Key words
osteocalcin - children - overweight - adipokines - leptin - adiponectin
Introduction
Osteocalcin (OC), also known as bone Gla protein (BGP), is one of the very few osteoblast-specific
proteins. It is synthesized by mature osteoblasts, odontoblasts and hypertrophic chondrocytes
during the process of bone formation. However there was no clear correlation of OC
serum concentration to bone mass. OC has several features of a hormone: it is synthesized
as a prepromolecule, it is a cell-specific molecule, it is secreted into the circulation,
but a receptor has not yet been identified. Posttranslationally, OC undergoes a vitamin
K dependent carboxylation. In healthy adults a relatively large fraction (20–30%)
of circulating OC is uncarboxylated, and this fraction in children is even larger
(up to 60%) [1]. In mice it was shown, that the carboxylated part has a higher affinity to hydroxyapatite
of the bone, whereas the uncarboxylated fraction of OC is possibly responsible for
putative effects on glucose homeostasis [2].
OC deficient mice develop a phenotype marked by higher bone mass and improved quality
of bone function, revealing OC as a determinant for bone formation [3]. Interestingly, these OC deficient mice had higher blood glucose levels, lower insulin
levels and they were obese [2]. In normal mice continuous OC administration affected insulin secretion, insulin
sensitivity and fat mass [4]. OC acts directly on pancreatic beta cells in culture and in vivo, leading to increased
beta cell mass and increased insulin secretion, respectively OC can ameliorate the
severity of obesity and type 2 diabetes in mice. In patients with type 2 diabetes
several studies showed that OC serum levels were reduced. Children suffering from
type 1 diabetes have been reported to have significantly lower OC levels than healthy
control subjects at the same pubertal stage [5]. Hence, it appears from these studies that OC is decreased in states of reduced
insulin secretion and resulting in hyperglycemia. On the other hand, in a recent study
it was shown that OC levels were lower in obese children and that they were related
to insulin resistance (IR) and leptin serum concentrations. It was therefore hypothesized
that OC could represent a new link between obesity and IR [6]. Accordingly, our studies are aimed to evaluate whether or not OC serum levels are
related to parameters of overweight and serum adipokine levels in a cross sectional
study in children and adolescents.
Subjects and Methods
Study population
The study population consisted of 497 healthy Caucasian children and adolescents from
the Leipzig Schoolchildren Project [7]. Informed consent was obtained from parents and subjects 12 years of age or older.
Pubertal stages were determined according to Marshall and Tanner (as described in
1969 and 1970). Subjects were selected and grouped according to gender and pubertal
stage; 45–54 individuals for each pubertal stage in each sex (Tanner stages, TS 1–5).
We grouped patients according to TS in early pubertal (TS 1 and 2), mid pubertal (TS
3) and late pubertal (TS 4 and 5) groups. Clinical examination and blood withdrawing
was performed between 0800 and 1 100 h. The study protocol was approved by the Ethical
Committee of the Medical Faculty of the University of Leipzig, Germany, Numbers 781
and 782.
Measurement of anthropometric parameters
Height and weight were determined using precision stadiometers and scales to the nearest
of 0.1 cm and 0.1 kg, respectively. For standardization of height and weight, reference
percentiles for central Germany were applied [8]. The body mass index (BMI) was standardized referring to national reference data
[9]. The waist to hip ratio (WHR) was calculated by measuring the waist at the smallest
circumference between hip and chest, and hip circumference at the widest using a nonstretchable
metric band.
Underweight, normal weight and overweight were defined as BMI below the 10th, between 10th and 90th and above the 90th percentile, respectively.
Biochemical analyses
Serum was separated by centrifugation and aliquots were stored at − 80°C for biochemical
analyses. Levels of OC were determined by the fully automated immunoassay system Advantage
(Nichols-Diagnostics). The sensitivity of the assay was calculated to be 0.1 ng/ml.
Intra- and inter-assay coefficients of variation (CV) were less than 9.5%.
