Blood samples of the patient and both parents as well as a transiliac bone and skin sample of the patient were obtained postmortem using standard techniques. A transiliac bone sample of an age-matched infant was identified from an ethically approved postmortem study on sudden infant death and used as a control for comparison of histomorphometric and BMDD outcomes. A skin biopsy was taken from an age- and gender-matched healthy infant. Written informed consent was obtained from the patient’s and control’s parents/guardian.

The transiliac bone sample was analyzed by histomorphometry and quantitative backscattered electron imaging (qBEI) at the Ludwig Boltzmann Institute of Osteology (LBIO) in Vienna using standard procedures (20–22).

In short, the sample was fixed in 70% ethanol, dehydrated in graded series of alcohol, and embedded in methyl-methacrylate. Static histomorphometric parameters of bone formation and resorption, as well as adipocyte surface in the total bone marrow space, were evaluated in undecalcified 3–5 μm thick sections stained with Goldner’s trichrome using a light microscope (Axiophot, Zeiss, Oberkochen, Germany) equipped with a digital camera (AxioCam HRc, Zeiss, Oberkochen, Germany) (22). For assessment of bone marrow adipocytes, optical images (1.5 mm²) from the sections were taken (n = 3) and analyzed with an in-house custom-made macro for ImageJ (version1.53) using the BioVoxxel Toolbox (23) to detect semiautomatically the adipocyte and to calculate the adipose tissue area as a percentage of the total marrow area. The minimal size of an adipocyte was defined at 525 µm². Structural histomorphometric parameters were calculated from the qBEI image. The residual bone sample block was prepared for qBEI. The surface plane containing bone tissue obtained after grinding and polishing was coated by carbon prior to analysis with a field emission scanning electron microscope (FESEM Supra 40, Zeiss, Oberkochen, Germany) equipped with a four-quadrant semiconductor backscatter electron detector. The Zeiss Supra 40 was operated with an electron energy of 20 keV. The gray levels reflecting the mineral/calcium content were calibrated by the material contrast of pure carbon and aluminum. Gray-level histograms were deduced and transformed into calcium weight percent (weight% calcium) histograms. The entire cross-sectional area of the transiliac bone sample was imaged with a spatial resolution of 1.8 μm per pixel. From the acquired image, parameters for histomorphometric indices of mineralized bone structure and the bone mineralization density distribution (BMDD) were evaluated. The BMDD parameters were measured in trabecular and cortical compartments separately. For cortical bone, the arithmetic mean was calculated from both cortices. All obtained parameters were compared with the bone sample of the age-matched control, as well as to available pediatric reference values (21, 22).

For the isolation and culture of skin primary fibroblast cells from the patient and the age- and gender-matched healthy individual, the skin samples were washed with 70% EtOH and 1× PBS. After that, they were cut into small pieces and placed in a culture dish containing Dulbecco’s modified Eagle’s medium (DMEM), supplemented with 10% fetal calf serum (FCS), 1% anti–anti (antibiotic–antimycotic; 10,000 U/ml), and 1% L-glutamine (200 mM). Fibroblasts grew out of the tissue within several weeks, the culture medium was changed once a week, and cells were split at 80% confluency. Before the treatment with specific recombinant proteins, passage-matched patient and control cells were serum starved. To induce the TGF-β/SMAD pathway, cells were treated with 10 ng/ml recombinant human TGF-β1 protein (Abcam, Amsterdam, NL) for 48 h. All cells were incubated at 37°C in a humidified incubator with 5% CO₂.

Genomic DNA extraction from peripheral blood leukocytes was performed using magnetic beads with the chemagic™ Prime DNA Blood 7k kit on a chemagic™ Prime 4/8 instrument (both Perkin Elmer, Waltham, MA, USA) according to the manufacturer’s instructions. The DNA concentration was assessed using the Qubit™ dsDNA BR Assay Kit and fluorometer (both Thermo Fisher Scientific, Waltham, MA, USA). Exonic regions were enriched with the SureSelect Human All Exon 60MbV6 Kit (Agilent, Santa Clara, CA, USA) and sequenced as 100-bp paired-end runs on an Illumina HiSeq4000 platform (Illumina, San Diego, CA, USA). Read alignment was performed to the reference genome (UCSC build hg19) using Burrows–Wheeler Aligner v.0.7.5a (24). SAMtools v.0.1.19 (25) and GATK v.3.8 (26) were used for the detection of small insertions and deletions (<200 nt) and single-nucleotide variants. Copy number variants (CNVs) were called using The R package ExomeDepth v.1.1.10 (27). For variant filtering and gene prioritization, a custom-build pipeline was used (28, 29).

