Elevated ceramides 18:0 and 24:1 with aging are associated with hip fracture risk through increased bone resorption

Clinical data using peripheral blood samples

The baseline characteristics of the study participants are shown in Table 1. Among 46 controls and 28 patients with fragility HF (mean ages, 72.1 ± 5.3 [range, 65–90] and 78.3 ± 9.1 [range, 65–95] years, respectively), 31 (67.4%) and 21 (75.0%), respectively, were women. There were no significant differences in weight, height, body mass index (BMI), smoking status, alcohol intake, history of diabetes, and serum 25-hydroxyvitamin D₃ (25-OH-D₃) between the groups. Bone mineral density (BMD) values at the lumbar spine and proximal femurs were significantly lower in patients with HF than in those without HF. Serum C-terminal telopeptide of the type I collagen (CTX) levels were markedly higher in patients than in controls, whereas the difference in serum osteocalcin (OSC) levels between the groups was not statistically significant.

The association of age with plasma ceramide levels was investigated by the Pearson correlation analysis (Supplementary Table 1). Age was positively correlated with circulating C16:0, C18:0, and C24:1 ceramide levels, while there was no association between age and plasma C14:0, C18:1, C20:0, and C24:0 ceramides.

Difference in plasma ceramide levels according to fragility HF status was analyzed by analysis of covariance (ANCOVA) (Figure 1). After adjustment for potential confounders, including sex, age, BMI, smoking status, alcohol intake, 25-OH-D₃, and diabetes, patients with fragility HF had 21.3%, 49.5%, 34.3%, and 22.5% higher plasma C16:0, C18:0, C18:1, and C24:1 ceramide levels than those without HF, respectively (statistically significant). On the other hand, there were no significant differences in plasma C14:0, C20:0, and C24:0 ceramide levels between the groups.

The risk for fragility HF according to plasma ceramide levels was determined through multiple logistic regression analysis (Table 2). After considering confounding factors, the odds ratios (ORs) per standard deviation (SD) increment in circulating C16:0, C18:0, C18:1, and C24:1 ceramide levels for fragility HF were 2.73, 3.20, 2.86, and 3.43, respectively. However, the ORs for HF in terms of plasma C14:0, C20:0, and C24:0 levels were not statistically significant.

Multiple linear regression analyses were performed to examine the relationship of plasma ceramides with BMD and bone turnover markers (BTMs) (Table 3). After adjustment for sex, age, BMI, smoking status, alcohol intake, 25-OH-D₃, and diabetes, plasma C18:0 and C24:1 ceramides were inversely associated with BMD values at the lumbar spine and total femur, respectively. However, other ceramides, including C14:0, C16:0, C18:1, C20:0, and C24:0, were not significantly correlated with BMD values at any site. Plasma C16:0, C18:0, C18:1, C24:0, and C24:1 ceramides, but not C14:0 and C20:0 levels, were positively related to serum CTX level, a bone resorption marker, whereas no circulating ceramides had a significant association with serum OCN level, a bone formation marker, in the multivariate adjustment model.

Clinical data using BM samples

To understand more accurate mechanisms explaining the detrimental effects of ceramides in human bone health, C16:0, C18:0, C18:1, and C24:1 ceramides, which showed the most convincing results in peripheral blood, were additionally measured in BM aspirates, directly reflecting the bone microenvironment.

Pearson correlation analysis showed that the peripheral blood and BM levels of all four ceramides, especially C18:0, C18:1, and C24:1, were significantly correlated (Supplementary Table 2).

After adjustment for sex, age, BMI, smoking status, alcohol intake, 25-OH-D₃, and diabetes, patients with fragility HF had 69.2%, 51.9%, 67.3%, and 38.0% higher BM C16:0, C18:0, C18:1, and C24:1 ceramides levels than those without HF, respectively, consistent with the results using peripheral blood (Figure 2).

