Whole Exome Sequencing Reveals Potentially Pathogenic Variants in a Small Subset of Premenopausal Women with Idiopathic Osteoporosis

Abstract

Osteoporosis in premenopausal women with intact gonadal function and no known secondary cause of bone loss is termed idiopathic osteoporosis (IOP). Women with IOP diagnosed in adulthood have profound bone structural deficits and often report adult and childhood fractures, and family history of osteoporosis. Some have very low bone formation rates (BFR/BS) suggesting osteoblast dysfunction. These features led us to investigate potential genetic etiologies of bone fragility.

In 75 IOP women (aged 20–49) with low trauma fractures and/or very low BMD who had undergone transiliac bone biopsies, we performed Whole Exome Sequencing (WES) using our variant analysis pipeline to select candidate rare and novel variants likely to affect known disease genes. We ran rare-variant burden analyses on all genes individually and on phenotypically-relevant gene sets. For particular genes implicated in osteoporosis, we also assessed the frequency of all (including common) variants in subjects versus 6,540 non-comorbid female controls.

The variant analysis pipeline identified 4 women with 4 heterozygous variants in LRP5 and PLS3 that were considered to contribute to osteoporosis. All 4 women had adult fractures, and 3 women also had multiple fractures, childhood fractures and a family history of osteoporosis. Two women presented during pregnancy/lactation. In an additional 4 subjects, 4 different relevant Variants of Uncertain Significance (VUS) were detected in the genes FKBP10, SLC34A3, and HGD. Of the subjects with VUS, 2 had multiple adult fractures, childhood fractures, and presented during pregnancy/lactation, and 2 had nephrolithiasis. BFR/BS varied among the 8 subjects with identified variants; BFR/BS was quite low in those with variants that are likely to have adverse effects on bone formation. The analysis pipeline did not discover candidate variants in COL1A1, COL1A2, WNT, or ALPL. Although we found several novel and rare variants in LRP5, cases did not have an increased burden of common LRP5 variants compared to controls. Cohort-wide collapsing analysis did not reveal any novel disease genes with genome-wide significance for qualifying variants between controls and our 75 cases. In summary, WES revealed likely pathogenic variants or relevant VUS in 8 (11%) of 75 women with IOP. Notably, the genetic variants identified were consistent with the affected women’s diagnostic evaluations that revealed histological evidence of low BFR/BS or biochemical evidence of increased bone resorption and urinary calcium excretion. These results, and the fact that the majority of the women had no identifiable genetic etiology, also suggest that the pathogenesis of and mechanisms leading to osteoporosis in this cohort are heterogeneous Future research is necessary to identify both new genetic and non-genetic etiologies of early-onset osteoporosis.

INTRODUCTION

Osteoporosis is uncommon in premenopausal women, and most affected young women have a secondary cause of bone loss or fragility^1–5. However, some young, otherwise healthy women with intact gonadal function and no known secondary cause of bone loss present with unexplained low trauma fractures while others have very low areal bone mineral density (aBMD) but no history of low trauma fractures. We have included such women in our characterization studies of and treatment studies for premenopausal idiopathic osteoporosis (IOP)^2,6,7.

We previously reported that women with IOP (n=64; 40 of whom are included in this genetics study) have severe bone microarchitectural abnormalities, with thinner cortices, and fewer, thinner, more widely separated and heterogeneously distributed trabeculae in comparison to healthy controls^2,8,9. Bone remodeling rate, whether based upon serum bone turnover markers or tetracycline-based quantitative dynamic histomorphometry, was heterogeneous. On average, bone remodeling did not differ between affected subjects and controls². However, one subgroup had high bone remodeling, hypercalciuria and elevated serum 1,25(OH)₂D levels consistent with idiopathic/primary hypercalciuria, while another subgroup had extremely low bone remodeling consistent with defects in osteoblast function.²

We have also previously reported that many women with IOP have a history of childhood fractures and more than half report a family history of osteoporosis⁴. These findings, together with their early and severe clinical presentation, and the evidence of osteoblast dysfunction in some and primary hypercalciuria in others, lead us to hypothesize that many women with IOP have a previously unrecognized primary form of osteoporosis that is genetic in etiology.

In this study, we aimed to define genetic characteristics of premenopausal women with IOP and to determine the frequency of identifiable genetic osteoporosis etiologies. We used Whole Exome Sequencing (WES) to identify high-impact variants in known candidate genes for early osteoporosis and collapsing analysis to discover new gene-disease associations. WES allows us to analyze the genetics of each sample individually, combine the samples into cohort-wide burden tests using our collapsing analysis framework, and compare our cohort to 6,540 non-comorbid female control samples as well as proband samples analyzed using identical methods at the Institute for Genomic Medicine (IGM). Additionally, we aimed to define clinical and medication response characteristics among those with identified variants.

METHODS

Premenopausal women, aged 18–48, were recruited at Columbia University Irving Medical Center (CUIMC), New York, NY and Creighton University, Omaha, NE by advertisement, and self- or physician referral. Women were recruited in two subgroups: FRACTURE subjects and LOW BMD subjects; when FRACTURE and LOW BMD groups are combined, they are called all “affected subjects”. Inclusion criteria and group definitions have been previously published^2,4,6. The FRACTURE subjects included women with a documented low-trauma fracture after age 18, regardless of whether areal BMD (aBMD) by DXA was low. The LOW BMD group included women with low aBMD by dual energy x-ray absorptiometry (DXA; T score ≤ −2.5 or Z score ≤ −2.0) at the spine or hip and no history of adult low trauma fracture. Fractures were ascertained by review of radiographs or reports and categorized as low trauma (i.e., trauma equivalent to a fall from a standing height or less) after review by a physician panel (ES, AC, RRR). Skull and digit fractures were excluded. Subjects were evaluated more than three months after their most recent fracture, and were normally menstruating/estrogen sufficient, more than 12 months postpartum and more than 6 months postweaning.

Affected subjects were enrolled at Columbia University, New York, NY or Creighton University, Omaha, NE. This study on genetics in IOP enrolled subjects who were participating in other characterization and treatment studies of IOP conducted at our institutions. The subjects enrolled in this genetics study had participated in two studies: (1) 37 were enrolled in AR049896, entitled “Idiopathic Osteoporosis in Premenopausal Women,” a cross-sectional characterization study, initiated in 2005, (2) 35 were enrolled in FD003902, entitled, “Phase 2 Study of Teriparatide for Treatment of Osteoporosis in Premenopausal Women,” a randomized controlled study of teriparatide in premenopausal women with IOP, initiated in 2011, and (3) 3 additional affected subjects had participated in both studies. All had transiliac bone biopsies after tetracycline labeling of bone forming surfaces. Bone formation rate / bone surface (BFR/BS) was calculated, as previously described^2,6. All bone remodeling data presented here are from pretreatment assessments.

