A novel variant of NOTCH2 causes skeletal fragility

Abstract

A 2-month-old female child presented with a low-trauma femoral fracture, low bone mass, bowing of long bones, loss of height of a thoracic vertebra, and Wormian bones. Exome sequencing revealed the presence of a novel heterozygous 4006G > C pG1336R mutation in exon 25 of NOTCH2 in the child and her father. In silico analysis considered the variant as likely deleterious. CRISPR/Cas9 was used to introduce the Notch2^4006G>C mutation into Notch2 to create Notch2^em1Ecan mutant mice. Homozygous Notch2^em1Ecan mutant mice were active, appeared healthy, had normal femoral length, but lower weights than controls. μCT of the distal femur revealed a 25 % decrease in trabecular bone volume, and a decrease in total, bone and marrow area, in periosteal perimeter and polar moment of inertia, revealing the presence of small and potentially fragile bones. Three-point bend testing demonstrated decreased toughness in Notch2^em1Ecan femurs. Cancellous bone histomorphometry demonstrated decreased eroded surface, and Raman spectroscopy revealed normal mineral to matrix ratios, carbonate:phosphate and collagen peak ratios. A structure homology model of NOTCH2 EGF33–36 repeats suggests that the G1336R mutation may disrupt the local structure, reducing the flexibility of the extracellular domain and thereby affecting receptor activation and signaling. Indeed, there was a modest decrease in Notch canonical target genes in osteoblasts from Notch2^em1Ecan mice. Osteoblast and osteoclast differentiation were diminished in cells from Notch2^em1Ecan mice. In conclusion, a novel mutation affecting the NOTCH2 extracellular domain is associated with small and apparently fragile bones, possibly due to altered Notch signaling.

1. Introduction

Notch receptors (Notch 1 to 4) have been established as important determinants of cell differentiation and function in multiple tissues and cell systems including bone, where they modulate skeletal development, and bone remodeling [1–9]. Notch receptors are activated following interactions with members of the Jagged and Delta-like family of ligands present in an adjacent cell. The extracellular domain of Notch is composed of multiple epidermal growth factor (EGF)-like repeats, and EGF repeats 11 and 12 interact with Notch ligands. Notch ligand interactions lead to the endocytosis of the ligand in the adjacent cell creating a pulling force sufficient to expose the negative regulatory region (NRR) of the receptor to enzymes and subsequent proteolytic cleavage [10]. This releases the Notch intracellular domain (NICD), which translocates to the nucleus where it interacts with recombination signal-binding protein for Ig of κ (RBPJκ) and mastermind-like to induce the transcription of target genes, such as members of the Hes and Hey families of genes [11–17]. Although interactions of Jagged and Delta-like with EGF repeats 11 and 12 are critical for Notch activation, other EGF motifs play a significant role in the control of Notch activation. In Drosophila, the abruptex region, EGF 24–29, interacts with EGF 11–12, and protects them from interactions with Jagged and Delta-like preventing Notch activation, as a way to control Notch signaling [18]. Importantly, much of the work on the function of the Notch extracellular domain has been conducted in Drosophila or has addressed the function of NOTCH1, and information about events leading to Notch activation in mammalian cells and other Notch receptors is limited.

Notch1, 2 and 3 transcripts are detected in skeletal cells, and the receptors encoded by these genes regulate osteoblastogenesis and osteoclastogenesis [8,19]. Pathogenic variants of NOTCH2 and NOTCH3 associated with mutations in exon 34 and 33, respectively, result in a gain of Notch function due to the stabilization of their respective NICD and present with serious clinical manifestations affecting multiple organs including the skeleton [20,21]. The NOTCH2 pathogenic variants are associated with Hajdu Cheney Syndrome, and the NOTCH3 variants with Lateral Meningocele Syndrome. Mouse models of both entities exhibit substantial cancellous bone osteopenia. Pathogenic variants of NOTCH2 that lead to a loss of function are associated with Alagille Syndrome, which is most often caused by JAG1 mutations [22,23]. Mouse models haboring a heterozygous Jag1 allele in the context of a Notch2 hypomorphic allele phenocopy the human Alagille Syndrome [24]. These observations verify that normal NOTCH2 signaling is required for the normal development of multiple organs, including the skeleton.

In the present study, we report a child that presented with skeletal fragility and harbored a novel NOTCH2 variant of uncertain significance in exon 25 affecting the extracellular domain of NOTCH2. To define the consequences of the variant in bone metabolism, a mouse model (Notch2^em1Ecan) harboring the identical mutation present in the child was created. The phenotype of Notch2^em1Ecan mice was established by determining femoral bone microarchitecture using microcomputed tomography (μCT), femoral bone histomorphometry, mechanical testing and Raman spectroscopy. Possible mechanisms involved were explored in cultures of osteoblasts, chondrocytes and osteoclasts from control and experimental mice.

