New Method for Raman Investigation of the Orientationof Collagen Fibrils and Crystallites in the HaversianSystem of Bone

GUILLAUME FALGAYRAC,* SEBASTIEN FACQ, GERARD LEROY, BERNARD CORTET,and GUILLAUME PENELUniv Lille Nord de France, F-59000 Lille, France (G.F., S.F., G.L., B.C., G.P.); IMPRT IFR 114, EA 4032, Pathophysiology and Treatment of

Calcified Tissues, UDSL, F-59000 Lille, France (G.F., S.F., G.L., B.C., G.P.); and Department of Rheumatology, Hopital Roger Salengro, CHRULille, 59000, F-59000 Lille, France (B.C.)

Knowledge of the organization of the components of bone is of primary

importance in understanding how this tissue responds to stresses and

provides a starting point for the design and development of biomaterials.

Bone structure has been the subject of numerous studies. The mineralized

fiber arrangement in cortical bone is either a twisted or orthogonal

plywood structure. Both mineral models coexist in compact bone. Raman

polarized spectroscopy offers definite advantages in the study of biological

samples, enabling the simultaneous analysis of mineral and organic

components and the determination of molecular orientation through the

polarization properties of the Raman scattering. In this study, we used the

Raman polarization approach to simultaneously investigate the orienta-

tion of collagen fibrils and apatite crystals in human cortical bone. Raman

bands ratios were monitored as a function of sample orientation. Specific

ratios were chosen—such as m3 PO4/m1 PO4, amide III (1271 cm�1)/amide

III (1243 cm�1), and amide I/amide III (1243 cm�1)—due to their

sensitivity to apatite-crystal and collagen-fibril orientation. Based on this

original approach, spatial changes were monitored as a function of

distance from the Haversian canal. The results revealed simultaneous

tilting in intra-lamellar collagen-fibril and mineral crystal orientations.

These results are consistent with a twisted plywood organization in the

Haversian bone structure at the lamellar level. But at molecular level, the

co-alignment of the collagen fibrils and the apatite crystal is observed in

the innermost lamellae and becomes gradually less ordered as the distance

from the Haversian canal increases. This work highlights the interest of

Raman spectroscopy for the multiscale investigation of bone structure.

Index Headings: Raman spectroscopy; Polarization; Human bone;

Structure; Cortical.

INTRODUCTION

Knowledge of the organization of the components of bone—its organic and mineral content—is of primary importance inunderstanding how this tissue responds to exogenous andendogenous stresses (mechanical or chemical). Furthermore, itprovides a starting point for the design and development ofbiomaterials to replace bone or induce bone formation. Themineral component of bone is carbonated calcium phosphate.The organic component comprises 85–90% type I collagenfibrils, which provide a supporting matrix in which the mineralcrystals grow.1 In any study of bone properties, investigationsmust be conducted at several levels of organization to gain athorough understanding of the influence of structure andcomposition on these properties. Bone and osteon structurehave been the subject of numerous studies. Small-angle X-rayscattering (SAXS), neutron diffraction, transmission electronmicroscopy (TEM), and polarized light microscopy are the

main techniques that have been used to investigate bonestructure.2–8 Carbonated apatite crystals are oriented in thesame direction as collagen fibers in bone.8 During thedevelopment of skeletal systems, fibers are organized inprecise directions within the supporting tissue to achieveoptimal performance of their mechanical functions.9 It has beenproposed that the mineralized fiber arrangement in corticalbone is either a twisted or orthogonal plywood structure. Bothmodels coexist in compact bone.10

Vibrational spectroscopy is an increasingly popular tech-nique for investigating bone properties. Infrared and Ramanspectroscopies have been used to investigate compositionalchanges associated with various bone diseases and aging.11–16

While both techniques are suited to the study of bonecomposition and structure, Raman spectroscopy offers definiteadvantages. Raman spectra can be collected with micrometer-scale spatial resolution and gives simultaneous informationabout the mineral and organic-matrix components of bone.Furthermore, Raman spectroscopy is noninvasive and nonde-structive and as such is a promising technique for in vivoanalysis.17,18 Additionally, Raman scattering is stronglydependent on molecular orientation.19 Using polarized Ramantechniques, the dependence of Raman intensity on c-axisorientation has been demonstrated in human enamel crystal-lite.20–22 This approach can also provide new informationabout dental decay.23