Inclusion and exclusion criteria were as previously reported^2,4,6. We defined premenopausal status as regular menses off hormonal contraception, and with early follicular phase follicle stimulating hormone (FSH) levels <20 mIU/mL. Secondary causes of osteoporosis were excluded by detailed history, physical and biochemical evaluation in subjects and controls⁴: estrogen deficiency, eating disorders associated with amenorrhea, endocrinopathies (including abnormal thyroid function, hypo- or hypercalcemia, hyperparathyroidism, and Cushing syndrome), celiac or other gastrointestinal diseases, abnormal mineral metabolism, marked hypercalciuria (>300 mg/gCr), and drug exposures. Women with serum 25-hydroxyvitamin D (25-OHD) levels below 20 ng/ml were excluded. Treatment naïve women were recruited for these studies.

All subjects provided written informed consent. The Institutional Review Boards of both institutions approved these studies.

WHOLE EXOME DATA ANALYSIS AND WORKFLOW:

Within the Institute for Genomic Medicine (IGM) at Columbia University Irving Medical Center, we analyzed exome data from 75 premenopausal women with IOP, which we compared to 18,746 male and female non-comorbid controls selected from other studies at the IGM for diagnostic analyses and 6,540 female controls for refined burden tests. All patient samples were sequenced at the IGM with Roche’s KAPA library. Samples were sequenced on Illumina HiSeq2500 next-generation sequencing machines (Illumina Inc, San Diego, CA, USA) using the Dynamic Read Analysis for GENomics (DRAGEN, Illumina Inc, San Diego, CA, USA) to align reads to the Genome Reference Consortium Human Build 37. Variants were called using Analysis Tool for Annotated Variants¹⁰ an IGM variant-calling pipeline, in keeping with the Genome Analysis Tool Kit (GATK 1.6–11¹¹ and utilizing read mapping (BWA-0.5.10 to GRCh37) and PCR duplicate removal (Picard-tools 1.59).

Patient-Specific Diagnostic Analyses

The IGM variant analysis pipeline highlights rare and novel variants which are likely to affect known disease genes. We exclude variants that are of low quality as defined by GATK best practices (i.e. have low read-depth or predicted to be incorrectly called) or are recurrently observed in more than five healthy controls within our in-house variant database and external control reference samples (Exome Aggregation Consortium - ExAc, Exome Variant Server - EVS). Candidate variants must satisfy specific inclusion criteria based on plausible genetic architectures and bioinformatic signatures, such as known pathogenic variants, rare (< 6.29e-5) missense variants adjacent to known pathogenic variants, protein-truncating variants in susceptible genes, compound heterozygous variants, and newly homozygous or hemizygous variants. Candidate variants were reviewed by IGM staff including clinicians and genetics researchers to assess phenotypic match, plausible pathogenicity, and to remove variants which satisfied the automated criteria but were unlikely to contribute to the patient phenotype. The remaining candidate variants which achieved an American College of Medical Genetics (ACMG) Likely Pathogenic or Pathogenic¹² classification, along with selected relevant variants of uncertain significance, were classified as Pathogenic by the clinical research team and validated using Sanger sequencing in a clinical diagnostic laboratory prior to disclosure to the subjects. Given the limited number of osteoporosis-associated genes, we also considered potentially novel dominant inheritance by investigating Variants of Uncertain Significance, as has been done in similar rare osteoporosis cohorts¹³. We focused on osteoporosis-relevant variants which achieved this ACMG category primarily due to a known autosomal recessive mode of inheritance in their associated gene with individual variants that are likely disruptive or otherwise satisfied multiple ACMG Pathogenicity criteria, especially for compound heterozygote variants which were paired with a putatively neutral/benign variant.

Analysis of Common Variants in Phenotypically-Relevant Genes

To identify any excess burden between cases and IGM controls (n=18,746), we tabulated the number of variants at all allele frequencies in 16 osteoporosis genes of interest that were also known to have some common variation in the human population in gnomAD the Genome Aggregation Database (ALPL, COL1A1, COL1A2, COL2A1, COL3A1, COL5A1, COL5A2, COL9A1, COL9A2, COL9A3, COL10A1, COL11A2, GBA, LRP5, PLS3, and WNT), with particular focus on LRP5 because a prior genetics study of young adults with primary osteoporosis found a high prevalence of variants in LRP5¹⁴.

Cohort Level Burden Tests

We followed the procedures in Povysil, 2019¹⁵ for cohort pruning and rare variant collapsing analyses. Collapsing analyses are burden tests in which the rates of variation in individual genes are combined or “collapsed” across samples and compared between cases and controls. Complex phenotypes are likely to be enriched for rare variants¹⁶. Because by definition these exact rare variants will not commonly be found in all the cases, looking across a gene increases statistical tests’ power and allows exploration of gene association. Several initial preprocessing and quality control steps help remove unwanted noise to ensure that the comparisons between cases and controls are valid. For these analyses, our control cohort consisted of 6,540 healthy, unrelated non-comorbid female samples. All samples were required to have below 8% contamination rate and to have at least 10x coverage in at least 90% of the Consensus Coding Sequences (CCDS) regions. We reviewed the full cohort for kinship status using KING relatedness software¹⁷ and found no relationships. We conducted principal component analysis using FlashPCA¹⁸ and used the Louvain method for community detection on the six most informative principal components¹⁹ to create clusters based on geographic ancestry. We employed UMAP²⁰ for dimensionality reduction in order to visualize the clusters (Supplemental Figure 1). We retained clusters that had at least 8 cases, keeping a European cluster with 55 cases and an admixed cluster with 8 cases. All further analyses are applied to each cluster at a time, with statistical test values collated across the clusters. In each cluster, we only considered sites with equivalent rates of coverage between cases and controls, removing sites in the CCDS (HG19, release 20) with a 7% or greater coverage difference between all cases and all controls (average 3.1MB sites removed), leaving an average of 30.4MB sites in each cluster to be used for analysis.