Leptin levels were measured as described previously [10]
[11]. Adiponectin levels were determined by RIA (Linco Research, Inc., St. Louis, MO).
The sensitivity of this assay was calculated to be 1 ng/ml. Intra- and inter-assay
CVs were less than 14%. Resistin levels were determined by an ELISA (Mediagnost, Reutlingen)
with a sensitivity of 12 pg/ml and intra- and inter-assay CVs between 2.4% and 6.8%.
Statistical analysis
Some investigated parameters did not adhere to the normal distribution. Comparison
of variables was performed using Mann-Whitney U and Wilcoxon’s signed rank test. Spearman’s
nonparametric correlations was utilized to determine relationships between variables.
Multiple stepwise linear regression analysis was performed for OC with age, gender,
pubertal stage, BMI standard deviation score (SDS), waist to hip ratio (WHR), fat
free mass (FFM), leptin, adiponectin and resistin. Age entered the calculation as
squared parameter, because the association of age and OC was non-linear. The selection
of variables used for the final regression model was obtained by applying backward
elimination to the initial multiple regression model. Repeatedly performing backward
elimination and removing the least significant parameter in each step, yielded the
removal of (in that order): FFM, WHR, BMI-SDS, adiponectin, resistin, leptin, waist circumference and hip circumference. So the risk for a possible multicolinearity
was minimized. The threshold for statistical significance was set at p<0.05. Statistical analyses were performed using R (Development Core Team, http://www.R-project.org). The library quantreg version 4.48 by Roger Koenker was used to estimate the age
dependent OC centiles as splines with 4 degrees of freedom (http://CRAN.R-project.org/package=quantreg).
If not otherwise mentioned values are given as mean ± standard deviation.
Results
We investigated OC serum levels and different anthropometric parameters in a large
and well characterized population of children and adolescents ([Table 1]).
Table 1 Anthropometric characteristics of 497 children differentiated by Tanner stage (TS)
and gender.
|
|
Age (years)
|
Height (cm)
|
Weight (kg)
|
BMI-SDS
|
WHR* (cm)
|
|
|
Mean±SD
|
Mean±SD
|
Mean±SD
|
Mean±SD
|
Mean±SD
|
|
All values presented as mean and standard deviation. TS1/2 = prepubertal, TS3 = midpubertal,
TS4/5 = late pubertal. BMI-SDS: body mass index – standard deviation score. WHR: waist
to hip ratio. SD: standard deviation. Note in some categories limited information
was present, thus the number of study subjects for the WHR was reduced (*)
|
|
total
|
Sum (n=497/*432)
|
12.33±2.48
|
155.40±14.56
|
47.55±13.73
|
0.16±1.00
|
0.80±0.05
|
|
Boys (n=247/*221)
|
12.77±2.57
|
158.99±15.74
|
50.27±14.77
|
0.12±1.08
|
0.82±0.05
|
|
Girls (n=250/*211)
|
11.90±2.30
|
151.86±12.31
|
44.87±12.03
|
0.20±0.92
|
0.78±0.05
|
|
TS1/2
|
Sum (n=205/*184)
|
10.20±1.52
|
143.30±10.64
|
37.60±10.66
|
0.18±1.03
|
0.82±0.04
|
|
Boys (n=103/*94)
|
10.39±1.66
|
144.84±11.39
|
39.23±11.72
|
0.22±1.09
|
0.83±0.04
|
|
Girls (n=102/*90)
|
10.01±1.35
|
141.15±9.47
|
35.95±9.19
|
0.15±0.96
|
0.81±0.04
|
|
TS3
|
Sum (n=99/*76)
|
12.26±1.47
|
157.90±8.23
|
49.15±10.93
|
0.28±1.04
|
0.80±0.06
|
|
Boys (n=49/*40)
|
13.04±1.53
|
161.91±8.15
|
53.79±12.20
|
0.33±1.21
|
0.82±0.06
|
|
Girls (n=50/*36)
|
11.50±0.89
|
153.96±6.14
|
44.60±6.99
|
0.23±0.83
|
0.78±0.05
|
|
TS4/5
|
Sum (n=193/*172)
|
14.63±1.45
|
167.30±9.12
|
57.31±10.09
|
0.07±0.95
|
0.78±0.05
|
|
Boys (n=95/*87)
|
15.21±0.91
|
172.83±7.52
|
60.43±10.00
|
− 0.10±0.95
|
0.80±0.04
|
|
Girls (n=98/*85)
|
14.06±1.65
|
161.93±7.12
|
54.28±9.21
|
0.24±0.92
|
0.76±0.05
|
In the total study population circulating OC increased during pubertal development
reaching highest values in midpuberty (TS3) ([Fig. 1]). The peak in the cross-sectional 50th centile of OC occured in girls at 11.9 years and hence earlier than in boys with
a peak at 13.2 years. However, the 3rd centile showed no distinct peak at all ([Fig. 2]).