Very few studies have been conducted using polarizedRaman techniques to assess human bone structure. Kazanci etal. explored modification of mineral/organic and carbonate/phosphate ratios as a function of polarization direction of theincident light at the micrometer scale.24 In a follow-up to thatstudy, they analyzed bone osteonal tissue in the longitudinaland transverse directions. This approach provided informationon variations in bone composition within the osteonal structure.They highlighted the importance of polarization direction withrespect to the structure.25 In this polarized Raman study, weidentified new characteristic Raman bands ratios that weresensitive to the mineral and collagen-fibril orientation ofhuman femoral bone. The aim of this study is to investigate therelative orientation of mineral and collagen fibril at themicrometric level within the Haversian system.

MATERIALS AND METHODS

Preparation of Samples. Animal specimen used in thisstudy was obtained under the agreement of the VeterinaryDepartment of the French Ministry of Agriculture (agreementNo: 59-350120). A 12 3 6 3 1 mm strip was cut from sheep (5

Received 4 January 2010; accepted 22 April 2010.* Author to whom correspondence should be sent. E-mail: Guillaume.falgayrac@univ-lille2.fr.

Volume 64, Number 7, 2010 APPLIED SPECTROSCOPY 7750003-7028/10/6407-0775$2.00/0

� 2010 Society for Applied Spectroscopy

years old, Colombia-Rambouillet, female) Achilles’ tendon andfixed in 4% formaldehyde. The strip was cut so as to reveal thelongitudinal orientation of collagen fiber. The z-axis and x-axiswere defined as the axes parallel and perpendicular to thecollagen fiber, respectively (Fig. 1).

Human specimen used in this study was obtained from thelaboratory of anatomy of the Medical School of the Universityof Lille (France) under the agreement of the ethics committee.A bone section was obtained from the cadaver of a healthy 69-year-old male. A femoral sample was cut from the middle ofthe diaphysis and fixed in 4% formaldehyde at roomtemperature. A longitudinal section was cut to prepare a 12 36 3 1 mm sample, which was ground using decreasing grainsizes (30, 3, 0.3 lm, Abrasives Center) and stuck with Araldite(Huntsman Advanced Materials Bale Switzerland) on amicroscope slide. The bottom section of Fig. 2 representshow the sample was oriented with respect to the z- and x-axes.The z-axis and x-axis were defined as the axes parallel andperpendicular, respectively, to the long axis of the femur.

Raman Spectroscopy. Raman spectroscopy was performedusing a Labram confocal microspectrometer (HORIBA Gr,Jobin Yvon, Lille, France). The Raman signal was processedby a spectrograph equipped with an air-cooled charge-coupleddevice (CCD) detector. The Raman spectra were excited usinga helium–neon laser (632.8 nm) with an output of 8 mWreaching the sample. The 1003 microscope objective (numer-ical aperture¼ 0.80) gave a spot size close to one micrometer.The spectral resolution was 2 cm�1. Spectral acquisitions wereperformed in the 800 to 1700 cm�1 range. For each spectrum,integration time was 100 s and 10 accumulations wereperformed. Acquisition data were processed using the LabSpecsoftware (HORIBA Gr, Jobin Yvon, Lille, France). Back-ground fluorescence was subtracted from the baseline compu-tation. A polynomial filtering was performed.

Three types of arrangements were used for spectralacquisition.

First, Raman analyses were carried out on sheep tendon andhuman femur to determine bands that were sensitive to theincident light polarization. Polarized Raman spectra wereacquired in two configurations, longitudinal and transversal,corresponding to polarization orientations that were parallel

(zz) and perpendicular (xx) to the z-axis, respectively. The zz(or xx) orientation corresponded to a polarization of theincident light in the z (or x) direction and collected scatteredlight polarized in the z (or x) direction.