2. Case report

A full-term baby born by cesarean section following a 40-week normal pregnancy presented with a low-trauma femoral fracture at 2 months of age. The fracture occurred while diapers were being changed by a parent. Past medical and surgical history, as well as review of systems, were negative. Family history revealed a healthy mother, but the child’s father had suffered wrist and fibula/tibia fractures, the latter post-traumatic, and had undergone a bone biopsy with inconclusive results. Physical exam was unremarkable except for swelling of the left thigh. Weight was 4.7 kg (Z = −1.42) representing a 10 % gain over a 25-day period, height 0.56 m (Z = −1.48), and body mass index 15.05 Kg/m². Serum (reference laboratory in parenthesis) calcium 10.3 mg/dl (8.8–10.2), phosphorous 5.5 mg/dl (3.0–7.5), alkaline phosphatase 217 international units/lt (80–380), 25 hydroxy-vitamin D 13 ng/ml (20–52), 1,25 dihydroxy-vitamin D 62 pg/ml (25–66), intact parathyroid hormone (PTH) 20.8 pg/ml (15.0–65.0), collagen type I c telopeptide 677 pg/ml [25]. Xray skeletal survey revealed an increased number of Wormian bones in the skull, loss of height of a thoracic vertebra and a left femoral fracture. Bone mineral density of antero-posterior views of lumbar (L) 1–4 was 0.155 g/cm². The child was seen in follow-up at 8 months of age, when no additional fractures were reported. Gene analysis did not reveal the presence of known mutations associated with osteogenesis imperfecta, and specifically, no mutations in COL1a1 or COL1a2 [26–28]. Instead, a novel heterozygous missense substitution c.4006G > C, pG1336R (chr 1.120468433 hg 19) in exon 25 of NOTCH2 was found (Fig. 1). Additional genetic studies revealed the presence of the same variant in the child’s father; the mother was not affected. The c4006G > C, pG1336R variant is not reported in ClinVar or the Human Gene Mutation Database, indicating that it is a novel mutation [29]. The allele frequency is 3 in 1,614,036 alleles in the Genome Aggregation Database (gnomAD) (Version 4) or 0.0002 % of the alleles, consistent with an exceptionally rare disease [30]. In silico analysis was used to predict the possible pathogenicity of the G1336R mutation. The Z score for the allele in gnomAD is 6.05, indicating gene constraint for missense changes based on the lower-than-expected occurrence of the missense variant. Genomic evolutionary rate profiling GERP score 5.84, and phylogenic p value PlyloP score 5.6, both considered that the mutation as deleterious [31]. Polypher considered the mutation probably damaging and Sorting Intolerant from Tolerant (SIFT) predicted it to be deleterious [32,33]. Splice site analysis using Berkeley Drosophila Genome Project revealed a moderate reduction in acceptor site predictions from a score of 0.92 in native splice site score to 0.69 in the variant splice site [34]. Overall, the in silico analysis reveals that the c4006G > C, pG1336R NOTCH2 variant is likely deleterious. This report was prepared in accordance with the ethical guidelines of the Yale Institutional Review Board/Human Investigation Committee.

Fig. 1.

NOTCH2 domains depicting the G1336R substitution transcribed from the Notch2^4006G>C mutation. Reproduced with permission and modified from Zanotti and Canalis, Endocrine Reviews, 37:223, 2016.

3. Materials and methods

3.1. Notch2^em1Ecan mouse model

CRISPR/Cas9 technology was used to introduce the Notch2^4006G>C mutation into exon 25 of Notch2, duplicating the mutation found in the child described in the case report. A single-strand DNA fragment containing the 4006G > C mutation, Notch2 sgRNA, designed to cleave exon 25 of Notch2 adjacent to a protospacer adjacent motif at the junction of intron 24 and exon 25, and Cas9/ribonucleoprotein complex were co-injected into C57BL/6 J one-cell embryos at the Center for Mouse Genome Modification, UConn Health. Targeted embryos were transferred into CD1 pseudo-pregnant foster females, and founders were identified by genotyping and the proper insertion of the 4006G > C mutation was verified by targeted DNA sequencing of founders and F1 pups (Fig. 2); these were crossed with C57BL/6 J mice to establish mutant mouse lines. Mice of both sexes were studied independently in a C57BL/6 J genetic background following crosses of heterozygous Notch2^em1Ecan mice with wild type mice for the generation of heterozygous Notch2^em1Ecan mutant and control mice and following intercrosses of heterozygous Notch2^em1Ecan mice for the generation of homozygous Notch2^em1Ecan mice and littermate sex-matched controls. Genotyping was conducted in tail (adult) or skin (fetal) DNA using primers CCCGTGAGCCATCAGCCCTC and GGCATCTCGCCCCAGAGAAACG for mutant and CATTTCCTTGCCGTGTCATACGGG and CTGGCACATCCTGACTCGCAG for wild type sequences. Studies were approved by the Institutional Animal Care and Use Committee of UConn Health.

Fig. 2.

Sequencing of a DNA fragment spanning the 4006G > C pG1336R knock-in mutation in exon 25 of the Notch2^em1Ecan mouse line.

3.2. Prenatal skeletal phenotype

To visualize skeletons during fetal development, embryos were skinned and eviscerated and fixed in 100 % ethanol, dehydrated in acetone and stained for 10 days in 2 volumes of 0.14 % Alcian blue 8GX in 70 % ethanol, 1 volume of 0.12 % Alizarin red S in 95 % ethanol, 8 volumes of 100 % Glacial Acetic Acid, and 50 volumes of 70 % ethanol (Sigma Aldrich, St. Louis, MO). Embryos were placed in 1 % KOH in 20 % glycerol (Sigma-Aldrich) and stored in an ethanol, glycerol benzyl alcohol solution [35].

3.3. Microcomputed tomography (μCT)

Femoral microarchitecture was determined using a μCT instrument (Scanco μCT 40, Scanco Medical AG, Bassersdorf, Switzerland), calibrated at periodic intervals with a phantom provided by the manufacturer [36,37]. Femurs from control and Notch2^em1Ecan mice were scanned in 70 % ethanol at high resolution, energy level of 55 peak kilovoltage (kVp), intensity of 145 μA, and integration time of 200 ms as reported [1,38]. A total of 100 slices at midshaft or 160 slices at the distal metaphysis were acquired at an isotropic voxel size of 216 μm³ and a slice thickness of 6 μm and selected for analysis. Microarchitecture images of cancellous femoral bone were evaluated starting ~1.0 mm proximal from the femoral condyles. Contours were drawn manually every 10 slices, a few voxels away from the endocortical boundary, to define the region of interest for analysis, whereas the remaining slice contours were iterated automatically. Total volume, bone volume, bone volume fraction, trabecular thickness, trabecular number, connectivity density, structure model index, material density were measured in trabecular regions using a Gaussian filter (σ = 0.8) and defined thresholds. A threshold of 240 permil equivalent to 355.5 mg of hydroxyapatite (HA)/cm³ was used [36,37]. For analysis of cortical bone, contours were iterated across 100 slices along the cortical shell of the femoral midshaft, excluding the marrow cavity. Analysis of bone volume/total volume, porosity, cortical thickness, total cross-sectional and cortical bone area, segmented bone area, periosteal and endosteal perimeter, material density and polar moment of inertia (pMOI) were conducted using a Gaussian filter (σ = 0.8, support = 1) with a threshold of 400 permil equivalent to 682.9 mg of HA/cm³.

