Bone 69 (2014) 89–97

Contents lists available at ScienceDirect

Bone

j ourna l homepage: www.e lsev ie r .com/ locate /bone

Original Full Length Article

Pediatric reference Raman data for material characteristics of iliactrabecular bone

S. Gamsjaeger a,1, B. Hofstetter a,1, N. Fratzl-Zelman a, P. Roschger a, A. Roschger a,c, P. Fratzl c, W. Brozek a,A. Masic c, B.M. Misof a, F.H. Glorieux b, K. Klaushofer a, F. Rauch b, E.P. Paschalis a,⁎a Ludwig Boltzmann Institute of Osteology at the Hanusch Hospital of WGKK and AUVA Trauma Centre Meidling, 1st Medical Department, Hanusch Hospital,Heinrich Collin Str. 30, A-1140 Vienna, Austriab Genetics Unit, Shriners Hospital for Children and McGill University, Montreal, Quebec H3G 1A6, Canadac Max Planck Institute of Colloids and Interfaces, 14424 Potsdam, Germany

⁎ Corresponding author at: Ludwig Boltzmann InstituteHeinrich Collin Str. 30, A-1140 Vienna, Austria.

E-mail address: eleftherios.paschalis@osteologie.at (E.1 SG, and BH contributed equally to this work.

http://dx.doi.org/10.1016/j.bone.2014.09.0128756-3282/© 2014 Elsevier Inc. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 12 May 2014Revised 9 September 2014Accepted 11 September 2014Available online 22 September 2014

Edited by: Robert Recker

Keywords:Healthy childrenRaman spectroscopyTissue ageSubject ageBone material propertiesNormative data

Bone material characteristics are important contributors in the determination of bone strength. Raman spectro-scopic analysis provides information onmineral/matrix ratio, mineralmaturity/crystallinity, relative pyridinoline(Pyd) collagen cross-link content, relative proteoglycan content and relative lipid content. However, publishedreference data are available only for adults. The purpose of the present study was to establish reference dataof Raman outcomes pertaining to bone quality in trabecular bone for children and young adults. To this end, tis-sue age defined Raman microspectroscopic analysis was performed on bone samples from 54 individuals be-tween 1.5 and 23 years with no metabolic bone disease, which have been previously used to establishhistomorphometric and bone mineralization density distribution reference values. Four distinct tissue ages,three well defined by the fluorescent double labels representing early stages of bone formation and tissue mat-uration (days 3, 12, 20 of tissue mineralization) and a fourth representing old mature tissue at the geometricalcenter of the trabeculae, were analyzed. In general, significant dependencies of the measured parameters on tis-sue age were found, while at any given tissue age, sex and subject agewere not confounders. Specifically, miner-al/matrix ratio, mineral maturity/crystallinity index and relative pyridinoline collagen cross-link content indexincreased by 485%, 20% and 14%, respectively between days 3 and 20. The relative proteoglycan content indexwas unchanged between days 3 and 20 but was elevated in the old tissue compared to young tissue by 121%.The relative lipid content decreasedwithin days 3 to 20 by−22%. Thus, themethod allows not only themonitor-ing of material characteristics at a specific tissue age but also the kinetics of tissue maturation as well. Theestablished reference Raman database will serve as sensitive tool to diagnose disturbances inmaterial character-istics of pediatric bone biopsy samples.

© 2014 Elsevier Inc. All rights reserved.

Introduction

In addition to bonemass and architecture, bonematerial quality is animportant determinant of bone strength, and there is an increasing inter-est in assessing properties of the extracellular matrix for the diagnosis ofbone diseases [1–3]. Connective tissue disorders like OsteogenesisImperfecta or Ehlers–Danlos syndrome are characterized by alteredcomposition and structure of the bone tissue [4]. Also metabolic disor-ders of mineral homeostasis like hyperparathyroidism [5], chronic kid-ney diseases [6] or diabetes [7] affect bone strength and lead tofragility fractures that are often not associated with changes in bonemineral density, and the same holds for idiopathic osteoporosis [8,9].

of Osteology, HanuschHospital,

P. Paschalis).

Due to continuous remodeling activity by the bone cells, materialcharacteristics of the organic and inorganic bone matrix are highlydynamic at the microscopic level. The mineralization process occurs in abiphasic way: a rapid phase where the organic matrix becomes mineral-ized to about 70% of its final valuewithin a fewdays (primarymineraliza-tion), followed by a second phase where the rate of mineralization ismuch lower taking months or years to achieve plateau level (secondarymineralization) [10–12]. Similarly, organic matrix components such astype I collagen and proteoglycans undergo significant post-translationalmodifications after having been synthesized by the relevant bone cells[13–15]. As a result, bothmineral and organicmatrix properties at themi-croscopic level are expected to be different depending on the relative tis-sue age or remodeling rates [11,16].

Children and adolescents have high bone turnover rates. Indeed,during growth, the skeleton is not only modeled to increase externalsize and shape of the bones but is also highly remodeled [17].

Table 1Summary of the technical variance for each parameter reported in the present study,expressed as percentage of coefficient of variation (% COV), calculated from 25 repeatedmeasurements of the exact same anatomical location.

Parameter % COV

Mineral/matrix 1.582Relative proteoglycan content 2.606Relative lipids content 5.202Mineral maturity/crystallinity (v1PO4 FWHH) 1.508Relative pyridinoline content 2.021

90 S. Gamsjaeger et al. / Bone 69 (2014) 89–97

However, in sharp contrast to the situation in high turnover osteoporo-sis, during growth there is a positive imbalance and bone formation pre-vails over bone resorption leading to a net increase in trabecularthickness after each remodeling cycle [17]. In children undergoing a“growth spurt”, a situation that is inherently linked to markedly in-creased bone turnover, a transient “weakening” of the bones has beenobserved [18]. All together these observations underline the necessityto gain further insights in bone material characteristics of the develop-ing skeleton.

