Correlation of Serum Osteocalcin Fractions with Bone Mineral Density ...

Calcif Tissue Int (1998) 63:375–379

© 1998 Springer-Verlag New York Inc.

Correlation of Serum Osteocalcin Fractions with Bone Mineral Density in Women During the First 10 Years after Menopause M. H. J. Knapen,1 A. C. Nieuwenhuijzen Kruseman,2 R. S. M. E. Wouters,2 C. Vermeer1 1 2

Department of Biochemistry, Maastricht University, P.O. Box 616, 6200 MD Maastricht, The Netherlands Department of Internal Medicine, University of Maastricht, The Netherlands

Received: 14 November 1997 / Accepted: 23 March 1998

Abstract. Serum immunoreactive osteocalcin (irOC) consists of two fractions that differ from each other by their affinity for hydroxyapatite. The high and low affinity fractions are referred to as irOCbound and irOCfree, respectively. To evaluate whether these fractions are determinants for different characteristics of bone or bone metabolism, we have performed a cross-sectional study among 212 apparently healthy women between 20 and 90 years of age. Bone mineral density (BMD) was determined at the lumbar spine, and the right femur neck, trochanter, and Ward’s triangle using dual-energy X-ray absorptiometry (DXA). Biochemical markers for bone formation and resorption were determined in serum and in urine. After classification according to menopausal age, an inverse correlation was found in the 1–10 years postmenopausal women between irOCfree and BMD, notably of the Ward’s triangle and femur neck. It is concluded that in 1–10 years postmenopausal women, irOCfree is an independent marker for BMD, but that in other age groups the association is less clear or is absent. Key words: Vitamin K — Osteoporosis — Bone markers — Gamma carboxyglutamate.

Bone metabolism may be investigated using serum and urine markers that originate from bone and reflect a specific step in either bone formation or bone resorption. One of these markers is osteocalcin, a small protein synthesized by the osteoblasts. Pro-osteocalcin contains 5 glutamate residues, 3 of which may be converted into ␥-carboxyglutamate (Gla) in a vitamin K-dependent carboxylation reaction during the posttranslational modification preceding cellular secretion [1, 2]. The Gla residues are high-affinity sites for calcium, notably in the form of hydroxyapatite crystals [3]. Osteocalcin is specifically produced by osteoblasts (and odontoblasts), and a major part of the de novo synthesized protein is bound to the hydroxyapatite matrix in bone; about 20%, however, is set free in the blood stream. Circulating osteocalcin is generally regarded as a specific marker for osteoblastic activity and for bone formation [4–6]. During recent years we have demonstrated that serum osteocalcin can be subdivided into two fractions that differ in their affinity for hydroxyapatite [7, 8]. In most of the

Correspondence to: C. Vermeer

studies published thus far, the differentiation between both fractions was based on the extraction of serum with hydroxyapatite powder, and comparing its osteocalcin content before and after the adsorption step. In this paper we will designate total immunoreactive serum osteocalcin as irOCtotal, the fraction with high affinity for hydroxyapatite as irOCbound, and the low affinity fraction as irOCfree. In the normal population, the number of Gla-residues per osteocalcin molecule is variable, and in this respect osteocalcin is different from other well-known Gla proteins (e.g., prothrombin and related blood coagulation factors) [9]. Because of its high affinity for hydroxyapatite, irOCbound may be expected to contain a high number of Gla residues, and there is little doubt that it originates from de novo synthesis in the osteoblasts. Although neither the structure nor the origin of irOCfree are clear at this time, its low affinity for hydroxyapatite suggests that it has a low Gla content. Recently, it was shown by Szulc et al. [10, 11], that the hip fracture risk in postmenopausal women is six-fold increased in subjects with increased levels of circulating irOCfree and that serum irOCfree levels are inversely correlated with bone mass [12]. The correlation coefficient between irOCfree and bone mass was not very high, however. In this paper we confirm the initial observations by Szulc, and we have investigated whether the correlation between bone mass and osteocalcin fractions in women of various ages is increased in certain well-defined chronological or postmenopausal age groups.

