Individuals presenting in Galway University Hospitals for an elective hip replacement surgery or as a result of traumatic injury were approached for marrow donation. Following the review of a patient information leaflet and the receipt of informed consent, surgically discarded marrow was collected for research purposes. Approximately 3 mL of bone marrow was combined with 500 µL of heparin (Wockhardt, Clonmel, Ireland) at collection. The MSCs were isolated within 72 h of marrow harvest.

The donors in both cohorts were aged populations who presented with typical comorbidities. Individuals over the age of 50 admitted to Galway University Hospitals following fragility fracture were identified as having OP [18]. Control cohorts were selected from samples isolated from individuals undergoing elective hip replacement surgeries, and for each assay were age- and gender- matched to the osteoporotic donor samples to the extent possible (Table S1). Though the majority of the donors receiving hip replacement for reasons other than frailty have arthritis, arthritis is a highly prevalent condition associated with age that is also present in the cohort of donors with OP. Prioritisation was given in this study to accessing bone marrow from an age-matched population due to the numerous age-related confounding factors associated with MSCs from younger donors, and the very low prevalence in the population of an aged cohort with no musculoskeletal morbidity. Further, accessing the bone marrow of donors in this age category without prior medical need would be of ethical concern. A sample of donor charts were surveyed for bone health-relevant medications and overall only 15% of donors were found to be taking any [7% age-matched control (AMC), 22% individuals with OP]. Bone health-relevant medications recorded were calcium and vitamin D₃ supplements and denosumab.

Our previous work identified that as a variable, gender was found to have no impact on MSC number or function in any of the assays performed [19], nevertheless considering that sufficient samples from men with OP were not possible to obtain (due to low prevalence), men were excluded from the analyses. Therefore, only women with and without OP are included in the analysis. Gender was self-identified, and no samples were received by individuals identifying as non-binary or other.

MSCs were isolated from the collected marrow as previously described [19]. The collected marrow was centrifuged and the pellet resuspended in phosphate buffered saline (PBS; Gibco via Biosciences, Dun Laoghaire, Ireland, 14190094). The number of mononuclear cells (MNCs) were quantified and the cells were plated in tissue culture flasks in expansion medium [minimal essential medium α (α-MEM) with GlutaMAX™; Gibco, 32561029] supplemented with 10% fetal bovine serum (FBS; HyClone via ThermoFisher, Ballycollen, Ireland SV30160.03), 1% antibiotic antimycotic solution (Merck, Dublin, Ireland, A5955), and 1 ng/mL recombinant human fibroblast growth factor (FGF)-basic (Peprotech, London, UK, 100-18B) at 37°C with 5% CO₂. Following the first passage (P0), the MSCs were lifted from the growth surface with 0.25% trypsin-ethylenediaminetetraacetic acid (EDTA; Gibco, 25200056), evenly distributed in freezing medium [10% dimethyl sulfoxide (DMSO; Sigma, Arklow, Ireland, D2650) with 90% FBS], and cryopreserved in liquid nitrogen.

Following expansion through one passage (P1), the BD Biosciences Stemflow Human MSC Analysis Kit (BD, Limerick, Ireland, 562245) was used to characterize the plastic-adherent cultures. They were profiled for expression of markers according to the International Society for Cellular Therapy criteria for the identification of human bone marrow-derived MSCs [20]. Individually conjugated CD73, CD90 and CD105 antibodies were employed to identify positive markers while a cocktail of phycoerythrin (PE)-conjugated antibodies targeting CD34, CD45, CD11b or CD14, CD19 or CD79α, and HLA-DR antibodies served to characterize expression of negative markers.

Total RNA was extracted from MSCs at P1 using the QIAGEN RNEasy Plus Mini Kit (Qiagen, Germany) and according to the manufacturer’s protocol. RNA quality and quantity was assessed with Nanodrop (Thermo Fisher, USA) and Agilent Bioanalyser 2100 with RNA 6000 Nano LabChip Kit (Agilent Technologies, USA). Optimal RNA characteristics for sequencing were set by LC Sciences, USA [RNA amount ≥ 2 μg, 260/280 > 1.8, 260/230 > 1.0, RNA integrity number (RIN) value ≥ 7.0] who were contracted to perform sequencing, quality control, mapping and assembly of reads, and assessment of differential expression of genes and transcripts, including Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) analysis. Genes were categorised as “novel” in the context of OP when neither the gene nor protein name in combination with the search term “osteoporosis” yielded any result on PubMed.