For establishing a diagnosis in an individual patient or for investigat-ing treatment follow-up, a comparisonwith normal values fromhealthysubjects is of crucial importance. A precise tool to allow the study ofbone metabolism in individual patient disorders is bone biopsies.Transiliac bone biopsies are traditionally used to perform bonehistomorphometry that allows the quantification of the size andamount of the mineralized and unmineralized bone tissue and evalua-tion of bone cell function [19,20]. Dynamic evaluation of cell function re-quires that the biopsy is fluorescent labeled i.e., the patient has receivedat least two courses of tetracycline or other fluorescent drugs prior to bi-opsy. Further information in particular about bone material quality canbe extracted from the available biopsy bone tissue applying differenttechniques. The tetracycline labels allow specific tissue ages in cancel-lous bone to be distinguished. Glorieux et al. presented reference datafor pediatric iliac bone histomorphometry data based on the evaluationof 58 bone biopsies from a healthy population from 1.5 to 23 years [21].These biopsieswere further used to assess the bonemineral density dis-tribution (BMDD) by quantitative backscattered electron imaging(qBEI) [21,22]. The combination of the histomorphometric and BMDDreference variables has raised the possibility of obtaining new insightswithin bone structural and material characteristics in the developingskeleton, as recently demonstrated in pediatric patients with geneticdisorders like osteogenesis imperfecta or in dialyzed patients to evalu-ate the skeletal effects of growth hormone therapy [23–27]. Howeveruntil now, such pediatric reference data assessing the organic compo-nent of the bone matrix is still lacking yet necessary as changes in theorganic bone composition have already been observed in fractureprone children [28,29].

Raman microspectroscopy is a vibrational spectroscopy techniquethat allows the determination of bone material characteristics (of bothmineral and organic matrix bone components) in bone biopsy blockswith a spatial resolution of ~1 μm, bymeasuring thewavelength and in-tensity of inelastically scattered light frommolecules [30,31]. In combi-nation with fluorescence labeling, this offers the capability to establishthese properties as a function of tissue age [30–34].

The purpose of the present study was to establish a reference data-base of trabecular bone with the Raman derived bone quality indicesfor growing children, to complement the already existing ones basedon histomorphometry and BMDDdata [21,22], in the same iliac crest bi-opsy cohort, as a function of sex-, subject-, and tissue-age. The tissue agenormalization is of particular interest as it minimizes the effect of boneturnover on the monitored bone quality indices and allows the descrip-tion of the kinetics of maturation of the material characteristics investi-gated in this normal subject cohort [30–33].

Material & methods

Subjects

The study population comprised 54 Caucasian subjects aged 1.5 to23 years (32 female, 22 male) without any known metabolic bone dis-order, as previously described [21,22]. Transiliac bone biopsy sampleswere acquired during surgery for various conditions, such as lowerlimb deformities, scoliosis, clubfeet and other conditions requiring cor-rective surgery. The bone specimens were collected on day 4 or 5 afterdual labeling with demeclocycline (15–20 mg per kg body weight perday) taken orally during two two-day periods separated by a 10-day-

free interval. All subjects were ambulatory, had normal renal function(assessed by serum creatinine measurements) and had no evidence ofany metabolic bone disease. None of the included subjects was eitherimmobilized prior to surgery or received medications known to affectbone metabolism.

Raman analysis

Raman microspectroscopic analyses employed a Senterra (BrukerOptik GmbH) instrument. A continuous laser beam was focused ontothe sample through a Raman fluorescence microscope (OlympusBX51, objective 50×) with an excitation of 785 nm (100mW) and a lat-eral resolution of ~0.6 μm [32,33]. The Raman spectra were acquiredfrom the biopsy block polished surface, using a thermo-electric–cooledcharge-coupled device (CCD) (Bruker Optik GmbH). All data analyseswere done with the Opus Ident software package (OPUS 6.5, BrukerOptik GmbH). Once acquired, the Raman spectra were baselinecorrected (rubber band, 5 iterations) to account for fluorescence, andthe following Raman parameters were calculated as published else-where [32,33]. The mineral/matrix ratio was expressed as the ratio ofthe integrated areas of the v2PO4 (410–460 cm−1) to the amide III(1215–1300 cm−1) bands. Thematurity/crystallinity of the boneminer-al apatite crystallites was approximated from the inverse of the fullwidth at half height (FWHH) of the v1PO4 (930–980 cm−1) band [31,35]. The relative pyridinoline (Pyd) content (a major trivalent collagencross-link) was calculated as the absorbance height at 1660 cm−1/area of the Amide I band (1620–1700 cm−1) [31,36–39]. The relativeproteoglycan (PG) content was defined as the PG/matrix ratio, whichwas calculated from the ratio of the integrated areas of the proteogly-can/CH3 (1365–1390 cm−1) band representative of glycosaminogly-cans (GAGs) [40–42], to the amide III (1215–1300 cm−1) bands,respectively. Finally, the relative lipid content was expressed as lipids(~1298 cm−1)/amide III [43].

The technical variance for each parameter monitored in the presentstudy was calculated through the acquisition of 20 repeated measure-ments of the same anatomical location, and expressed as % coefficientof variation (% COV) (Table 1).

Anatomical area selection criteria

For each biopsy, three trabeculaewith clearly discernible double tet-racycline labels were analyzed in the following regions (see Fig. 1): Be-tween the 2nd label and mineralizing front (2 μm before 2nd label)corresponding to a mineralized tissue age of 1–3 days, between thetwo labels corresponding to a tissue age of about 4–20 days and rightbehind the 1st label (2 μm behind 1st label) corresponding to a tissueage of slightly greater than 20 days, related to the tissue that wasformed just before the first dose of demeclocycline was taken. In eachof these regions three measurements were obtained, for a total of ninemeasurements per individual per anatomical area. Additionally, threemeasurements in the geometrical center of each trabecula utilizedwere also acquired, representative of old tissue (tissue agemuchgreaterthan 20 days) with prolonged secondary mineralization designated as“CENTER”.