We ran a variety of collapsing models, each of which considers specific properties for qualifying variants. Variants were filtered for varying rates of frequency, predicted effect, intolerance scores, and pathogenicity scores¹⁵. We retained variants that passed GATK’s recommended variant filters. Each model defines carrier status, allowing us to compare, in every gene, rates of cases carrying a qualifying variant to rates of controls. We used the exact two-sided Cochran-Mantel-Haenszel (CMH) statistical test to compare the carrier rates and create a p-value for our test statistic.

For specific hypotheses about potentially associated pathways and known gene sets, we collapsed onto specific gene sets using a list of 184 primary osteoporosis genes, a more general list of 305 potentially associated genes (from literature reviews^21–23 and scraping OMIM), and on 48 genes related to adrenocortical hormones (gene lists are presented in Table 1). We ran these analyses both on our primary cohort and upon a subset of 33 IOP samples from subjects with BFR/BS below the median (BFR/BS < 0.0091 mm²/mm/year). For each of the varying collapsing model filters, we compared the rates of cases with at least one qualifying variant in any of the gene set members versus the same rates in controls, resulting in a p-value for each gene set.

TABLE 1:

GENE SETS USED

Description	Number of Genes	Genes
Osteoporosis genes, from literature review.	184	‘WNT3A, CYP17, TNFSF11, SOST, HDAC5, FKBP10, TGFB1, ITGA1, FOXL1, BMND, RANK, CATK, ITGB3, CALCR, VDR, REL, ADAMTS18, SOX6, COL, SPTBN1, SOX5, IL6, SOST, TMEM38B, SFRP4, DPYD, TYROBP, TUBA1B, CCNE1, TCIRG1, STARD3NL, RIL, FOXC2, P3H1, MARK3, TGFB3, MEF2C, LRP4, PLS3, FAM3C, PDGFD, IFITM5, OSTM1, MTHFR, CTNNB1, CSTA, CLCN7, MHC, SP7, SEC24D, MEPE, SERPINF1, TNFRSF11A, TNFRSF11B, RANKL, DCDC5, DCDC1, OPN, LEPRE, BMP2, BMP1, LRP5, WNT1, ESR1, DKK1, ZBTB40, CTR9, OPG, SPP1, SERPINH1, CRHR1, FLJ42280, PPIB, GPR177, CRTAP, RIZ1, OCIL, TIMP2, ERb, WNT10b, ER-a, CALM1, PRL, WNT10B, ARHGEF3, VEGF, PIR, WNT7b, IL-6, MDR1, LHB, ALOX5, DMP1, ENPP1, LRP1, NR1I3, FGFR2, ROR2, FGFR1, BMP7, IL-15, RUNX2, NPY, DBP, WNT7B, NFATC1, TNFRII, CYP19, Runx2, HOXA, TSHR, PTN, PTH, IL-23, WISP3, ANXA6, THSD7, THSD4, IGFBP2, MSTN, CNR2, APC, P2X 7, PPAR-g, TNFRSF1B, CYP19A1, SHBG, FLNB, FLNB FLT1, GHRH, CTR, IL-23R, FABP3, ADCY10, TGF-b1, FLT1, WNT3a, FZD1, THSD7A, CYP1B1, FZD6, Smad6, PTHRLH, SPARC, TNF-a, CASR, SRD5A2, HMGA2, COMT, LEPR, SFRP2, SFRP1, LRP6, P2X7, ALOX15, PLOD, SREBF1, COL1A2, COL1A1, NOG, CAR, PTHR2, PTHR1, CYP17A1, BMPR1B, GR, TWIST1, PTHLH, GDF5, HSD11B1, NOS3, ESRRA, LTBP2, LHCGR, CD40, CA10, AR, IL6R, PBX1, CA8, DKK2, MMP2, IL-6R, ALPL
Any bone-related genes, from literature reviews and scraping OMIM.	305	ABCC9, ACAN, ACVR1, ADAMTS2, ADAMTSL2, AGC1, AGXT, ALDH18A1, ALPL, AMER1, ANKH, ANO5, ANTXR2, AP2S1, APC, ARMC5, ASPN, ASXL2, ATL1, ATL3, ATP6V0A2, ATP7A, ATP7B, B3GALT6, B3GAT3, B4GALT7, BANF1, BDMR, BMND7, BMND8, BMP1, BMP2, BRAF, C1S, CA2, CALCR, CANT1, CASR, CAVIN1, CBS, CCT5, CHEK2, CHST14, CLCN5, CLCN7, COL10A1, COL11A2, COL1A1, COL1A2, COL2A1, COL3A1, COL5A1, COL5A2, COL9A1, COL9A2, COL9A3, COMP, COX4I2, CREB3L1, CRIPT, CRTAP, CTC1, CTSC, CTSK, CYBA, CYBB, CYP27A1, DCHS1, DDX58, DGKZ, DHCR24, DKC1, DNMT1, DSPP, DUSP6, DVL, DVL1, EED, EIF2AK3, ERCC6, ESR1, FAM111A, FAM123B, FAM134B, FAM20A, FAM20C, FAT4, FBN2, FERMT3, FGF17, FGF23, FGFR1, FGFR3, FKBP10, FKBP14, FLNA, FLNB, FLRT3, FRZB, FWS, G6PC, G6PT1, GALNS, GBA, GCM2, GLB1, GLIS3, GNAS, GNAS-AS1, GNAS1, GNASAS, GNPAT, GNPTAB, GORAB, GZF1, HECW2, HFE, HLA-DQA1, HLA-DQB1, HNRNPA1, HPGD, HRAS, HS6ST1, HSD17B4, HSPG2, IARS2, ICK, IER3IP1, IFIH1, IFITM5, IFT122, IGF1, IL17RD, IL1RN, IRX5, KCNJ1, KISS1R, KRAS, LBR, LEMD3, LEPRE1, LIFR, LMNA, LRP4, LRP5, LRP6, MAFB, MALT1, MAP2K1, MAP2K2, MAPRE2, MATN3, MET, MGAT2, MIFT, MIR2861, MITF, MMP13, MMP2, MPV17, MTAP, NAGA, NCF1, NCF2, NDN, NGF, NGFB, NOLA3, NOP10, NOTCH2, NPR2, NRAS, NSMF, NTRK1, OCRL, OCRL1, OI16, ORC1, OSTM1, P3H1, P4HB, PAPSS2, PCCA, PCCB, PCNT, PCNT2, PDB4, PDE11A, PDGFRB, PDLIM4, PEX12, PEX2, PHEX, PIGT, PIGY, PLOD, PLOD1, PLOD2, PLOD3, PLS3, PMM2, POF1B, POLD1, PORCN, PPIB, PPNAD4, PRKACA, PRKAR1A, PROK2, PRS, PSMB8, PTDSS1, PTH1R, PTHR1, PTRF, PYCR1, RAB3GAP1, RAB7A, RB1, RCHTS, RECQL2, RECQL4, RETREG1, RIN2, RNF125, RNU4ATAC, RPL11, RPS19, RSPRY1, RUNX2, SC5D, SC5DL, SEC23A, SEC24D, SEDL, SERPINH1, SFRP4, SH3BP2, SH3PXD2B, SHOX, SKI, SLC12A1, SLC17A5, SLC26A2, SLC34A1, SLC37A4, SLC39A13, SLC39A8, SLC40A1, SLC7A7, SLC9A3R1, SLCO2A1, SMAD3, SMAD4, SMARCAL1, SMARCD2, SMPD1, SMS, SNRPN, SNX10, SOST, SP7, SPARC, SPDT, SPRY4, SPTLC1, SPTLC2, SQSTM1, SRC, STAT1, STN1, STX16, TAPT1, TBC1D24, TBCE, TCBE, TCIRG1, TERT, TGFB3, TGFBR2, TINF2, TMEM165, TNFRSF11A, TNFRSF11B, TNFSF11, TNXB, TP53, TRAPPC2, TREM2, TRMT10A, TRPS1, TRPV4, TYROBP, UFSP2, UROS, VDR, VPS53, WBS, WISP3, WNK1, WNT1, WRN, ZBTB20, ZMPSTE24, ZNF687
Adrenocortical, from scraping OMIM.	48	AAAS, ABCD1, AIRE, ARNT2, BCOR, CACNA1H, CDKN1C, CYP11B1, CYP11B1, CYP11B2, GK, EBP, GCCD3, GLI3, GNAS, GNAS1, HESX1, HPE1, HSD17B4, HSD3B2, HYLS1, ICK, KCNJ5, KIAA0196, LHX3, LHX4, LMNA, MCM4, MKS1, MRPS7, NFKB2, NR0B1, NSDHL, PEX5, PEX6, POMC, POR, PRKAR1A, PROP1, RBM28, SAMD9, SGPL1, SIX3, STEAP3, WASHC5, WNT3, WNT4, ZMPSTE24,

