The NOTCH3 score: a pre-clinical CADASIL biomarker in a novel human genomic NOTCH3 transgenic mouse model with early progressive vascular NOTCH3 accumulation

Generation of NOTCH3 transgenic mice

For transgenesis, a 142,63 kb BAC clone was used (RP11-456 N16 BAC, Bacpac resources, Oakland, USA) (Ensemble release 59). The BAC contains the full-length human genomic NOTCH3 gene and 44 kb of upstream and 67 kb of downstream sequence, including flanking genes SYDE1, ILVBL, EPHX3 and a part of the BRD4 gene (Fig. 1a). The c.544C > T (p.Arg182Cys) mutation was introduced using two-step Red-mediated recombination as previously described [20]. BAC constructs were injected into fertilized C57BL/6 J Ico oocytes. Positive transgenic founder mice were identified by PCR on DNA isolated from mouse ears using human specific primers (for primer sequences see Additional file 1: Table S1). The presence of the mutation was confirmed by direct Sanger sequencing analysis of PCR products (Fig. 1b). Five transgenic mouse strains were generated: one carrying the wild-type NOTCH3 transgene (tgN3^WT) and four carrying the mutant NOTCH3 transgene (tgN3^MUT). In each strain, integration of the BAC was confirmed by PCR analysis of NOTCH3 and the flanking genes SYDE1, ILVBL and EPHX3 (for primer sequences see Additional file 1: Table S1). All transgenic mouse strains bred normally. All experiments described in this study were approved by the local ethical committee for animal experimentation.

NOTCH3 expression analysis in NOTCH3 transgenic mice

Total RNA was extracted from a brain hemisphere using RNA-Bee (Tel-test Inc., Friendswood, USA). For RT-PCR analysis, first-strand cDNA was synthesized using oligo (dT) primers. RT-PCR analysis was performed with primers across the human NOTCH3 transcript (exons 2–4, exons 14–16, exons 30–32, exons 32–33, and the 3′UTR). For qPCR analysis, cDNA synthesis was performed with random hexamer primers, using the Revert Aid H Minus first strand kit (Thermo Scientific, Waltham, USA). Quantitative PCR was performed in four 10-week-old male and female transgenic mice and non-transgenic littermates, using both human- (exons 7–9, 216 bps) and mouse-specific (exons 6–8, 220 bps) primers. Mouse Gapdh was used as a reference gene, and human NOTCH3 expression levels were calculated relative to endogenous mouse Notch3 expression levels. A possible effect of differences in primer efficiencies of human- and mouse-specific NOTCH3/Notch3 primersets was excluded by LinregPCR [21].

NOTCH3 immunohistochemistry and quantification in NOTCH3 transgenic mice

Vascular NOTCH3 accumulation in brain was analysed prospectively in groups of three mice, at age 4, 6, 12, 24, 52 and 82 weeks. In addition, vascular NOTCH3 accumulation was assessed in heart, aorta, liver, kidney, skin and tail at age 20 months. NOTCH3 immunohistochemistry was performed on cryosections that were fixated in acetone and incubated overnight with an antibody directed against the human NOTCH3^ECD (Novus Biologicals, Littleton, USA; dilution 1:2000). The following day, sections were incubated with peroxidase labelled polymer conjugated to anti-rabbit immunoglobulins (Envision kit, Dako, Glostrup, Denmark), and developed using 3,3′-diaminobenzidine. Quantification of the vascular NOTCH3 accumulation load was performed on brain cryosections. Per time point (6, 24, 52 and 82 weeks), three tgN3^MUT350 mice and three tgN3^MUT150 mice were analysed. For each mouse, four frontal lobe brain sections were NOTCH3 immunostained simultaneously. Sections were scanned using the Ultra Fast Scanner (Philips, Eindhoven, The Netherlands), from which 10 images representative for the NOTCH3 accumulation observed in that mouse were obtained. To exclude a possible bias in image selection, a second, blinded observer was asked to obtain images independently (Additional file 1: Figure S1). ImageJ analysis was performed as follows: the image was converted into an 8-bit image, and filtered using the unsharp mask filter (radius 1, mask 0.60). Next, a threshold was set to a signal intensity of 150 to determine the NOTCH3-positive area. Within the NOTCH3-positive area, individual NOTCH3 particles were identified based on size and circularity (size = 0-30, circularity = 0.50-1.00). Finally, the NOTCH3 score was determined by quantifying the total area of the NOTCH3 positive particles within the image. The average of the three mice per time point was plotted and used for further statistical analysis.