Finally, we investigated the association of ceramide levels with bone biochemical markers measured in BM samples. In the multivariate adjustment model, higher BM C16:0, C18:0, C18:1, and C24:1 ceramides levels were significantly associated with higher BM tartrate-resistant acid phosphatase-5b (TRAP-5b) levels, whereas only C18:0 ceramide was inversely correlated with the BM bone-specific alkaline phosphatase (BSALP) levels (Table 4). The BM C16:0, C18:0, and C24:1 ceramides levels had a positive association with the BM receptor activator of nuclear factor-κB ligand (RANKL) levels, and the BM C18:0 ceramide levels showed an inverse correlation with the BM osteoprotegerin (OPG) levels. Consequently, higher levels of all four ceramides measured in BM aspirates were significantly associated with higher BM RANKL/OPG ratio after adjustment for confounders.

vitro data

To further support our observations in humans, we treated the lineages of osteoclasts and osteoblasts with the ceramides within the in vitro system. Among various ceramides, we specifically focused on C18:0 and C24:1 ceramides, because these were related most strongly to fragility HF (Table 2) as well as age (Supplementary Table 1) in humans.

When primary mouse BM macrophages (BMMs) were treated with or without various concentrations of C18:0 ceramide in the presence of macrophage colony-stimulating factor (M-CSF) and RANKL, C18:0 markedly stimulated the formation of differentiated osteoclasts, as shown by the TRAP staining results (Figure 3A), and increased osteoclastogenesis by C18:0 also was observed in the coculture system with mouse calvaria osteoblasts (Figure 3B). In addition, C18:0 enhanced bone resorption by approximately 4.4-fold compared to the untreated controls (Figure 3C). Consistently, the expression levels of osteoclast differentiation markers, such as Trap, Ctr, Mmp9, and Catk, were significantly increased by C18:0 (Figure 3D). To determine if the effects of C18:0 on osteoclasts were mediated by the production of RANKL and/or OPG in osteoblasts, primary mouse calvaria osteoblasts were incubated with or without C18:0 ceramide; however, there were no differences in the RANKL/OPG ratio between the two groups (Figure 3E), indicating that C18:0 ceramide ay directly stimulate osteoclastogenesis and the resultant bone resorption rather than through the increase of RANKL/OPG ratio from osteoblasts, at least within the in vitro system. Next, when we assessed the effects of C18:0 on osteoblast biology, the viability (Supplementary Figure 1A), ALP activity (Supplementary Figure 1B), and expression of osteoblast differentiation markers (Supplementary Figure 1C) were not changed by C18:0 treatment in mouse calvaria osteoblasts.

As observed in C18:0 treatment in the in vitro system, C24:1 ceramide also increased the osteoclastogenesis (Figures 4A, 4B), bone resorption (Figure 4C), and expression levels of osteoclast differentiation markers Figure 4D), and its effects on viability (Supplementary Figure 2A), ALP activity (Supplementary Figure 2B), and expression of osteoblast differentiation markers (Supplementary Figure 2C) were not observed in mouse calvaria osteoblasts. However, C24:1 treatment significantly increased the RANKL/OPG ratio in osteoblasts (Figure 4E), suggesting that C24:1 ceramide can stimulate bone resorption not only by the direct increase in osteoclast differentiation, but also the indirect regulation of RANKL/OPG system in osteoblasts in the in vitro system.

The in vitro experiments above were performed using BMMs isolated from young (6-weeks old) mice, as is widely used in bone research [15]. In addition to these, we investigated the effects of C18:0 and C24:1 ceramides on osteoclast differentiation from BMMs collected from old (24-months old) mice and confirmed that both ceramides markedly increased osteoclastogenesis (Supplementary Figure 3).