RESULTS

CHARACTERISTICS OF THE STUDIED SUBJECTS: ALL AFFECTED SUBJECTS

Seventy five women enrolled in AR049896 or FD003902 consented to and provided samples for genetic analysis by WES; all underwent tetracycline-labelled transiliac bone biopsy. Characteristics of the affected subjects are shown in Table 2. Affected subjects ranged in age from 20 to 49 years. The great majority self-categorized race as white/Caucasian. Fracture subjects were included on the basis of low trauma spine, hip, pelvis, rib, and extremity fractures. Among the 59 subjects included on the basis of fracture history, over 60% had a history of multiple fractures, and 29% had a history of vertebral fractures. Approximately a quarter of the fracture patients had a history of fracture occurring in association with pregnancy or lactation. Many had a history of childhood fractures and the majority reported a family history of osteoporosis.

TABLE 2:

Characteristics of the Cohort Studied: 75 premenopausal women with idiopathic osteoporosis

	FRACTURE SUBJECTS N = 59	LOW BMD SUBJECTS N = 16	ALL AFFECTED SUBJECTS N = 75
Race (NIH Categories)
-White	58	15	73
-Black African American	1	0	1
-American Indian/Alaska Native	0	0	0
-Asian	0	0	0
-Native Hawaiian/Other Pacific Islander	0	0	0
-More than one race reported	0	1 (White and Native Hawaiian/Other Pacific Islander)	1
Ethnicity (NIH Categories)
Non-Hispanic	56	15	71
Hispanic	3	1	4
Age (years)	37 ± 7	41 ± 7	38 ± 7
Ht (cm)	163 ± 7	163 ± 6	163 ± 7
Wt (kg)	60 ± 11	56 ± 8	59 ± 11
BMI (kg/m2)	22.6 ± 4.0	20.8 ± 2.0	22.2 ± 3.7
BMD Z scores
LS	−1.9 ± 1.2	−2.1 ± 0.7	−1.9 ± 1.1
TH	−1.3 ± 0.9	−1.7 ± 0.7	−1.3 ± 0.9
FN	−1.5 ± 0.9	−2.0 ± 0.7	−1.6 ± 0.9
DR	0.1 ± 0.9	0.1 ± 0.9	0.1 ± 0.9
Number of adult fractures	3.0 ± 2.5	0	2.3 ± 2.5
History of multiple fractures in adulthood	37/59= 63%	NA	37/75=49%
History of childhood fractures	24/59 = 41%	4/16 = 25%	28/75 = 37%
History of pregnancy or lactation associated fractures	14/59 = 24%	NA	14/75 = 19%
History of vertebral fractures	17/59 = 29%	0/16 = 0%	17/75 = 23%
Family history of osteoporosis	36/59 = 61%	12/16 = 75%	48/75 = 64%
Bone formation rate (histomorphometry of bone biopsy sample; n=71); mm2/mm/year	0.010 ± 0.008	0.011 ± 0.005	0.010 ± 0.008
Low bone turnover as previously defined	15/55 = 27%	2/16 = 13%	17/71 = 24%

We previously categorized IOP women with tissue-level bone formation rate analyzed on transiliac bone biopsy samples (BFR/BS) < 0.006 mm²/mm/year as “low-turnover IOP”²; 24% were classified as low-turnover IOP.

The biopsy studies enrolled treatment naïve IOP subjects. As previously reported⁶, brief and distant treatment with oral bisphosphonates or teriparatide was allowed for study entry. Seven of the 75 women had received prior treatment (5 with oral bisphosphonate). None of these subjects were among the 24% classified as low-turnover IOP.

GENETICS FINDINGS

Diagnostic Analyses

We did not find any ACMG Likely Pathogenic compound heterozygous or newly homozygous variants in genes with strong phenotypic matches to osteoporosis. WES revealed four different heterozygous variants of potential clinical significance in four of the 75 subjects, all of which were ACMG Likely Pathogenic variants in genes associated with dominant inheritance. WES also revealed four different plausibly relevant heterozygous variants, individually categorized as ACMG Pathogenic or Likely Pathogenic, in genes associated with recessive inheritance in four additional subjects. Because subjects were heterozygous for these variants associated with recessive inheritance, we considered them variants of uncertain significance (VUS). No subject had ACMG Pathogenic variants or any VUS found in COL1A1 or COL1A2, the genes responsible for over 90% of cases of osteogenesis imperfecta (OI).

