Abstract
Spinal muscular atrophy (SMA) is a fatal neurodegenerative disease of newborns and children caused by mutations or deletions of the survival of motoneuron gene 1 resulting in low levels of the SMN protein. While neuromuscular degeneration is the cardinal symptom of the disease, the reduction of the ubiquitously expressed SMN additionally elicits non-motoneuron symptoms. Impaired bone development is a key feature of SMA, but it is yet unknown whether this is an indirect functional consequence of muscle weakness or caused by bone-intrinsic mechanisms. Therefore, we radiologically examined SMA patients in a prospective, non-randomized cohort study characterizing bone size and bone mineral density (BMD) and performed equivalent measurements in pre-symptomatic SMA mice. BMD as well as lumbar vertebral body size were significantly reduced in SMA patients. This growth defect but not BMD reduction was confirmed in SMA mice by μCT before the onset of neuromuscular symptoms indicating that it is at least partially independent of neuromuscular degeneration. Interestingly, the number of chondroblasts in the hypertrophic zone of the growth plate was significantly reduced. This was underlined by RNAseq and expression data from developing SMA mice vertebral bodies, which revealed molecular changes related to cell division and cartilage remodeling. Together, these findings suggest a bone intrinsic defect in SMA. This phenotype may not be rescued by novel drugs that enhance SMN levels in the central nervous system only.
Introduction
Spinal muscular atrophy (SMA) is a progressive neurodegenerative disease primarily characterized by the degeneration of α-motoneurons in the anterior horn of the spinal cord. As a consequence, SMA patients develop muscle weakness, progressive paralysis and muscle atrophy and show changes of the musculoskeletal system including progressive spinal deformity and premature death in the most severe cases (1). SMA is a monogenic disease, and all patients harbor deletions or mutations in the survival of motoneuron gene 1 (SMN1) (2). However, humans possess the very similar SMN2 gene, which codes for the same protein but includes a critical C to T transition within exon 7. This mutation is translationally silent but important for SMN2 pre-mRNA splicing as it is part of a splice enhancer region. The pre-mRNA is incompletely spliced and SMN2 produces about 10% of the SMN protein compared to SMN1 (3,4). Therefore, SMN2 is not able to fully rescue the loss of the SMN1 gene in patients, and SMA is caused by low SMN levels.
The number of SMN2 gene copies is the most important genetic modifier of the disease with an inverse correlation of gene copy number and severity (5). This leads to a clinical spectrum which has been categorized in five clinical subtypes by disease onset, life expectancy, and motor function milestones (6–8). The most severe SMA type 0 is characterized by a congenital onset, and untreated patients die before or within the first month after birth (9). SMA type I patients represent the most frequent subtype with symptom onset in the first 3 months, inability to sit or to gain head control and an early death within the first 2–3 years of life (8,10). Intermediate type II patients typically develop the first symptoms between the sixth and eighteenth months of age, are never able to stand and have a marked reduction of life expectancy (7). Disease onset in mild SMA type III patients occurs after the eighteenth month, but patients are able to stand and walk independently (7,8,11), while there are only mild muscle-weaknesses during adulthood in type IV patients (7).
The SMN protein is ubiquitously expressed and its deficiency elicits non-motoneuron symptoms such as metabolic changes (12) and defects in vasculature (13,14), heart (15), muscle (16,17), pancreas (18) and liver (19) conceiving SMA a multisystem disorder (20,21). It has been hypothesized that specific SMN thresholds are needed for the proper function of an organ and that severe SMA type I patients with very low SMN levels are more susceptible to peripheral defects compared to milder affected patients (22). Impaired bone structure and strength is one of the key features of SMA resulting in low bone mineral density (BMD), fragile bones, pathological fractures and progressive spinal deformities (23–26). It is still unclear whether impaired bone structure is solely a secondary phenomenon due to neurological impairment or if it occurs independent of the neuromuscular phenotype.
Here, we hypothesized that part of the SMA bone phenotype is independent of the neuromuscular impairments and an intrinsic characteristic of the disease at least in severe cases. To address this, we characterized bone morphology in SMA children directly compared to pre-symptomatic SMA mice. The use of SMA mice before the onset of muscle impairments allowed us to uncouple neuromuscular defects from the bone phenotype. Besides a reduced BMD, we found an impaired bone growth in SMA children. This growth defect but not the mineralization defect was evident in pre-symptomatic SMA mice. An in-depth analysis in these mice revealed an unimpaired bone mineralization and trabecular remodeling indicating a normal osteoblast and osteoclast activity. However, the hypertrophic zone of the growth plate exhibited a reduced number of chondrocytes. Chondroblasts at the growth plate are highly mitotic cells, which become hypertrophic chondrocytes that secrete the extracellular matrix components of the hyaline cartilage such as collagens and proteoglycans. Thereby, they critically drive longitudinal bone growth during endochondral ossification. Our findings of an impaired chondrogenesis in SMA mice additionally became evident on the level of gene expression. Bones of SMA mice showed changed molecular networks associated with extracellular matrix and cartilage formation as well as cell cycle control and mitosis.