Total RNA was isolated from patient-derived cultured skin fibroblasts with the RNeasy Mini Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions. The RNA integrity number (RIN) was assessed using the RNA 6000 Nano Kit and Agilent 2100 Bioanalyzer (both Agilent Technologies, Santa Clara, CA, USA). 1 µg of total RNA, quantified with the Qubit RNA BR Assay Kit and fluorimeter (both Thermo Fisher Scientific, Waltham, MA, USA), was used for library preparation with the TruSeq Stranded mRNA Sample Prep LS Protocol (Illumina, San Diego, CA, USA) described in detail in (30). RNA libraries were sequenced as 100-bp paired end runs on an Illumina HiSeq 4000 platform (Illumina, San Diego, CA, USA). After demultiplexing, read mapping to the hg19 genome assembly was performed with STAR v.2.7.0a (31) using default parameters. Detection of novel splice junctions was allowed by setting twopassMode to “Basic.” RNA reads per gene were counted using GenomicAlignments (32), and genes with a 95th percentile FPKM below 1 were considered as not expressed and removed from the dataset. Events of aberrant RNA expression were detected using the R package OUTRIDER (33) in a dataset of 269 samples.

Quantitative proteomics from cultured fibroblasts was performed as previously described in detail (34) with two minor changes. Peptides were labeled using Tandem Mass Tag (TMT) 11-plex, not 10-plex reagents (Thermo Fisher Scientific, Waltham, MA, USA). Peptide separation prior to mass spectrometric analysis was performed using high-pH reverse-phase and not-trimodal mixed-mode chromatography. Each TMT batch consisted of nine samples and two reference samples, which were kept identical between batches to allow for subsequent inter-batch data normalization. Protein expression outliers were computed by running one sample against all others in a moderated t test approach using the R/Bioconductor package limma (35) in a dataset of 71 samples from 8 individual TMT batches.

A 130-bp PCR product was amplified using the following primers, which are flanking the deletion ASCC1-del-D (5′-gccaaactctttttcacagagg-3′) and ASCC1-del-R2 (5′-gctcggcttgttctgttttc-3′) to confirm the mutation in the patient-derived cells. In order to confirm an intact ASCC1 sequence in the control cells, a 215-bp PCR product was amplified using the following primers: ASCC1-del-D (5′-gccaaactctttttcacagagg-3′) and ASCC1-WT-R (5′-tggctaacaagcagaactgg-3′). The PCR was performed with Q5 Hot Start High-Fidelity 2x Master Mix (New England Biolabs, Frankfurt, Germany) and under the following thermocycling conditions: 30 s at 98°C followed by 31 cycles of 98°C for 10 s, 60°C for 30 s, and 72°C for 15 s, and another 7 min at 72°C. The amplification products were analyzed for their purity and size by agarose gel electrophoresis. After that, they were treated with Exo-CIP Rapid PCR Cleanup (New England Biolabs, Frankfurt, Germany) and sent to Eurofins for Sanger sequencing.

Total mRNA from the cultured age-, gender-, and passage-matched patient and healthy control skin primary fibroblasts was isolated using Monarch^® Total RNA Miniprep Kit (New England Biolabs, Frankfurt, Germany). Reverse transcription (RT) was done by using High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Vienna, Austria). Analysis of ASCC1, RUNX2, SERPINF1, MMAA, CACFD1, RABEPK, and HPRT1 gene expression levels was performed on the LightCycler^® 480 System (Roche Diagnostics, Vienna, Austria) using LightCycler^® FastStart DNA Master HybProbe (Roche Diagnostics, #12239272001) and the following TaqMan^® probes: Hs00418608_m1 (ASCC1), Hs01051148_g1 (RUNX2), Hs00369340_m1 (SERPINF1), Hs02800695_m1 (HPRT1). Cycling conditions for quantitative PCR (qPCR) were 1 min at 95°C followed by 50 cycles of 95°C for 10 s and 60°C for 45 s. The delta–delta Ct method (2^–ΔΔCt) was used to calculate the relative fold gene expression, whereby HPRT1 was used as a housekeeping gene.

Total mRNA isolation from hMSC and cDNA synthesis was performed as described in (36). In short, RNA extraction was done using TRIzol reagent and cDNA was synthesized using RevertAid First Strand cDNA Synthesis (Thermo Fisher Scientific). Real-time qPCR was performed on the QuantStudio™ 7 Flex Real-Time PCR System using Promega GoTaq™ qPCR Master Mix (Promega), and the primers are shown in Supplementary Table 1 . Cycling conditions for qPCR were 10 min at 95°C followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. The delta–delta Ct method (2^–ΔΔCt) was used to calculate relative fold gene expression, whereby 36B4 was used as a housekeeping gene.

