Here we aimed to identify rare genetic variants in a list of candidate bone-related genes in order to broaden our understanding of the genetic complexity underlying a quantitative bone trait such as BMD. For this purpose, an in-house bone gene panel was designed and DNA samples from postmenopausal women with extreme lumbar spine BMD values and Z-scores were subjected to next-generation sequencing. After filtering by predicted functionality, six genetic variants were identified. The only one cataloged as pathogenic was a heterozygous variant in the COL1A2 gene.

This variant, p.(Gly751Ser) in COL1A2, has been reported previously. De Paepe et al. [24] identified it in two sibs with severe osteogenesis imperfecta, who were homozygous for the missense change, and in other family members who were heterozygous and presented mild clinical manifestations of OI. Similar to our case, Spotila et al. [32] found this same variant in heterozygosity in a woman with postmenopausal osteoporosis. These cases evidence the phenotypic and genotypic overlap between osteoporosis and mild osteogenesis imperfecta and point to the milder character of the p.(Gly751Ser) variant in heterozygosis, which has been consistently associated with osteoporosis rather than OI.

Osteogenesis imperfecta (OI) is a phenotypically and genetically heterogeneous skeletal dysplasia characterized by bone fragility, growth deficiency, and skeletal deformity. The most frequent mutations causing OI are single-nucleotide substitutions that replace glycine residues or exon splicing defects in the COL1A1 and COL1A2 genes that encode the α1(I) and α2(I) collagen chains. Mutant collagen is partially retained intracellularly, and after secretion, it assembles in disorganized fibrils, altering mineralization. Alternatively, truncating mutations lead to nonsense-mediated mRNA decay and subsequent quantitative defects (about 50% reduction) of otherwise normal collagens. OI is characterized by a wide range of clinical outcomes ranging from mild forms of skeletal fragility (often related to quantitative defects) to lethal phenotypes [33, 34]. The genetic spectrum of OI includes at least 16 other genes, most of them playing a pivotal role in synthesis, post-translational modification, and processing of type I collagen. All of this heterogeneity leads to a variety of inheritance patterns spanning from autosomal dominant and recessive as well as X-linked recessive which further complicate the disease classification and clinical characterization [19, 35].

Another gene associated with OI for which we found a rare variant is SEC24D, which encodes a component of the COPII complex involved in protein export from the endoplasmic reticulum (ER). Mutations in this gene have been associated with Cole-Carpenter syndrome, a recessive disorder affecting bone formation, resulting in craniofacial malformations and bones that break easily, as well as a syndromic form of osteogenesis imperfecta. Functional studies of fibroblasts from a Cole-Carpenter syndrome patient displayed moderate accumulation of collagen in a significantly enlarged ER, indicative of a collagen export defect. The patient, a 7-year-old boy, had moderately reduced BMD (− 2.0 SD in a DXA whole body measurement), while the BMD of his heterozygous carrier parents, measured by DXA and peripheral quantitative computed tomography, was in the normal age-adjusted range [28, 36]. The variant found in our study, p.(Leu526Phe), is cataloged as likely pathogenic in HGMD, but the fact that it is present in heterozygosity would make the patient an asymptomatic carrier. Alternatively, we might speculate that the variant, together with other variants that may have gone undetected (such as second mutation in SEC24D or in a SEC23A), might indeed contribute to explain the low BMD (− 2.73) of our patient. Further description of the BMD status of additional heterozygous carriers of this rare form of OI may help clarify the involvement of SEC24D in low bone mass phenotypes.

The woman with the lowest LS Z-score of our cohort carried the variant p.(Ile721Asn) in the T-cell immune regulator 1 gene (TCIRG1), which encodes the α3 subunit of the vacuolar ATPase proton pump involved in acidification of the osteoclast resorption lacuna and in secretory lysosome trafficking [37]. Mutations in this gene are mainly involved in autosomal recessive osteopetrosis (ARO)—or exceptionally in autosomal dominant osteopetrosis (ADO)—due to the inefficient bone resorption by nonfunctional osteoclasts [25, 38], and in this scenario, it is difficult to reconcile the mutation in heterozygosity in our patient with her severe low bone mass (LBM). Nevertheless, variants in TCIRG1 have been previously described in patients with LBM phenotype which experienced atypical femoral fractures [16]. Additionally, the p.(Ile721Asn) variant described here falls precisely in the proton pumping channel of the protein [37], and it is tempting to speculate that it might cause the channel to remain permanently open, acidifying the osteoclast lumen space and increasing bone resorption. Functional assays would be required to prove this hypothesis.

