Study population

The Study of Health in Pomerania (SHIP-TREND) is a population-based study located in West Pomerania, a rural region in northeast Germany [11]. A stratified (age, sex and city/county of residence) random sample of 8826 adults aged 20–79 years was drawn from population registries. Sample selection was facilitated by centralization of local population registries in the Federal State of Mecklenburg-West Pomerania. Baseline examinations were conducted between 2008 and 2012. In total, 4420 subjects chose to participate (50.1% response). All participants gave written informed consent before taking part in the study. The study was approved by the ethics committee of the University of Greifswald and conformed to the principles of the declaration of Helsinki. SHIP data are publically available for scientific and quality control purposes. Data usage can be applied for via www.community-medicine.de to ensure compliance with all legislation.

For a subsample of 995 participants without self-reported diabetes, plasma as well as urine metabolomome data based on MS (see below) were acquired. Participants exhibiting at least one of the following characteristics were excluded from the study sample (overlap exists): missing values in exposure or confounder (n = 13), renal failure (estimated glomerular filtration rate (eGFR) <60 mL/min/1.73m² n = 7), intake of medication influencing the serum OCN concentration (bisphosphonates and combinations, selective estrogen receptor modulators, parathyroid hormones and analogues or calcitonin, corticosteroids and combinations, strontium ranelate, vitamin D and analogues; n = 16), newly diagnosed diabetes (n = 30) or extreme parathyroid hormone concentrations (>120 pg/mL; n = 7). Finally 931 participants, comprising 414 men and 517 women, were included in the present study.

Laboratory measurements and phenotypic characterization

Smoking status (current, former or never smokers), daily alcohol consumption and physical activity (≥1 h training a week) were assessed using computer-aided personal interviews. Waist circumference (WC) was measured to the nearest 0.1 cm using an inelastic tape midway between the lower rib margin and the iliac crest in the horizontal plane. Hypertension was defined as an increased blood pressure (BP) (systolic BP of ≥140 mmHg or diastolic BP of ≥90 mmHg) or the use of antihypertensive medication. The intake of oral contraceptives (OC; ATC: G03A) or postmenopausal hormone therapy (PHT; ATC: G03C, G03D and G03F, N = 31) were defined based on ATC codes.

Fasting blood samples were taken from the cubital vein of participants in the supine position between 7.00 a.m. and 12.00 p.m. In the same time span spot urine samples were taken. All samples were either analyzed immediately or stored at −80°C. Serum OCN concentrations were measured with the IDS-iSYS N-Mid Osteocalcin assay on the IDS-iSYS Multi-Discipline Automated Analyser (Immunodiagnostic Systems Limited, Frankfurt am Main, Germany) according to the instructions for use. This assay detects the intact OCN polypeptide (amino acids 1–49) and the N-terminal-Mid OCN fragment (amino acids 1–43). The measurement range of the assay was 2–200 ng/mL. The limits of blank and detection were 0.27 ng/mL. The limit of quantitation was 1.57 ng/mL. As recommended by the manufacturer, three levels of control material were measured. During the course of the study, the coefficients of variation were 13.4% at low, 15.0% at medium, and 17.0% at high serum OCN concentrations in the control material. Serum cystatin C concentrations were measured using a nephelometric assay (Dimension VISTA, Siemens Healthcare Diagnostics, Eschborn, Germany) with a functional sensitivity of 0.05 mg/L. The cystatin C-based eGFR was calculated using the CKD-EPI cystatin C equation: eGFR = 133 × min(serum cystatin C / 0.8, 1)^-0.499 × max(serum cystatin C / 0.8, 1)^-1.328 × 0.996^age [× 0.932 if female] [12].

Metabolomics measurements

Non-targeted metabolomics analysis for metabolic profiling was conducted at the Genome Analysis Center, Helmholtz Zentrum München, Germany. A detailed description of metabolite measurements, annotations and data processing is given in the appendix (S1 Appendix). Briefly, two separate LC-MS/MS analytical methods were used, as previously published [13], to obtain broad, untargeted plasma and urine metabolite spectra. After preprocessing, 475 plasma and 558 urine metabolites remained for the statistical analyses. Some of these metabolites could not be unambiguously assigned to a chemical identity and are therefore referred to with the notation “X” and a unique number.