Recently, a novel generation of SMN-enhancing drugs such as nusinersen and Zolgensma was approved for the treatment of SMA. Nusinersen is a non-blood-brain barrier penetrant antisense oligonucleotide, which is primarily effective in the central nervous system (CNS) (27). Zolgensma is an adeno-associated virus 9 delivering an SMN cDNA (28). However, adeno-associated virus (AAV)-delivered cDNA becomes diluted in daughter cells during mitosis (29). Thus, both drugs have a limited or unknown effect on mitotic cells outside the CNS (30). Our findings suggest an impaired longitudinal bone growth associated with impaired chondrogenesis. This may affect children with centrally restored SMN levels and argue in favor of chondroblasts as additional future treatment targets.
Results
Impaired bone in SMA is characterized by low BMD leading to fragile bones, pathological fractures and progressive spinal deformities (23–26). However, it is still unknown to which extent this phenotype is caused by neuromuscular degeneration or by bone intrinsic mechanisms. Here, we hypothesize that at least parts of the bone phenotype are caused by a deficiency of the SMN protein within bone cells such as osteoblasts or chondroblasts. Therefore, we employ the strategy to (i) characterize the bone phenotype in SMA children and (ii) explore to which extent these bone abnormalities occur in pre-symptomatic SMA mice devoid of any neuromuscular phenotype.
SMA patients have reduced BMD and smaller vertebral bodies
In a prospective non-randomized cohort study, radiological data of 42 SMA children were compared to age-matched controls (Supplementary Material, Table S1). The size of the lumbar vertebral body 1 (L1) was determined on radiographs from 31 patients measuring the anterior height (cranio-caudal extent) and the depth of the upper endplate (dorso-ventral extent) (Fig. 1A and B). There was no correlation between either the SMA type or the treatment status (with the SMN-enhancing drug nusinersen) with any of the radiological outcomes (data not shown). However, both the anterior height as well as the depth of the upper endplate were significantly smaller in the SMA patient cohort with a stronger effect on the dorso-ventral extent of the vertebral body (Fig. 1C and D). Therefore, vertebra L1 was significantly smaller in SMA patients compared to age-matched healthy children. Furthermore, we performed computer tomography (CT) analyses on a sub-cohort of 24 SMA patients for the evaluation of the BMD within the trabecular bone of the L1 vertebral body (Fig. 2A–C). Again, there was no correlation between SMA type or treatment status and BMD values (data not shown). However, BMD values were significantly lower in SMA children compared to age-matched healthy controls (Fig. 2D), and the average z-score of SMA patients was reduced while the z-scores of healthy age-matched controls were in the normal range (Fig. 2E) in accordance with the literature (31,32).
Radiological examination of the lumbar vertebral body 1 (L1) in control and SMA patients. Lateral X-ray of L1 of a healthy control (A) and an SMA patient (B). The anterior height (C) and depth of the upper endplate (D) were calculated. Both anterior height and depth of the upper endplate were significantly reduced in SMA children compared to controls (C, D). Bars show mean ± SEM with individual data points. Significance was tested using an unpaired Student’s t-test with ***P < 0.001.
BMD of the lumbar vertebral body 1 (L1) in control and SMA patients. For BMD analysis, the range of interest was defined on three planes: axial (A), sagittal (B) and frontal (C). BMD values were extracted for a central position. BMD values of SMA patients were significantly reduced in comparison to healthy age-matched controls (D). Z-scores were significantly lower in SMA compared to healthy children (E). Bars show mean ± SEM with individual data points. Significance was tested using an unpaired Student’s t-test with ***P < 0.001.
Pre-symptomatic SMA mice have smaller vertebral bodies and shortened femora
To compare our findings in SMA patients to a corresponding animal model, we analyzed Taiwanese SMA mice compared to control littermates. These mice are homozygous for the murine Smn1 knock-out allele and have a single allele with two copies of the human SMN2 transgene (33). First symptoms occur at post-natal day five at the earliest with weight loss and impaired motor functions. None of these symptoms are manifested at post-natal day three when SMA-mice are phenotypically indistinguishable from control littermates and have the same motor functions and body weight (34). We radiologically examined pre-symptomatic mice by measuring the vertebrae as well as the femur. Both, the cumulative height of the L2-L5 vertebrae and the femur were shortened in SMA mice (Supplementary Material, Fig. S1). To maximize comparability between human and mice data, we selected the murine Th13 vertebra for a detailed μCT analysis since it is equivalent to the human L1 vertebra (Fig. 3A–F). Matching the patient situation in L1, Th13 was smaller in SMA mice compared to control littermates as indicated by a reduced vertebral height, surface area and volume (Fig. 3G-I). Compacta and trabecular bone were not clearly distinguishable from each other, and we measured the BMD in both bone compartments. However, there was no difference between BMD measured in cortical and trabecular bone between control and SMA mice (Fig. 3J).