OLDER LABEL

YOUNGEST

20 µm

b

a

c

50 µm +

CENTER

Wavenumber cm-1

Wavenumber cm-1

400 600 800 1000 1200 1400 1600 1800

Ram

an In

tens

ity

(abs

ortp

ion

units

)

between the two labels

center of the trabeculum

behind the 1stlabel

behind the 2ndlabel

amide IIIv2PO4

v1PO4

dPGs

1250 1300 1350 1400 1450 1500

050

100

150

Ram

an In

tens

ity (

abso

rptio

n un

its)

lipids (CH2)

Fig. 1. (a) Fluorescence light photomicrograph (objective lens 50×) of a forming trabecular surfacewith the fluorescent double label evident showing the positions of the 3firstmeasurements(red crosses), (b) a picture through the 20− lens showing where the positionwhere the 4thmeasurement was obtained (red cross; although the picture was obtainedwith the 20× lens forpresentation purposes, the actual measurements were obtained through the 50× lens so as to ensure identical volume of analysis with the 3 first measurements). Typical Raman spectra areshown in (c) with the spectral regions of interest appropriately labeled, while the spectral region of ~1200–1500 cm−1 is isolated and zoomed for better visualization of the peaks that wereutilized for proteoglycan and lipids considerations (d).

91S. Gamsjaeger et al. / Bone 69 (2014) 89–97

Statistical analysis

For each individual, the nine values/microanatomical location,representative of every tissue age specific region, were averaged andthe resultant value was treated as a single statistical unit for the appro-priate anatomical location and tissue age. The data (with the exceptionof FWHH) were not normally distributed (as per Kolmogorov–Smirnovtest), thus comparisons betweenmale and female groups at the four dis-tinct tissue ages considered were made using Mann–Whitney U-test.Comparisons as a function of tissue age were made using one-wayANOVA without assumption of Gaussian distribution (Kruskal–Wallis),followed by Dunn's post-hoc comparison. Statistical significance wasassigned to p b 0.05. Correlations amongst the parameters monitoredand subject age were analyzed by Spearman's nonparametric test.

Results

Information on the technical variance for each parameter reported isprovided in Table 1, as % COV.

Raman spectra were obtained at tissue areas of four different ages(Fig. 1). Fig. 1a shows the fluorescent labels and the three positionsclose to the fluorescence labels (tissue that is 1–3, 4–20, and N20 daysold) where Raman spectra were acquired as visualized by the light mi-croscope (fluorescence mode) of the Raman instrument. AdditionalRaman spectra were obtained in the geometrical center of these trabec-ulae (Fig. 1b), representative of older tissue age (CENTER) compared tothat of the first 3 measurements. Fig. 1c is a stack of typical Raman

spectra at the 4 different tissue ages in the same trabeculum, with thepeaks of interest appropriately labeled. In Fig. 1d the area from 1200to 1500 cm−1 of the 1–3 day-old tissue Raman spectra is enlarged tomark the proteoglycan Raman band (1376 cm−1; CH3) and the Ramanvibration (CH2) from the lipids.

A summary of all Raman spectroscopic outcomes pooled for bothsexes and all subject ages (median, 25th and 75th percentiles) at the dif-ferent tissue ages is provided in Table 2. Sex differences as tested byMann–Whitney comparison (data not shown) were not significant forany of the Raman outcomes. Subject age as tested by Spearman Rankorder correlation (Figs. 2a to 6a) was not a confounder for Raman out-comes with one exception. For the less well-defined oldest analyzedtissue age (CENTER) mineral maturity/crystallinity index revealed aweak significant positive regression (Spearman r = 0.360, p = 0.005)(Fig. 3a). As a consequence, the data were pooled for subsequentanalyses.

The mineral/matrix ratio significantly increased continuouslyat early tissue ages from day 3 to day 20 by about 485%. A further in-crease (83%) occurred with respect to age of the central bone region(Fig. 2b).

Mineral maturity/crystallinity also significantly increased as a func-tion of early tissue age from days 3 to 20 by about 20%. However,there was no difference between day 20 and beyond (center region ofthe trabeculae) (Fig. 3b).

Relative Pyd content was significantly increased as a function ofearly tissue ages (~14% from days 3 to 20) (Fig. 4b), but there was nofurther increase between older (day 20) and beyond (center region ofthe trabeculae).

Table 2Summary of intrinsic material characteristics (median, and 25th and 75th percentile values) of children (age range 1.5–23 years, sample size N = 54) monitored at each specific tissuesite. Since the tissue age of the trabecular geometrical centers is not precisely known, it is denoted by anatomical/geometrical location.

Tissue age(days)

Mineral/matrix Mineral maturity/crystallinity(1/FWHH v1PO4)

Relative pyridinolinecontent

PG/matrix Lipids/matrix

1–3 Median 0.085 0.038 0.014 0.057 0.09125th percentile 0.064 0.036 0.012 0.045 0.08175th percentile 0.153 0.040 0.016 0.069 0.099

4–20 Median 0.176 0.043 0.015 0.041 0.08625th percentile 0.10 0.042 0.014 0.030 0.07575th percentile 0.285 0.046 0.017 0.058 0.098

** *** *N20 Median 0.497 0.047 0.016 0.059 0.071

25th percentile 0.391 0.046 0.016 0.045 0.06275th percentile 0.622 0.047 0.017 0.082 0.077

*** ### *** ### *** # *** ###Center Median 0.911 0.047 0.016 0.126 0.057

25th percentile 0.792 0.046 0.016 0.101 0.04775th percentile 1.025 0.047 0.016 0.151 0.063

*** ### §§§ *** ### *** *** ### §§§ *** ### §§§

*p b 0.05, **p b 0.01, ***p b 0.0001, vs 1–3 day-old tissue.#p b 0.05, ##p b 0.01, ###p b 0.0001, vs 4–20 day-old tissue.§p b 0.05, §§p b 0.01, §§§p b 0.0001, vs N20 day-old tissue.

92 S. Gamsjaeger et al. / Bone 69 (2014) 89–97

Relative proteoglycan content expressed as PG/Amide III ratio wassimilar between the 3 early tissue ages, but was significantly higher atthe oldest tissue (CENTER) compared to the younger tissues (~121%)(Fig. 5b).

Relative lipid content significantly decreased as a function of tissueage from 1–3 to ~20 days old (~−22%) and from N20 days old toCENTER further ~−20% (Fig. 6b). Evidence for the specificity for lipidsof the Raman peak used, and its independence from tissue organiza-tion/orientation is provided in Appendix A.