Fig. 1 Osteocalcin serum levels (y-axis, ng/ml) in boys (white) and girls (grey) depending
on Tanner stage (TS). Significance level was set as p≤0.05 (*), p≤0.01 (**), p≤0.001
(***). The effect of gender was compared in each TS ((1) to (5)). OC differences were
compared in boys (B1–B4) and girls (G1–G4) for adjacent Tanner stages. The data were
analyzed using U-test.
Fig. 2 Centiles for osteocalcin serum levels (p3, p50, p97): depending on age for boys (dotted)
and girls (dashed).
OC levels were significantly higher in boys compared to girls if subjects with pubertal
stage 1, 4 and 5 were compared (p<0.05). If boys and girls were separately analyzed,
all OC levels were significant different in adjacent TS ([Fig. 1]) with the exception of the comparison between TS 2 and 3. The resulting centiles
for OC in boys and girls depending on age are shown in [Fig. 2].
This study did not reveal a strong linkage of OC to overweight/obesity or significantly
higher values of OC in lean subjects. Only in early puberty OC levels were significantly
lower in overweight (BMI>90th percentile) compared to normal weight peers (BMI 10–90th percentile) (p=0.01). If stratified for gender this was only significant for early
pubertal boys (p=0.01) but not for girls ([Fig. 3]). Interestingly, in pubertal stage 3 underweight peers (BMI<10th percentile) demonstrated significant lower levels of OC if compared to the normal
weight group. This was obviously detectable for the mid pubertal combined group of
boys and girls (p=0.03) as well as for midpubertal boys alone (p=0.04) but not for
girls (data not shown).
Fig. 3 Osteocalcin serum levels (y-axis, ng/ml) compared to BMI. I=BMI 10–90th percentile, II=BMI >90th percentile. a Serum OC levels depending on BMI and pubertal stages in girls. b OC depending on BMI and pubertal stages in boys. c OC depending on BMI and pubertal stages in girls.
WHR, a marker for visceral fat in overweight subjects, showed a positive correlation
to serum OC levels in the total study population (p<0.0001, r=0.202, n=432). But if
stratified for pubertal stage there was only a significant correlation observed between
serum OC and WHR in the late pubertal combined group of boys and girls (p<0.0001,
r=0.297, n=172) as well as in girls alone (p=0.0008, r=0.358, n=85) but not in boys.
As expected, there was a strong relation between WHR and BMI.
The present study demonstrated no significant correlation between serum OC levels
and adiponectin or leptin, neither in the total study population nor in the subsets
stratified for gender and BMI. As expected leptin serum levels were elevated in overweight
children. Only, in late puberty the combined group of boys and girls showed a significant
negative correlation between leptin and OC levels (p=0.005, r= − 0.20, n=193).
A significant negative correlation was detectable between OC and resistin serum levels
within the total study population (p<0.0001, r= − 0.330, n=186). This correlation
was also significant in each pubertal stage, but if stratified by gender it was only
significant in early pubertal boys (p=0.008, r= − 0.424, n=38). The serum levels for
OC and measured adipokines were summarized in [Table 2] dependent on gender and pubertal stage.
