Contents - Previous - Next

Biochemical markers for assessing skeletal growth

1. Introduction
2. Biosynthesis of fibrillar collagens
3. Markers for bone and cartilage turnover
4. Bone resorption markers
5. Bone formation markers
6. Future studies
7. Concluding remarks

S.P. Robins

Biochemical Sciences Division, Rowett Research Institute, Bucksburn, Aberdeen AB2 9SB, Scotland, UK

Many of the biochemical markers for assessing skeletal turnover are based on the unique metabolism of fibrillar collagens. Intracellular modifications lead to the formation of hydroxyproline and hydroxylysine glycosides, both of which have been used as markers of collagen degradation. However, hydroxyproline is metabolised extensively in the liver and both components may be derived from several different tissue sources. The pyridinium crosslinks of collagen have been shown to provide more specific and sensitive markers of collagen degradation, since these compounds are only present in the mature, insoluble fibrils. In addition, pyridinium crosslinks are unaffected by diet and are not metabolised in the body. Following development of HPLC methods for the quantification of urinary crosslinks, these techniques have been validated as indices of bone resorption in studies of a wide range of metabolic bone diseases. Subsequently, the proportion of free crosslinks in urine was shown to be relatively consistent in different individuals, allowing development of a simple, direct immunoassay. The excretion of crosslinks in children was related to growth velocity and, in studies of malnourished children, the values before treatment were related to the child's growth response.

For measuring bone formation, the serum concentrations of the C-terminal propeptide of procollagen type I (PICP) appear to reflect the activity of the osteoblasts, but additional information on physiological variations is necessary. The major non-collagenous components of bone in serum, osteocalcin or bone Gla protein, has long been used as a marker of bone formation, but there are a number of factors that complicate interpretation of the results. These include variations in the immunochemical reactivity, the possible presence of degradation fragments in serum and the dependence on vitamin K status for adequate enzymatic carboxylation. Nevertheless, assays for intact osteocalcin have been shown to be related to growth velocity in children.

There are few suitable serum or urinary indices for cartilage metabolism and development of more specific markers, particularly for growth plate cartilage, are required to distinguish between linear growth and bone remodelling. Assessments of skeletal metabolism should, wherever possible, include a combination of different markers so that the balance between formative and resorptive events can be adequately evaluated.

1. Introduction

The purpose of this review is to assess the recent developments in biochemical markers for skeletal growth. There is a large literature on this topic and, due to limitations of space, this account must necessarily be selective. In attempting to present a balanced view of the usefulness of the different techniques available, however, it is intended to provide a critical evaluation so that the limitations in interpreting data from these techniques can be appreciated.

Measurements of growth involve the determination of both synthesis and degradation rates. For the former, the techniques are based either on quantitative analysis of specific, newly-formed cellular products or on the determination of amounts or activities of specific enzymes; a prime example is the bone-specific isoform of alkaline phosphatase. Other enzymes such as tartrate resistant acid phosphatase may be used for measurement of degradation rates, but this is more commonly achieved by analysis of the serum or urinary concentrations of specific components or fragments removed from the tissue. Considerable advances have been made recently in defining more suitable degradation products and these will be discussed in more detail later in this review. Many of the components used to monitor skeletal metabolism are based on collagen, which comprises over 90% of the protein in bone and about 70% of that in cartilage. It is pertinent therefore intitially to review briefly the stages of collagen biosynthesis which give rise to the different markers used.

2. Biosynthesis of fibrillar collagens

To date, about 15 genetically distinct types of collagen are known (for review, see van der Rest & Garrone, 1991), but many of these are quantitatively minor with specific functions often achieved through association with the main fibrillar types. The latter are represented by collagen type I, the main constituent of bone and skin collagens, type II present in all cartilages, and collagen type III which is essentially absent from bone but which is widely distributed in most soft tissues.