Analyses of vertebral body Th13 of pre-symptomatic SMA and control mice. Mice were prepared at pre-symptomatic post-natal day 3 (P3), and the vertebral column was subjected to μCT analysis. (A, B) μCT images of the thoracic vertebral body 13 (Th13) were 3D-reconstructed and are depicted from a ventral view. (C, D) Digital frontomedial sections as well as (E, F) horizontal sections reveal mineralized bone (red) with non-mineralized bone marrow in the center. Scale bars, 0.5 mm. (G) The height, (H) the surface area and (I) the volume of the Th13 vertebral body were quantified in control mice (tgSMN2tg/0, mSmn1+/−) and SMA littermates (tgSMN2tg/0, mSmn1−/−). (J) The mineralization was measured as mean gray value within the mineralized volume (C—F, colored). Bar graphs show mean ± SEM and individual data points. Significance was tested with an unpaired Student’s t-test with ns = non-significant, *P < 0.05 and **P < 0.01.
To further address this, we analyzed the femur of pre-symptomatic SMA mice compared to controls. The femur is the largest murine bone of which the cortical and trabecular bone are easily distinguishable (Fig. 4). Again, the femur size was smaller in SMA mice compared to controls: the length of the femur (Fig. 4A, B and G), the diameter (Fig. 4E, F and H) and the trabecular bone volume (Fig. 4C–F and I) were reduced. However, trabeculae were not changed as indicated by an unaffected trabecular thickness and mineralization (Fig. 4J and K). Similarly, the cortical thickness was the same in SMA mice and controls (Fig. 4L). The absence of an effect on trabecular and cortical thickness as well as on BMD indicates that there is no global effect of SMN reduction on osteoblasts and osteoclasts in pre-symptomatic SMA mice. Indeed, the mineralization activity of primary osteoblasts from SMA mice was not reduced compared to control mice (Supplementary Material, Fig. S2).
μCT analyses of pre-symptomatic SMA mice and control femora. The femora of post-natal day 3 (P3) mice were analyzed with μCT allowing a (A, B) 3D reconstruction, (C, D) identification of trabecular bone in digital longitudinal or (E, F) cross sections (green area in both). Scale bars are equal to 1 mm (A–D) and 0.25 mm (E, F). (G) The length of the femora and (H) the diameter were quantified from control mice (tgSMN2tg/0, mSmn1+/−) and SMA littermates (tgSMN2tg/0, mSmn1−/−). (I) The trabecular bone volume was measured according to a gray value excluding the highly mineralized cortical bone and the growth plate (green area in C–F). Areas of higher mineralization within this volume are trabeculae. (J) Mean thickness and (K) mineralization were quantified. (L) The thickness of the highly mineralized cortical bone was quantified at the longitudinal center of the femora. Bar graphs show mean ± SEM and individual data points. Significance was tested with an unpaired Student’s t-test with ns = non-significant and *P < 0.05.
Reduced longitudinal bone growth is associated with chondroblast abnormalities
The impaired bone growth but not the mineralization defect was evident in pre-symptomatic SMA mice, which indicates that this effect occurs independent of the neuromuscular phenotype. But which mechanism may selectively impair bone growth but spare mineralization and remodeling? To address this question, we stained non-decalcified slices of control and SMA mice femurs with Movat’s pentachrome, allowing to analyze the detailed structure of the growth plate. The growth plate facilitates longitudinal bone growth (Fig. 5A and B). Chondrocytes primarily drive this process with (i) proliferation at the distal end, (ii) formation of columns and secretion of the hyaline cartilage extracellular matrix in the zone of maturation and (iii) becoming hypertrophic and apoptotic in the hypertrophic zone at the junction to the area of mineralized trabecular bone. Any changes in chondrocyte biology should be detectable at the last step of the chondrocyte life-cycle within the hypertrophic zone (Fig. 5A’ and B’). Comparing SMA mice and controls, there was no change in the average thickness of the hypertrophic zone and only a weak tendency for an increased chondrocyte size (Fig. 5C and D). Importantly, we found a reduced chondrocyte density in SMA mice, which is in accordance with the reduced longitudinal bone growth (Fig. 5A’, B’ and E).