MSCs at 80% confluence were cultured in osteogenic medium consisting of 1 nmol/L dexamethasone (Merck, D4902), 100 μmol/L ascorbic acid 2-phosphate (Merck, A8960), and 10 mmol/L β-glycerophosphate (Merck, G9422) in low-glucose Dulbecco’s modified eagle medium (DMEM; Merck, D6046) with 10% FBS and 1% penicillin-streptomycin (PS; Gibco, 15140-122). Negative, undifferentiated controls were maintained in expansion medium. In both conditions, the medium was changed twice weekly for up to 17 days. Calcium deposited into the monolayer’s extracellular matrix was stained on ice with 2% Alizarin Red S (Merck, A5533) in water following fixation in 95% methanol. Images of Alizarin Red stained cultures were obtained with a 4× objective lens on an Olympus BX43 brightfield microscope fitted with HD Chrome camera (1/.8”) and 0.5× C-mount adapter. The percentage surface area stained or occupied by Alizarin Red was calculated using Fiji (version 2) [21, 22]. A colour threshold was applied to each image, establishing what delineated red staining within the image. The mask created was then transformed into a binary image and percentage surface area of stain occupying the image was quantified.

Fresh marrow was plated at 3 × 10⁵–1.5 × 10⁶ MNC per well in 3 replicate wells of a 6-well plate with expansion medium for 8–10 days. Following expansion, colony forming unit-fibroblast (CFU-F) and colony forming unit-osteoblast (CFU-O) were assessed. CFU-Fs were identified by rinsing the cultures with PBS, followed by incubation in 10% neutral buffered formalin (Sigma, HT501128) and staining with 0.8% crystal violet (Sigma, C0775) in methanol. CFU-Os were identified following incubation with p-nitroblue tetrazolium chloride/5-bromo-4-chloro-3-indolyl phosphate (NBT/BCIP; Merck, B1911) to visualize alkaline phosphatase activity without fixation. Stained colonies were counted with an Olympus BX43 brightfield microscope. The total number of CFU-F per 100,000 MNCs was recorded following P0.

Knowing the number of CFU-Fs in a given quantity of MNCs, the total number of MSCs in the marrow isolate were calculated at P0. From P1–P5, the cumulative population doublings were quantified by recording the number of MSCs plated and harvested in each passage [19]. Knowing the number of days in a passage, the cumulative doublings over time were recorded.

Adipogenic differentiation was initiated on a confluent MSC monolayer by introducing an induction medium containing 1 μmol/L dexamethasone, 10 μg/mL human recombinant insulin (Merck, 11376497001), 200 μmol/L indomethacin (Merck, I7378), and 500 μmol/L 3-isobutyl-1-methyl-xanthine (Merck, I7018) in high glucose DMEM (Gibco, D5796), with 10% FBS and 1% PS. Three days induction periods were alternated with one day maintenance phases, when the cells were cultured in α-MEM, with 10% FBS, 10 μg/mL insulin and 1% PS. Following three induction cycles, the cells were stabilized in a 7-day maintenance period before harvesting.

Oil Red O (Sigma, O0625) staining was used to visualize and quantify lipid deposition. The culture was washed in PBS, preserved in 10% neutral buffered formalin, then exposed to 0.18% Oil Red O and the residual stain removed by rinsing the culture in 60% isopropanol (Sigma, 109827). Finally, the monolayer was counterstained with Harris Modified Haematoxylin Solution (Merck, HHS16) before brightfield imaging was carried out as described above.

To quantify the amount of Oil Red O stain retained in the culture, the monolayer was washed with 99% isopropanol to extract the stain. The extract’s absorbance was measured at 490 nm using a multimode plate reader (Victor X3, Perkin Elmer, Dublin, Ireland).