Collagen is unique in the number of post-ribosomal modifications that occur during biosynthesis and processing to form insoluble collagen fibrils. One of the early modifications, hydroxylation of proline and lysine residues to form hydroxyproline and hydroxylysine, occurs in the rough endoplasmic reticulum and is closely followed by glycosylation of certain hydroxylysine residues to form O-linked glycosides. The fibrillar collagens are synthesized as larger precursors, procollagens, with extension polypeptides at each end of the molecule. Association of the three procollagen chains and winding up of the helix is driven by the formation of disulphide bonds within the C-terminal propeptide. Helix formation prevents any further enzyme reactions and the procollagen molecule is transported to the Golgi where additional carbohydrate additions occur in the non-helical portions of the molecule. Following secretion of the procollagen, an active process involving cytoskeletal elements, the N- and C-terminal propeptides are removed by specific proteases and the helical molecules spontaneously associate into fibrils containing a quarter-staggered array which gives rise to the characteristic banding pattern evident in the electron microscope. The extension peptides, particularly the C-terminal portion, are released into the blood and can be measured as indicators of the rate of collagen production (vice infra).

Following the extracellular assembly of collagen, the final enzymatic modification is the oxidative deamination of specific lysine or hydroxylysine residues at the ends of the molecules by the copper-dependent enzyme, lysyl oxidase (Robins, 1988). As summarized in Fig. 1, the tissue specificity of collagen cross-linking is governed entirely by whether the oxidized residue is lysine, as in skin and some tendons or hydroxylysine, as in bone and cartilage. The former pathway leads to the formation during maturation of histidine-based crosslinks, whereas hydroxylysine aldehydes give rise to keto-amine intermediates that are converted on maturation to trifunctional, pyridinium crosslinks. The time taken for 'maturation' of the crosslinks also appears to vary between tissues and this term should not be confused with physiological age. Thus, conversion of 'intermediate' to 'mature' crosslink is an on-going process which could take hours or days in cartilage but may take several weeks in skin (Reiser & Last, 1986). Tissues with a high turnover or from growing individuals will have a higher proportion of intermediate crosslinks but some mature crosslinks will be present in young tissue.

3. Markers for bone and cartilage turnover

Bone undergoes continual turnover to repair minor fissures and to maintain its structural integrity, such that, even in adults, about 10% of the skeleton is renewed each year. This process occurs as localized episodes of bone resorption by osteoclasts followed by replacement of the bone by osteoblasts. As outlined in Fig. 2, the resorptive and formative phases are tightly coupled by a series of mechanisms that have yet to be fully elucidated. Also summarized in Fig. 2 are the main biochemical markers that are available for assessing bone turnover, most of which are based on collagen metabolism and have already been introduced. Osteocalcin (also known as Bone Gla Protein) is the most abundant non-collagenous protein in bone (Price et al., 1976) and its extensive use as a serum marker for bone formation will be discussed in more detail later in this review. There are many other non-collagenous components in bone that have potential as markers of skeletal metabolism; these include osteopontin, osteonectin, bone sialoprotein and a2HS-glycoprotein. In general, however, these components are either not present in sufficient amounts or are not specific for bone (Heinegard & Oldberg, 1989).

Fig. 1. Formation and maturation of collagen crosslinks from different tissues.

Tissue specificity of collagen crosslinking is governed primarily by the presence of either lysine or hydroxylysine residues in the oxidisable telopeptide positions, giving rise to Schiff base or keto-amine intermediates that are transformed to the mature crosslink structures shown.

In comparison with bone, relatively few markers have been developed to monitor cartilage metabolism. A number of specific components are potentially useful (Heinegard & Oldberg, 1989; Saxne & Heinegard, 1989; Saxne & Heinegard, 1992), even though in most cases the functions of these components are uncertain.

In addition, measurements of the glycosaminoglycan side chains of the large aggregating proteoglycans may reflect cartilage damage (Thonar et al., 1985; Heinegard & Saxne, 1991). The C-propeptide of collagen type II, a 110 kDa protein termed chondrocalcin (Hinek, Reiner & Poole, 1987) is partially retained within the cartilage matrix (Niyibizi, Wu & Eyre, 1987) but can be measured in synovial fluid as an indicator of synthetic activity.

Fig. 2. Biochemical markers of bone metabolism.

The main markers available for measuring the coupled processes of bone formation and resorption are indicated, together with other potential markers.