ACMG Likely Pathogenic Variants Associated with Known Dominant Inheritance

The four Likely Pathogenic variants were either protein-truncating variants (PTVs), which are expected to severely disrupt gene function, or were missense variants occurring very close to known Pathogenic variants (Table 3). One subject had a variant in PLS3; the other 3 subjects had 3 different variants in LRP5 (Table 3). Each of the reported variants were identified in only a single individual. All four subjects with ACMG Likely Pathogenic variants were in the FRACTURE subgroup.

Table 3:

LIKELY PATHOGENIC GENETIC VARIANTS IDENTIFIED

	Subject #1	Subject #2	Subject #3	Subject #4
Gene	PLS3	LRP5	LRP5	LRP5
OMIM Inheritance	X-linked dominant	AD, AR	AD, AR	AD, AR
Variant ID	X-114863640-GA	11-68205958-GC-G	11-68157455-G-C	11-68197163-G-T
HGVS	ENST00000420625.2:c.367+1G>A	ENST00000294304.7:c.4157delC ENSP00000294304.6:p.Ile1387Serf sTer52	ENST00000294304.7:c.1519G>C ENSP00000294304.6:p.Gly507Arg	ENST00000294304.7:c.3758G>T ENSP00000294304.6:p.Cys1253Phe
Effect	Loss-of-Function: splice-donor variant	Loss-of-Function: Frame-shift	Missense	Missense
Zygosity	Heterozygous	Heterozygous	Heterozygous	Heterozygous
Allele occurrence s in gnomAD	0	0	0	1
ACMG	Likely Pathogenic	Likely Pathogenic	Likely Pathogenic	Likely Pathogenic
Nearby Pathogenic variants			PMID 22456437: c.1519G>A p.G507S (HGMD)	PMID 20340138: Likely pathogenic – rs768615287

The PLS3 variant (c.367+1G>A) and one of the LRP5 variants (p.Ile1387SerfsTer52) were flagged by our pipeline because they are rare PTVs in known disease genes. They were classified as ACMG Likely Pathogenic variants because the genetic architecture of PLS3 includes X-linked dominant pathogenic PTVs causing osteoporosis with fractures²⁴ and because LRP5 has an established dominant loss-of-function genotype. The other two LRP5 variants were missense variants that were flagged for being at the same site as previously confirmed pathogenic variants. One of the missense variants (p.Cys1253Phe) had been previously reported as Likely Pathogenic in ClinVar²⁵; the other (p.Gly507Arg) was a different amino acid change at the exact site as a previously confirmed disease-causing mutation (DM) in HGMD²⁶. Prior evidence that missense changes at these sites contributes to osteoporosis differentiated these variants from other missense variants in osteoporosis associated genes and resulted in their classification as ACMG Likely Pathogenic variants.

Variants of Uncertain Significance

Four subjects’ samples were heterozygous for likely damaging variants in genes known only to have recessive inheritance (SLC34A3, FKBP10, HGD), and were categorized as variants of uncertain significance (VUS). Each of the variants reported were identified in only a single individual. Subject #5 was heterozygous for a frameshift variant²⁷ in FKBP10 that is considered Pathogenic in ClinVar. Subject #6 was heterozygous for a missense mutation in SLC34A3²⁸ that is considered Pathogenic in ClinVar and also had another SLC34A3 variant that is very common among controls. The variant included in Table 4 is the variant found in ClinVar; due to the inheritance associated with SLC34A3 and observation of the common allele in many healthy controls, the variant was classified as a VUS. Two subjects had different VUS identified in HGD: Subject #7 had a frameshift mutation that is considered Likely Pathogenic and Subject #8 had a missense mutation²⁹ that is considered Pathogenic in ClinVar. Both Subjects #7 and #8 were heterozygous for variants considered Pathogenic in ClinVar and also additionally contained the same common HGD allele. Due to the prevalence of this common allele, it is unlikely that it is significantly disruptive; however, it may be mildly disruptive and contribute to Pathogenicity when paired with other disruptive alleles. Additional details are shown in Table 4. While these variants did not satisfy the genetic architectures and bioinformatic signatures used by our variant analysis pipeline to highlight them as potential diagnostic variants, the accompanying subject phenotypes and predicted variant effects motivated further investigation since these relevant variants may have increased susceptibility for osteoporosis in some way¹³.

TABLE 4:

VARIANTS OF UNCERTAIN SIGNIFICANCE IDENTIFIED

	Subject #5	Subject #6	Subject #7	Subject #8
Gene	FKBP10	SLC34A3	HGD	HGD
OMIM Inheritance	AR	AR	AR	AR
Variant ID	17-39975558-T-TC	9-140127675-C-T	3-120369585-GATGGGCATGTCCTTCCCTAGAACTGAGCCACTTACCTGTTCTCCATGGAGGTATTGC-G	3-120352080-T-C
HGVS	ENST00000321562.4:c.824_825insC ENSP00000317232.4:p.Gly278ArgfsTer95	ENST00000538474.1:c.575C>T ENSP00000442397.1:p.Ser192Leu	ENST00000283871.5:c.413_434+35delGCAATACCTCCATGGAGAACAGGTAAGTGGCTCAGTTCTAGGGAAGGACATGCCCAT	ENST00000283871.5:c.1102A>G ENSP00000283871.5:p.Met368Val
Effect	Frameshift	Missense	Frameshift	Missense
Zygosity	Heterozygous	Compound Heterozygous with a Common Allele	Compound Heterozygous with a Common Allele (3-120389316-T-A; c.240A>T; p.Gln80His)	Compound Heterozygous with a Common Allele (3-120389316-T-A; c.240A>T; p.Gln80His)
Allele occurrences in gnomAD	30	99	1	51
ACMG	Pathogenic	Pathogenic	Likely Pathogenic	Pathogenic
Additional information	This variant is DM in HGMD and Pathogenic in ClinVar for Osteogenesis Imperfecta AR PMID 20362275.	ClinVar RSID 199690076. This variant is Pathogenic in ClinVar. A variant at the same site with a different nucleotide change is also Pathogenic in Clinvar. PMID 16358215. Patient was comphet for this one (c.575C>T) and c.304+2T→C		HGD gene was formerly named AKU DM in HGMD and Pathogenic in ClinVar for alkaptonuria. PMID 9529363.