Growth plate morphology in pre-symptomatic SMA mice. The femora of (A) control and (B) SMA mice were dissected. Non-decalcified slices were stained with pentachrome, which visualizes non-mineralized cartilage (yellow), mineralized cartilage (blue-green), muscle (red) and nuclei (black). The area between collagen rich (yellow) and osteoblast/bone marrow-rich (black nuclei) was defined (A’, B′, lines) and is characterized by larger hypertrophic chondroblasts (A’, B′, asterix). (C) The area and the length of the hypertrophic zone were measured, and the mean height was calculated. (D) The chondroblast size was measured, and (E) the number of chondroblasts was counted for a calculation of their density within the hypertrophic zone. Scale bars show 200 μm (A, B) or 50 μm (A’, B′). Bar graphs show mean ± SEM and individual data points. Significance was tested with an unpaired Student’s t-test with ns = non-significant and *P < 0.05.
Molecular networks associated with hyaline cartilage formation are downregulated in SMA mice
The pre-symptomatic changes in SMA bone development seem to be restricted to chondroblast-regulated longitudinal growth. Indeed, we could not detect extensive expressional changes in markers of bone development with the exception of VEGFR2 and PDGFRα both of which were downregulated in SMA mice vertebral bodies Th3–13 (Supplementary Material, Fig. S3). However, both receptors are expressed on chondroblasts, which may explain these findings (35,36). To further elucidate the underlying molecular changes, we performed an RNAseq screening with pooled bone samples of SMA and control mice from post-natal day 1 (Fig. 6A). The fold-change for each detected gene was calculated and plotted against the sum of reads in control and SMA samples as an indicator for the expression (Fig. 6A). Genes with a low expression displayed higher fold-changes representing a poor signal-to-noise ratio, which was addressed by proper thresholding (Fig. 6A, dotted lines) to identify regulated genes (Fig. 6A, red dots; Supplementary Material, Table S2). Importantly, all of these transcripts were downregulated in SMA mice compared to controls (Fig. 6A). A gene ontology (GO) enrichment analysis revealed that these genes are involved in biological processes such as cell fate commitment, skeletal system morphogenesis and brain development (Fig. 6B). Moreover, they preferentially act extracellularly on collagen as indicated by a significant enrichment of the cellular component GO categories extracellular matrix and collagen containing extracellular matrix (Fig. 6C). The involved reactome pathways point towards a specific role of the altered genes in extracellular matrix organization and Keratan sulfate metabolism—a component of hyaline cartilage (Fig. 6D). Additionally, we performed a network analysis on the regulated genes identifying six functional clusters (Fig. 6E). Besides clusters involved in cartilage remodeling (collagen and collagen remodeling, hyaluronan and extracellular matrix binding), downregulated genes in the vertebral bodies of SMA mice are involved in cell cycle and chromosome remodeling (Fig. 6E). Cell cycle control and mitosis of chondroblasts as well as collagen synthesis critically modulate cartilage formation and longitudinal bone growth during endochondral ossification. Therefore, we validated the altered expression of key-molecules via qRT-PCR in pooled post-natal day 1 samples from SMA and control mice (Fig. 6F). Indeed, collagen type XII α1 (Col12a1) and type V α2 (Col5a2) as well as the collagen crosslinking enzyme lysyl oxidase and the small leucine-rich proteoglycan osteoglycin were downregulated in SMA. Moreover, the cell proliferation marker Ki67 (Mki67) was reduced in early post-natal SMA mice compared to controls (Fig. 6F), which may relate to the reduced number of chondrocytes in SMA mice at later pre-symptomatic stages (Fig. 5).
RNAseq analyses of pre-symptomatic control and SMA mice vertebral bodies. (A) The fold change between SMA and control pooled samples from post-natal day 1 was calculated, and the log2 was generated for each mRNA target. This was plotted against the sum of SMA and control expression levels for each gene (cumulative expression). SMA-dependent candidate mRNAs were above a cumulative expression level of 27 and above log2 = 1 or below log2 = −1 (dotted lines). (B) GO enrichment analysis of candidate mRNAs against a reference list of mRNAs (A, all mRNAs above a sum expression of 27) for the GO-category biological process, (C) cellular component and (D) reactome pathway. (E) EMBL-STRING network analysis with edges formed on data sources ‘experiments’, ‘databases’ and ‘co-expression’ with line-thickness representing interaction probability. Bars show fold enrichment with FDR-corrected Fisher’s exact test *q < 0.05, **q < 0.01, ***q < 0.001. (F) The expression of key-molecules was measured in vertebral bodies of pooled samples from post-natal day 1 SMA and control mice via qRT-PCR. Bars show mean ± SEM of three technical replicates.