4. Bone resorption markers

For many years, urinary hydroxyproline has been used as a marker of bone degradation, although the limitations of the technique have become increasingly apparent. These have been reviewed previously (Robins, 1982) and will not therefore be discussed in detail. Essentially, the presence of hydroxyproline in proteins other than collagen, such as the complement component C1q, acetylcholinesterase and lung surfactant proteins, gives rise to a lack of specificity. Moreover, the fact that about 90% of the protein is metabolised in the liver leads to poor sensitivity of the technique. The hydroxylysine glycosides are not metabolized in the body and their rates of excretion in the urine can provide a more quantitative assessment of degradation (Segrest & Cunningham, 1970). The mono- and di-saccharide forms were initially thought to provide some tissue specificity since the disaccharide, glucosyl-galactosyl-hydroxylysine, was more prevalent in skin whereas the monosaccharide, galactosyl-hydroxylysine, predominated in bone (Pinnell, Fox & Krane, 1971). However, the presence of the disaccharide in C1q, a component with a very rapid turnover, has confounded this interpretation (Krane et al., 1977). Nevertheless, galactosyl-hydroxylysine appears to be relatively specific for bone and HPLC assays to quantify this component in urine have recently been applied in several different bone diseases (Moro et al., 1988; 1993).

The pyridinium crosslinks are specifically located in collagen and because their formation occurs at the final stage of maturation of the collagen, these crosslinks provide good candidates as markers only of insoluble collagen degradation. Unlike hydroxyproline and hydroxylysine glycosides, the crosslinks are unaffected by the relatively high levels of collagen degradation, both intracellularly and at subsequent stages of processing (Bienkowski et al., 1978).

4.1. Pyridinium crosslinks as collagen degradation markers

Although an enzyme-linked immunosorbent assay (ELISA) had been developed to measure pyridinoline in urine hydrolysates (Robins, 1982), determination of both crosslinks in urine was possible only after the development of an HPLC technique based on pre-fractionation of urinary hydrolysates by partition chromatography and quantification using their natural fluorescence (Black, Duncan & Robins, 1988). Pyridinoline (Pyd), also known as hydroxylysyl-pyridinoline or HP, is widely distributed in different tissues (Robins, 1983; Eyre, Koob & VanNess, 1984). The analogue deoxypyridin oline (Dpd), also referred to as lysyl-pyridinoline or LP, was initially thought to be present only in bone and dentine (Eyre, Koob & VanNess, 1984) but more recent analyses have revealed a wider tissue distribution (Robins, Duncan & Riggs, 1990) as shown in Table 1. Since other tissues containing Dpd, such as aorta, are known to have very slow turnover rates (Nissen, Cardinale & Udenfriend, 1978), the presence of Dpd in urine can be regarded as a specific marker of bone turnover. Although Pyd is present at very high concentrations in cartilage collagen, the small pool size relative to bone suggests that very little of the urinary output is derived from cartilage. In fact there is indirect evidence that the majority of both crosslinks are derived from bone collagen. Thus, the Pyd/Dpd ratio in human bone collagen, which is in the range 3-4 (Eyre, Dickson & VanNess, 1988), is very similar to the ratio observed in normal adult urine (Seibel, Duncan & Robins, 1989), a relationship that is also true for other species in which the Pyd/Dpd ratio may vary from unity in rats to about 10 in sheep.

Table 1. Hydroxy-pyridinium crosslink content of human tissues






Articular cartilage


1.47 0.23


Cortical bone


0.35 0.09

0.08 0.02

Trabecular bone


0.26 0.08

0.06 0.02



0.30 0.07

0.07 0.01

Intervertebral disc


1.14 0.11




0.47 0.35

0.05 0.03

Synovial tissue (RA)


0.48 0.08

0.03 0.01

The HPLC methods have been taken up by several groups and there is now a considerable body of evidence to support the validity of the pyridinium crosslinks as markers of bone resorption in a wide range of different bone disorders, arthritic diseases and malignancies (Editorial, 1992; Eyre, 1992; Delmas, 1992; Demers, 1992; Seibel et al., 1992).

4.2. Crosslinks, growth and nutrition

As shown in Fig. 3, crosslink excretion in children is up to 20-fold higher than in adults when expressed relative to urinary creatinine. Similar relationships have been obtained using 24h collections (Beadsworth, Eyre & Dickson, 1990). The values do, however, show considerable variations that appear to be more marked for boys (Fig. 3).