Discussion

The purpose of the present study was to establish reference valuesof bonematerial characteristics for Caucasian children and young adultsas can be accessed by Raman microspectroscopy from transiliac bonebiopsy sample blocks as a function of sex, subject, and tissue age. Thetissue age normalization is expected to minimize any effects ofdivergent bone turnover rates amongst the individuals investigated.

a

Fig. 2. (a) Scatter plots show that mineral/matrix ratio, expressed as the ratio of the integratedinvestigated in the present study. The linear regression line is also shown for the 4 different tissas a function of tissue age (**p b 0.01, ***p b 0.0001).

The fluorescence markers of routine tetracycline labeling were used todefine age and the early maturation stages of the bone tissue (1–3 to20 days). We did not observe any dependency of the Raman outcomesat these defined tissue ages on subject age or gender. Therefore, asingular Raman data set is necessary to characterize the materialcharacteristics of the reference cohort containing 54 subjects (aged 1.5to 23 years).

In thepresentwork,five differentmaterial parameters of bone tissuewere considered:

1) The mineral/matrix ratio provides information on the amount ofmineral normalized for the amount of organic matrix within thevolume of bone material analyzed. The results of the presentstudy indicate that this ratio is very sensitive to tissue age, inagreement with previously published reports on mineral massfractions or ratios at different bone tissue ages in humans and an-imal models [11,44–47]. The mineral/matrix ratio increased fromdays 1–3 to day 20 and was the highest at the oldest tissue at

b

1-3

4-20 >2

0

CENTER0.0

0.5

1.0

1.5

areas of the v2PO4 and amide III bands, was independent of subject age at all 4 tissue agesue ages considered. (b) On the other hand, themineral/matrix ratio significantly increased

a

b

Fig. 3. (a) Scatter plots indicate that themineral maturity/crystallinity, estimated from the inverse of the FullWidth at Half Height (FWHH) of the v1PO4 bandwas independent of subjectage at the 3 younger tissue ages analyzed. It was however dependent on subject age only at the oldest of the ages examined, namely at the trabecular center (Spearman r = 0.360, p =0.005). It was also significantly dependent on tissue age. The linear regression line is also shown for the different 4 tissue ages considered. (b). As can be seen, this parameter significantlyincreases as a function of tissue age (b) (***p b 0.0001).

93S. Gamsjaeger et al. / Bone 69 (2014) 89–97

central trabecular area in agreement with the well-known timecourse of bone matrix mineralization, which is rapid in its initialphase (primary mineralization) and slows down with prolongedduration of mineralization (secondary mineralization) [10,11,48].Interestingly, in recent reports comparing groups of adult patientswith similar BMD but divergent fracture history, this mineral/matrix ratio measured between double fluorescent labels was

a

Fig. 4. (a) Scatter plots show that the relative pyridinoline collagen cross-link content was indepfor the different 4 tissue ages considered. On the other hand, it significantly increased as a func

significantly different between these groups (fracture vs. non-fracture) [9,49,50] emphasizing its relevance in predicting bonestrength.

2) Bone mineral maturity/crystallinity is a measure reflectingthe chemical makeup of the apatite crystallites and by extrapolationtheir size [3,16,51]. Theoreticalmodels predict that larger crystallitesin general result in compromised mechanical behavior [3,52–54].

1-3

4-20 >2

0

CENTER0.012

0.013

0.014

0.015

0.016

0.017 *

* * *

* * **

Tissue Age (days)

Pyd

/ A

mid

eI

b

endent of subject age at all 4 tissue ages analyzed. The linear regression line is also showntion of tissue age (b) (*p b 0.05; ***p b 0.0001).

1-3

4-20 >2

0

CENTER

0.00

0.05

0.10

0.15

0.20

Fig. 5. (a) Scatter plots indicate that the relative proteoglycan content was independent of subject age at all 4 tissue ages considered in the present work. (a) The linear regression line isalso shown for the different 4 tissue ages considered.When the data were considered as a function of tissue age (b), the relative proteoglycan content at the center (oldest tissue age)wassignificantly greater compared to the 3 younger ages (***p b 0.0001).

94 S. Gamsjaeger et al. / Bone 69 (2014) 89–97

Additionally, its considerable contribution in the determinationof bone strength may be inferred from the changes described inmineral maturity/crystallinity index in disease and in response totherapies [16,50,55–59]. Moreover, in cases where divergencebetween BMD and fracture incidence is noted, this parameter corre-lated with the latter [50]. In the present study, mineral maturity/crystallinity exhibited a distinct increase with tissue maturation(early tissue ages). There was no further increase from tissue ageday 20 on to the oldest analyzed tissue at the geometrical center oftrabeculae. The values of these oldest analyzed tissue regionsdisplayed a weak subject age-dependency. The reason for this is

Fig. 6. (a) Scatter plots show that the relative lipid content was independent of tissue age at all 4considered), while it significantly decreased as a function of tissue age (b) (***p b 0.0001).

unclear, and no definite argument may be put forth due to the un-certainty in the exact tissue age of the specific anatomical locationconsidered (there may be differences ranging from months toyears depending on the individual).

3) The relative pyridinoline content (normalized to organicmatrix con-tent) is part of one of the most distinct feature of type I collagen inmineralized tissues, namely its cross-linking chemistry and molecu-lar packing structure [15]. The intermolecular crosslinking providesthefibrillarmatriceswith variousmechanical properties such as ten-sile strength and viscoelasticity. The importance of collagen inter-molecular crosslinks to the mechanical performance of bone is also

******

******

***

tissue ages analyzed (the linear regression line is also shown for the different 4 tissue ages

95S. Gamsjaeger et al. / Bone 69 (2014) 89–97

very apparent in the pyridoxine deficient chick animal model[60–62], as well as in lathyrism [38,63,64]. Changes in thesecrosslinks, even at confined anatomical locations, are sufficientto influence bone strength in the absence of alterations in eithermineral quantity or quality [38]. Finally, as is the case with min-eral maturity/crystallinity, in cases where deviation betweenBMD values and fracture incidence is encountered, collagencrosslinks correlate with the latter in animal models and inhumans [9,50,65,66]. Pyridinoline is a mature, non-reducible,trivalent collagen crosslink abundant in mineralizing type I col-lagen [15]. In the present study, relative pyridinoline contentwas found to increase within early tissue ages, but from day20 on to old tissue (central area) no further increase wasobserved.