3.4. Bone Histomorphometry

Bone histomorphometry was conducted in 1-month-old mice injected with calcein 20 mg/kg and demeclocycline 50 mg/kg at a 48-h interval and sacrificed 48 h following the administration of demeclocycline. Dissected femurs were fixed in 70 % ethanol, embedded in methyl methacrylate, sectioned at a thickness of 5 μm along the sagittal plane on a Microm microtome (Richards-Allan Scientific, Kalamazoo, MI), and stained with 0.1 % toluidine blue. Static and dynamic parameters of bone morphometry were measured in a defined area between 0.35 mm and 2.16 mm from the growth plate at a magnification of 100× using an OsteoMeasure morphometry system (Osteometrics, Atlanta, GA). Stained sections were used to draw bone tissue contours and to measure trabecular separation, number and thickness, osteoid and eroded surface, as well as to count osteoblast and osteoclast number. Mineralizing surface per bone surface and mineral apposition rate were measured on unstained sections visualized under UV light and a triple diamidino-2-phenylindole/fluorescein/Texas red set long pass filter, and bone formation rate was calculated [39].

3.5. Mechanical testing

Three-point bend tests were performed on femurs from control and Notch2^em1Ecan mice using a Mach-1 mechanical testing system (Biomomentum, Laval, Quebec, Canada). Femurs that had been preserved in 70 % ethanol were placed in the 3-point bend system with a 6 mm span, such that the anterior face underwent tensile loading. Bones were loaded at a rate of 0.1 mm/s until failure. The forces were measured using a 25 kg load cell. Post-failure, the average cortical thickness and bone diameter at the point of fracture were measured using calipers. These values were used to calculate the second moment of inertia and average centroid of the bone assuming that it was a hollow cylinder tube. Force, displacement, and cross-sectional data were input into custom Matlab code developed in the Deymier laboratory to calculate structural (maximum force, yield force, stiffness and work) and material (maximum stress, resilience, modulus, toughness) mechanical properties.

3.6. Bone tissue composition by Raman spectroscopy

Raman spectroscopy was performed on femurs from Notch2^em1Ecan and control littermates, embedded in methyl methacrylate as described under bone histomorphometry and sectioned along the coronal plane into ~20 μm thick sections. Raman analysis was performed on a Witec alpha 300 Raman Spectrometer (Witec, Ulm, Germany) using a 785 nm laser at a power of ~40 mW with an acquisition time of 4 s × 25 acquisitions using a 50× objective [40,41]. For each slide, one section was selected for analysis. Measurements were made at 8–10 locations around the bone cortex, and the center of the cortical bone was analyzed whenever possible. The laser focus was manually adjusted at each location to maximize the quality of the acquired spectra. Following data acquisition, spectra were corrected for cosmic ray removal (setting 4/5) and background subtraction (shape 100) using the Witec Control 5 program. Five peaks were selected for analysis: 960 $\Delta {\text{cm}}_{\text{v1}}^{-1}$ phosphate in apatite peak, 1000 Δcm⁻¹ phenylalanine in collagen peak, 1070 Δcm⁻¹ carbonate in apatite peak, 1450 Δcm⁻¹ CH₂ bending in collagen peaks, and the 1670 amide I in collagen peak. The peaks were fit using a Lorentzian peak fit in Witec Control 5. Spectra were batch-fit by slide to minimize variations, and the peak center, area, and width were recorded.

3.7. Calvarial-derived osteoblast cell cultures

Osteoblast-enriched cells were isolated from parietal bones from 3- to 5-day-old Notch2^em1Ecan and control littermate mice. Bones were exposed to five consecutive reactions with 1.2 units/ml Liberase TL (Sigma-Aldrich, St. Louis, MO) for 20 min at 37 °C [42,43]. Cells from the last three digestions were considered enriched in osteoblasts, pooled and seeded at a density of 10,000 cells/cm² and cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with non-essential amino acids (both from Life Technologies, Thermo Fisher Scientific, Grand Island, NY), ascorbic acid and 10 % heat-inactivated fetal bovine serum (FBS, Atlanta Biologicals, Norcross, GA) in a 5 % CO₂ atmosphere at 37 °C. In one experiment, cells were cultured on culture plates coated with Dll1-IgG2A fragment crystallizable (Fc) recombinant fusion protein 125 ng/cm² (R&D Systems, Minneapolis, MN), as described [44]. Cell extracts were obtained for RNA determinations.

3.8. Chondrocyte cultures

Chondrocytes were isolated from the epiphyseal cartilage of long bones of the hind and fore limbs from 3- to 4-day-old Notch2^em1Ecan and control littermate mice under a Unitron Z850 stereo microscope (Commack, NY). Tissue was collected in high-glucose-containing DMEM and digested with 0.25 % trypsin/EDTA (Life Technologies, Thermo Fisher Scientific), and 200 U/ml of collagenase type II (Worthington Biochemical Corporation, Lakewood, NJ) at 37 °C as described [45]. Following straining through a 70 μm membrane, cells were collected by centrifugation, seeded at a density of 110,000 cells/cm² and cultured in DMEM in the presence of 10 % heat inactivated FBS at 37 °C in a humidified 5 % CO₂ incubator [46,47].

3.9. Bone marrow-derived macrophage (BMM) cultures and osteoclast formation

BMMs were obtained by flushing the marrow of Notch2^em1Ecan mice and littermate controls and subsequent lysis of erythrocytes as described previously [48]. Cells were recovered by centrifugation and suspended in α-minimum essential medium (α-MEM, Life Technologies) containing 10 % FBS and cultured on plastic petri dishes at a density of 8 × 10⁶ cells/cm² in the presence of recombinant human macrophage-colony stimulating factor (M-CSF) at 30 ng/ml for 3 to 4 days. M-CSF cDNA and expression vector were obtained from D. Fremont (St. Louise, MO), and M-CSF was purified as previously reported [49]. Cells were collected following exposure to 0.25 % trypsin/EDTA and seeded on tissue culture plates at a density of 300,000 cells/cm² in α-MEM with 10 % FBS, M-CSF at 30 ng/ml and recombinant murine receptor activator of NF-κB ligand (RANKL) at 10 ng/ml. Tnfsf11 encoding RANKL cDNA and expression vector were obtained from M. Glogauer (Toronto, Canada), and RANKL purified as described [50]. Cultures were conducted until the formation of multinucleated cells positive for tartrate resistant acid phosphatase (TRAP), determined using a commercial kit (Sigma-Aldrich, St. Louis, MO).