Study participants

The study population comprised consecutive patients 65 years or older who underwent hip surgery at the Department of Orthopedic Surgery, Asan Medical Center (AMC; Seoul, Korea) between November 2012 and December 2013. All patients underwent hip surgery because of fragility HF or other causes, such as osteoarthritis. Fragility fractures result from mechanical forces that would not ordinarily cause fracture, known as low-level (or low-energy) trauma [30, 31]. The World Health Organization has quantified this as forces equivalent to falls from a standing height or less [32]. Exclusion criteria were drugs that could affect bone metabolism taken for >6 months or within the previous 12 months before hip surgery, such as bisphosphonate, systemic glucocorticoids, or hormone replacement therapy; diseases that might cause secondary osteoporosis, such as hyperthyroidism or rheumatoid arthritis; fever (oral temperature ≥ 38.0°C) or abnormal complete blood count findings of leukocytes (<4.0 or >10.0 × 10⁹/L) or platelets (<150 or >350 × 10⁹/L); and abnormal liver or kidney functions. These criteria were used to eliminate systemic illness. Finally, subjects with fractures not caused by low-energy trauma, such as motor vehicle accidents or falls above a standing height, were also excluded. Among the 102 eligible participants not satisfying any of the detailed exclusion criteria, we could contemporaneously collect blood and BM samples during hip surgery from 74 participants with their consent. Consequently, 46 controls and 28 patients without and with fragility HF, respectively, were finally enrolled in our study.

A patient questionnaire was used to assess smoking (current smoker), alcohol intake (≥3 U/day), history of medication use, previous medical or surgical procedures, and reproductive status (including menstruation). This study was approved by the AMC institutional review board. Written informed consent was provided by all enrolled participants.

Measurement of ceramides in peripheral blood and BM aspirates in humans

Venous blood and BM samples were collected during hip surgery. After sample centrifugation at 3000 revolutions per minute (rpm) for 5 minutes at 4°C, we carefully collected the supernatants to exclude cell components. All samples with hemolysis or clotting were discarded. Human peripheral and BM plasma were mixed well, and internal standard solution (500 nM C17 ceramide) was added to the samples before extraction. Ceramides were extracted by Bligh and Dyer method [33]. After lipid extraction, organic solutions containing lipids were dried using a vacuum centrifuge and stored at −20°C until liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. The dried matter was reconstituted with methanol (MeOH) and injected into the LC-MS/MS system. All lipid standards, including internal standards, were purchased from Avanti-Polar Lipids (Alabaster, AL, USA) and Sigma-Aldrich Corp. (St. Louis, MO, USA).

Lipid levels were determined using an LC-MS/MS system equipped with 1290 HPLC (Agilent, Waldbronn, Germany) and QTRAP 5500 (AB Sciex, Toronto, Canada). A reverse-phase column (Pursuit5 C18, 150 × 2.1 mm) was used with mobile phase A (5 mM ammonium formate/MeOH/tetrahydrofuran [500/200/ 300]) and mobile phase B (5 mM ammonium formate/MeOH/ tetrahydrofuran [100/200/700]). The LC was run at 200 μL/minute and 35°C. The LC gradient was as follows: 50% of A at 0 minutes, 50% of A for 5 minutes, 50%–30% of A for 3 minutes, 30% of A for 7 minutes, 30%–10% of A for 7 minutes, 10% of A for 3 minutes, 10%–50% of A for 0.1 minutes, and 50% of A for 4.9 minutes. Multiple reaction monitoring was performed in the positive ion mode, and the extracted ion chromatogram corresponding to the specific transition for each lipid was used for quantification. The calibration range for each lipid was 0.1–1000 nM (r² ≥ 0.99). Data were analyzed using Analyst 1.5.2 software.

Biochemical measurement in humans

To measure biochemical BTMs, fasting blood samples were obtained in the morning. Serum CTX levels were measured using an electrochemical-luminescence immunoassay (Roche Diagnostics GmbH, Mannheim, Germany), with intra- and inter-assay coefficients of variation (CVs) of 1.0%–4.6% and 1.6%–4.7%, respectively. Serum OSC concentration was also measured using an electrochemical-luminescence immunoassay (Roche Diagnostics GmbH), with intra- and inter-assay CVs of 1.2%–4.0% and 1.7%–6.5%, respectively. Serum concentration of 25-OH-D₃ was measured by radioimmunoassay (Cobra II Auto-γ Counting System; Packard Instruments, Downers Grove, IL), with a lower limit of detection of 0.6 ng/mL (1.5 nmol/L) and intra- and inter-assay CVs of < 3.5%.