Table 2 Osteocalcin, leptin, resistin and adiponectin serum levels of 497 children, differentiated
by Tanner stage (TS).
|
|
Osteocalcin (ng/ml)
|
Leptin (ng/ml)
|
Resistin* (ng/ml)
|
Adiponectin** (ng/ml)
|
|
|
Mean±SD
|
Mean±SD
|
Mean±SD
|
Mean±SD
|
|
All values are given as mean and standard deviation. TS1/2 = prepubertal, TS3 = midpubertal,
TS4/5 = late pubertal. Note in some categories limited information was present, thus
the number of study subjects for resistin (*) and adiponectin (**) serum levels was
reduced
|
|
total
|
Sum
(n=497/*186/**197)
|
9.78±6.61
|
6.67±5.89
|
12.77±6.07
|
7.02±2.15
|
|
Boys
(n=247/*92/**97)
|
10.86±6.44
|
5.05±5.57
|
11.59±6.10
|
6.73±2.20
|
|
Girls
(n=250/*94/**100)
|
8.72±6.61
|
8.29±5.75
|
13.92±5.82
|
7.30±2.06
|
|
TS 1/2
|
Sum
(n=205/*77/**80)
|
9.40±5.30
|
6.42±6.14
|
12.65±7.07
|
7.71±1.92
|
|
Boys
(n=103/*38/**40)
|
9.96±5.29
|
6.05±6.49
|
11.15±6.88
|
7.96±1.85
|
|
Girls
(n=102/*39/**40)
|
8.83±5.24
|
6.81±5.74
|
14.12±6.94
|
7.47±1.96
|
|
TS 3
|
Sum
(n=99/*37/**39)
|
14.45±7.91
|
7.12±5.42
|
11.20±4.56
|
6.52±2.34
|
|
Boys
(n=49/*20/**20)
|
14.69±7.48
|
6.27±5.90
|
10.65 ±4.91
|
5.93±2.37
|
|
Girls
|
(n=50/*17/**19)
|
14.22±8.30
|
7.96±4.76
|
11.85±4.01
|
7.14±2.14
|
|
TS 4/5
|
Sum
(n=193/*72/**78)
|
7.80±5.94
|
6.71±5.83
|
13.70±5.39
|
6.56±2.09
|
|
Boys
(n=95/*34/**37)
|
9.86±6.24
|
3.34±3.48
|
12.64±5.65
|
5.83±1.80
|
|
Girls
(n=98/*38/**41)
|
5.80±4.87
|
9.97±5.79
|
14.65±4.96
|
7.22±2.11
|
Multiple stepwise linear regression analysis of OC with age, gender, pubertal stage,
BMI-SDS, waist circumference, hip circumference, WHR, and adipokines revealed only
gender, pubertal stage and age as independent predictors ([Table 3]). Thereby, age entered the calculation as squared parameter, because the association
of age and OC was non-linear. This result was found if the whole study cohort but
also if subsets of normal weight peers (BMI 10–90th percentile) and overweight peers (BMI > 90th percentile) were analysed.
Table 3 Independent predictors of OC serum levels determined by multiple stepwise linear
regression analyses: Pubertal stage and gender were coded as factors. As the relation
between OC levels and age was nonlinear, age entered the calculation as squared parameter.