NOTCH3 immunohistochemistry and quantification in human material

We used paraffin embedded frontal lobe brain sections from three deceased CADASIL patients (I: female age 59, p.Arg153Cys; II: female age 57, p.Arg153Cys; III: male age 70, p.Cys446Phe) and three deceased controls with no known cerebrovascular disorders (I: male age 67, II: male age 58, III: male age 53). Sections were de-waxed, rinsed with ethanol and blocked with methanol/H₂O₂. After heat-induced antigen retrieval in 0.01 M citrate buffer pH 6, slices were washed three times with PBS, and incubated overnight at room temperature with a 1:1 cocktail of anti-NOTCH3^ECD (dilution 1:500) and anti-CD31 (Dako, Glostrup, Denmark; dilution 1:50). The following day, sections were washed and incubated for 1 hour at room temperature with a 1:1 cocktail of anti-rabbit Envision/HRP (Dako) and goat anti-mouse alkaline phosphatase (Vector Laboratories, Burlingame, CA, USA; dilution 1:25). Finally, sections were sequentially developed with 3,3′-diaminobenzidine solution and Vector Blue (Vector laboratories). Per individual, four images were taken at a 400× magnification on a Leica IM 500 microscope and analysed using ImageJ software. The vessel area was selected manually based on a positive CD31 staining. Within the vessel area, the NOTCH3 score was calculated using an intensity threshold of 100.

Electron microscopy in NOTCH3 transgenic mice

Brain tissue was fixed in 1.5 % glutaraldehyde and 1 % paraformaldehyde in 0.1 M cacodylate buffer, post-fixed in a solution of 2 % osmium tetroxide and 2 % potassium ferrocyanide, dehydrated and embedded in epon 812 (LX112). After selection of areas of interest on 1 μm toluidine stained sections, ultrathin sections were cut, contrasted with 3 % uranylacetate and Reynolds lead citrate and examined with a JEOL JEM-1011 electron microscope (Advanced Microscopy Techniques, Woburn, USA).

Statistical analysis

Statistical analyses were performed using Graphpad Prism. Differences in NOTCH3 score between the two mouse strains at a given time point were analysed using the unpaired student’s t-test. Differences in NOTCH3 score between time points were analysed using One-Way ANOVA and Fishers least significant difference post-hoc analysis. Differences in slope (i.e. the rate of increase of the NOTCH3 score) over time were analysed using linear regression.

Brain MRI of NOTCH3 transgenic mice

Brain MRI was performed in 15 mice at 20 months of age; six TgN3^MUT350 mice, five tgN3^WT mice, and four non-transgenic littermates. Mice were anesthetised by inhalation of 2 % isoflurane in a 1:1 mixture of oxygen and air. Respiration rate was monitored with a respiratory pad and kept between 50 and 80 respirations per minute by adjusting the isoflurane concentration. T2 weighted imaging was performed on a 7 Tesla Bruker Pharmascan using a 23 mm quadrature coil with the following parameters: TE = 12 ms, RARE factor = 8, effective TE = 48 ms, TR = 4 s, 8 averages. Field-of-view = 19x19 mm, matrix = 196x196, resulting in an in-plane resolution of 97 μm. Slice thickness = 0.5 mm, with 32 slices.

Analysis of brain parenchyma in NOTCH3 transgenic mice

After MRI, anesthetised mice were sacrificed using cardial perfusion with ice-cold phosphate buffered saline. One brain hemisphere was formalin fixated and paraffin embedded. Sections were stained with hematoxylin and eosin (H&E) to analyse the presence of infarctions, with Kluver Barrera Luxol fast blue to visualise myelin and with Perl’s iron to assess the presence of microbleeds. Astrogliosis was analysed using an anti-glial fibrillary acidic protein (GFAP) antibody (rabbit anti-GFAP, Dako; dilution 1:1000), which was incubated overnight at room temperature. As a secondary antibody, biotin labelled swine-anti Rabbit (Dako; dilution 1:600) was used, this was incubated 1 hour at room temperature. Finally, sections were incubated with avidin-biotin complex (Vectastain ABC-Elite Kit, Vector Lab, Burlingame, USA) for 30 minutes at room temperature and developed in 3,3′-diaminobenzidine solution. The detection of macrophages and myelin was performed using the Animal Research Kit peroxidase (Dako). Biotinylated primary antibodies against CD68 (anti-CD68 clone KP-1, Dako; dilution 1:1000) and myelin proteolipid protein (anti-PLP, clone plpc-1, Serotec, Kidlington, UK; dilution 1:500) were incubated overnight after heat-induced antigen retrieval in 0.01 M EDTA pH 8.0. The following day, sections were incubated with HRP- conjugated streptavidin for 30 minutes at room temperature and developed in 3,3′-diaminobenzidine solution.

Four mutant human NOTCH3 transgenic mouse strains with distinct NOTCH3 expression levels

Using qPCR analysis on RNA isolated from brain of 10-week old mice, we found that the four mutant NOTCH3 p.Arg182Cys transgenic mouse strains (tgN3^MUT) had human NOTCH3 expression levels of 100, 150, 200 and 350 %, respectively, compared to endogenous mouse Notch3 expression (Fig. 1c). Human NOTCH3 expression in the lowest expressing tgN3^MUT strain, was comparable to that in the wild-type strain (tgN3^WT), i.e. ~100 %. RT-PCR analysis with multiple NOTCH3 primer sets spanning the complete transcript, showed that the complete human NOTCH3 cDNA was present in the transgenic transcript (Additional file 1: Figure S2). There was no difference in endogenous mouse Notch3 expression between transgenic and non-transgenic mice (Additional file 1: Figure S3).