4) The relative proteoglycan content (normalized to organic ma-trix); in vitro, in situ, and in vivo experiments have shown pro-teoglycans to be negative modulators of mineralization[67–69]. Proteoglycans have also been identified in perilacunarmatrix around the osteocyte lacunae, and around the canaliculi[70] in compact lamellar rat and human bone. Recently, it wasproposed that a plausible role of these osteocyte-related proteo-glycans (and in particular perlecan/Hspg2 (PLN)) was to preventmineralization in the pericellular space of the lacunocanalicularnetwork so as to ensure uninhibited interstitial fluid movement[71]. In the present study, significant differences were observedbetween the oldest tissue age and the 3 younger ones exam-ined. This increase may be due to several possibilities. Differentproteoglycans may be present at the formation surfaces and inthe older tissue, having different numbers of glycosaminoglycanchains/molecule. Moreover, proteoglycans are known to under-go extensive post-translational modifications including additionof glycosaminoglycan chains as well as N- and O-linked oligo-saccharides along other changes as a function of age [13]. Unfor-tunately, Raman microspectroscopic analysis methods to dateare not discriminant enough to identify whether this increasein glycosaminoglycans is due to different proteoglycans, or dueto the same proteoglycans that have undergone different post-translational modifications. Another possible explanation forthe difference in relative proteoglycan content as determinedby the glycosaminoglycan concentration normalized to thetotal amount of organic matrix between the oldest and the 3younger tissue ages may be due to differences in canalicularnetwork density between actively bone forming surfaces anddensely mineralized bone.

5) Relative lipids content (normalized to organicmatrix); lipids havebeen implicated in the literatureasmineralnucleators, responsibleformineralization of collagenfibers. A layer of round lipidparticleshas been identified to mediate collagen calcification in compactbone formation, and calcium-acidic phospholipid-phosphatecomplexes have been shown to increase in concentration duringcartilage calcification and early bone formation [72–75]. In thepresent study, the relative lipid content was significantly depen-dent on tissue age, with the highest values encountered in youn-gest bone formed, in agreement with what has been reported inthe literature concerning their role.

In general the variability in the parameters monitored in thepresent study (as seen in the interquartile range) was greater atthe sites of younger tissue age versus those of older tissue age(center site). This is most likely due to the fact that in zone ofyoung tissue changes of tissue properties are rapidly occurringdue to maturation processes. For example, close behind the min-eralization front the mineral concentration in the bone matrix israpidly increasing within few days from 0% up to 70% of fully min-eralized bone tissue (primary mineralization) [11,45]. In contrastin the center zone only secondary mineralization occurs with

only a relatively slow increase in mineral concentrations towardsfull mineralization (in scales of months to years) [10,11,44,45].Thus the zone with relatively old tissue age is less heterogeneous.On the other hand, establishing values at the trabecular centersmay prove useful in the future, in situations where unlabeledspecimens are available and need to be compared to a normativedatabase.

An important finding of the present study is thatwhen similar tissueages are considered, the monitored parameters are statistically similarregardless of the subject age (in the present study the subject agerange was 1.5–23 years). This fact suggests that the osteoblasts of thesubjects having different ages are producing a bone matrix, which is ofsimilar quality with respect to identical tissue/matrix ages. Further,it highlights the need for tissue age normalization when microscopictechniques such as Raman and Fourier transform infrared micro-spectroscopy or Imaging are used to investigate bone material withinsingle bone structural units (BSUs, bone packets and osteons), irrespec-tive of subject age. Additionally, the information obtained through suchtechniques,while complementary,may not be in concert with the infor-mation obtained through techniques utilizing the whole bone tissue,because it is a composition of numerous BSUs exhibiting a certain agedistribution strongly dependent on the bone formation/turnover rateof the subject. Consequently it is important to consider and compareeither the whole tissue, or tissue-age normalized anatomical areas.Finally, there was no dependence of the intrinsic bone material proper-ties on sex.

A limitation of the present work is that, although Raman spectro-scopic analysis is capable of providing information on proteoglycanand lipid concentrations, it is not discriminant enough to date to readilydistinguish between different proteoglycans and different lipids. On theother hand, since Raman analysis of PGs is based on GAGs, it should bekept in mind that in bone, chondroitin 4-sulfate constitutes ~90% of thetotal GAG content and is found predominantly in biglycan and decorin[76]. Another limitation of the present study is that the measurementswere performed in iliac crest biopsies from Caucasian children only(as theywere the only ones available to us at the present time), and fur-ther studies are needed to decipherwhether these observations hold forother ethnicities as well.

In conclusion, the results of the present study indicate that bonema-terial characteristics likemineral/matrix ratio, mineralmaturity/crystal-linity, relative pyridinoline collagen cross-link content, relativeproteoglycan, and relative lipids content in trabecular bone of childrenand young adults (1.5–23 years old) are highly dependent on tissueage,while at any given tissue age, sex or subject age is not a confoundingfactor. As it is state of the art to perform bone biopsieswith prior doubletetracycline fluorescence labeling for differential diagnosis, the present-ed method of Raman microspectroscopic examination can be appliedadditionally to such biopsy samples. The established reference databaseallows now the detection of disturbances in material characteristics ofpediatric bone biopsy samples.

Acknowledgments

This study was supported by the AUVA (Research funds of theAustrian workers compensation board), by the WGKK (Viennese sick-ness insurance funds) and the Shriners of North America.

Appendix A

The relative lipid contentwas expressed as the ratio of the integratedarea of the lipid band ~1298 cm−1/amide III [43]. To further demon-strate the selectivity of the band ~1298 cm−1 for lipids, Raman spectraof fatty tissue from porcine skin (obtained at a local slaughterhouse),purified type I bovine bone collagen (kindly provided by Dr. S. Robins,Rowett Institute, UK), and commercially available biglycan (Sigma, Al-drich) are shown in the figure below.