Age-, gender-, and passage-matched fibroblasts from the patient and the healthy control, untreated and treated, were washed with cold 1× PBS and lysed by using RIPA buffer (Cell Signaling, Leiden, NL) supplemented with 100× protease inhibitor cocktail (Sigma-Aldrich, Vienna, Austria) and 100× phosphatase inhibitor cocktail (Sigma-Aldrich, Vienna, Austria) and 200× PMSF (Cell Signaling, Leiden, NL). Proteins were isolated by centrifugation at 14,000x g for 10 min at 4°C. Protein concentrations were measured using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Vienna, Austria). Equal amounts of proteins (~30 µg) were loaded onto 4%–20% gradient gels (Bio-Rad, Vienna, Austria), separated by gel electrophoresis, and electrotransferred onto PVDF membranes (Bio-Rad, Vienna, Austria). After incubating the membranes with a blocking reagent (Bio-Rad, Vienna, Austria), they were probed with the following primary antibodies diluted in blocking buffer overnight: monoclonal anti-rabbit COL1A1 (1:1,000, Cell Signaling, Leiden, NL), polyclonal anti-rabbit COL1A2 (1:1,000, Abcam, Amsterdam, NL), monoclonal anti-rabbit actin (1:2,000, Cell Signaling, Leiden, NL), monoclonal anti-rabbit pSmad3 (Ser423/425, 1:1,000, Cell Signaling, Leiden, NL), and monoclonal anti-rabbit Smad3 (1:1,000; Cell Signaling, Leiden, NL). Membranes were washed with 1× TBS 0.1% Tween. Detection was carried out using the corresponding peroxidase-conjugated (HRP) secondary antibodies (1:15,000, Cell Signaling, Leiden, NL) diluted in a blocking buffer with incubation for 1 h at room temperature. The membranes were washed again with 1× TBS 0.1% Tween and developed using SignalFire Elite ECL Reagent (Cell Signaling, Leiden, NL). Quantification of the bands was done with Bio-Rad Image Lab 6.1 software to calculate the expression levels. All band intensities were normalized to those of actin.

Human bone marrow-derived mesenchymal stromal cells (hMSCs) were obtained from Lonza (#PF-2501, Donor #4266). HMSCs were cultured and differentiated as described by Bruedigam et al. (36). Briefly, cells were seeded (20,000 cells per well of a 12-well plate) and maintained for 2 days in alpha minimum essential medium (α-MEM, Gibco) supplemented with 10% heat-inactivated FCS, penicillin–streptomycin, HEPES, and calcium chloride and adjusted to a pH of 7.5. Osteogenic differentiation was initiated using a differentiation medium containing α-MEM supplemented with 10 mM β-glycerophosphate and 100 nM dexamethasone added for a period of 25 days. Adipogenic differentiation was initiated using α-MEM supplemented with 100 nM dexamethasone, 500 µM 3-isobutyl-1-methylxanthine, and 60 µM indomethacin for 17 days. Cells at passage 8 were used in all experiments, and the differentiation medium was refreshed twice a week. All experiments were performed in at least triplicates per condition.

The two constructs of short hairpin RNAs (shRNAs) targeting endogenous ASCC1 ID12209 (renamed as #1 shRNA ASCC1) and ID12211 (renamed as #2 shRNA ASCC1) and the non-targeting shRNA with a scrambled sequence serving as a control (Scr) were purchased from the TRC-Hs1.0 library (Sigma-Aldrich) ( Supplementary Table 1 for sequences). Lentivirus (LV) was produced by transient transfection into 293FT cells using a standard calcium phosphate precipitation method as described previously (37). After 48 h, LV-containing supernatants were harvested and used immediately for hMSC transduction (24 h after attachment). One day later, the medium was replaced with osteogenic or adipogenic differentiation induction medium and cells were cultured until further analysis.