Age at onset of vascular NOTCH3 protein accumulation correlates with NOTCH3 expression levels

To analyse the presence and onset of a CADASIL vascular phenotype, NOTCH3 immunohistochemistry was performed on brain slices from mice between the ages of 4 weeks and 20 months. This showed that all tgN3^MUT strains developed cerebrovascular NOTCH3 accumulation, as seen by a positive, granular NOTCH3 staining of the vessel wall (Fig. 2a), similar to that which is seen in CADASIL patients (Fig. 2b). There was a considerable difference in age at onset of positive NOTCH3 staining per mouse strain, ranging from 6 weeks in tgN3^MUT350 mice to 12 months in tgN3^MUT100 mice. The age at onset directly correlated with the level of human NOTCH3 RNA expression for all four tgN3^MUT strains i.e. the higher the NOTCH3 RNA expression level, the earlier the onset of NOTCH3 accumulation (Table 1). Furthermore, in each mutant strain, the positive NOTCH3 immunostaining became progressively more intense and granular with age (Fig. 2a). The individual granular NOTCH3 deposits increased not only in number, but also in size. This was most prominent in mice with the highest NOTCH3 RNA expression level (tgN3^MUT350), in which the NOTCH3 protein accumulation progressively evolved to a vessel wall packed with intense and big granular NOTCH3 deposits at age 20 months. Characteristic granular NOTCH3 staining was also present in arterioles of the heart, liver, kidney, skin and tail, but not in the aorta (Fig. 2c, data not shown). Overall, the NOTCH3 accumulation observed in the extra-cerebral arterioles was less pronounced than in the brain. Electron microscopy of brain arterioles revealed characteristic electron dense deposits within the basement membrane (Fig. 2d) reminiscent of GOM deposits seen in CADASIL patients (Fig. 2e). Neither GOM nor increased cerebrovascular NOTCH3 staining was found in tgN3^WT mice at 20 months of age (Fig. 2a). Taken together, these analyses show that transgenic human NOTCH3 p.Arg182Cys mice develop an early and progressive systemic arteriopathy which closely resembles the vascular pathology seen in CADASIL patients, with age-at-onset correlating with the respective levels of mutant human NOTCH3 expression.

Mouse strain	NOTCH3 expression levela	Age at onset NOTCH3 accumulationb
tgN3MUT 350	350 %	6 weeks
tgN3MUT 200	200 %	3 months
tgN3MUT 150	150 %	5 months
tgN3MUT 100	100 %	12 months

^a mRNA NOTCH3 expression levels relative to endogenous mouse Notch3 expression levels ^b first sign of positive, granular NOTCH3 staining in brain vessels, as determined by an experienced neuropathologist (S.v.D)

Development of the NOTCH3 score, a quantitative biomarker for CADASIL

As we observed such an early and clear age- and NOTCH3 expression level- dependent vascular NOTCH3 accumulation load, we set out to objectify this by developing a quantitative measure for NOTCH3 staining. This was accomplished by capturing and measuring the surface area of CADASIL specific granular NOTCH3 deposits within brain sections using ImageJ software (Fig. 3a), which we called the ‘NOTCH3 score’. This quantification was first performed in tgN3^MUT350 mice, which clearly showed that the NOTCH3 score increased with age, confirming our qualitative observations (Fig. 3b). Next, we validated the NOTCH3 score in a second mouse strain, tgN3^MUT150, in which the same age-dependent increase in the NOTCH3 score was seen. At each time-point, the NOTCH3 score was lower for the tgN3^MUT150 mice compared to the tgN3^MUT350 mice, reflecting the correlation between NOTCH3 expression level and NOTCH3 accumulation load (score at age 20 months: 659 ± 51 vs. 1150 ± 107, p = 0.002) (Fig. 3c). Furthermore, progression of NOTCH3 accumulation was slower in tgN3^MUT150 mice compared to tgN3^MUT350 mice, as shown by a significant difference in the slope of the NOTCH3 scores between the two mouse strains (11.1 ± 0.5 vs. 6.2 ± 0.3, p = 0.002). Finally, we tested the approach in brain sections of three unrelated CADASIL patients. Measurement of the NOTCH3 accumulation load using the NOTCH3 score was technically feasible in human tissue (Fig. 3d) and showed a significantly higher NOTCH3 score in patients than in controls (score 3.81 ± 1.85 vs. 0.24 ± 0.17, p = 0.02) (Fig. 3e).

No clear brain parenchyma phenotype in tgN3^MUT mice

Finally, to determine whether we could correlate the NOTCH3 score to a brain phenotype in tgN3^MUT mice, we performed brain MRI and histopathology in mice aged 20 months (Additional file 1, Figure S4, S5). In two of the six tgN3^MUT350 mice, hyperintensities were seen on T2 weighted images, cranial to the corpus callosum and around the ventricles in the frontal lobe. However, similar hyperintensities were found in one non-transgenic littermate. Histopathological examination did not show any signs of astrogliosis or white matter lesions in the mutant mice. The absence of consistent and specific brain abnormalities in tgN3^MUT mice prohibited the testing of a potential correlation between NOTCH3 score and brain phenotype.