96 S. Gamsjaeger et al. / Bone 69 (2014) 89–97

We have also shown previously that certain Raman bands are sensi-tive to tissue organization/orientation [47]. To examine whether theratio used in the present paper to determine the relative lipids contentis dependent on tissue orientation, we calculated the lipids/amide IIIratio in spectra previously obtained in the mineralizing turkey leg ten-don (a model system for highly oriented collagen and apatite crystal-lites) animal model as a function of Raman laser polarization and theresults are presented in the graph below:

As may be seen, this ratio is independent of Raman laser polariza-tion, thus independent of tissue organization/orientation. The instru-ment set-up was the same as in the published study and for eachpolarization, 5 Raman spectra were obtained and calculated.

References

[1] Fratzl P, Gupta HS, Roschger P, Klaushofer K. Bone nanostructure and its relevancefor mechanical performance, disease and treatment. Nanotechnology 2010;2:345–60.

[2] Seeman E. Structural basis of growth-related gain and age-related loss of bonestrength. Rheumatology (Oxford) 2008;47(Suppl. 4):iv2–8.

[3] Fratzl P, Gupta H, Paschalis E, Roschger P. Structure and mechanical quality of thecollagen–mineral nano-composite in bone. J Mater Chem 2004;14:2115–23.

[4] Eyre DR, Weis MA. Bone collagen: new clues to its mineralization mechanism fromrecessive osteogenesis imperfecta. Calcif Tissue Int 2013;93:338–47.

[5] Zoehrer R, Dempster DW, Bilezikian JP, Zhou H, Silverberg SJ, Shane E, et al. Bonequality determined by Fourier transform infrared imaging analysis in mild primaryhyperparathyroidism. J Clin Endocrinol Metab 2008;93:3484–9.

[6] Malluche HH, Porter DS, Monier-Faugere MC, Mawad H, Pienkowski D. Differencesin bone quality in low- and high-turnover renal osteodystrophy. J Am Soc Nephrol2012;23:525–32.

[7] Saito M, Marumo K. Bone quality in diabetes. Front Endocrinol (Lausanne) 2013;4:72.

[8] Malluche HH, Porter DS, Mawad H, Monier-Faugere MC, Pienkowski D. Low-energyfractures without low t-scores characteristic of osteoporosis: a possible bone matrixdisorder. J Bone Joint Surg Am 2013;95:e1391–6.

[9] Misof B, Gamsjaeger S, Cohen A, Hofstetter B, Roschger P, Stein E, et al. Bonematerialproperties in premenopausal womenwith idiopathic osteoporosis. J BoneMiner Res2012;27(12):2551–61.

[10] Boivin G, Meunier PJ. Changes in bone remodeling rate influence the degree of min-eralization of bone. Connect Tissue Res 2002;43:535–7.

[11] Roschger P, Paschalis EP, Fratzl P, Klaushofer K. Bonemineralization density distribu-tion in health and disease. Bone 2008;42:456–66.

[12] Ruffoni D, Fratzl P, Roschger P, Phipps R, Klaushofer K, Weinkamer R. Effect of tem-poral changes in bone turnover on the bone mineralization density distribution: acomputer simulation study. J Bone Miner Res 2008;23:1905–14.

[13] GrzesikWJ, Frazier CR, Shapiro JR, Sponseller PD, Robey PG, Fedarko NS. Age-relatedchanges in human bone proteoglycan structure. Impact of osteogenesis imperfecta.J Biol Chem 2002;277:43638–47.

[14] Robins SP. Biochemistry and functional significance of collagen cross-linking.Biochem Soc Trans 2007;35:849–52.

[15] Yamauchi M. Collagen: the major matrix molecule in mineralized tissues. In:Anderson JJB, Garner SC, editors. Calcium and phosphorus in health and disease.New York: CRC Press; 1996. p. 127–41.

[16] Paschalis EP, Mendelsohn R, Boskey AL. Infrared assessment of bone quality: a re-view. Clin Orthop Relat Res 2011;469:2170–8.

[17] Parfitt AM, Travers R, Rauch F, Glorieux FH. Structural and cellular changes duringbone growth in healthy children. Bone 2000;27:487–94.

[18] Cooper C, Dennison EM, Leufkens HG, Bishop N, van Staa TP. Epidemiology ofchildhood fractures in Britain: a study using the general practice research database.J Bone Miner Res 2004;19:1976–81.

[19] Recker RR, Kimmel DB, Dempster D, Weinstein RS, Wronski TJ, Burr DB. Issues inmodern bone histomorphometry. Bone 2011;49:955–64.

[20] Rauch F. Bone biopsy: indications and methods. Endocr Dev 2009;16:49–57.[21] Glorieux FH, Travers R, Taylor A, Bowen JR, Rauch F, NormanM, et al. Normative data

for iliac bone histomorphometry in growing children. Bone 2000;26:103–9.[22] Fratzl-Zelman N, Roschger P, Misof BM, Pfeffer S, Glorieux FH, Klaushofer K, et al.

Normative data on mineralization density distribution in iliac bone biopsies ofchildren, adolescents and young adults. Bone 2009;44:1043–8.

97S. Gamsjaeger et al. / Bone 69 (2014) 89–97

[23] Fratzl-Zelman N, Morello R, Lee B, Rauch F, Glorieux FH, Misof BM, et al. CRTAPdeficiency leads to abnormally high bone matrix mineralization in a murinemodel and in children with osteogenesis imperfecta type VII. Bone 2010;46:820–6.

[24] Lindahl K, Barnes AM, Fratzl-Zelman N, Whyte MP, Hefferan TE, Makareeva E, et al.COL1 C-propeptide cleavage site mutations cause high bone mass osteogenesisimperfecta. Hum Mutat 2011;32:598–609.