Calcium measurements and Alizarin Red S staining at day 25 of osteogenic differentiation were carried out as previously described (36). For the calcium quantification, cell extracts were prepared using PBS containing 0.1% Triton X-100 and further sonication. These samples and the residual extracellular matrix material in the plates were incubated overnight with 0.24 M HCl at 4°C. The calcium concentration was determined photometrically at 595 nm using a calcium assay reagent by combining reagent I (1 M ethanolamine buffer pH 10.6) with reagent II (0.35 mM O-cresolphthalein; 9.8 mM 8-hydroxyquinoline in 0.6 M HCl) in a ratio 1:1. Measurements were performed using the Perkin Elmer Victor X2 multilabel Microplate Reader. For the Alizarin Red S staining, cells were washed in PBS, fixed with ice-cold 70% ethanol for 1 h, and then stained for 5–10 min with a solution of 0.6% Alizarin Red S (pH 4.1).

Oil Red O staining at day 17 of adipogenic differentiation was performed, as previously described (36). Briefly, cells were flushed with PBS, fixed with 10% formalin overnight at 4°C, and rinsed with 60% 2-propanol. The cells were then stained with Oil Red O working solution (three parts of saturated Oil Red O solution and 2 parts of Milli-Q water) for 10–30 min. Cells were further incubated with DAPI (100 µg/ml in PBS) for 5 min in the dark. Imaging of all the wells was done. For the extraction of the lipid droplets, the plates were dried and incubated with IGEPAL working solution (4% IGEPAL in 100% 2-propanol) for 30 min. The quantification of the lipid droplets was done by photometric measurements at 490 nm.

Results were assessed using two-sided Student’s t-tests to compare two groups and expressed as means ± SEM. A value of P < 0.05 was considered significant. All statistical evaluations were undertaken using GraphPad Prism 5 software.

This girl was born prematurely by cesarean section at 34 + 1 weeks of gestation to consanguineous parents of Pakistani descent. At birth, the patient demonstrated severe congenital muscular hypotonia, a virtual absence of spontaneous movements, inability to breath and swallow, and required cardiopulmonary resuscitation and ventilation. She acquired a left-sided femur fracture during the resuscitation process including an intraosseous needle placement at the tibia. Additionally, she had arachnodactyly, gracile long bones, and very slender ribs on X-rays ( Figure 1A ). Her parents were reported to be healthy except for hearing loss in the mother. At birth, the girl also had a severe vitamin D deficiency with hypocalcemia as a consequence of maternal vitamin D deficiency. Her calcium and vitamin D levels normalized promptly following supplementation. Additionally, she suffered transient thoracic chyle leak and intestinal hypomobility and developed progressive brain atrophy. Due to progressive respiratory insufficiency, the patient died at the age of 3 months, immediately after the withdrawal of care with parental consent.

Bone histology showed highly disordered collagen organization typical for primary woven bone and a remarkably high quantity of bone marrow adipocytes ( Figures 1B–I ), normally absent in infants since at birth bone marrow is entirely hematopoietic. However, bone marrow adipocytes were not homogenously distributed throughout the bone sample: the marrow area in the highly trabecularized cortical plates and toward the central trabecular area was nearly completely replete by adipocytes ( Figures 1B–E ), whereas scattered adipocytes were viewed in the central part of the trabecular compartment ( Figures 1G–I ). In total, the adipocytes covered 19.9% of the bone marrow area of the patient ( Table 1 ). In contrast, no adipocyte was viewed in the sample of the age-matched control infant ( Figure 1J ). Active bone formation was observed on the periosteal surfaces ( Figure 1F ). Furthermore, irregularly shaped osteocyte lacunae were detected in the trabecular and cortical bone ( Figure 1K ) in accordance with the woven bone conformation ( Figure 1I ). Structural histomorphometric analysis revealed that trabecular bone volume was reduced by around 50% compared with reference values from older children, the age-matched control sample ( Table 1 ), and previously reported infants (38). Osteoclasts were mostly viewed on cortical surfaces leading to the observed trabecularization of the cortical plate ( Figure 2A ), but since osteoblasts and osteoclasts were both highly decreased in the central trabecular area, there was no proper bone remodeling into lamellar bone ( Figures 2A , 1G ). As a result, the thin and isolated trabecular features consisted mainly of immature woven bone rather than lamellar bone ( Figures 1H, I ). Despite the young age of the child, the bone sample did not show any residual mineralized cartilage.

Compared with the reference values of healthy older children, the bone mineralization density distribution (BMDD) curves of the patient in both cancellous and cortical bone were shifted to the left, toward a lower mineral content of the matrix ( Figure 2B , Table 2 ). This was in line with a reduced average degree of mineralization (CaMean) in both bone compartments. The fraction of lowly mineralized bone tissue (CaLow) was elevated in the cortical and trabecular bone, whereas the fraction of highly mineralized bone tissue (CaHigh) was below the normal range. However, bone matrix mineralization was slightly increased compared with the age-matched control bone sample consistent with the presence of woven bone observed by polarized light microscopy, which is known to have higher mineralization than lamellar bone. Thus, the combination of high woven bone and the increased matrix mineralization in the patient’s bone compared with the age-matched control suggests a failure of proper remodeling into mature lamellar bone and proper endochondral ossification.