Discussion
Spinal muscular atrophy has been considered as a neuromuscular disease with the bone phenotype being secondary to muscle weakness and atrophy. In general, the reduction of mechanical force results in reduced BMD, while muscle training enhances mineralization (37,38). The degeneration of the paraspinal muscles in SMA patients induces deformations such as scoliosis, pelvic obliquity, kyphosis and lordosis, which often require surgical interventions (39). Here, we report a significant reduction of BMD and in the mean z-score of the lumbar vertebra, which is in line with previous results in SMA type II children (23–26). This effect is even more pronounced in SMA type I (23), and the mineralization deficiency relates to an increased fracture risk (23,25). A bone intrinsic origin of the osteoporotic phenotype related to altered activity of osteoclasts has been previously suggested by findings in 4 week-old SMA mice (40). However, we could not confirm impaired osteoblast or osteoclast activity in pre-symptomatic mice, since there were no abnormalities in mineralization, cortical thickness or trabecular morphology in vertebral body and femur. Moreover, intramembranous ossification of flat bones requires a coordinated osteoclast activity, and there are no reports about defective skull development in SMA patients. Together, this indicates that the mineralization defect is indeed secondary to the neuromuscular degeneration.
SMA children showed reduced vertebral growth in comparison to healthy controls. Longitudinal bone growth is highly influenced by mechanical force during normal development either being suppressive or promotive dependent on the stimulus (41). Although it is likely that muscle weakness and impaired ambulatory ability substantially contribute to the reduced size of the vertebral bodies, our results in SMA mice revealed a reduced longitudinal bone growth, which was independent of neuromuscular defects. This defect was associated with a reduced number of chondroblasts in the hypertrophic zone of the growth plate. Chondroblasts are crucial for the longitudinal, endochondral bone formation as they continuously form the cartilaginous matrix for the later mineralization and remodeling by osteoblasts and osteoclasts (42). In accordance with a reduced cell number, we report a reduced gene expression of cell cycle-related genes such as Ki67. Chondrogenesis is critically regulated by the activity of the RhoA-ROCK pathway. Increased RhoA-ROCK stabilizes stress fibers associated with chondrocyte de-differentiation (43) and inhibits early chondrogenesis as well as later hypertrophic differentiation (44–46). Interestingly, we and others previously reported an altered activity of this pathway in SMA (34,47–50). The SMN protein directly interacts with the actin-binding protein profilin2a, which is a direct target of the ROCK pathway (47). The reduction of the SMN protein results in an increased profilin2a-ROCK interaction and activation (51). However, whether this mechanism influences chondrogenesis in SMA remains to be determined in future studies. Another possible mechanism may relate to SMN’s role in the resolution of co-transcriptional R-loops. SMN interacts with the zinc finger protein 1 (ZPR1) (52), which prevents the formation of pathogenic R-loop structures formed during transcription. A reduction of the SMN protein results in activation of DNA damage response pathway leading to genomic instability and motoneuron death (53). Interestingly, peripheral tissues such as heart, lung, liver, and muscle displayed a dysregulation of ZPR1 indicating a systemic impact of this mechanism (53).
The downregulation of the matrix components such as collagen type V, which nucleates collagen fibrils, and collagen type XII, which locates to the growth plate (54,55), may thus be a consequence of impaired ROCK-related chondrogenesis or ZPR1-related cell death of mitotic chondroblasts (56). Despite these developmental aspects, chondrocytes are crucial for the function of articular cartilage, and their impairment is related to joint diseases such as Ehlers–Danlos syndromes. Interestingly, collagen type V and XII mutations cause classical as well as myopathic Ehlers–Danlos syndromes (57), and hypermobile joints in the upper extremities are characteristic for SMA patients (58). Recently, impaired collagen metabolism has been reported for SMA mice kidneys (59). Therefore, dysregulated collagens may be a common mechanism for peripheral organ defects in SMA.
Peripheral phenotypes become increasingly important in times of treatments focusing on the restoration of the SMN levels in the CNS with an unclear or absent effect on SMN levels in the periphery (30). Recently, two SMN-enhancing drugs, nusinersen, an intrathecally applied antisense oligonucleotide and a gene-replacement therapy with an adeno-associated virus (AAV9), have been approved for the treatment of SMA (27,28,60–62). However, nusinersen is not blood-brain barrier penetrant and the AAV-delivered SMN cDNA dilutes in mitotic cells (27,29). Therefore, both drugs preferentially target the CNS or post-mitotic cells including motoneurons in the spinal cord. However, the effect on mitotic cells in the periphery is limited. Since SMA patients benefit from SMN-enhancing drugs with better motor functions and survival, peripheral phenotypes will become more dominant during the progression of treated SMA (30). Clinical studies as well as experimental studies in ‘treated’ SMA mice are needed to further address peripheral phenotypes in SMA. However, our results in pre-symptomatic SMA mice argue in favor of highly mitotic chondroblasts as an important therapeutic target for bone development and joint function in future combinatorial treatment regimens of SMA.