CLINICAL PHENOTYPE OF THE EIGHT SUBJECTS WITH GENETIC FINDINGS

Subjects with ACMG Likely Pathogenic Variants:

All four subjects were in the FRACTURE subgroup and had sustained adult low trauma fractures at sites rich in trabecular bone. Three had aBMD by DXA that was below the expected range for their age at the spine (Z score < −2; Table 5). Two subjects had presented with fractures (spine, hip) in the context of pregnancy or lactation, a presentation called pregnancy and lactation associated osteoporosis (PLO)³⁰. The subject with the PLS3 variant also had a history of renal stones. Details of the clinical phenotypes are shown in Table 5.

Table 5:

CHARACTERISTICS OF THE SUBJECTS WITH LIKELY PATHOGENIC GENETIC VARIANTS IDENTIFIED and THOSE WITH VARIANTS OF UNCERTAIN SIGNIFICANCE

	Subjects with Likely Pathogenic Variants				Subjects with Variants of Uncertain Significance
	Subject #1	Subject #2	Subject #3	Subject #4	Subject #5	Subject #6	Subject #7	Subject #8
Age (years)	31	44	44	38	43	29	34	41
Ht (cm)	154	162	161	167	167	168	164	154
Wt (kg)	56	74	62	64	53	51	67	39
BMI (kg/m2)	23.6	28.1	23.9	23.0	19.1	18.2	24.8	16.5
BMD Z scores
LS	−3.8	−2.1	−3.1	−1.7	−2.5	−3.2	−1.9	−1.2
TH	−2.4	−1.0	−1.6	−0.8	−1.7	−1.4	−1.4	−1.8
FN	−2.2	−1.5	−2.6	−0.7	−2.5	−1.6	−1.3	−2.6
DR	−0.8	−0.8	−0.1	0.4	0.3	−2.1	−0.3	0.1
Number of adult fractures	7	2	1	2	0	2	10	0
Fracture Types (in adulthood)	wrist, rib, 5 vertebrae (thoracic)	ankle and wrist	Rib	Right hip × 2	N/A	2 vertebrae (thoracic)	Multiple tibia and femur, 2 vertebrae (thoracic and lumbar)	N/A
History of pregnancy or lactation associated fractures	YES	NO	NO	YES	NO	YES	YES	NO
History of vertebral fractures	YES	NO	NO	NO	NO	YES	YES	NO
History of multiple fractures in adulthood	YES	YES	NO	YES	NO	YES	YES	NO
History of childhood fractures	YES	NO	YES	YES	NO	YES	YES	NO
Family history of osteoporosis	YES	YES	YES	NO	YES	NO	NO	NO
History of Renal Stones	YES	NO	NO	NO	NO	YES	NO	YES
Bone formation rate (histomorphometry of bone biopsy sample; mm2/mm/year)	0.000 (undetectable)	0.000 (undetectable)	0.007	0.002	0.003	0.010	0.013	0.023
Low bone turnover as previously defined	YES	YES	NO	YES	YES	NO	NO	NO
Responded to teriparatide, a bone formation stimulating medication. Response defined as spine BMD increase > Hologic least significant change (0.026 g/cm2) at 12 months	YES	N/A	N/A	NO	NO	YES	YES	YES

Subjects with VUS:

Among the four subjects with VUS, two had very low BMD without a history of adult low trauma fracture (LOW BMD subgroup), and two had severe PLO with multiple vertebral fractures (FRACTURE subgroup). Three had aBMD by DXA that was below the expected range for their age at the spine (Z score < −2; Table 5). Subjects #6, #7 and #8 had VUS in genes associated with nephrolithiasis. Two subjects (one with a variant in SLC34A3, and one with a variant in HGD) had a history of renal stones. Subject #6 (SLC34A3 variant) had severe recurrent nephrolithiasis and hypercalciuria. Subject #8 reported prior episodes of symptomatic nephrolithiasis. Neither subject #7 or #8 (both with HGD VUS) had hypercalciuria at baseline (both 24h urine calcium measures < 160 mg/gCr). Additional details of the clinical phenotypes are shown in Table 5.

BONE TURNOVER AND RESPONSE TO TERIPARATIDE IN THE EIGHT SUBJECTS WITH GENETICS FINDINGS

Bone remodeling rates differed in the eight subjects with genetic findings. The five subjects with variants in PLS3, LRP5, and FKBP10 had low tissue level bone remodeling rates; 4 of the 5 met our previously defined² criteria for low turnover IOP based on BFR/BS. In contrast, Subjects #6, 7, and 8 all had VUS that, in the homozygous state, would be hypothesized to lead to increased bone resorption and also had higher bone remodeling rates at the tissue level. None of the eight subjects with genetic findings had received any prior osteoporosis therapy.

Six of the eight subjects with identified genetic variants took part in studies of teriparatide for premenopausal IOP^6,7. Enrollment occurred > 12 months postpartum and > 6 months postweaning in all subjects. Four of the subjects, those with variants in PLS3, SLC34A3 and HGD, responded well to teriparatide. Subject #1 (PLS3 variant) had a robust response to teriparatide with lumbar spine BMD increases of 13% at 12 months, and 18% at 24 months, while her femoral neck BMD increased by only 3% at 24 months. The 3 subjects with VUS related to nephrolithiasis (SLC34A3 and HGD) also had robust BMD response to teriparatide (lumbar spine BMD increase 6–12% at 12 months). In contrast, two subjects with variants in LRP5 and FKBP10 had poor responses to teriparatide. Subject #4 (the only treated subject with an LRP5 variant) experienced much smaller increases in lumbar spine BMD of 3% at 12 months and 6% at 24 months, while femoral neck BMD increased by 5% by 24 months. Subject #5 (with FKBP10 VUS) had an increase in lumbar spine BMD of less than 2% at 12 months. Both subjects #4 and #5 were considered teriparatide “nonresponders” based on 12M LS response for analyses conducted for teriparatide studies^6,7.