|
ß coefficient
|
standard error
|
p-value
|
|
TS = Tanner stage. TS1+2 = prepubertal, TS3 = midpubertal, TS4+5 = late pubertal
|
|
intercept
|
9.524
|
0.644
|
<2e-16
|
|
TS 1+2
|
6.579
|
0.859
|
8.00e-14
|
|
TS 3
|
10.111
|
0.919
|
<2e-16
|
|
TS 4+5
|
4.943
|
1.145
|
1.87e-05
|
|
female
|
− 2.544
|
0.501
|
5.20e-07
|
|
age (squared)
|
− 0.026
|
0.006
|
7.27e-06
|
Discussion
Our study population demonstrated clearly that OC serum levels were correlated with
pubertal development. This correlation is supported by the results of a couple of
studies [12]
[13], whereas other authors did not find such a link [6]
[14]. The detected OC peak levels in TS3 were presumably due to the onset of growth spurt
at this pubertal stage. Girls reached the OC peak level at a younger age compared
to boys as they enter puberty earlier and onset of growth spurt starts at younger
age. OC values were significantly higher in boys compared to girls if subjects with
pubertal stage 1, 4 and 5 were compared. For this reason a higher bone mass in boys
than in girls could be the explanation [14].
However, there was no major relationship between OC and indicators of body weight,
although we can not exclude a minor impact of OC on the very obese phenotype. The
significant negative correlation between serum OC levels and indicators of body weight
in underweight boys and girls in mid puberty may indicate a later onset of growth
spurt in underweight adolescents.
Interestingly and opposite to our expectations, girls with increased WHR have higher
OC serum levels than girls with normal WHR. This could be explained by the size of
the study subgroups, particularly the low number of obese girls with an increased
WHR.
These results are partially in contrast with other studies reporting that obese children
had significantly lower OC levels compared with their normal weight peers [15]
[16]. These findings in children was underlined by a recent study showing that substantial
weight loss was also associated with an increase in OC [6]. Most of these studies that demonstrated a clear relation between OC and obesity
investigated a larger number of overweight subjects with a broader range in the distribution
of fat mass compared to our cohort. The presented cohort included subjects with a
rather normal distribution of BMI and so the number of overweight subjects was distinctly
lower than in other studies [6,]
[]
[15,16].
As adipokines are indicators of adipose tissue mass, we hypothesized to detect an
association between OC and serum adipokine levels of adiponectin, leptin and resistin
[17]. It is well known that obesity and parameters of the metabolic syndrome are associated
with high leptin and low adiponectin serum levels [18]. Leptin produced by the adipose tissue could be the link between obesity and low
OC serum levels observed in obesity. Moreover, leptin has important effects on bone
metabolism [19] and leptin-deficient ob/ob and leptin-resistant mice have elevated OC serum levels
[20]. In our study, only in late puberty, the combined group of boys and girls showed
a significant negative correlation between leptin and OC levels. Thus, the relation
of OC and leptin appears to be dependent on pubertal status.
In rodent models, it was additionally shown, that OC can influence insulin sensitivity
and secretion. Thereby, the impact of OC on insulin sensitivity and visceral fat mass
probably depends on an enhanced expression of adiponectin in adipocytes [2]
[4]. However, in this study no correlation between OC and adiponectin serum levels was
observed.
Resistin initially was studied as a possible link between adiposity and insulin resistance
in humans. However, the association between circulating resistin levels and obesity
and type 2 diabetes-related phenotypes is unclear [21]
[22]. Currently, resistin is considered as a marker of inflammation processes related
to obesity [23]. From the teleological point of view, low resistin should favor bone growth in the
inflammation free status.
This study has a few potential limitations. First, BMI percentiles were used to classify
overweight individuals, but BMI is considered to be only one surrogate marker of fat
mass. Second, WHR is an acceptable marker for overweight adolescents, but the information
in younger individuals is only very limited and somewhat questionable. Unfortunately,
the collected data on fat mass, FFM and homeostasis model assessment (HOMA) was insufficient
for gender and Tanner stage specific analyses. Third, a cross sectional study of subjects
from a community was conducted and the number of overweight or obese subjects was
distinctly lower than the number of lean ones. This “mismatch” between obese and normal
weight subjects can cause a potential bias in the analysis.
We conclude that there is no major relationship between OC and metabolism, but we
can not exclude minor relations between OC and metabolism. The negative relationship
with serum resistin levels might rather point to a link between OC and inflammatory
states. Longitudinal studies are needed for a better understanding of putative interactions
between OC and fat tissue during the transition from normal weight to obesity.