The following bone biochemical markers were measured in BM samples with immunoassay kits: TRAP-5b (Cat #SB-TR201A; Immunodiagnostic Systems, Scottsdale, AZ, USA), BSALP (Cat #MBS262250; Mybiosource, San Diego, CA, USA), OPG (Cat #ab100617; Abcam, Cambridge, MA, USA), and RANKL (Cat #K1016; Immunodiagnostic Systems). The intra- and inter-assay CVs for each assay were as follows: TRAP-5b <9% and <9%, BSALP <7% and <12%, OPG <10% and <12%, and RANKL <3.5% and <9.3%, respectively.

BMD measurement in humans

Areal BMD (g/cm²) was measured at the lumbar spine (L1–L4) and proximal femurs (femoral neck and total femur) by dual-energy x-ray absorptiometry using the Lunar system (Prodigy; Madison, WI, USA) software version 9.30.044. The precision of the equipment, presented as CVs, was 0.67% and 1.25% for the lumbar spine and femoral neck, respectively, in 17 volunteers who were not enrolled in the study. Each volunteer underwent five scans on the same day, getting on and off the table between examinations. Bone density was expressed as the Z-score, which is defined as the number of SDs above or below mean for healthy subjects of the same sex and age.

Reagents for in vitro experiments

C18:0 and C24:1 ceramides were purchased from Cayman Chemical (Ann Arbor, MI, USA). Soluble RANKL was purchased from R&D System, Inc. (Minneapolis, MN, USA), and M-CSF was purchased from PeproTech EC (London, UK). Fetal bovine serum (FBS) and α-minimum essential medium (α-MEM) were purchased from Gibco (Grand Island, NY, USA) and Wel Gene (Daegu, Korea), respectively. Prostaglandin E₂ and 1α,25-dihydroxyvitamin D₃ (1α,25-OH[2] D₃) were purchased from Sigma-Aldrich Corp.

Cell cultures

Primary osteoclasts were generated as previously described [15]. Primary mouse BM cells (BMCs) were obtained by flushing the femur and tibia of 6-week-old ICR or 24-month-old C57/BL6 mice (Orient, Seongnam, Korea) and cultured at 37°C in α-MEM containing 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin in a humidified atmosphere with 5% CO₂. After a 24-hour culture, nonadherent cells were collected and cultured in 96-well plates at a density of 4 × 10⁴ cells/well. The BMMs were induced to differentiate into osteoclasts with 30 ng/mL M-CSF and 100 ng/mL soluble RANKL for more than four days, with the culture medium being changed every two to three days.

Primary mouse osteoblasts were isolated by sequential collagenase digestion of calvaria obtained from newborn mice and maintained in α-MEM containing 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. Mature osteoblasts were generated from mouse calvaria osteoblasts in the presence of 50 μg/mL ascorbic acid and 10 mM β-glycerophosphate for seven days.

For the coculture system, primary mouse calvaria osteoblasts were cultured for one day in 48-well plates at a density of 1 × 10⁴ cells/well. Primary mouse BMMs (1 × 10⁵ cells) were added to each well, and BMMs were cocultured with osteoblast for ten days in the presence of 10⁻⁸ M 1α,25-OH(2) D₃ and 10⁻⁶ M prostaglandin E2, changing the culture medium every two to three days.

TRAP staining and in vitro resorption assay

After culture in an osteoclast-inducing medium, adherent cells were fixed and stained with TRAP, an enzymatic marker of osteoclasts, using a leukocyte acid phosphatase kit (Sigma-Aldrich Corp.) in accordance with the manufacturer’s instructions. TRAP-positive multinucleated cells containing ≥3 nuclei were considered to be osteoclasts and were counted under a light microscope (Olympus Optical; Tokyo, Japan).