Analysis of Common Variants in Phenotypically-Relevant Genes

A prior genetics study of young adults with primary osteoporosis found a high prevalence of variants in LRP5¹⁴. Therefore, we compared the burden of all LRP5 variants between cases and controls across different allele frequencies. We did not find any significant excess of variants. While 34 subjects (45%) had a qualifying variant, this rate was quite similar to the rate in the reference control cohort (36%, 6,762/18,746, p=0.12) with the primary difference explainable by the rare, confirmed diagnostic variants described above. We investigated the common LRP5 variants highlighted in other studies (Supplementary Table 2) and found notable differences among the tabulated allele frequencies between our controls and other resources, particularly for variants near regions of low sequencing complexity which may introduce methodological differences in alignment and variant calling software that are not typical when considering rare variation. We additionally did not find increased burden in common variants across all 16 implicated osteoporosis genes described above.

Cohort Level Burden Tests

Collapsing onto phenotypically relevant gene sets did not reveal significant increased burden between cases and controls. Our monogenic collapsing analysis showed notable but non-genome-wide significant increased burden in COL6A6 within our protein-truncating variant (PTV) model. Five cases (7.9%) had a PTV rarer than 0.1% in gnomAD and evaluated as real by the Loss-Of-Function Transcript Effect Estimator (LOFTEE³¹), while only eight (0.37%) of controls did (CMH p-value 2.6e-05). The quantile-quantile plot displaying the distribution of expected versus observed p-values from the CMH test (Supplemental Figure 2) shows no inflation. COL6A6 encodes a component of the extracellular matrix (Collagen VI)³², as do known osteoporosis-associated genes COL1A1 and COL1A2 (along with implicated collagen genes COL2A1, COL9A1, COL9A3, and COL11A2). COL6A6 has also been associated with atopic dermatitis³³ and lung adenocarcinoma³⁴, suggesting this collagen may not be exclusively associated with osteoporosis and instead is a risk factor that may contribute to susceptibility to osteoporosis.

DISCUSSION

In this group of 59 premenopausal women with unexplained low trauma adult fractures and 16 premenopausal women with unexplained very low BMD, WES found increased PTV signal in COL6A6 and identified ACMG Likely Pathogenic variants in four women. One was found to have a heterozygous variant in PLS3, while three were found to have different heterozygous variants in LRP5. In addition, heterozygous VUS were identified in 4 other subjects that were potentially relevant to their clinical presentations. These unique data allowed us to relate patient-specific genetic findings in relevant genes in these women to both tissue-level bone remodeling and to clinical characteristics including history of nephrolithiasis and response to the bone-anabolic medication, teriparatide.

PLS3 is located on the X chromosome and codes for plastin 3. Variants have been associated with significant bone fragility in hemizygous males, whereas heterozygous females have been reported to have a milder presentation or normal bone density^24,35. In this case, we identified a severely disruptive heterozygous variant in a female with significant bone fragility (fractures) and a presentation consistent with PLO. The patient’s father also had a history of low trauma fractures and was later found to carry the hemizygous, X-linked variant.

The LRP5 gene encodes low-density lipoprotein receptor-related protein 5, which plays a key role in skeletal homeostasis through several potential mechanisms including (1) as part of a transmembrane complex that binds Wnt signaling proteins leading to increased osteoblast differentiation and survival³⁶ and (2) via inhibition of serotonin pathways resulting in an increase in osteoblast proliferation^37,38. Gain-of-function mutations lead to increased bone mass³⁹, while homozygous loss-of-function mutations are associated with osteoporosis-pseudoglioma syndrome⁴⁰. Heterozygous mutations have been associated with bone fragility in the absence of ophthalmic findings in both children⁴¹ and adults¹⁴, including both protein-truncating variants and missense variants. Relevant to the clinical presentation of subject #4, several studies have documented heterozygous and compound heterozygous LRP5 variants in patients with PLO^13,42,43. The three different Likely Pathogenic variants we confirmed in our cohort accord with these established genotypes.

Heterozygous VUS were found in 4 additional subjects, one with a variant in FKBP10, which is associated with type XI OI, and 3 with variants in genes associated with hypercalciuria and/or increased bone resorption and renal stone disease (SLC34A3 and HGD). The variants found were individually classified (ACMG) as Pathogenic or Likely Pathogenic. However, because only autosomal recessive inheritance is known in the associated genes and these subjects were heterozygous for these variants, we classified the variants as VUS.

Homozygous mutations in FKBP10 are associated with osteogenesis imperfecta Type XI. Subject #5 was heterozygous for a frameshift variant²⁷ in FKBP10 that is considered Pathogenic in ClinVar and occurs several times in gnomAD controls. FKBP10 variants are considered to be Pathogenic with autosomal recessive inheritance and in the absence of another Pathogenic allele, this variant was classified as a VUS. Our subject had very low tissue level BFR/BS and poor response to teriparatide (Table 5). It is possible that the heterozygous state contributed to these characteristics.

Homozygous or compound heterozygous mutations in SLC34A3, encoding the sodium (Na⁺)-dependent phosphate cotransporter 2c (NPT2c), are associated with hypophosphatemic rickets with hypercalciuria. In kindred studies, heterozygous carriers have been shown to have milder biochemical profiles and an increased frequency of renal calcifications and renal stones^44,45. A heterozygous carrier state (for a different heterozygous SLC34A3 variant) has been documented in a patient with PLO¹³. Subject #6 had severe recurrent nephrolithiasis, hypercalciuria and a high tissue level bone remodeling rate, suggesting that the pathogenesis of her osteoporosis may be related to hypercalciuria. Given both her severe osteoporosis/PLO and severe nephrolithiasis, as well as the prior data on heterozygous carrier state, we considered this VUS to potentially contribute to her phenotype.

Homozygous mutations in HGD are associated with alkaptonuria, a condition characterized by accumulation of homogentisic acid in many tissues. Clinical features include arthropathy, tendinopathy, osteoporosis, extraskeletal calcifications, renal stones and high bone resorption that has been attributed to inflammatory effects of homogentisic acid accumulation⁴⁶. Subjects #7 and #8 had different HGD variants. While we considered both HGD variants to be VUS, the fact that one subject had a history of nephrolithiasis and both had relatively high tissue level bone remodeling rates consistent with increased bone resorption supports our hypothesis that these variants could contribute to their clinical presentation.

While the genetic characteristics of the VUS do not allow us to be certain that there is a causal relationship between the finding and the clinical presentation, we consider it possible that the VUS could increase susceptibility to the onset of osteoporosis in early adulthood and could influence the clinical characteristics or response to teriparatides in these subjects. Therefore, we included information on our study participants with these potentially relevant VUS, as has been done in other early-onset osteoporosis cohorts¹³. Supporting our inclusion of these findings, several aspects of these women’s clinical presentation are consistent with the predicted pathogenetic mechanisms of the variants of interest.