Statistical analysis

For descriptive analyses continuous data were expressed as median (25th; 75th quartile) and nominal data as percentage. For bivariate comparison of men and women, the Mann–Whitney U test (continuous data) and the χ2 test (nominal data) were used. Associations between OCN concentrations as exposure and the plasma and urine metabolome as outcome were tested by linear regression analyses controlling for age, sex, WC and physical activity. Since the first results revealed a high number of significantly associated metabolites, including markers of kidney function, we further adjusted for the eGFR, the intake of OC or PHT as well as smoking behavior in a second model. The association between the serum OCN concentration and the eGFR was tested using partial correlation coefficients, controlling for age, sex and WC. To account for multiple testing, we adjusted the p-values by controlling the false discovery rate (FDR) at 5% using the Benjamini-Hochberg procedure. This type of correction allows restricting the occurrence of false-positive findings among all nominal significant findings to a certain threshold. In our study maximally 5% of the presented findings might be false positives.

The interplay between bone, OCN and metabolism

The most pronounced association with serum OCN was found for PHP in plasma and urine. PHP is a collagen-derived peptide and against this background PHP was discussed as potential new, fast-responding marker for alterations in bone status or metabolism [16]. A study among postmenopausal women [16] confirmed highly significant correlations of serum OCN with urinary PHP as well as of both markers with other markers of bone turnover or dual X-ray absorptiometry (DXA). These results are not surprising as during bone turnover both the collagenous and the non-collagenous parts of the organic matrix are resorbed and newly synthesized, releasing degradation intermediates like PHP. The association of the serum OCN concentration with PHP is further supported by the observed positive association between serum OCN and the urinary glycylproline concentration. Glycylproline is a dipeptide and part of the primary structure of collagen. This finding is most likely due to the concurrent release of both OCN and glycylproline during bone remodeling. For X-18927, an unknown metabolite, no further information is currently available. However, the strong correlation of X-18927 with PHP and glycylproline suggests a functional link between these three metabolites.

Other detected metabolites, for example kynurenine, seem to be linked to OCN and bone metabolism in a rather indirect manner. Kynurenine is a degradation product of tryptophan and was positively associated with OCN in the present study. The Hordaland Health Study [17] discovered a relation between tryptophan metabolism and bone mineral density (BMD) whereby kynurenine showed a positive association to BMD in middle-aged women and men and older women. However, an in vitro study revealed opposite findings by reporting that tryptophan increased OCN expression in bone mesenchymal stem cells (BMMSC) and mediates increased proliferation of BMMSC, while kynurenine had no influence on OCN expression and exerts inhibitory effects on BMMSC proliferation [18]. The associations between tryptophan metabolism and BMD or OCN may further be modulated by inflammation. The key step of tryptophan degradation is catalyzed by an enzyme called indoleamin 2,3-dioxygenase 1 (IDO1). The enzyme activity seemed to be increased by inflammation, especially by IFN-γ [17]. In in vitro experiments the blockage of IDO led to decreased osteoblastogenesis and IDO1 knockout (IDO^-/^-) mice showed osteopenia with decreased amounts of osteoblasts and increased amounts of osteoclasts [19]. Possibly, even subclinical inflammation could be able to increase kynurenine concentrations by changing expression of IDO1 and reduce BMD by inducing bone resorption. Increased bone resorption, in turn, could be a potential promoter of increased serum OCN.

Citrulline, which was positively associated with OCN, is a non-proteinogenic amino acid that is either ingested or produced endogenously. It is a side product during arginine degradation in the synthesis of nitric oxide (NO). NO is an important signaling molecule in almost every tissue, including bone [20] and in vitro studies indicated a reliance of NO availability for proliferation of osteoblast-like cells [21]. NO also improves blood circulation [22], which may enhance bone nutrition and oxygen supply. Additionally, several previous studies demonstrated a correlation between bone health and cardiovascular disease [23]. Up to now, the mechanisms behind this interplay are not fully understood. The relation between OCN and NO might provide a possible explanation and the basis for future research. Another possible explanation for the association between plasma citrulline and OCN might be found in energy metabolism. As mentioned above, OCN seems to have protective effects towards development of diabetes and the metabolic syndrome. Citrulline treatment in vitro, using hepatic cell lines [24], or in vivo, in rodent models fed a high-fat diet, improved insulin sensitivity and decreased circulating insulin concentrations [25]. Citrulline also seems to be a predictor for the metabolic syndrome [26] and a mice model simulating type 1 diabetes mellitus showed increased plasma citrulline concentrations [27]. Altogether the anti-diabetic effects of OCN and citrulline seem congruent with each other. However, whether the involved mechanisms are interlinking with each other is speculative. Previous studies revealed that OCN is influencing K⁺_V-channels [28] and GPRC6A [29] in pancreatic islet cells, but the exact role of citrulline in energy metabolism remains unclear.