To measure the resorption area, BMMs (4 × 10⁴ cells/well in 96-well plates) were seeded onto dentine discs with 30 ng/mL M-CSF and 100 ng/mL soluble RANKL. After 10 days, the cells on the dentine discs were removed completely by wiping with a cotton swab, and then the dentine slices were stained with hematoxylin (Sigma-Aldrich Corp.) for one minute. The area of resorbed pits was analyzed using Image-Pro Plus software (MediaCybernetics, Silver Spring, MD, USA).

Quantitative reverse-transcription PCR (qRT-PCR)

Total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA, USA), according to the manufacturer's protocol. First-strand cDNA was synthesized with the Superscript III First-Strand Synthesis System (Invitrogen) using oligo dT primers. qRT-PCR was performed using the Light Cycler 480 (Roche; Mannheim, Germany) with Probes Master Mix (Roche). The primers were Trap (Mm00475698_m1), Ctr (Mm00432282_m1), Mmp9 (Mm00442991_m1), CatK (Mm00484039_m1), Rankl (Mm00441906_m1), Opg (Mm01205928_m1), Col1 (Mm00801666_g1), Alp (Mm00475834_m1), and Ocn (Mm03413826_mH) from Applied Biosystems (Foster city, CA, USA). The threshold cycle (Ct) value for each gene was normalized to the Ct value of 18S (Hs03928990_g1).

Cell viability

Cell viability was measured using the Cell Counting Kit-8 (Dojindo; Kumamoto, Japan), according to the manufacturer’s instructions. Briefly, 10 μL WST-8 dye (2-[2-methoxy-4-nitrophenyl]-3-[4-nitrophenyl]-5-[2,4-disulfophenyl]-2H-tetrazolium, monosodium salt) was added to each well in a 96-well plate for one hour at 37°C, and the absorbance at 450 nm was then read using a microplate reader SPECTRAmax 340PC (Molecular Devices, Palo Alto, CA, USA) with a reference wavelength of 650 nm.

ALP activity

Primary mouse osteoblasts were seeded at a density of 5 × 105 cells/well in 12-well plates and were differentiated into osteoblasts for seven days. The cells were washed with PBS, and the ALP activity was measured using the p-nitrophenyl phosphate hydrolysis method [34]. The ALP activity in each sample was normalized relative to total cellular protein content, which was determined by the BCA method (Pierce, Rockland, IL, USA).

Statistical analysis

All data are presented as means ± SD or as numbers and percentages, unless otherwise specified. The baseline characteristics of the study population according to the status of fragility HF were compared using Student’s t-tests for continuous variables and χ² tests for categorical variables. Relationships between age and plasma ceramide levels and between peripheral plasma and BM ceramide levels were investigated using Pearson’s correlation analysis. The multivariable-adjusted least-square mean levels (95% confidence intervals [CIs]) of ceramides in peripheral blood and BM samples in terms of the presence of fragility HF were estimated and compared by ANCOVA after adjustment for potentially confounding factors, including sex, age, BMI, smoking status, alcohol intake, 25-OH-D₃, and diabetes. These confounding variables were selected based on their clinical applicability. To generate ORs per SD increment in plasma ceramide levels for fragility HF, we performed multiple logistic regression analyses after adjustment for confounding variables. The associations of peripheral blood and BM ceramide levels with BMD, BTMs, and bone-related markers in BM were investigated through multiple linear regression analyses.

All in vitro data were expressed as means ± standard error of mean (SEM) of at least three independent experiments conducted with triplicate measurements, unless otherwise specified. The significance of differences between two groups was assessed using the Mann-Whitney U test, whereas differences between ≥ 3 groups were tested using the analysis of variance (ANOVA) with posthoc analysis via Tukey's honest significance test.

All statistical analyses were performed using SPSS version 18.0 (SPSS, Inc., Chicago, IL, USA). P < 0.05 was considered statistically significant.