Despite extensive analysis via WES, we found no variants of direct clinical significance in 89% of the subjects, and reached a confident diagnosis in 5% of our subjects (4/75). Our results differ from the larger proportion recently reported in a similar group of patients with primary osteoporosis studied using candidate gene sequencing (Collet et al.)¹⁴ and are similar to the 9% rare and novel variant diagnostic rate in the same study (Collet et al.). Similar to Collet et al., we observed elevated rates among a few common LRP5 variants compared to the Genome Aggregation Database (gnomAD). In contrast, however, we found that these variants were also elevated in our internal controls compared to gnomAD; this may possibly be due to population stratification, since our control cohort has higher rates of Caucasians (58%) and Middle Eastern (10%) than gnomAD. We observed that these common variants, which occurred at notably different allele frequencies between our controls and gnomAD, contained local regions of low sequence complexity that may introduce methodological differences in alignment in variant calling software. While we cannot confirm the exact source of these differences in allele frequencies, none of the common LRP5 variants highlighted in other studies were found to be enriched among our IOP cohort compared to controls. The samples in Collet et al.¹⁴ were enriched for novel and rare LRP5 variants (binomial p-value = 2.2e-5) and the common variant Val667Met (binomial p-value = 1.1e-3) compared to either gnomAD or the IGM controls. We did not see this trend among our samples. This may reflect a more specific or severe phenotype among the Collet et al. samples.

Our findings also differ from the finding of “relevant genetic variants” (which include VUS) in 50% of women in a cohort with PLO¹³, a group notable for a very severe presentation of adult early-onset osteoporosis, often with multiple vertebral fractures. In our cohort of 75 women with IOP, 14 had PLO. In this PLO subgroup, we found relevant genetic variants (including VUS) in a comparable proportion of 29% (4/14).

Among the four PLO subjects with relevant variants, two had Likely Pathogenic variants and two had VUS. As noted above, tissue-level BFR/BS varied among these four subjects with different variants. We have previously reported that the PLO subgroup within our IOP cohort have a more severe clinical presentation and lower BFR/BS, on average, than the group as a whole⁴⁷. Other studies in women with PLO have documented both new osteogenesis imperfecta diagnosis⁴⁸ and LRP5 mutations^13,42,43. While these data suggest that genetic variants leading to bone formation defects would be most likely in the PLO subgroup, our findings here suggest that several mechanisms may contribute to risk for PLO. The severity of the PLO presentation suggests a primary or genetic etiology of their bone fragility. Yet, our diagnostic pathway identified Pathogenic Variants or VUS in only four of the 14 women classified as PLO within this cohort. There is clearly a need for further investigation of potential genetic etiologies of PLO, including both those that may affect bone formation and those that may affect renal calcium handling and/ or bone resorption.

Of note, the four subjects with PLS3 and LRP5 defects and the subject with FKBP10 VUS had low tissue level bone remodeling, which suggests that their osteoporosis could be due to defective bone formation rather than excessive bone resorption. Identification of specific genetic etiologies that predict impaired bone formation may have therapeutic implications. While treatments that stimulate bone formation might be expected to be efficacious in such women^6,7, hree subjects with PLS3, LRP5 and FKBP10 variants and low tissue level bone remodeling varied in their response to teriparatide. This observation suggests that drugs that stimulate bone formation may be an effective therapy for some causes of osteoblast dysfunction, but not for others. Responsiveness may depend on whether the osteoblast is inherently normal but insufficiently stimulated or whether it lacks the capacity to be stimulated by osteoanabolic drugs. Additionally, some studies suggest that genetic defects in bone formation may be associated with increased risk of atypical femoral fractures, both in treatment naïve and bisphosphonate treated patients⁴⁹. These potential clinical consequences suggest that genetic testing should have a role in the care of young adult subjects with primary forms of osteoporosis.

This study has several limitations. There may have been bias in that more severely affected women or women with a family history of early osteoporosis may have been more likely to enroll in biopsy, treatment and genetic studies for their osteoporosis. This may have made it more likely to find an identifiable genetic etiology. There are additional limitations related to our diagnostic analysis methods. The IGM variant analysis pipeline focuses on highlighting novel and rare variants within established disease genes or variation intolerant genes which may exclude common alleles contributing to IOP from consideration. This diagnostic analysis is empowered when considering de novo variants and comparisons between the family genetics and any parental phenotype. However, our diagnostic yield here may be reduced since these patients were all analyzed without access to any parental genetic data. These diagnostic criteria are intentionally strict and may fail to provide diagnoses for patients with only a few pathogenicity indicators, such as the VUS presented here that are paired with common alleles. The small sample size of this study underpowers our cohort analysis for detecting variant enrichments and may fail to identify recurrent variants as enriched among cases.

CONCLUSION

In conclusion, in a cohort of 75 premenopausal women with early onset idiopathic osteoporosis, WES genome-wide analysis revealed Pathogenic variants in 4 subjects and VUS in genes relevant to osteoporosis in an additional 4 subjects. Variants in LRP5 were most common. All 4 Pathogenic variants were either protein-truncating/loss-of-function variants or occurred at protein sites that have known pathogenic variants in ClinVar or HGMD. The 4 VUS variants were considered likely damaging, but only autosomal recessive inheritance has been previously described in their associated genes. This study also provides unique information on consistent relationships between the subjects’ genetic findings, their tissue level bone remodeling and their responses to teriparatide: BFR/BS was quite low in subjects with variants that would be predicted to affect bone formation and higher in subjects with variants predicted to increase bone resorption, with or without associated hypercalciuria and/or nephrolithiasis. Additionally, subjects with variants in both LRP5 and FKBP10 did not respond to teriparatide. In 89% of affected subjects, no pathogenic genetic variants were found by WES, despite their having clinical presentations that were often more severe and histomorphometric characteristics that were similar to those in whom clinically significance variants were identified. This suggests that many genetic etiologies of bone fragility have yet to be elucidated. Further research is needed to identify new genetic causes of osteoporosis that present in young adulthood.

Highlights:

• We analyzed exome data from 75 premenopausal women with idiopathic osteoporosis (IOP)

• WES revealed likely pathogenic variants or relevant VUS in 8 (11%) of 75 women with IOP

• Variants in LRP5 were most common; variants in PLS3, SLC34A3 and HGD were also found

• Subjects’ genetic findings related to tissue level bone remodeling and teriparatide response