3.10. Quantitative reverse transcription–polymerase chain reaction (qRT-PCR)

Total RNA was extracted from cells with the RNeasy kit (Qiagen, Valencia, CA), in accordance with manufacturer’s instructions, as previously reported [1,42,51,52]. Equal amounts of RNA were reverse transcribed using the iScript RT-PCR kit (Bio-Rad, Carlsbad, CA) and amplified in the presence of specific primers (Integrated DNA Technologies, IDT, Coralville, IA) (Table S1) with SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) at 60 °C for 35 cycles. Transcript copy number was estimated by comparison with a serial dilution of cDNA for Alpl, Bglap, Col2a1, Hes1 (American Type Tissue Culture Collection, ATCC, Manassas, VA), Hey1 and Hey2 (T. Iso, Gunma University, Gunma, Japan), Heyl (from D. Srivastava, Gladstone Institute of Cardiovascular Disease, San Francisco, CA or Dharmacon, Lafayette, CO), Acan, Notch2 and Sox9 (Thermo Fisher Scientific), Tnfsf11 (Source BioScience, Nottingham, UK) [53,54]. Amplification reactions were conducted in a CFX96 qRT-PCR detection system (Bio-Rad), and fluorescence was monitored during every PCR cycle at the annealing step. Data are expressed as copy number corrected for Rpl38 (from ATCC) [55].

To measure Notch2^4006G>C mutant transcripts, total RNA was reverse transcribed, as described, and Notch2 and Notch2^4006G>C cDNA was amplified by PCR in the presence of Sso Advanced Universal Probes Supermix (BioRad) gene expression assay mix, and HEX labeled Notch2 (TTGTCCCCCAGGGTTC) and FAM labeled Notch2^4006G>C (TCCCCCACGGTTCTC) probes (BioRad) at 95 °C for 10 s then 60 °C for 30 s and repeated for 45 cycles [56]. Notch2 or Notch2^4006G>C mutant transcript copy number was estimated by comparison to a serial dilution of a 100 to 200 base pair (bp) synthetic DNA fragment (IDT) with or without the mutation in the Notch2 locus cloned in pcDNA1.3(−) and copy number was corrected for Rpl38 [57].

3.11. Collagen analysis

To analyze the collagen composition of cells from Notch2^em1Ecan mice, osteoblast-enriched cells were cultured to confluence, and medium and cell extracts were precipitated with ammonium sulphate (Sigma Aldrich), as described [58,59]. Collagen was digested with pepsin 0.1 μg/ml in 0.5 M acetic acid (Sigma Aldrich) and precipitated with NaCl (0.9 M) in 0.5 M acetic acid, lyophilized, resuspended in sample buffer with the addition of Dylight 550 NHSEster dye (Life Technologies) and fractionated by electrophoresis on 6 % polyacrylamide gels (PAGE) following denaturation. Collagen chains were visualized by fluorescence on a Chemidoc MP Imaging System (BioRad).

3.12. Protein modeling

Structural models of wild-type and mutant NOTCH2 were generated using AlphaFold 3 (https://alphafoldserver.com/) [60]. The input sequence corresponded to the EGF33–36 repeats of NOTCH2 (Uniprot ID: Q04721; residues 1264–1412), in both wild-type and mutant forms. Five models were predicted for each protein sequence, and only the top-ranked model is included. Model accuracy was assessed using the predicted local distance difference tests (pLDDT) metric, a per-atom confidence score on a 0–100 scale, with higher values indicating greater confidence. Structural superimposition and molecular rendering were conducted with PyMOL (Schrödinger, New York, NY).

3.13. Statistics

Data are expressed as individual sample values and means ± SD. Data represent biological replicates except for cell cultures, which represent technical replicates. Statistical differences were determined by unpaired Student’s t-test for pairwise comparisons. The chi-square test was applied to compare the observed genotypes with the expected ratios according to a Mendelian model.

4. Results

4.1. Generation and general appearance of Notch2^em1Ecan mutant mice

CRISPR/Cas9 technology was used to introduce the 4006G > C mutation into the murine Notch2 locus to reproduce the NOTCH2 mutation described in the case report. The presence of the mutation was verified by DNA sequencing, as shown in Fig. 2. The variant leads to a G1336R substitution in EGF 34 of NOTCH2 (Fig. 1). The general appearance of Notch2^em1Ecan mice was comparable to that of wild-type controls; however, a small (13 %) decrease in weight of Notch2^em1Ecan homozygous mice was noted. The femoral length of homozygous male and female Notch2^em1Ecan mice was not different from that of control mice (Fig. 3). Notch2^4006G>C transcripts were detected in tibiae from Notch2^em1Ecan homozygous mice and not in control mice; these expressed Notch2 wild-type mRNA.

Fig. 3.

Weight, femoral length, and identification of Notch2^em1Ecan alleles by genotyping of tail DNA and mRNA levels in bone extracts. Left panels, body weight and femoral length of 1-month-old homozygous Notch2^em1Ecan (black bars) and sex-matched controls (white bars); males n = 11 control and n = 7 Notch2^em1Ecan; females n = 19 control and n = 13 Notch2^em1Ecan. On the right panels, tail DNA was obtained from Notch2^em1Ecan and control mice for genotyping, and total RNA was obtained from tibiae of Notch2^em1Ecan (black bars) and sex- and age-matched control (white bars) mice. Bars represent means ± SD and individual values are shown as open circles. *Significantly different between Notch2^em1Ecan and control by unpaired t-test, p < 0.05.

Following the intermatings of Notch2^em1Ecan heterozygous mice, genotyping was conducted at weaning, except for select cell culture experiments where genotyping was done at days 3–5 postnatally. Genotyping of 41 litters (375 pups) demonstrated a lower-than-expected yield of homozygous mice; 27 (7 %)were homozygous and 218 (58 %) were heterozygous for the G1336R mutation, and 130 (35 %) were wild type mice. The observed genotype is significantly different from the expected Mendelian ratios of 25 %, 50 % and 25 %, respectively, p < 0.05, χ² 24.9. Although there was no obvious evidence of postnatal lethality, mice were not observed continuously and postnatal lethality cannot be excluded with certainty. Because of the low litter yield of homozygous mutant Notch2^em1Ecan mice, and to exclude whether the Notch2^4006G>C mutation had an impact on prenatal skeletal development, that could influence the postnatal phenotype, control and Notch2^em1Ecan embryos were obtained and stained with alizarin red and alcian blue. Following two rounds of Notch2^em1Ecan heterozygous intermatings, 23 embryos were obtained; 2 homozygous and 13 heterozygous for the mutation and 8 wild type embryos for a ratio of 9 %, 56 % and 35 %, respectively. Resorbed embryos were not noted. The ratios are similar to those observed prenatally, suggesting that loss of homozygous Notch2^em1Ecan mice occurred during embryonic life. There were no appreciable differences between either heterozygous or homozygous Notch2^em1Ecan embryos at 17 days of embryonic life (not shown) or 20 days and control littermates (Fig. 4), indicating that there was not a major embryonic developmental impact of the Notch2 mutation on the skeleton.