Graphpad Prism (San Diego, USA) and R (version 4.0.2) with R Studio (Boston, USA) were used to analyse the resultant data from all outcome measures. Normal distribution of data was determined in Graphpad Prism by P > 0.05 on both Shapiro-Wilk and Kolmogorov-Smirnov tests. Normally distributed data were assessed by t test or Analysis of Variance (ANOVA), whilst non-normal data were assessed by Mann-Whitney. Data variance was also compared between groups using F test in Graphpad Prism, with all normally distributed datasets found to have P > 0.05.

MSCs were successfully isolated from bone marrow samples donated by people diagnosed with OP and an AMC cohort of donors unaffected by OP. Summary statistics of donor age are provided in Table S1. MSC identity was evidenced by surface marker expression, differentiative and culture phenotype (Figure 1). Flow cytometry confirmation was according to The International Society for Cellular Therapy criteria [20], i.e., ≥ 95% cells positive for CD73, CD90, and CD105, and ≤ 5% of cells positive for CD34, CD45, CD11b or CD14, CD19 or CD79α, and HLA-DR (Figure 1A–D). Confirmation was provided by functional phenotyping of the cells’ ability to form colonies, to adhere to plastic and their multilineage potential along with their appropriate fibroblastic shape (Figure 1E).

RNA sequencing (RNA-seq) was performed on MSCs from both groups to identify pathways modified in the setting of OP. MSCs from people with OP exhibited altered transcriptional expression when compared with those from AMC. Only 32 genes were differentially expressed with adjusted P value (q value) ≤ 0.05, including 17 which were down-regulated, and 15 which were up-regulated. GO pathway analysis highlighted the biological processes that were significantly altered in MSCs from people with OP (Figure 2A). These pathways included skeletal system development, extracellular matrix organisation, and cell-cell interaction. We hypothesised that these gene expression changes may play a role in the capacity of MSCs to support bone homeostasis. A heatmap was generated of those differentially expressed genes which reported a consistent intragroup effect as assessed by individual analysis (Figure 2B). These 16 genes [protein kinase CGMP-dependent 1 (PRKG1), MBL associated serine protease 1 (MASP1), neurofilament light chain (NEFL), adherens junctions associated protein 1 (AJAP1), pyrimidinergic receptor P2Y6 (P2RY6), TRNA isopentenyltransferase 1 (TRIT1), family with sequence similarity 193 member B (FAM193B), ephrin B2 (EFNB2), prostaglandin D2 receptor (PTDGS), actin remodeling regulator (NHS), Rho GTPase activating protein 32 (ARHGAP32), EPH receptor B6 (EPHB6), Zinc finger protein 555 (ZNF555), telomerase associated protein 1 (TEP1), UBX domain protein 8 (UBXN8), and HAUS augmin like complex subunit 5 (HAUS5)] were then categorised according to functional involvement in processes relevant to OP; the processes of osteogenesis, proliferation, cell membrane function, cell fate choice and neuromuscular function were identified, with the majority of genes having known roles in osteogenesis (Figure 2C). Those genes which were identified by the initial bulk fold-change analysis with q ≤ 0.05 but which demonstrated P > 0.05 upon individual analysis due to high intragroup variability are presented in Figure S1.

Considering the significant alteration to the skeletal system development GO pathway, and the significantly different expression of genes related to osteogenesis observed, a targeted interrogation of further genes known to be involved in osteogenesis was performed. We found a statistically significant reduction in the expression of WAVE (also known as WASF1 or WASP family member 1; Mann-Whitney, P = 0.005), Fos proto-oncogene (c-fos, Mann-Whitney, P = 0.035), SPARC related modular calcium binding 1 (SMOC1; t test, P = 0.014) and natriuretic peptide receptor 3 (NPR3; t test, P = 0.009), and increased transcript levels of Reelin (Mann-Whitney, P = 0.011) and oxytocin receptor (OXTR; t test, P = 0.048) in people with OP (Figure 3A–F). There was also a trend to reduced Osterix expression (Mann-Whitney, P = 0.05) in MSCs from individuals with OP as compared to AMCs (Figure 3G). To assess the impact of the described transcript changes on the capacity of these primary cultures to differentiate to osteoblasts, MSCs from donors with or without OP were cultured in osteogenic medium and assessed for calcium deposition using Alizarin Red staining. MSCs from both donor cohorts successfully deposited calcium in the extracellular matrix in response to differentiation factors, with both groups exhibiting an increase in comparison to MSCs not exposed to differentiation medium. Modified differentiation capacity was identified by quantification of Alizarin Red staining (Figure 3H and I). This was assessed by both two-way ANOVA (P = 0.046) in Prism and binomial generalised linear model (GLM) in R (P < 0.001) and indicated a reduced osteogenic capacity associated with MSCs being derived from a person with OP. No correlation between age and osteogenesis was observed (data not shown).