In collaboration with the Nutrition Institute in Rome and the University of Aberdeen, these markers have been applied to study the response to treatment of a group of 47 malnourished children (Branca et al., 1992). Clearly, expression of the crosslink excretion relative to creatinine was inappropriate in these studies and, as a surrogate for skeletal mass, the results were expressed as nmoles per hour relative to an exponent of the child's height. A logarithmic plot of the excretion against height revealed an exponent of 2 for these studies (Branca et al., 1992).

Fig. 3. The variations with age in Pyd, expressed relative to creatinine, are shown for young (a) females and (b) males in comparison with the mean values (2 SD) for adults (shaded bar).

Changes with age in pyridinoline (Pyd) excretion. (a)

Changes with age in pyridinoline (Pyd) excretion. (b)

Subsequently, further studies of normal children have shown that the crosslink excretions are best expressed as nmoles/h/m3. Crosslink excretion was significantly lower in the malnourished child compared with that after recovery, and there were positive correlations between the crosslink excretion and the rate of height gain. Multiple regression analysis indicated that there were significant relationships with the crosslink excretion at admission, age and weight-for-height, which together accounted for 44% of the variance in height velocity of the children. The crosslink excretion at admission therefore gave some indication of the likely response to treatment in terms of height gain. Comparison of the Pyd/Dpd ratios in urine showed that there was no significant difference between the values at admission (4.5 1.0) and at discharge (4.4 0.6), despite very different growth rates at these stages. These results therefore confirm the insensitivity of the current techniques to changes in growth cartilage turnover which are presumably occurring during this period.

4.3. Recent advances in methodology

In recent years there have been a number of improvements in methodolgy and the purpose of this section is not to deal with these in detail but simply to illustrate the types of assays that are now available.

Although nearly all studies to date have been performed on hydrolysed urine, analysis of urine directly without hydrolysis (Fig. 4) showed the presence of both Pyd and Dpd, together with an additional component that has been identified as a glycosylated derivative (Robins, Duncan & Riggs, 1990). Major improvements in the precision of the HPLC assay have been achieved by the introduction of a synthetic internal standard which has facilitated the complete automation of the assay (Pratt et al., 1992). Analysis of total and free crosslinks in adult urine samples from normal volunteers and from patients with a wide range of different diseases showed that the proportions of free crosslinks were relatively consistent; these were about 40% for Pyd with slightly higher values for free Dpd (Robins et al., 1991). Thus, measurement of free crosslinks gives similar information to that obtained by measuring the total amounts. The ELISA for Pyd originally developed (Robins, 1982) was shown not to react with the free crosslinks in urine (Robins et al., 1986). Separation of the Pyd isolated from bone hydrolysates by ion-exchange chromatography showed that the ELISA reacted only with diastereoisomers produced during the hydrolysis step and not with the single natural isomer of Pyd isolated from urine (Fig. 5). These observations have led to the development of a direct immunoassay for the crosslinks using antibodies raised against the natural isomer of Pyd (Seyedin et al., 1993). Preliminary evaluations of the assay have been performed in adults but the question of whether these types of assay would be applicable to growth studies had not until recently been addressed.

Fig. 4. Free crosslinks in urine.

The HPLC chromatogram shows the elusion positions of the pyridinium crosslinks, pyridinoline (Pyd) and deoxypyridinoline (Dpd), and the glycosylated derivative, glucosylgalactosyl-pyridinoline (Glc.Gal-Pyd), obtained by direct analysis of urine (80 ml) with fluorescence monitoring of the natural fluorescence of the crosslinks.

4.4. Free crosslinks in children

The main question governing the applicability of this technique is whether there is a consistent proportion of free crosslinks in children as was found in adults. Analyses of the free and total crosslinks by HPLC (Table 2) had shown that the proportions of both free Pyd and Dpd in healthy, growing children were similar to the corresponding values in adults. Also, there were no differences in the proportions of free crosslinks with age between 2 and 14 years. How ever, analyses of urines from a group of 54 children with various growth abnormalities (primarily idiopathic short stature) showed that the amounts of free crosslinks were slightly lower in comparison with the other two groups (Table 2). It may be concluded therefore that the direct immunoassays for crosslinks are probably applicable to growth studies in children although some caution must be exercised until more information is gained on the patterns of crosslink excretion under different metabolic states.