Materials and Methods
SMA patients and controls
After ethics committee approval, 42 children with SMA and spinal deformity were included in a prospective non-randomized cohort study. All participants were informed about the purpose of the study, and a voluntary oral and written consent was obtained. The majority (81%, n = 34) presented with SMA type II and 9.5% had SMA type I (n = 4) or type III (n = 4) each. At the time of investigation, 40% (n = 17) were treated by intrathecal nusinersen injection. The others (60%, n = 25) did not receive nusinersen treatment. None had taken part in a gene therapy survey. Radiographic analysis of the anterior height and depth at the upper endplate was measured for the first lumbar vertebra (L1) on standardized sitting lateral radiographs using Centricity Enterprise Web Version 3.0 (GE Healthcare Medical Systems, Chicago, United States, 2006) (Fig. 1). These measurements were performed for a subgroup of 31 SMA patients at an average age of 12.8 +/− 2.0 years. An age-matched healthy group of 16 children (average age 13.2 +/− 0.4) served as a control for these radiographic measures. Quantitative vertebral trabecular BMD of L1 was measured on pre-calibrated computed tomography (CT) scans (Somatom Definition AS, Siemens, Erlangen, Germany) with 0.6 mm slice thickness of the whole spine as part of pre-surgical investigations before definite spinal fusion. A subgroup of 24 SMA patients could be analyzed at an average age of 12.5 +/− 1.6 years. Pre-calibrated spinal CT scans without pathology of 22 age-matched children served as a control (12.8 +/− 3.8 years). Using QCTpro® (Mindways software), BMD and Z-score parameters (63) of L1 in a central position were extracted using pre-calibrated CT data (Fig. 2). Z-score parameters were derived by comparison to a pediatric reference population on a standard deviation scale (31,32). These values were compared to data from the age-matched healthy control group.
Animals
All animal experiments were conducted in accordance with the German animal welfare law and approved by the Ministry of Food, Agriculture and Consumer Protection of Lower Saxony (LAVES file no. 33.12-42502-04-15/1774). Taiwanese SMA mice ([FVB.Cg-Tg(SMN2)2Hung Smn1tm1Hung/J]) (33) were bred by a well-established breeding scheme (64) resulting in litters with half SMA mice (tgSMN2tg/0, mSmn1−/−) and half control mice (tgSMN2tg/0, mSmn1+/−). For analysis, animals were decapitated on post-natal day 3, and a tail tip biopsy was taken for genotyping as described previously (65). For osteoblast preparation, the skin was removed, and the calvariae were prepared and stored in Hank’s balanced salt solution (HBSS, Gibco). For mRNA analysis, the total spinal column was removed, and the thoracic vertebral bodies 3–13 were prepared removing the vertebral arch as well as any connective tissue. For morphological analyses, the inner organs were removed, and the animals were pinned on a Sylgard 184 elastomer (Dowsil)-coated 6 cm dish. Animals were fixed with 4% (w/v) paraformaldehyde in PBS over night at 4°C and stored in 80% (v/v) ethanol.
Mouse X-ray analyses
Fixed skeletons were removed from the petri dish and placed on an X-ray imaging film (Dürr Dental) together with a Glow ‘n Tell Tape scale. Skeletons were irradiated from a distance of 13 cm with an X-ray machine (Heliodent Plus, Dentsply) using 60 kV and an exposure time of 0.05 s. The imaging film was scanned and digitalized with the VistaScan Perio Plus. Right femur lengths and lumbar vertebrae L2-L6 heights were measured with ImageJ in a blinded manner.
Mouse μCT analyses
Left femora and the vertebrae Th12-L6 were removed from the fixed skeletons and dried at room temperature overnight. Bone samples were placed in polypropylene straws where they were fixed in position using plastic pellets. The Inveon μCT device (Siemens) was used to scan the samples with 80 kV and 500 μA using a 0.8 mm carbon filter. A total of 720 tomographic cross sections were scanned from each sample in a 360° sequence with an average digital slice thickness of 8.6 μm. Exposure time was 4000 ms. Inveon Workplace software (Siemens) was used for further analysis. The length of the femora was measured by the multi-modal 3D analysis tool as the maximal longitudinal extent. A digital cross-sectional image was generated at the exact center of this maximal longitudinal extent. The cortical bone thickness and the cross-sectional length were measured multiple times from the cross-sectional image, and mean values were calculated. Trabecular bone was analyzed in the diaphysis defined by 0.8 mm distance from each end of the femur omitting the epiphysis. A voxel gray value below 350 served as a threshold for the identification of the trabecular bone volume, which was fine-tuned by hand. Within this volume, all structures with a voxel gray value above 280 were defined as trabecular bone. Voxels below this value were identified as bone marrow. These volumes of interest (VOIs) were used for the morphometric analyses of bone (66) and for the calculation of mean voxel grey values for each VOI giving the mineralization grade. Mineralization of thoracic vertebral bodies was measured in a VOI identified by a voxel gray value greater than 350. This VOI included compacta as well as trabecular bones, which were not clearly distinguishable from each other. The height of thoracic segment 13 (Th13) was measured in the frontal view as the maximal cranio-caudal extent.