Considering the observed transcriptional changes in genes related to proliferation we next assessed functional proliferative capacity. We found no difference between groups (Figure 4A and B). P0 population doublings which were calculated using MNC plating density and the known number of CFU-F/100,000 MNC were approximately equal for MSCs from both people with OP and AMC (Figure 4A). Cumulative doublings were recorded from P1 through to P5 and identified no impact of OP on proliferative capacity (Figure 4B). This adequate proliferation ex vivo was not reflected in the numbers of MSCs present in the bone marrow of people with OP as measured by counting CFU-F and CFU-O plated from fresh bone marrow samples (Figure 4C–E). Fewer MSCs were recorded from people with OP than from the AMC cohort as a proportion of the MNC population (Figure 4C). Of these MSC colonies, natively osteogenic (i.e., CFU-Os) were categorised by positive staining for alkaline phosphatase activity. CFU-O were also reduced in people with OP compared to AMC donors (Figure 4D). As the CFU-O population is a subset of the CFU-F population, we assessed the effect of OP on the proportion of CFU-O to non-osteogenic CFUs (nO-CFU, i.e., the remaining CFU-F without native osteogenic capacity). Binomial GLM analysis in R identified that the proportion of CFU-O to nO-CFU was significantly altered according to OP status of donor (using OP as the variable P = 0.014 on n = 12 OP and n = 5 AMC). There was a smaller proportion of CFU-O within the CFU-F population in samples derived from people with OP (median: 0.5/1.6, i.e., 31%) compared to AMC (median: 2.2/3.1, i.e., 71%) (Figure 4E). These data indicate that though both osteogenic and non-osteogenic MSCs are reduced in number in people with OP compared to people without OP, the osteogenic population is disproportionately reduced.

Previous studies have suggested preferential adipogenesis of bone marrow MSCs as a mechanism for reduced MSC osteogenesis in OP development. Since MSCs are the common progenitor for adipocytes and osteoblasts, we hypothesised that this switch in cell fate choice may be driving the observed osteogenic modifications. We therefore interrogated our RNA-seq dataset for potential transcriptional modifications in genes associated with adipogenesis and found transcript level changes in CEBPα, LEPR, ID4 and STEAP4 (Figure 4F–I). Surprisingly, each of these genes involved in promoting adipogenesis had lower expression in MSCs from people with OP versus those from AMC. To further investigate, we progressed to assess the adipogenic potential of MSCs isolated from donors living with and without OP. MSCs were induced to differentiate with adipogenic medium, and their lipid content observed using Oil Red O staining. Overall, the lipid content per cell appeared to be comparable in both cultures but with fewer cells entering the adipogenic lineage in cultures from donors with OP in comparison to the homogenous distribution of lipid-filled MSC-derived adipocytes in induced cultures from AMC donors (Figure 4J). Quantification of Oil Red O was by absorbance measurements of extracted stain, and it demonstrated that MSCs isolated from donors living with OP had less Oil Red O retention than MSCs isolated from AMC donors (Figure 4K). Considering the high variability in the range of adipogenic capacity in both cohorts, a relationship to donor age was assessed. Increased age was associated with a reduced adipogenic potential in MSCs from donors with OP only (Figure 4L and M). These findings indicate that MSC adipogenesis is reduced in people with OP in an age-dependent manner and that those changes are stable through ex vivo expansion.