Fig. 4.

Representative images of Notch2^em1Ecan embryos. Twenty-day-old embryos from heterozygous and homozygous Notch2^em1Ecan and control littermates stained with alcian blue and alizarin red are shown. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4.2. Skeletal microarchitecture and bone Histomorphometry of Notch2^em1Ecan

Femoral microarchitecture, determined by μCT, of 1-month-old male homozygous Notch2^em1Ecan mice revealed a ~ 25 % decrease in trabecular bone volume/total volume (BV/TV), a decrease in trabecular thickness, and an increase in structure model index, indicating a tendency toward rod-like trabeculae (Fig. 5, Table 1). μCT of 1-month-old homozygous female mice revealed no alterations in cancellous bone when compared to control mice (Table 1). In contrast, cortical bone architecture revealed that both male and female mice had decreased total area, bone area, and marrow area as well as cortical thickness and periosteal perimeter, indicating that bones from Notch2^em1Ecan mice were smaller than controls. Density of material and pMOI were decreased, indicating that bones from Notch2^em1Ecan mice were potentially fragile [61]. Since the body weight of homozygous Notch2^em1Ecan mice was modestly lower than that of controls, correction of the microarchitectural parameters described was minimized if corrected for body weight. μCT analysis of heterozygous male and female mice did not reveal alterations in either cancellous or cortical bone when compared to control littermate mice (not shown).

Fig. 5.

Representative μCT of femurs from 1-month-old male control and homozygous Notch2^em1Ecan mice. A sagittal cut is shown for cancellous bone, and a cross-sectional cut at mid-diaphysis is shown for cortical bone.

Table 1.

Femoral microarchitecture assessed by μCT of 1-month-old homozygous Notch2^em1Ecan mutant mice and sex-matched littermate controls.

	Males		Females
	Control n = 11	Notch2em1Ecan n = 7	Control n = 19	Notch2em1Ecan n = 13
Distal Femur Trabecular Bone
Bone Volume/Total Volume (%)	12.7 ± 3.7	9.4 ± 2.6*	9.9 ± 1.3	9.4 ± 2.1
Trabecular Separation (μm)	196 ± 19	208 ± 32	208 ± 12	210 ± 33
Trabecular Number (1/mm)	5.2 ± 0.5	4.9 ± 0.7	4.8 ± 0.3	4.9 ± 0.6
Trabecular Thickness (μm)	35 ± 4	32 ± 2*	32 ± 2	31 ± 1
Connectivity Density (1/mm3)	357 ± 95	271 ± 94	311 ± 55	310 ± 88
Structure Model Index	1.9 ± 0.3	2.2 ± 0.2*	2.0 ± 0.2	2.2 ± 0.2
Density of Material (mg HA/cm3)	786 ± 15	777 ± 18	791 ± 13	786 ± 19
Femoral Midshaft Cortical Bone
Bone Volume/Total Volume (%)	84.1 ± 1.6	81.2 ± 3.3	83.7 ± 1.7	81.7 ± 2.2*
Porosity (%)	15.9 ± 1.6	18.8 ± 3.3	16.3 ± 1.7	18.3 ± 2.2*
Cortical Thickness (μm)	102 ± 11	89 ± 11+	95 ± 7	87 ± 8*
Total Area (mm2)	1.5 ± 0.1	1.2 ± 0.1*	1.4 ± 0.1	1.3 ± 0.1*
Bone Area (mm2)	0.5 ± 0.07	0.4 ± 0.05*	0.5 ± 0.04	0.4 ± 0.09*
Marrow Area (mm2)	1.0 ± 0.07	0.8 ± 0.1*	0.9 ± 0.06	0.8 ± 0.12*
Periosteal Perimeter (mm)	4.3 ± 0.1	3.9 ± 0.2*	4.2 ± 0.1	4.0 ± 0.2*
Endocortical Perimeter (mm)	3.5 ± 0.1	3.2 ± 0.2*	3.4 ± 0.1	3.3 ± 0.2*
Density of Material (mg HA/cm3)	976 ± 19	949 ± 25*	983 ± 28	956 ± 25*
pMOI (mm4)	0.176 ± 0.03	0.111 ± 0.03*	0.148 ± 0.04	0.118 ± 0.02*

μCT was performed on distal femurs for trabecular bone and midshaft for cortical bone. Values are means ± SD

^*Significantly different from control by unpaired t-test, p < 0.05

⁺p 0.06.

Because a trabecular bone phenotype was present only in male mice, cancellous bone histomorphometry was conducted in femurs from male mice. Histomorphometry confirmed a decrease in bone volume and trabecular thickness observed with the μCT analysis and demonstrated a decrease in eroded surface and osteoclast number/bone area (Table 2).

Table 2.

Cancellous femoral bone histomorphometry of 1-month-old homozygous Notch2^em1Ecan mutant male mice and sex-matched littermate controls.

	Control	Notch2em1Ecan
Bone Volume/Tissue Volume (%)	18.4 ± 8.8	10.7 ± 5.5+
Trabecular Number (1/mm)	4.6 ± 1.5	3.6 ± 1.3
Trabecular Thickness (μm)	38.7 ± 6.4	28.3 ± 4.8*
Osteoblasts/Bone Perimeter (1/mm)	15.1 ± 4.2	12.7 ± 2.2
Osteoblasts/Trabecular Area (1/mm2)	141 ± 71	88 ± 25
Osteoid Surface/Bone Surface (%)	1.3 ± 1.5	0.8 ± 0.6
Osteoclasts/Bone Perimeter (1/mm)	3.1 ± 0.6	2.7 ± 0.7
Osteoclasts/Trabecular Area (1/mm2)	26.7 ± 4.0	18.3 ± 4.9*
Eroded Surface/Bone Surface (%)	1.9 ± 0.6	1.2 ± 0.4*
Mineral Apposition Rate (μm/day)	1.8 ± 0.3	1.6 ± 0.3
Mineralizing Surface/Bone Surface (%)	4.6 ± 1.5	3.6 ± 1.3
Bone Formation Rate (μm3/μm2/day)	0.04 ± 0.02	0.05 ± 0.02

Bone histomorphometry was performed on sagittal sections of distal femurs. Values are means ± SD; n = 6 for control, n = 5 for Notch2^em1Ecan mutant.