Non-mineralized histology
Right femora were dissected from PFA-fixed mice and incubated as follows: 1st day in 70% (v/v) ethanol at room temperature, 2nd day in a desiccator at 0.5 bar and room temperature, 3rd day in isopropanol at room-temperature, 4th day in a fresh vial of isopropanol at room temperature, 5th day in xylene at room temperature, 6th day in a fresh vial of xylene at room-temperature, 6th–9th day in pre-infiltration solution (Technovit 9100, Kulzer, 64 715 444) at 4°C and 10th–17th day in infiltration solution (Technovit 9100, Kulzer, 64 715 444) at 4°C. After each solution change, femora were incubated in a desiccator at 0.5 bar for infiltration. Bones were embedded in polymerization solution (Technovit 9100, Kulzer, 64 715 444), and polymerization was performed at 4°C. 5 μm sections were produced with a rotational microtome (RM 2165, Leica) and pressed at 37°C for 7 days on poly-L-lysine-coated object slides. Slices were stained with Movat’s pentachrome stain as follows: dehydration with [1] 15 min xylene, [2] 15 min xylene, [3] 10 min 2-methoxyacetate, [4] 2 min 100% (v/v) ethanol, [4] 2 min 100% (v/v) ethanol, [5] 2 min 90% (v/v) ethanol, [6] 2 min 80% (v/v) ethanol, [4] 2 min 70% (v/v) ethanol and five washing steps with distilled water. The staining followed a 14-step protocol: [1] 10 min alcian blue, [2] 5 min tube water, [3] 60 min alkaline ethanol (10% (v/v) ammonium hydroxide in 96% (v/v) ethanol), [4] 10 min flowing tube water, [5] 2 min distilled water, [6] 10 min Weigert’s iron hematoxylin stain, [7] 2 min distilled water, [8] 15 min flowing tube water, [9] 15 min Brilliant Crocein R—acid fuchsin, [10] 1 min 0.5% (v/v) acetic acid, [11] 20 min 5% (w/v) phosphotungstic acid, [12] 2 min 0.5% (v/v) acetic acid, [13] wash three times in 100% ethanol 5 min each step, [14] 60 min 6% (w/v) Safran du Gâtinais in ethanol, [15] wash three times in 100% ethanol 3 min each step, [16] 2 min xylene, [17] mounting in Eukitt (Chem-Lab Nv, 04.0503.0500).
Quantification of chondroblasts
Multiple image aligned photomicrographs were taken from the Movat’s stained non-decalcified femora with 20x objective equipped BX51 microscope (Olympus, Germany) with an automated microscope stage (MBF Bioscience) using the Stereo Investigator software version 11.07 (MBF Bioscience). ImageJ was used for the identification of hypertrophic zone as a region of interest defined by the absence of cartilage (yellow staining) and osteoblasts/bone marrow (black nuclei). The cross-sectional length of the hypertrophic zone was measured alongside the cartilage frontier. The mean height was calculated dividing the hypertrophic zone area by its cross-sectional length. The hypertrophic chondroblasts were identified and marked by hand to measure mean chondroblast size and number. Division of chondroblast number by the hypertrophic zone area resulted in the chondroblast density. All evaluations were performed in a blinded manner.
RNAseq analysis
Thoracic vertebral bodies Th3-Th13 were snap-frozen in liquid nitrogen and stored at −80°C. A tungsten stainless steel 5 mm bead was incubated together with the sample for 30 min at −80°C and used for disruption of bone and cartilage in the TissueLyser LT (Qiagen) for 13 min at 50 Hz. The RNeasy Plus Mini Kit (Qiagen) was used for RNA isolation according to the manufacturer’s guidelines. Vertebral RNA of four SMA mice and four control samples were pooled, respectively, with a total amount of 250 ng RNA each pool. mRNA was enriched with the NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB) and the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (NEB) according to the manufacturer’s guidelines. Nucleotides, primers and enzymes were removed by the Agencourt AMPure XP (Beckman Coulter), and the cDNA library was synthesized with seven PCR cycles using the NEBNext Multiplex Oligos for Illumina Kit. cDNA size distribution was controlled with a Bioanalyzer High Sensitivity DNA Assay (Agilent) and quantified with a Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific). After NaOH-denaturation, 1.3 mL of a 1.8 pm sample were analyzed with an Illumina NextSeq 550 Sequencer (Illumina) with a read length of 76 bp. The nfcore/rnaseq software (National Genomics Infrastructure Plattform, SciLifeLab) was used for data processing and mapping to the mouse genome (GENCODE.org, GRCm38.p6). The DESeq2 Software package was used for data normalization (67).