Materials and Methods Subjects In a cross-sectional study, 850 Caucasian women between 20 and 90 years of age were randomly selected from the city population files of Maastricht. Inclusion criteria for subjects participating in the study included apparently good health, no recent (30 years since menopause (ysm). If plotted in this way the decrease in BMD is rather constant for the femur neck and trochanter, but for the lumbar spine and Ward’s triangle it is most pronounced in the 1–10 ysm group, with a constant, but slower loss of BMD in the higher age groups (Fig. 2). For a number of data analyses we have therefore subdivided the total cohort of 212 women into three subpopulations: premenopausal (n ⳱ 85), 1–10 ysm (n ⳱ 33), and >10 ysm (n ⳱ 94). The baseline characteristics of these groups are summarized in Table 1, and show that in the 1–10 ysm group the BMD values for lumbar spine, femoral neck and Ward’s triangle are significantly lower than in the premenopausal group. The differences between the 1–10 ysm group and the >10 ysm group were statistically significant at all sites measured. Biochemical markers for bone metabolism were all elevated in the 1–10 ysm group and remained high at older age. The increase in irOCtotal resulted from roughly proportional inclines of both irOCbound and irOCfree. Only the increase in urinary Gla did not reach the level of significance. In the total group of participants, we tested the correla-

M. H. J. Knapen et al.: Serum Osteocalcin as a Marker for BMD

377

Table 1. Baseline characteristics in women classified according to menopausal age Postmenopausal Premenopausal

1–10 ysm

>10 ysm

85 39.6 ± 0.9

33 56.8 ± 0.6c 6.0 ± 0.4

94 68.1 ± 0.8e 19.4 ± 0.8e

Anthropometric parameters Height (m) Weight (kg) BMI (kg/m2)

1.65 ± 0.01 66.2 ± 1.3 24.5 ± 0.5

1.60 ± 0.01b 68.2 ± 1.4 26.5 ± 0.6

1.59 ± 0.01b 68.2 ± 1.1 26.9 ± 0.4

Bone mineral density (g/cm2) Lumbar spine (L2-L4) Femur neck Ward’s triangle Trochanter

1.24 ± 0.02 0.98 ± 0.01 0.87 ± 0.02 0.80 ± 0.01

1.08 ± 0.03c 0.92 ± 0.02b 0.76 ± 0.02c 0.78 ± 0.02

64.4 ± 2.3 30.6 ± 1.2 2.43 ± 0.10 1.54 ± 0.07 0.90 ± 0.05

81.8 ± 4.0c 41.3 ± 2.0c 3.66 ± 0.2c 2.49 ± 0.1c 1.18 ± 0.09b

84.1 ± 2.0c 43.7 ± 1.3c 3.92 ± 0.1c 2.66 ± 0.08c 1.26 ± 0.06c

0.25 ± 0.02 13.8 ± 0.6 4.20 ± 0.2

0.30 ± 0.02 16.6 ± 0.9a 4.75 ± 0.2

0.41 ± 0.03b,d 16.8 ± 0.7a 4.74 ± 0.2

Number Age (yr) Y.s.m. (yr)

Biochemical parameters Serum Total AP (U/liter) Bone-specific AP (U/liter) irOCtotal (ng/ml) irOCbound (ng/ml) irOCfree (ng/ml) Urine Ca/creat (mmol/mmol) OHPro/creat (␮mol/mmol) Gla/creat (␮mol/mmol)

1.02 ± 0.02c 0.81 ± 0.01c,e 0.64 ± 0.01c,e 0.72 ± 0.01b,d

All values are given as the mean ± SEM Abbreviations: BMI: body mass index; ysm: years since menopause; AP: alkaline phosphatase Significance between pre- and postmenopausal groups: aP < 0.05; bP < 0.005; cP < 0.0005. Significance between 1–10 ysm and >10 ysm groups: dP < 0.05; eP < 0.0005

tion between BMD and the various anthropometric and biochemical variables. For each variable an inverse correlation was found with BMD (data not shown). To test the association of each variable with the expected BMD at the various skeletal sites we performed multiple linear regression analysis in which we took into consideration irOCfree, irOCbound, B-AP, calcium/creatinine, OHPro/creatinine, Gla/creatinine, age, weight, and BMI. As can be concluded from the high T-values (T > 5), the observed correlations were mainly explained by age and either weight or BMI (Table 2). This may be expected in a group with a large age range. The contribution of biochemical markers was only small; those not mentioned in the table did not contribute significantly to the expected BMD. Similarly we analyzed the three subpopulations defined by menopausal age, and found that in both the premenopausal and 10 ysm group the fraction of variance was mainly explained by age and weight, with little contribution of one of the biochemical variables (data not shown). In the 1–10 ysm women, however, irOCfree and to a lesser extent also irOCbound, were found to be independent biochemical variables showing a linear correlation (P < 0.05) with BMD. The mutual correlation between irOCfree and irOCbound was weak (r ⳱ 0.3) and statistically not significant (P > 0.05). Contributions to BMD of B-AP, calcium, OHPro, and Gla were small (if any) and non-significant (P > 0.05). The highest R2adj values were found for the BMD of the Ward’s triangle, with a contribution of approximately 70% of age and irOCfree. This correlation was increased further at

all sites of bone densitometry, when subjects with a high BMI (>2 standard deviations above the age group average) were excluded from the calculations (data not shown). Using the equations thus derived, we calculated the predicted BMD values for each subject in the 1–10-years postmenopausal group, and compared the data obtained with the observed values. In Figure 3 we plotted the actual BMD at the sites of the right femur neck and the right Ward’s triangle as a function of the predicted values. The regression lines closely approximate the equation y ⳱ x, with correlation coefficients between actual and predicted BMD values of 0.80 (for the femur neck) and 0.88 (for the Ward’s triangle). Discussion