^*Significantly different from control by unpaired t-test, p < 0.05

⁺p 0.1.

4.3. Mechanical testing of femurs from Notch2^em1Ecan mice

Three-point bend tests of femurs revealed that the toughness of the bones was significantly reduced (45 %) in samples from Notch2^em1Ecan female mice as compared to controls (Table 3). There was a trending reduction in the yield strain in femurs from Notch2^em1Ecan mice as compared to controls, but there were no differences in the max force, field force, stiffness, work, max stress, modulus, resilience, or yield stress between control and Notch2^em1Ecan mice.

Table 3.

Mechanical testing assessed by 3-point bend testing of 1-month-old homozygous female Notch2^em1Ecan mutant mice and sex matched controls.

	Control (n = 4)	Notch2em1Ecan (n = 3)
Maximal Force (N)	6.6 ± 0.8	5.8 ± 0.4
Stiffness (N/m)	23.9 ± 7.1	23.4 ± 3.6
Yield Force (N)	4.9 ± 0.5	4.3 ± 1.1
Work (Nm)	0.6 ± 0.2	0.4 ± 0.2
Max Stress (MPa)	199.1 ± 28.6	189.3 ± 14.9
Modulus (MPa)	1733.7 ± 940.2	2112.7 ± 236.4
Resilence (MPa)	7.9 ± 4.0	5.0 ± 2.3
Toughness (MPa)	24.4 ± 5.8	13.2 ± 1.3*
Yield Strain (–)	0.10 ± 0.03	0.06 ± 0.02+
Yield Stress (MPa)	148.7 ± 5.5	141.8 ± 32.5

Three-point bend tests were performed on femurs from control and Notch2^em1Ecan mice. Values are means ± SD.

^*Significantly different from control by unpaired t-test, p < 0.05

⁺p 0.065.

4.4. Raman spectroscopy from femurs from Notch2^em1Ecan

Comparison of the spectra of the homozygous Notch2^em1Ecan bone samples and controls revealed an overlay of the average spectra for the mutant and control groups normalized to the height of the 960 peak. There were no apparent differences between control and Notch2^em1Ecan and control samples in the relative height and shape of the peaks (not shown). Quantitative analysis confirmed the initial assessment. The 960 peak location and peak width were not different between control and Notch2^em1Ecan samples, suggesting that there was no change in the maturity of the crystalline matrix in Notch2^em1Ecan, and both control and experimental samples exhibited mineral with a high level of crystallinity. The carbonate to phosphate ratio was not different in Notch2^em1Ecan and control samples, suggesting that there was no significant difference in the rate of bone turnover between the sample groups.

To ensure that possible modifications in collagen would have no effect on the mineral:matrix values, they were measured using ratios of the 960 peak to the 1000 phenylalanine, 1450 CH₂, and1675 Amide I collagen peaks. None of the mineral:matrix values were different between the control and Notch2^em1Ecan bones, suggesting that the mineralization level was not affected (Fig. 6). This suggests that the decrease in the mineral density by μCT scanning is not indicative of a change in the matrix but is instead caused by structural differences. The relative levels of phenylalanine, CH₂, and Amide I were obtained by taking relative ratios of the collagen peaks and no significant differences were found between the Notch2^em1Ecan and control samples, suggesting that there were no significant alterations in the collagen structure of bones from Notch2^em1Ecan mice.

Fig. 6.

Raman spectroscopy showing mineral:matrix peak ratios (left) and collagen peak ratios (right) in cortical bone from homozygous Notch2^em1Ecan male and control littermates. Values means ± SD and ranges; n = 4 biological replicates.

4.5. Osteoblast-enriched cell cultures from Notch2^em1Ecan mice

To determine the direct effect of the Notch2^4006G>C variant in cells of the osteoblast lineage, calvarial osteoblasts from Notch2^em1Ecan and control mice were isolated and cultured for 14 days. Wild type Notch2 transcripts were not detected in Notch2^em1Ecan cells and lower levels of the Notch target genes Hes1 and Hey2 mRNA were expressed in mutant than in control cells, suggesting decreased canonical Notch signal activation (Fig. 7). Osteoblastogenesis was decreased by the Notch2^4006G>C variant and Alp1, Bsp1 and Col1a1 mRNA levels were lower in Notch2^em1Ecan osteoblasts than in control cultures (Fig. 7). Although Col1a1 mRNA expression was decreased, collagen chains, examined by PAGE, were not affected by the Notch2^4006G>C variant and were detected in the cell layer but mostly in the culture medium, indicating normal collagen secretion (Fig. 7). Tnfsf11, encoding RANKL, was reduced in Notch2^em1Ecan osteoblasts. To ensure Notch activation, calvarial osteoblasts were cultured on Dll1-IgG2A and under these conditions Notch2^em1Ecan expressed lower levels of the Notch target genes Hes1, Hey1 and Hey2 suggesting decreased canonical Notch signal activation in Notch2^em1Ecan cells (Fig. 8).

Fig. 7.

Osteoblast-enriched cells harvested from 3- to 4-day-old homozygous Notch2^em1Ecan (black bars) and control littermate mice (white bars) cultured for 14 days following confluence (Day 0). Total RNA was extracted and gene expression determined by qRT-PCR. Data are expressed as Notch2, Notch2^4006G>C, Hes1, Hey1, Hey2, Alpl, Bglap, Col1a1 and Tnfsf11 copy number corrected for Rpl38. Bars represent means ± SD and individual values are shown as open circles.; n = 4 technical replicates. *Significantly different between Notch2^em1Ecan and control by unpaired t-test, p < 0.05. In the right lower corner, an image of α1 and α2 collagen type I chains fractionated by PAGE and visualized by fluorescence obtained from the cell layer and medium of control and Notch2^em1Ecan osteoblasts.

Fig. 8.