Given this striking bone phenotype, of high bone marrow adiposity and low osteoblast number, and since osteoblasts and adipocytes are both derived from hMSCs but are reciprocally regulated, we hypothesized that the patient has an intrinsic defect at early stages of osteogenic differentiation.

By exome sequencing, we identified a novel homozygous deletion in the ASCC1 gene (MIM: 614215; Homo sapiens chromosome 10, GRCh38.p14 Primary Assembly NC_000010.11: g. 72138925_72163147del24223; NM_001198799.2: c. 574-1471_831-5742del24223). This deletion of around 24 kb within the 10q22.1 chromosome band encompasses exons 7 and 8 of the ASCC1 gene (see integrated genome viewer (IGV) screenshot in Supplementary Figure 1B ). The variant was absent from GnomAD (gnomad.broadinstitute.org) in homozygous and heterozygous states. Both parents were heterozygous carriers for the variant confirming recessive inheritance ( Supplementary Figure 1A, B ). The variant was additionally confirmed by Sanger sequencing in the patient ( Supplementary Figure 1C ). The variant is predicted to lead to a nonsense-medicated mRNA decay and subsequent absence of the functional protein.

We analyzed the whole transcriptome and proteome in patient-derived fibroblasts by RNA sequencing and mass spectrometry and found ASCC1 to be hardly expressed, either on RNA or on protein level ( Figures 3A, B ). Based on proteomics analysis, it was also shown that ASCC2 is highly downregulated. Moreover, RNAseq and proteomics displayed significant downregulation of a few other genes apart from ASCC1 such as RABEPK and MMAA. The downregulation of MMAA, CACFD1, and RABEPK at the RNA level was confirmed by qPCR ( Supplementary Figure 2 ). Since these genes are not known to be associated with any bone disease or with ASCC1, they were not further analyzed. By qPCR, we confirmed the profound downregulation of ASCC1 mRNA and showed that the expression of RUNX2 and SERPINF1 was reduced by more than 50% (p < 0.05; Figure 3C ).

In order to study the effect of the novel ASCC1 mutation on TGF-β/SMAD signaling, an important bone pathway and patient-derived and control fibroblasts were treated with the TGF-β1 ligand. This treatment lead to a dysregulation of TGF-β/SMAD signaling ( Figure 4 ). The patient cells showed a significant reduction in the phosphorylation of SMAD3 of more than 40% (p < 0.01) compared with the treated healthy control cells. In addition, we analyzed the protein expression of type I collagen, which revealed no notable difference in COL1A1 or COL1A2 ( Supplementary Figure 2 ).

To determine the role of ASCC1 in human osteogenic and adipogenic differentiation, we used shRNAs targeting endogenous ASCC1 to knock down its expression in hMSCs. Two separate shRNAs (#1 shRNA ASCC1, #2 shRNA ASCC1) substantially decreased ASCC1 mRNA expression at day 6 of osteogenic and adipogenic differentiation by approximately 70% (p < 0.001) and 60% (p < 0.001), respectively ( Figures 5A, B ), compared with the scrambled control cells (Scr). ASCC1 expression remained downregulated until the end of osteogenic and adipogenic differentiation ( Figures 5A, B ). While the knockdown of ASCC1 resulted in the inhibition of osteogenesis, it promoted adipogenesis based on mRNA expression of osteogenic and adipogenic markers. ALPL and RUNX2 using one of the two shRNAs and CTNNB1 using shRNA #1 were substantially downregulated at day 20 of osteogenic differentiation ( Figure 5A ), whereas FASN and PPARγ revealed a pronounced upregulation in particular PPARγ with a 2.5-fold increase on day 17 of adipogenic differentiation ( Figure 5B ). The phenotype of restrained osteogenesis was reinforced by a massive reduction (>90%) in the amount of incorporated calcium in hMSC-derived osteoblasts at day 25 ( Figure 6A ). This was further confirmed by Alizarin Red S staining ( Figure 6B ). On the contrary, oil red O staining in hMSC-derived adipocytes at day 17 of differentiation showed an increased amount of Oil red O positive cells by approximately 50%. This enhanced adipogenesis was further underpinned by the photometric quantitative measurement of extracted lipid droplets, showing also an increase of around 50% ( Figure 6C ).