[25] Ben Amor IM, Edouard T, Glorieux FH, Chabot G, Tischkowitz M, Roschger P, et al.Low bone mass and high material bone density in two patients with Loeys–Dietzsyndrome caused by transforming growth factor beta receptor 2 mutations. J BoneMiner Res 2012;27:713–8.

[26] CheungM, Roschger P, Klaushofer K, Veilleux LN, Roughley P, Glorieux FH, et al. Cor-tical and trabecular bone density in X-linked hypophosphatemic rickets. J ClinEndocrinol Metab 2013;98:E954–61.

[27] Nawrot-Wawrzyniak K, Misof BM, Roschger P, Panczyk-Tomaszewska M,Ziolkowska H, Klaushofer K, et al. Changes in bone matrix mineralization aftergrowth hormone treatment in children and adolescents with chronic kidney failuretreated by dialysis: a paired biopsy study. Am J Kidney Dis 2013;61:767–77.

[28] Tamminen IS, Mayranpaa MK, Turunen MJ, Isaksson H, Makitie O, Jurvelin JS, et al.Altered bone composition in children with vertebral fracture. J Bone Miner Res2011;26:2226–34.

[29] Tamminen IS, Misof BM, Roschger P, Mayranpaa MK, Turunen MJ, Isaksson H, et al.Increased heterogeneity of bonematrix mineralization in pediatric patients prone tofractures: a biopsy study. J Bone Miner Res 2013;29:1110–7.

[30] Gamsjäger S, Kazanci M, Paschalis E, Fratzl P. Raman application in bone imaging. In:Amer M, editor. Raman spectroscopy for soft matter applications. Hoboken, NJ, USA:John Wiley & Sons, Inc.; 2009. p. 227–69.

[31] Morris MD, Mandair GS. Raman assessment of bone quality. Clin Orthop Relat Res2011;469:2160–9.

[32] Gamsjaeger S, Buchinger B, Zwettler E, Recker R, Black D, Gasser JA, et al. Bone ma-terial properties in actively bone-forming trabeculae in postmenopausal womenwith osteoporosis after three years of treatment with once-yearly zoledronic acid.J Bone Miner Res 2010;26:12–8.

[33] Hofstetter B, Gamsjaeger S, Phipps RJ, Recker RR, Ebetino FH, Klaushofer K, et al. Ef-fects of alendronate and risedronate on bonematerial properties in actively formingtrabecular bone surfaces. J Bone Miner Res 2012;27:995–1003.

[34] Gamsjaeger S, Brozek W, Recker R, Klaushofer K, Paschalis EP. Transmenopausalchanges in trabecular bone quality. J Bone Miner Res 2013;29(3):608–17.

[35] Kazanci M, Fratzl P, Klaushofer K, Paschalis EP. Complementary information onin vitro conversion of amorphous (precursor) calcium phosphate to hydroxyapatitefrom Raman microspectroscopy and wide-angle X-ray scattering. Calcif Tissue Int2006;79:354–9.

[36] Carden A, Morris MD. Application of vibrational spectroscopy to the study of miner-alized tissues (review). J Biomed Opt 2000;5:259–68.

[37] Carden A, Rajachar RM, Morris MD, Kohn DH. Ultrastructural changes accompanyingthe mechanical deformation of bone tissue: a Raman imaging study. Calcif Tissue Int2003;72:166–75.

[38] Paschalis EP, Tatakis DN, Robins S, Fratzl P, Manjubala I, Zoehrer R, et al. Lathyrism-induced alterations in collagen cross-links influence the mechanical properties ofbone material without affecting the mineral. Bone 2011;49:1232–41.

[39] Paschalis EP, Verdelis K, Doty SB, Boskey AL, Mendelsohn R, Yamauchi M. Spectro-scopic characterization of collagen cross-links in bone. J Bone Miner Res 2001;16:1821–8.

[40] Bansil R, Yannas IV, Stanley HE. Raman spectroscopy: a structural probe of glycos-aminoglycans. Biochim Biophys Acta 1978;541:535–42.

[41] Ellis R, Green E,Winlove C. Structural analysis of glycosaminoglycans and proteogly-cans by means of Raman microspectrometry. Connect Tissue Res 2009;50:29–36.

[42] Quilès F, Balandier J-Y, Capizzi-Banas S. In situ characterisation of a microorganismsurface by Raman microspectroscopy: the shell of Ascaris eggs. Anal Bioanal Chem2006;386:249–55.

[43] Zhang G, Moore D, Flach C, Mendelsohn R. Vibrational microscopy and imaging ofskin: from single cells to intact tissue. Anal Bioanal Chem 2007;387:1591–9.

[44] Akkus O, Polyakova-Akkus A, Adar F, Schaffler MB. Aging of microstructural com-partments in human compact bone. J Bone Miner Res 2003;18:1012–9.

[45] Fuchs RK, Allen MR, Ruppel ME, Diab T, Phipps RJ, Miller LM, et al. In situ examina-tion of the time-course for secondary mineralization of Haversian bone using syn-chrotron Fourier transform infrared microspectroscopy. Matrix Biol 2008;27:34–41.

[46] Fuchs RK, Faillace ME, Allen MR, Phipps RJ, Miller LM, Burr DB. Bisphosphonates donot alter the rate of secondary mineralization. Bone 2011;49:701–5.

[47] Gamsjaeger S, Masic A, Roschger P, Kazanci M, Dunlop JW, Klaushofer K, et al. Corti-cal bone composition and orientation as a function of animal and tissue age in miceby Raman spectroscopy. Bone 2010;47:392–9.

[48] Bala Y, Farlay D, Boivin G. Bone mineralization: from tissue to crystal in normal andpathological contexts. Osteoporos Int 2013;24:2153–66.

[49] Gourion-Arsiquaud S, Lukashova L, Power J, Loveridge N, Reeve J, Boskey AL. Fouriertransformed infra-red imaging of femoral neck bone: reduced heterogeneity ofmineral-to-matrix and carbonate-to-phosphate and more variable crystallinity in

treatment-naive fracture cases compared to fracture-free controls. J Bone MinerRes 2012;28(1):150–61.