Fig. 5. Chromatography of native and hydrolysed pyridinoline (Pyd).

Urinary (native) and hydrolysed Pyd were separated by ion-exchange chromatography with fluorescence monitoring. An inhibition ELISA for Pyd (hatched bars) showed reaction primarily with the diastereoisomers produced on hydrolysis, and virtually no reaction with the single, native component in urine.

Two alternative techniques based on the collagen crosslinks have recently been described: a urinary assay based on antibodies raised against peptides associated with the crosslinks (Hanson et al., 1992), and a measurement in serum of crosslink-containing peptides derived from collagen type I (Risteli et al., 1993). No detailed information is yet available, however, on the application of these assays in growth studies.

Table 2. Free crosslinks in children's urine



% Free crosslink



Healthy controls (aged 2-6 years)


37.1 5.0

39.7 4.5

Healthy controls (aged 7-14 years)


36.2 4.7

41.4 6.1

Children with growth abnormalities (aged 3-14 years)


29.7 5.0

33.4 6.0

5. Bone formation markers

Of the main formation markers listed in Fig. 2, discussion will be limited to the matrix components, osteocalcin and the procollagen peptides.

5.1. Osteocalcin

The schematic representation of the biosynthesis and metabolism of osteocalcin shown in Fig. 6 illustrates the issues to be considered in utilizing this component as a plasma or serum marker of bone formation. Osteocalcin synthesis is essentially specific to osteoblasts with only small amounts being produced by odontoblasts. In common with most secreted proteins, osteocalcin has a signal sequence that is removed in the rough endoplasmic reticulum to give pro-osteocalcin containing a 26 amino acid propeptide N-terminal to the 49-residue osteocalcin chain (Hauschka et al., 1989). Before secretion from the osteoblast, specific glutamic acid residues are carboxylated by a vitamin K dependent enzyme to form g-carboxyglutamic acid (Gla); human osteocalcin contains a maximum of three Gla residues per molecule (Hauschka et al., 1989). After cleavage of the propeptide and secretion, a large proportion of the native osteocalcin is incorporated into the mineralizing matrix, assisted by the calcium binding properties of the Gla residues. In patients treated with the anti-coagulant, Warfarin, inhibition of the carboxylase leads to the secretion of non-carboxylated osteocalcin, and a similar situation may, to a lesser degree, occur in individuals with vitamin K deficiency (Plantalech et al., 1991). The under-carboxylated molecules will, of course, be less likely to be incorporated into the matrix and a higher proportion will therefore be released into the blood. Most immunoassays for osteocalcin recognize both the native and non-carboxylated molecule equally and this heterogeneity with respect to carboxylation can therefore complicate interpretation of the results. Some attempts to establish the degree of osteocalcin carboxylation in serum have been made using hydroxyapatite binding in vitro (Price, Williamson & Lothringer, 1981; Knapen, Hamulyak & Vermeer, 1989), but the results appear to be very dependent on the precise conditions used for these estimations (Merle & Delmas, 1990).

Fig. 6. Metabolism of osteocalcin

Biosynthesis of osteocalcin giving rise to the native, calcium-binding molecule and the propeptide in serum, together with osteocalcin fragments from degradation of the matrix-bound material. In addition, lack of adequate carboxylation can produce non-carboxylated molecules reactive with most antibodies to osteocalcin.

A further complication for the interpretation of osteocalcin determinations in blood is that fragments of osteocalcin released from the matrix can also appear in the blood, and may react with osteocalcin antibodies in some assays (Tracy et al., 1990). These considerations may to some extent explain the large variations between different centres in the determination of osteocalcin (Delmas et al., 1990). In recent years there has been a move to the development of two-site assays which are more likely to detect only intact osteocalcin and not the degraded fragments. A number a such assays have been described recently (Kanzaki et al., 1992; Deftos et al., 1992). Measurement of the 26-residue propeptide of osteocalcin in blood has been achieved (Kanzaki et al., 1992) and this assay appears to provide a valid indication of osteoblastic activity without any resorptive component.