OCN and surrogates of lifestyle

In the present study, we observed a negative relation between OCN and urinary riboflavin, known as vitamin B2, a metabolite that cannot be synthesized endogenously. A previous study investigating more than 5300 elderly subjects proposed that a higher dietary riboflavin intake leads to slightly higher BMD in the femoral neck and lumbar spine but not to a reduced fracture risk [30] which is in contrast to our findings. However, the inverse associations observed here might be due to individual capacities of vitamin B2 uptake from the intestine or successful reabsorption within the kidneys which would finally lead to a decreased vitamin B2 excretion. Unfortunately, plasma concentrations of vitamin B2 were not assessed in the present study and hence we can only speculate about the endogenous availability of vitamin B2.

OCN was also inversely associated with urinary hydroxycotinine, a metabolite derived from tobacco metabolism [31]. Moreover, OCN was also linked to S-(3-hydroxypropyl)mercapturic acid (HPMA) another metabolite related to smoking. HPMA, as intermediate in acrolein metabolism, is a toxic and carcinogenic agent common in tobacco, food and the environment [32]. A Turkish study discovered decreased serum OCN concentrations in pregnant smoking women as well as in umbilical blood of their newborns. Interestingly, the effect on newborns is even demonstrable among mothers passively exposed to smoke [33] but an explanation is still outstanding. With respect to bone health, previous studies showed that smoking is a risk factor for osteoporosis [34] and is related to impaired BMD in older men [35] and premenopausal women [36]. Several mechanisms were proposed [34] which might explain the association between smoking and impaired BMD. Cigarette smoking impairs collagen synthesis and alters the circulating amount of several growth factors. Moreover, smoking influences the production and secretion of sexual, calciotropic and adrenocortical hormones. Altered concentrations of these hormones are also known to be related to an impaired BMD.

Beside the xenobiotic hydroxycotinine, the urinary concentration of 1-MU was positively related with OCN. 1-MU is a metabolite of caffeine or theophylline and therefore detectable in coffee or tea drinkers [37]. Coffee consumption seems to have multiple effects on bone metabolism. Long-term coffee consumption was associated with a slightly impaired BMD among Swedish women, but not with an increased fracture rate [38]. An in vivo investigation in rodents showed that caffeine also seems to increase osteoclast numbers [39]. These data are consistent with the present result, considering that higher circulating OCN concentrations may be related to a reduced BMD in elderly subjects [40, 41].

OCN associations distinct from bone metabolism

The inverse relation between OCN and pipecolate cannot be explained directly, but via its source lysine [42]. Lysine seems to boost osteoblast proliferation but an immediate influence of lysine on OCN could not be detected [43]. Apart from that, lysine is an activator of GPRC6A, the OCN receptor [44]. Decreased plasma concentrations of pipecolate might thus indicate diminished lysine degradation which in turn might provoke an attenuated induction of GPRC6A and hence higher OCN concentrations. However, plasma lysine concentration was not related to the serum OCN concentration in the present study and further studies are required to elucidate a possible mechanism behind this association.

Besides pipecolate, serum OCN was inversely related to intermediates of valine degradation, namely 3-hydroxyisobutyrate (3HIB) and 3-methyl-2-oxobutyrate (3M2O), pointing towards an implication on energy metabolism and glucose turnover. 3HIB was shown to promote trans-endothelial fatty acid (FA) transport and subsequent accumulation of fat in muscle tissue favoring local insulin resistance [45]. OCN deficient mice demonstrated decreased glucose tolerance and accumulation of body fat [3], which can also be provoked by insulin resistance. Nevertheless, OCN seems to have rather global effects on energy metabolism by influencing pancreatic beta cells via the GPRC6A [29] receptor and inhibiting K⁺_V-channels [28]. These channels are inhibited by ATP in times of sufficient glucose supply and therefore essential for the regulation of insulin release. Further research is necessary to investigate whether the systemic effects of OCN and the local effects of 3HIB are functionally related. 3M2O is located further upstream in the valine degradation pathway than 3HIB. This strengthens the indication towards a possible functional relation between serum OCN and valine degradation.

Strengths and limitations

Our study has strengths and some potential limitations. Metabolomics is a powerful tool for clinical research, since it is able to compress influences from genetics, environment, health behavior and interventions in intermediate phenotypes, which consist of final downstream products of various pathways. Since our findings depend on study region and measurement procedures they should be regarded as hypothesis generating. Replication in other cohorts and pharmacological intervention studies are needed to confirm our results and to obtain possible causal relations. Furthermore, epidemiological studies, especially cross-sectional analyses, are generally not suitable to prove causal relations.