[50] Paschalis EP, Glass EV, Donley DW, Eriksen EF. Bone mineral and collagen quality iniliac crest biopsies of patients given teriparatide: new results from the fracture pre-vention trial. J Clin Endocrinol Metab 2005;90:4644–9.

[51] Camacho NP, Rinnerthaler S, Paschalis EP, Mendelsohn R, Boskey AL, Fratzl P. Com-plementary information on bone ultrastructure from scanning small angle X-rayscattering and Fourier-transform infrared microspectroscopy. Bone 1999;25:287–93.

[52] Gao H, Ji B, Jager IL, Arzt E, Fratzl P. Materials become insensitive to flaws at nano-scale: lessons from nature. Proc Natl Acad Sci U S A 2003;100:5597–600.

[53] Gupta HS, Seto J, Wagermaier W, Zaslansky P, Boesecke P, Fratzl P. Cooperative de-formation of mineral and collagen in bone at the nanoscale. Proc Natl Acad Sci U S A2006;103:17741–6.

[54] Jager I, Fratzl P. Mineralized collagen fibrils: amechanicalmodel with a staggered ar-rangement of mineral particles. Biophys J 2000;79:1737–46.

[55] Boskey AL, DiCarlo E, Paschalis E, West P, Mendelsohn R. Comparison of mineralquality and quantity in iliac crest biopsies from high- and low-turnover osteoporo-sis: an FT-IR microspectroscopic investigation. Osteoporos Int 2005;16:2031–8.

[56] Fratzl P, Roschger P, Eschberger J, Abendroth B, Klaushofer K. Abnormal bone miner-alization after fluoride treatment in osteoporosis: a small-angle X-ray-scatteringstudy. J Bone Miner Res 1994;9:1541–9.

[57] Fratzl P, Paris O, Klaushofer K, Landis WJ. Bone mineralization in an osteogenesisimperfecta mouse model studied by small-angle X-ray scattering. J Clin Invest1996;97:396–402.

[58] Paschalis EP, Betts F, DiCarlo E, Mendelsohn R, Boskey AL. FTIR microspectroscopicanalysis of human iliac crest biopsies from untreated osteoporotic bone. Calcif TissueInt 1997;61:487–92.

[59] Durchschlag E, Paschalis EP, Zoehrer R, Roschger P, Fratzl P, Recker R, et al. Bonema-terial properties in trabecular bone from human iliac crest biopsies after 3- and 5-year treatment with risedronate. J Bone Miner Res 2006;21:1581–90.

[60] Masse PG, Rimnac CM, Yamauchi M, Coburn SP, Rucker RB, Howell DS, et al. Pyridox-ine deficiency affects biomechanical properties of chick tibial bone. Bone 1996;18:567–74.

[61] Masse P, Vuilleumier JP, Weiser H. Pyridoxine status as assessed by the concentra-tion of B6-aldehyde vitamers. Int J Vitam Nutr Res 1989;59:344–52.

[62] Masse P, Vuilleumier JP, Weiser H. Is pyridoxine an essential nutrient for bone? Int JVitam Nutr Res 1988;58:295–9.

[63] Oxlund H, BarckmanM, Ortoft G, Andreassen TT. Reduced concentrations of collagencross-links are associated with reduced strength of bone. Bone 1995;17:365S–71S.

[64] Oxlund H, Mosekilde L, Ortoft G. Reduced concentration of collagen reducible crosslinks in human trabecular bone with respect to age and osteoporosis. Bone 1996;19:479–84.

[65] Blank RD, Baldini TH, Kaufman M, Bailey S, Gupta R, Yershov Y, et al. Spectroscopi-cally determined collagen Pyr/deH-DHLNL cross-link ratio and crystallinity indicesdiffer markedly in recombinant congenic mice with divergent calculated bone tissuestrength. Connect Tissue Res 2003;44:134–42.

[66] Paschalis EP, Shane E, Lyritis G, Skarantavos G, Mendelsohn R, Boskey AL. Bone fra-gility and collagen cross-links. J Bone Miner Res 2004;19:2000–4.

[67] Boskey AL, Spevak L, Doty SB, Rosenberg L. Effects of bone CS-proteoglycans, DS-decorin, and DS-biglycan on hydroxyapatite formation in a gelatin gel. Calcif TissueInt 1997;61:298–305.

[68] Mochida Y, Duarte WR, Tanzawa H, Paschalis EP, Yamauchi M. Decorin modulatesmatrix mineralization in vitro. Biochem Biophys Res Commun 2003;305:6–9.

[69] Nielsen KL, Allen MR, Bloomfield SA, Andersen TL, Chen XD, Poulsen HS, et al.Biglycan deficiency interferes with ovariectomy-induced bone loss. J Bone MinerRes 2003;18:2152–8.

[70] Sauren YM, Mieremet RH, Groot CG, Scherft JP. An electronmicroscopic study on thepresence of proteoglycans in the mineralized matrix of rat and human compact la-mellar bone. Anat Rec 1992;232:36–44.

[71] Thompson WR, Modla S, Grindel BJ, Czymmek KJ, Kirn-Safran CB, Wang L, et al.Perlecan/Hspg2 deficiency alters the pericellular space of the lacunocanalicular sys-tem surrounding osteocytic processes in cortical bone. J Bone Miner Res 2011;26:618–29.

[72] Boskey AL, Posner AS, Lane JM, Goldberg MR, Cordella DM. Distribution of lipids as-sociated with mineralization in the bovine epiphyseal growth plate. Arch BiochemBiophys 1980;199:305–11.

[73] Boskey AL, Reddi AH. Changes in lipids during matrix: induced endochondral boneformation. Calcif Tissue Int 1983;35:549–54.

[74] Goldberg M, Boskey AL. Lipids and biomineralizations. Prog Histochem Cytochem1996;31:1–187.

[75] Raggio CL, Boyan BD, Boskey AL. In vivo hydroxyapatite formation induced by lipids.J Bone Miner Res 1986;1:409–15.

[76] Smith AJ, Singhrao SK, Newman GR, Waddington RJ, Embery G. A biochemical andimmuno-electron microscopical analysis of chondroitin sulphate-rich proteoglycansin human alveolar bone. Histochem J 1997;29:1–9.