There is conflicting evidence on the applicability of osteocalcin as a growth marker. Analyses in children aged 2-19 years showed that serum osteocalcin parallelled growth velocity (Delmas et al., 1986), but in infants no correlation was found between osteocalcin and linear growth velocity (Michaelson et al., 1992). Using a two-site assay for intact osteocalcin, measurements in children aged 4-15 years revealed a marked increase during the pubertal growth spurt in boys, although this was less clear for the girls (Kanzaki et al., 1992). In children with short stature without any growth hormone deficiency, osteocalcin levels were lower than in an age-matched control group (Colle, Ruffie & Ruedas, 1988). A number of studies have measured osteocalcin in children undergoing growth hormone therapy and have concluded that the changes during the first few weeks of treatment predict the linear growth response over a much longer time period (Johansen et al., 1990; Kanzaki et al., 1992).

5.2. Procollagen peptides

Although antibodies against both the N- and C-terminal propeptides of collagen type I have been available for some time (Taubman, Kammerman & Goldberg, 1976), not until the recent work of Risteli and colleagues (Risteli, 1990) has the potential usefulness, particularly of the C-propeptide, become realized. Using antibodies raised against human propeptide, these workers have shown that the C-terminal portion circulated in the blood as a single 100 kDa species (Melkko et al., 1990) and that the serum concentrations provided a marker of bone formation (Hassager, 1991). To date, however, there are comparatively few studies of the normal physiological variations and metabolism of these peptides (Smedsrod et al., 1990; Hassager et al., 1992). Analyses of changes with age using the Procollagen type I C-Propeptide assay (PCIP) revealed very similar patterns (Saggese et al., 1992) to those observed for osteocalcin. The serum concentrations in children were about 4-fold higher than adult values, although in neonates and very young infants much higher concentrations were recorded (Saggese et al., 1992).

Collagen type III is widely distributed in most soft tissues but is essentially absent from bone, except in blood vessels and tendon attachment sites (Keene, Sakai & Burgeson, 1991). Thus, measurement of the procollagen type III N-propeptide (PIIINP) may be particularly useful in comparing and distinguishing soft tissue and skeletal growth. The changes with age in serum PIIINP concentrations (Trivedi et al., 1985) were similar to those found subsequently for PICP. More recent studies using the PIIINP assay in children treated with growth hormone have revealed positive correlations between the changes over 5 weeks and the growth velocity over 6 or 12 months (Tapanainen et al., 1988).

6. Future studies

Many of the biochemical markers discussed in this review have already proved their applicability and usefulness in growth studies. All of these assays, however, indicate the combined effects of the various processes occurring during growth at all sites of the body. Increased amounts of the crosslink markers, for example, will emanate from the linear and appositional bone growth as well as from the remodelling component. These crosslinks may also be derived from the cancellous and cortical compartments of bone, or from other body tissues. Recent experiments studying the effects of nutritionally altered growth rates in lambs have demonstrated significant differences in pyridinium crosslink excretion between fast and slow growing animals despite similar rates of linear bone growth (Scott D, Abu Damir H & Robins SP, unpublished observations). Thus, the crosslink markers cannot always provide information on the rates of linear growth, and other more specific assays are required for this purpose.

Collagen type X is a non-fibrillar variant that is synthesized exclusively in the growth plate by hypertrophic chondrocytes (Kielty et al., 1985). The precise function of this collagen is unclear but its close association with the collagen type II of the growth plate (Chen et al., 1992) suggests a role in the mineralization of cartilage and its subsequent replacement by the collagen type I of bone. The restricted expression of collagen X in the growth plate and its complete removal during growth of the bone makes collagen X an excellent candidate as a marker specifically for linear growth. Current studies are therefore designed to develop suitable immunoassays to detect this component or its fragments to assess growth plate activity, initially in organ culture but subsequently in serum or plasma.

7. Concluding remarks

Considerable advances have been made in recent years in developing biochemical markers to assess growth. The pyridinium crosslinks provide much more specific markers of bone resorption than urinary hydroxyproline, and the availability of simple, direct immunoassays increases the applicability of this technique. Rarely, however, can one marker provide sufficient information and, particularly for bone metabolism, it is important to consider the balance between formation and resorption markers. At present there are few markers that relate specifically to linear growth and there is a need for further developments in this area.

Acknowledgements - Funding from the Scottish Office of Agriculture and Fisheries Department is gratefully acknowledged.

Contents - Previous - Next