We have demonstrated that irOCfree is an independent determinant for BMD in a well-defined group of women. Our data are consistent with earlier reports from Szulc et al. [10–12] who found a weak correlation (r ⳱ 0.2) between undercarboxylated osteocalcin and BMD in the 60–80-years group. In the present study, we have found a much stronger correlation (R2adj ⳱ 0.5–0.7) during the first 10 years after the menopause. Therefore, menopausal age, rather than chronological age, should be used to classify women in this kind of study. An uncertainty in the post-hoc demonstration of the predictive value of a bone marker for the measured BMD is the fact that the equations used are derived from the

378

M. H. J. Knapen et al.: Serum Osteocalcin as a Marker for BMD

Table 2. Multiple linear regression analysis between regional BMD values and biochemical and anthropometrical variables in 1-10-year postmenopausal women BMD region

Equation

Total group of women (n = 212) LS ⳱ 1.39 − 0.066*Age − 0.066*irOCfree + 0.0024*Weight Ne ⳱ 1.06 − 0.0053*Age + 0.0030*Weight − 0.097*Ca/creat − 0.039*irOCfree Tr ⳱ 0.77 − 0.0034*Age + 0.0090*BMI − 0.0032*OHPro/creat Wa ⳱ 0.95 − 0.0071*Age + 0.0042*Weight − 0.027*irOCbound − 0.076*Ca/creat

T

P

−9.34 −3.30 +2.47 −10.2 +4.35 −2.91 −2.79 −6.47 +5.14 −2.60 −11.2 +5.52 −2.45 −2.04

0.0001 0.001 0.015 0.0001 0.0001 0.004 0.006 0.0001 0.0001 0.010 0.0001 0.0001 0.015 0.043

1–10 years postmenopausal women (n = 33) −2.91 LS ⳱ 1.35 − 0.107*irOCbound Ne ⳱ 2.60 − 0.023*Age −6.18 − 0.14*irOCfree −5.90 − 0.0034*Weight −2.18 Tr ⳱ 1.56 − 0.120*irOCfree −3.65 − 0.011*Age −2.26 Wa ⳱ 2.59 − 0.027*Age −6.03 − 0.127*irOCfree −4.19 − 0.060*irOCbound −2.71

0.0067 0.0001 0.0001 0.038 0.0010 0.031 0.00001 0.0003 0.011

R2adj

0.380

0.488 0.274

0.570 0.195 0.633 0.295 0.719

Abbreviations: BMD, bone mineral density; LS, lumbar spine; Ne, right femur neck; Tr, right trochanter; Wa, right Ward’s triangle. Student’s t tests statistics were performed to derive the T-values. The two-tailed levels of significance (p) were used to test the hypothesis that BMD is not correlated to the biochemical and anthropometric variables

measurements, and need confirmation from a separately recruited, independent cohort. In this respect it is promising that our data confirm those of Szulc et al., so that the predictive value of irOCfree for hip fracture has been found in at least two independent cohorts of elderly women. The correlation between BMD and osteocalcin (both irOCtotal and irOCfree) was much better than that with other markers. The testing of both irOCfree and irOCbound has the disadvantage that it is complicated by the hydroxyapatite extraction. Quick and direct tests for both antigens separately will facilitate the evaluation of the physiological importance and the potential diagnostic value of this variable. Whereas irOCbound is assumed to consist mainly of intact, fully carboxylated osteocalcin, the fraction defined as irOCfree is generally considered as undercarboxylated osteocalcin, but obviously it may also contain denatured osteocalcin as well as osteocalcin fragments. It is unknown whether the latter forms are present in the circulation, and if so, in which amounts. The fact that in fresh-frozen serum the ratio between irOCfree and irOCbound was found to be closely similar in competitive radioimmunoassays and in ‘‘sandwich’’ immunoradiometric and enzyme-linked assays suggests that the main part of this fraction consists of fulllength osteocalcin [15]. Strikingly, it was found that irOCbound was poorly correlated with BMD. Since this fraction is generally thought regarded to represent intact, nor-