GO enrichment and network analyses
Regulated genes were identified according to the described threshold strategy (Fig. 6A). This list of regulated genes was compared to a list of reference genes for a significant enrichment of GOs. All genes detected with the microarray and a cumulative expression > 27 were included in the reference set. The PANTHER overrepresentation test (68) and the GO database (released 2019-12-09) were employed for the GO analyses with an FDR corrected Fisher’s exact test. The Search Tool for the Retrieval of Interacting Genes/Proteins (STRING, v11) (69) was used for the network analysis. Edges were defined by experiments, databases and co-expression and displayed in the probability mode with an interaction score of 0.4. Disconnected nodes were hidden, and no secondary shell with non-regulated genes was added.
qRT-PCR from vertebral bodies
1 μg RNA of each sample were reversely transcribed with the M-MLV reverse transcriptase (Invitrogen) and random hexamer primers (ThermoFisher Scientific) according to the manufacturer’s guidelines. The qPCR was performed as described previously (65). Briefly, 5 μL cDNA dilution (1:100 in Millipore water), 2 μL forward and reverse primer dilution (1.75 μM each primer) and 7 μL Power SYBRgreen (Applied Biosystems) were mixed in the MicroAmp reaction plate (Applied Biosystems). qPCR was performed using the StepOnePlus thermocycler (Applied Biosystems): initial 10 min with 95°C were followed by 40 cycles (15 s 95°C, 1 min 60°C) as well as a melt curve analysis. Primer sequences are given in Supplementary Material, Table S3.
Primary osteoblast culture
Each calvaria was incubated for 10 min at 37°C in 500 μL HBSS containing 0.25% (w/v) trypsin (Gibco, 25 050-014). After washing with DMEM (Gibco, 31 966–021), the calvaria was incubated with 600 μL DMEM containing 0.2% (w/v) collagenase type IV (Worthington Biochemical, LS004210) for 30 min at 37°C. The supernatant was discarded, and the calvaria was incubated with fresh 0.2% (w/v) collagenase type IV solution for 60 min at 37°C. The supernatant was collected in a falcon tube and pooled with 5 mL DMEM, which has been used to wash the remaining calvaria for a complete removal of the osteoblasts. Osteoblasts were pelleted for 5 min at 200 g, dissolved in culture medium containing DMEM (Gibco, 31 966-021), 10% (v/v) fetal bovine serum (FBS, Gibco, 10 500-064) and 1% (v/v) Pen/Strep (Gibco, 15 140-122) and seeded in a 6-well dish and grown to 80% confluency in a 75 cm2 culture flask. A total of 25,000 cells were seeded in a 6-well dish at day of in vitro 1 (DIV1) containing culture medium with additional 50 μg/ml ascorbic acid and 10 mM β-glycerol phosphate for differentiation. At DIV28, cells were fixed in methanol at −20°C and washed with tube water. For staining of the mineralized area, fixed cells were incubated with 5% (w/v) silver nitrate and incubated for 30 min under light. After three washing steps with water, the staining was fixed with 10% (w/v) sodium thiosulfate solution and washed with water. An Olympus BX70 equipped with a 4x objective was used to take photomicrographs of the central area of each 6-well dish. Eight bit images were analyzed with ImageJ measuring the area with grey values higher than 200.
Statistical analyses
Statistical analysis was performed using Excel Version 2010 (Microsoft Cooperation, Redmond, Washington, USA) and Graph Pad Prism Version 8.0 (GraphPad Software Inc. San Diego, California, USA). Statistical significance was defined with levels as P < 0.05 (*), P < 0.01 (**) and P < 0.001 (***). Details are reported in the figure legends.
Acknowledgements
We thank the Research Core Unit Genomics (RCUG) and Dr. O. Dittrich-Breiholz of the Hannover Medical School for transcriptome analyses. The authors (KAL, LB, KT, HML and AKH) are members of the European Reference Network for Rare Neuromuscular Diseases (ERN EURO-NMD). This work was supported by SMA Europe (to NH) and by the Deutsche Muskelstiftung (to NH and PC).
Conflict of Interest statement. The authors declare that they have no conflict of interest.
References
Author notes
Contributed equally as senior authors.