Another woman from the LZ group carried a missense mutation in ANO5. Mutations in this gene are responsible for gnathodiaphyseal dysplasia (GDD), a rare skeletal syndrome characterized by osteopetrosis-like sclerosis of the long bones and fibrous dysplasia-like cemento-osseous lesions of the jaw bone. People with this condition have reduced BMD which causes the bones to be unusually fragile. As a result, they typically experience multiple bone fractures during childhood, often from mild trauma or with no apparent cause [29]. The ANO5 gene encodes a Ca2+-activated Cl− channel involved in bone remodeling through the functional regulation of osteoclasts. Interestingly, most of the mutations in ANO5 leading to anoctamin-5 deficiency, representing loss-of-function, are causative of several forms of limb-girdle muscular dystrophy (LGMDL2L), while missense mutations in anoctamin-5 are causative of GDD which follows an autosomal dominant mode of transmission, reflecting a gain-of-function phenotype. Significantly, patients with LGMDL2L exhibit no pathogenic phenotype related to the bone, which, conversely, is evident in patients with GDD. Of note, the genetic variant reported in the present study, p.Cys804Ser, was previously described in compound heterozygosity in patients diagnosed with LGMD instead of GDD. Our patient did not display characteristic GDD features nor a clinical LGMD diagnosis, but had severe osteoporosis (spinal T-score <  − 3.5 associated with a couple of postmenopausal fragility fractures). All things considered, it is unlikely that the variant explains this low bone mass phenotype. However, functional analyses and-or discovery of additional carrier patients would be necessary to get a definitive statement.

It is well known that alterations of the Wnt pathway have profound effects on bone properties, both increasing bone mass (HBM) or decreasing it (osteoporosis). LRP4 mediates SOST-dependent inhibition of bone formation by facilitating the inhibitory action of sclerostin on LRP5 [39]. It has been demonstrated that missense mutations in LRP4 that prevent its interaction with SOST, decrease the inhibitory function of sclerostin in bone formation, generating an HBM phenotype [27, 40]. On the other hand, a Lrp4-deficient mutant mice revealed shortened total femur length, reduced cortical femoral perimeter, and reduced total femur bone mineral content and BMD [41] suggesting that mutations in different domains of LRP4 may also have different effects on BMD [42]. According to HGMD, the variant found in our study was previously described as a secondary mutation in a patient with 46,XY disorder of sex development [27] and its putative contribution to the low bone density phenotype of our patient is presently unclear [26].

Finally, the variant found in the LMNA gene in the only woman with HBM might hardly explain her bone density phenotype. It is well known that mutations in this gene are often associated with Emery-Dreifuss muscular dystrophy (for which no bone density defects have been described) [43] or premature aging syndromes (which typically lead to skeletal deformities and/or osteoporosis through activation of non-canonical pathways of osteoclastogenesis) [31]. Most probably, other genetic variants are involved in the HBM phenotype of our patient and this variant is a secondary finding.

The main limitation of this study is that the gene panel was mostly focused on genes involved in low bone mass phenotypes rather than those associated with HBM since the primary aim was to describe rare variants related to OI enriched in women with LBM. Therefore, mutations involved in HBM are underrepresented in our HZ group as well as the problem of studying recessive inheritance linked to the X chromosome only in the sample of women. Another limitation is the number of samples whereby we could not obtain statistically significant differences in the burden test. An additional limitation is that the lumbar spine was selected as the phenotyping site instead of the total hip, the latter being more reproducible in terms of BMD measurement. Unfortunately, total hip BMD was not extensively measured in our center at the time of DXA acquisition. However, the mean age of our study population was 55 years, an age in which spinal artifacts are not so common as in older individuals. In any case, all DXA scans were reviewed by an expert clinician to investigate the presence of artifacts. Also, in the early postmenopausal period, alterations in BMD occur quicker at the lumbar spine than at the hip [44], making it an adequate phenotyping site for the purpose of our study. Finally, the family of the member carrying the COL1A2 variant was not willing to collaborate for a genetic study and we could not reach a conclusion about the implications of this variant in this family. One strength of this study is that using extreme-truncate selection designs could increase the power for detecting genetic variants associated with BMD. It is demonstrated by a previous GWAS performed by Duncan et al. (2011), where they replicated 21 of 26 known BMD-associated genes and additionally reported six new genes, most of them included in our gene panel [45]. Unlike the study by Duncan et al. which used a GWAS approach, we focused on sequencing a selected gene panel in order to identify rare genetic variants. Hence, both studies are complementary and contribute to deciphering the BMD genetics in postmenopausal women.