Osteoblast-enriched cells harvested from 3- to 4-day-old homozygous Notch2^em1Ecan (black bars) and control littermate mice (white bars) cultured for 3 days following confluence on Dll1-IgG_2a coated culture plates. Total RNA was extracted and gene expression determined by qRT-PCR. Data are expressed as Hes1, Hey1, and Hey2, copy number corrected for Rpl38. Heyl was not detected. Bars represent means ± SD and individual values are shown as open circles; n = 4 technical replicates. *Significantly different between Notch2^em1Ecan and control by unpaired t-test, p < 0.05.

4.6. Chondrocyte-enriched cell cultures from Notch2^em1Ecan mice

To determine whether the Notch2^4006G>C variant modified chondrogenesis, chondrocytes from Notch2^em1Ecan and control mice were cultured for 28 days. Wild type Notch2 transcripts were not detected in Notch2^em1Ecan cells that instead expressed Notch2^4006G>C mRNA. Confirming the results observed in osteoblasts, Hey1, Hey2 and Heyl mRNA levels were not increased in chondrocytes from Notch2^em1Ecan mice when compared to control cells, except for a modest and transient increase in Hey2 (Fig. 9). The mRNA expression of Sox9 was suppressed during the initial phase of the culture, whereas Col2a1 and Acan transcripts were decreased throughout most of the culture period in Notch2^em1Ecan chondrocytes when compared to controls indicating a level of impairment in chondrocyte differentiation.

Fig. 9.

Chondrocyte-enriched cells harvested from epiphyseal cartilage from 3- to 4-day-old homozygous Notch2^em1Ecan (black bars) and control littermate (white bars) mice were cultured for 28 days following confluence (Day 0). Total RNA was extracted and gene expression determined by qRT-PCR. Data are expressed as Notch2, Notch2^4006G>C, Hey1, Hey2, Heyl, Sox9, Col2α1 and Acan copy number corrected for Rpl38. Bars represent means ± SD and individual values are shown as open circles; n = 4 technical replicates. *Significantly different between Notch2^em1Ecan and control by unpaired t-test, p < 0.05.

4.7. Bone marrow-derived macrophages from Notch2^em1Ecan mice

To determine the contributions of the Notch2^4006G>C variant to osteoclastogenesis, bone marrow-derived macrophages (BMM) from Notch2^em1Ecan and wild type mice were cultured in the presence of M-CSF and RANKL. Multinucleated TRAP positive cells differentiated in control and mutant BMMs, although a lower number of mature osteoclasts was detected in Notch2^em1Ecan than in control cells (Fig. 10).

Fig. 10.

Bone marrow-derived macrophages (BMMs) from 4-week-old homozygous Notch^em1Ecan (black bars) and control littermate (white bars) mice were cultured for 4 to 5 days in the presence of M-CSF at 30 ng/ml and RANKL at 10 ng/ml in cell culture plates, stained with TRAP and counted. TRAP-positive cells with more than 3 nuclei were identified as osteoclasts. Bars represent means ± SD and individual values are shown as open circles.; n = 4 technical replicates for control and Notch2^em1Ecan mice. *Significantly different between Notch2^em1Ecan and control by unpaired t-test, p < 0.05. To the left, representative images of osteoclasts from Notch2^em1Ecan and control. Bar 200 μm.

4.8. Protein structural predictions of the NOTCH2 EGF33–36 repeats

The structure of the NOTCH2 EGF34 repeat has not been determined experimentally, and the mechanism by which the G1336R mutation alters NOTCH2 activity remains unclear. This was addressed by using AlphaFold 3 to generate structural models of the wild-type and mutant EGF33–36 repeats of NOTCH2 (Fig. 11). The predicted 3D models of both wild-type and mutant repeats retained the near-linear domain organizational characteristic of EGF repeat-containing structures [62]. In the wild-type model, G1336 is positioned in a β-hairpin loop between two short β strands (Fig. 11A). Glycine at this position confers backbone flexibility, whereas arginine reduces flexibility and alters the hinge region between EGF repeats 34 and 35 (Fig. 11B). While the overall fold of the EGF34 repeat is preserved, the G1336R substitution restricts the bending angle between adjacent repeats, reducing the bent conformation observed in the wild-type repeats and producing a more rigid, extended arrangement of the repeats (Fig. 11B). As a consequence, EGF36, the last EGF repeat of the extracellular domain of NOTCH2, adopts a distinct orientation relative to EGF33 (Fig. 11C). This structural difference may underlie the functional consequences of the mutation described. The extracellular-transmembrane junction of NOTCH2 contains a NRR, which harbors the buried metalloprotease cleavage site S2 essential for receptor activation [63]. In the resting state, NRR is maintained in a protease-resistant conformation, sandwiched between the EGF repeats and the transmembrane domain. Ligand binding triggers a large conformational movement to expose the S2 site for cleavage. By restricting hinge flexibility between EGF34–35–36, the G1336R substitution may limit reorientation of EGF36, and thereby impair the conformational transition required for NRR activation. Further experimental validation will be required to test this structural hypothesis.

Fig. 11.

Structural models of wild-type and G1336R mutant EGF33–36 repeats of human NOTCH2. AlphaFold 3 model of wildtype (A) and G1336R mutant (B) are colored based on the pLDDT per position. Mapped pLDDT values range from blue, which are associated with high confidence predictions, progressing through yellow and red for less confident predictions. The G1336 (A) and R1336 (B) are shown in space-filling representation. (C) Superimposition of wildtype (cyan) and mutant (sand) models by alignment of the N-terminal EGF33 repeat. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

5. Discussion

In the present study, we report a child with a novel heterozygous NOTCH2 variant presenting with skeletal fragility. The NOTCH2 variant was transmitted by autosomal dominant inheritance by an affected father; the mother was not affected. The 4006G > C variant leads to a G1336R substitution in epidermal growth factor (EGF) 34 present in the extracellular domain of NOTCH2 (Fig. 1). Glycine to arginine mutations are common causes of diseases associated with pathogenic variants and may cause significant structural changes in transmembrane and matrix proteins [64,65]. In silico analysis revealed that the NOTCH2 variant described is deleterious. A structure homology model of the wild-type and mutant EGF33–36 repeats of NOTCH2 suggests that the G1336R substitution may alter the orientation of EGF36, thereby disrupting the conformational change in the NRR that is required for NOTCH2 activation and downstream signaling.

