelling evidence that, despite the overt resemblance to loss of NOTCH3 function, CADASIL-associated arterial SMC loss is not driven by a reduction in NOTCH3 activity or by an increase in NOTCH3 activity, but rather by the abnormal accumulation of Notch3ECD — a process in which both mutant NOTCH3 and WT NOTCH3 participate.

Elimination of one copy of WT Notch3 mitigates arterial pathology in TgNotch3R169C mice. (A) Representative images of retinal arteries from non-Tg Notch3+/+ (nonTg;N3+/+), non-Tg Notch3+/– (nonTg;N3+/–), TgNotch3R169C Notch3+/+ (TgN3R169C;N3+/+), and TgNotch3R169C Notch3+/– (TgN3R169C;N3+/–) mice aged 6 months, stained for α-SMA and perlecan. (B and C) Quantification of SMC defects (B) and SMC numbers (C) in retinal arteries from mice aged 6 months (n = 10–15 mice per genotype). Data were analyzed by Kruskal-Wallis test followed by Dunn’s post hoc test (B) and by 1-way ANOVA followed by Tukey’s post hoc test (C). (D) Number of TgN3R169C;N3+/+ and TgN3R169C;N3+/– mice without or with SMC gaps in retinal arteries. Data were analyzed by Fisher’s exact test. (E) Relative mRNA levels of murine Notch3, Grip2, and Nrip2 in TgN3R169C;N3+/+ and TgN3R169C;N3+/– mice, measured by quantitative reverse transcriptase PCR (n = 8 per genotype). Data were analyzed by Student’s t test. Scale bar: 20 μm (A).