Fig. 3. Predicted versus actual BMD of femur neck and Ward’s triangle in 1–10 years postmenopausal women. The predicted BMD was calculated by the multiple regression equations given in Table 2. Linear regression between predicted and actual BMD resulted in the equation y ⳱ 0.0088 + 1.01x (r ⳱ 0.80) for the right femur neck (䊊) and y ⳱ 0.0071 + 0.99x (r ⳱ 0.88) for the Ward’s triangle (䊉).

mally carboxylated osteocalcin originating from de novo bone formation, it seems likely that irOCbound will correlate with bone turnover (or rate of bone loss) rather than with BMD. This may become clear from prospective studies in which the bone loss is assessed by repeated bone densitometry. Obviously, longitudinal follow-up studies give more information about potential mechanisms than the crosssectional analysis used in the present study. For this reason a second screening with a 3-year time interval is foreseen in the same group. Since the ratio between irOCfree and irOCbound can be greatly changed by increasing a subject’s vitamin K intake, our study raises the question of whether prolonged supplementation with vitamin K may contribute to bone loss prevention in postmenopausal women. Preliminary data from a placebo-controlled Japanese trial suggests that this is indeed the case [16]. Acknowledgment: This study was supported by grant 28-2388 from the Prevention Fund. References 1. Vermeer C (1990) Gamma-carboxyglutamate-containing proteins and the vitamin K-dependent carboxylase. Biochem J 266:625–636 2. Price PA (1988) Role of vitamin K-dependent proteins in bone metabolism. Annu Rev Nutr 8:565–583 3. Hauschka PV, Carr SA (1982) Calcium-dependent alphahelical structure in osteocalcin. Biochemistry 21:2538–2547 4. Delmas PD (1992) Clinical use of biochemical markers of bone remodeling in osteoporosis. Bone 13:S17–S21 5. Eastell R, Robins SP, Colwell T, Assiri AMA, Riggs BL, Russell RGG (1993) Evaluation of bone turnover in type I osteoporosis using biochemical markers specific for both bone formation and bone resorption. Osteopor Int 3:255–260 6. Akesson K, Ljunghall S, Jonsson B, Sernbo I, Johnell O, Ga¨rdsell P, Obrant KJ (1995) Assessment of biochemical markers of bone metabolism in relation to the occurrence of fracture: a retrospective and prospective population-based study of women. J Bone Miner Res 10:1823–1829 7. Knapen MHJ, Hamulya´k K, Vermeer C (1989) The effect of vitamin K supplementation on circulating osteocalcin (bone Gla-protein) and urinary calcium excretion. Ann Int Med 111: 1001–1005 8. Knapen MHJ, Jie K-SG, Hamulya´k K, Vermeer C (1993)

M. H. J. Knapen et al.: Serum Osteocalcin as a Marker for BMD

9. 10. 11. 12.

Vitamin K-induced changes in markers for osteoblast activity and urinary calcium loss. Calcif Tissue Int 53:81–85 Cairns JR, Price PA (1994) Direct demonstration that vitamin K-dependent bone Gla protein is incompletely ␥-carbolylated in humans. J Bone Miner Res 9:1989–1997 Szulc P, Chapuy M-C, Meunier PJ, Delmas PD (1993) Serum undercarboxylated osteocalcin is a marker of the risk of hip fracture in elderly women. J Clin Invest 91:1769–1774 Sculz P, Chapuy MC, Meunier PJ, Delmas PD (1996) Serum undercarboxylated osteocalcin is a marker of the risk of hip fracture: a three-year follow-up study. Bone 18:487–488 Szulc P, Arlot M, Chapuy M-C, Duboeuf F, Meunier PJ, Delmas PD (1994) Serum undercarboxylated osteocalcin correlates with hip bone mineral density in elderly women. J Bone Miner Res 9:1591–1595

379

13. Behr W, Barnert J (1986) Quantification of bone alkaline phosphatase in serum by precipitation with wheat-germ lectin: a simplified method and its clinical plausibility. Clin Chem 32:1960–1966 14. Kuwada M, Katayama K (1983) An improved method for determination of gammacarboxy-glutamic acid in proteins, bone and urine. Analyt Biochem 131:173–179 15. Knapen MHJ, Eisenwiener H-G, Vermeer C (1996) Osteocalcin detection in aging serum and whole blood: stability of different osteocalcin fractions. Clin Chim Acta 256:151–164 16. Orimo H, Shiraki M, Fujita T, Onomura T, Inoue T, Kushida K (1992) Clinical evaluation of menatetrenone in the treatment of involutional osteoporosis—a double-blind multicenter comparative study with 1␣ hydroxy vitamin D3. J Bone Mineral Res 7:S122