To explore the consequences of the NOTCH2 variant, we created and validated a knock-in mouse model harboring the Notch2 4006G > C mutation and termed Notch2^em1Ecan. The murine and human NOTCH2 are highly conserved allowing the introduction of the human (NOTCH2^4006G>C) mutation into the mouse genome. Although heterozygous Notch2^em1Ecan mice did not exhibit a phenotype, male and female homozygous Notch2^em1Ecan mutant mice exhibited small and potentially fragile bones, reproducing functional outcomes of the human disease and validating the model for the study of the novel NOTCH2 variant. Importantly, the phenotype affected primarily cortical bone and cancellous bone osteopenia was modest and observed in male but not female Notch2^em1Ecan mice. Accordingly, cancellous bone histomorphometry revealed modest changes except for a decrease in eroded surface. Since the variant was transmitted by autosomal dominant inheritance, the lack of a phenotype in heterozygous mutant mice would suggest a greater degree of phenotypic penetrance in humans than in rodents.

The phenotype revealed significant decreases in pMOI in Notch2^em1Ecan mice indicating possible skeletal fragility [61]. Mechanical testing performed on femurs from female Notch2^em1Ecan mice confirmed a significant decrease in bone toughness indicating that the bone was more susceptible to brittle fracture and unable to absorb energy before breaking. To exclude prenatal developmental influences of the Notch2^4006G>C mutation, skeletons from Notch2^em1Ecan and control embryos were stained with alcian blue and alizarin red to detect cartilage and mineralized tissue, but we found no differences between the Notch2^em1Ecan mutant and control mice. However, the prenatal and postnatal yield of homozygous Notch2^em1Ecan mice following heterozygous intermatings was low, suggesting embryonic lethality.

There was no evidence of Notch gain-of-function in cells from Notch2^em1Ecan mice and osteoblast cultures revealed decreased expression of canonical Notch target genes when osteoblasts were cultured on plates coated with Dll1-Fc to ensure Notch signal activation. The results are in agreement with the structure homology model demonstrating that the GI336R substitution makes the NOTCH2 receptor rigid, possibly impairing the exposure of the NRR to proteolytic enzymes and consequent Notch activation. It is also possible that the rigidity of the extracellular domain affects its contact with ligands of the Jagged and Delta-like families or that the Notch2 mutation led to the activation of alternate non-canonical pathways, but this possibility was not explored in the current study.

Although the clinical features of the case reported are compatible with the diagnosis of osteogenesis imperfecta, exome sequencing did not reveal the presence of any known gene mutation associated with the disease, and specifically, no mutations in COL1a1 or COL1a2 were found [26–28]. Raman spectroscopy and confocal microscopy of femoral sections from Notch2^em1Ecan mice using backscattered second harmonic generation modules did not detect abnormal collagen structure or fibers (not shown), and fractionation of collagen chains secreted by osteoblasts did not detect any abnormalities. This would indicate that the NOTCH2 4006G > C mutation does not cause a classic form of osteogenesis imperfecta. Whereas osteogenesis imperfecta is most often associated with mutations in COL1A1 or COL1A2, occasionally pathogenic variants of unrelated genes have been reported to be associated with the disease [26–28]. Indeed, ~20 % of cases are associated with mutations in alternate genes, some without an effect on collagen synthesis, maturation or processing [28,66]. They constitute forms of non-classical osteogenesis imperfecta and contribute to the heterogeneity of the disease.

NOTCH2 enhances osteoclastogenesis, through its effects in cells of the myeloid lineage, mediated in part by HES1, and indirectly by inducing RANKL in osteoblasts; consequently, NOTCH2 gain-of-function mutations are associated with increased osteoclast number and bone resorption [4,67,68]. Cultures of BMMs from Notch2^em1Ecan mice exhibited decreased osteoclastogenesis, and osteoblast cultures expressed decreased Tnfsf11, encoding RANKL. These findings could possibly be responsible for the decreased eroded surface detected by femoral histomorphometry and are consistent with a decrease in NOTCH2 canonical signaling. Osteoblast and chondrocyte cultures from Notch2^em1Ecan revealed a decrease in the expression of genes associated with cell differentiation and maturation. The mechanism was not explored, but if indeed the Notch2^4006G>C mutation results in an impairment of Notch canonical signaling, the finding would suggest that optimal NOTCH2 signaling is required for osteoblastogenesis and chondrogenesis. It is also possible that Notch2^4006G>C variant resulted in the activation of non-canonical pathways responsible for the phenotype observed.

There are some limitations to the work described, including the description of the case reported which was limited due to the information available, and was associated with a dominant mutation, whereas Notch2^em1Ecan heterozygous mice did not manifest a phenotype. The Notch2^em1Ecan mouse line was studied only at 1 month of age and the number of mice studied was limited due to the finite availability of homozygous mutant mice and difficulties in maintaining a viable homozygous line. The basis for the reduced viability was not defined. Another limitation of the work was the fact that only samples from female mice were available for mechanical testing, resulting in underpowering of the testing, and femurs had been preserved in 70 % ethanol, possibly influencing the parameters tested, albeit equally in control and experimental samples [69].

In conclusion, a novel NOTCH2 deleterious variant is expected to be associated with skeletal fragility, and mouse lines harboring the mutation exhibit small and apparently fragile bones possibly due to alterations in Notch activation and canonical signaling.

Supplementary Material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.bone.2025.117702.

Abbreviations

α-MEM

α-minimum essential medium

bp

base pair

BMM

bone marrow-derived macrophages

DMEM

Dulbecco’s modified Eagle’s medium

EGF

epidermal growth factor

FBS

fetal bovine serum

Fc

fragment crystallizable

HA

hycroxyapatite

kVp

kilovoltage

M-CSF

macrophage-colony stimulating factor

μCT

microcomputed tomography

NICD

Notch intracellular domain

NRR

negative regulatory region

PTH

parathyroid hormone

pMOI

polar moment of inertia

PAGE

polyacrylamide gel electrophoresis

pLDDT

predicted local difference test

qRT-PCR

quantitative reverse transcription-polymerase chain reaction

RANKL

receptor activator of NF-κB ligand

RBPJΚ

recombination signal-binding protein for Ig of Κ

SIFT

Sorting Intolerant from Tolerant

TRAP

tartrate resistant acid phosphatase

Data availability

All data are available from the corresponding author upon a reasonable request.