Second, other acquisitions were carried out on the femursample with a goniometer mounted on the microscope stage toevaluate intensity variations in Raman bands as a function ofangular variations in the sample at the same point. Acquisitionswere made at different angles, from 08 to 1808 by steps of 108.The 08 and 908 angles corresponded to the zz and xxconfigurations, respectively. These orientations are shownschematically in Figs. 2a and 2b, respectively.

Finally, the femoral bone was mounted on the XY motorizedstage. Raman spectra were acquired over 35 lm along the x-axis by steps of 1 lm, from the Haversian canal to the rim ofthe osteon. This distance permitted covering an average of fourlamellae. The polarization direction of the incident light waskept longitudinal during all acquisitions. Variations in Ramanband intensity ratios were monitored along the x-axis. Tendifferent osteons were analyzed using this procedure.

The Raman signal is affected by both the composition andorientation of the sample. Because bone is an especiallyheterogeneous calcified tissue, direct comparison of bandintensities is hazardous. On the other hand, Raman bandintensity ratios of the same compound (mineral or organic) areprincipally affected by the polarization effect and minimally byvariations in composition. As such, Raman band intensityratios for the mineral and organic matrix were calculated andstudied as a function of sample orientation and location.

Statistical Analysis. Linear regressions were performed toassess the significance of relationships between intensity ratiovariations and the distance from the Haversian canal.Regressions were performed for each lamellae and werepooled over the ten osteons. The relation was consideredsignificant at the level of p � 0.05.

FIG. 1. 12 3 6 3 1 mm strip of sheep tendon cut in two for Raman analysis.The picture shows the longitudinal orientation of collagen fiber.

FIG. 2. Representation of the experimental setup carried out on the femoralsample as a function of sample orientation: (a) zz orientation; (b) xxorientation. The lower part shows the localization of the bone sample and howit is oriented as a function of the z- and x-axis. (BS) Beam splitter; (LP) linearpolarizer; (Obj) objective; and (To Spec.) to spectrograph.

776 Volume 64, Number 7, 2010

RESULTS

Raman Acquisitions with xx and zz PolarizationDirections of the Incident Light. Sheep Tendon. Ramanspectra acquired in the zz and xx configurations are shown inFig. 3. The Raman spectra exhibit characteristic bands of type Icollagen. The amide I position corresponding to the C¼Ostretching mode of the peptide linkage is observed at 1669cm�1. The CH2 deformation is represented by a Raman band at1451 cm�1. Amide III bands (1243 and 1271 cm�1) arise fromthe combination of N–H bending and C–N stretching of thepeptide group.19,26 The C–C stretching vibrations of the proteinbackbone can be observed at 920 and 937 cm�1.27 The band at1004 cm�1 is identified as a phenylalanine band.18

Comparison of the spectra revealed Raman bands sensitiveto light polarization orientation: 937, 1271, and 1669 cm�1.The amide III band at 1243 cm�1 remained constant. Othersbands displayed a slight dependence on light polarizationorientation.

Human Femur. Identical data were obtained for all theprobed osteons. Typical spectra of both polarizations arepresented in Fig. 4. The principal Raman bands from bonewere identified.25 The m1 phosphate vibration at 960 cm�1 isthe strongest marker for bone mineral. The m3 phosphatevibrations are visible at 1032 and 1046 cm�1. The strong bandat 1071 cm�1 indicates type B carbonate substitution in thebone specimen.28 The band at 1004 cm�1 is identified as a

phenylalanine band. In this configuration, the variation in lightpolarization orientation does not influence the intensity of them1 phosphate vibration. The m3 phosphate vibrations areinfluenced by the polarization direction. The Raman bandintensities at 1032 cm�1 and 1046 cm�1 increase and decrease,respectively, when the polarization direction switches from thezz to the xx configuration. Bands characteristic of collagen, asdescribed above, are also observed. Amide III (1273 cm�1) andamide I (1669 cm�1) exhibited the same behavior as a functionof the light polarization orientation. Their intensities increasedwhen the polarization direction of the incident light switchedfrom the zz to the xx polarization configuration.

The intensities of characteristic bands from the organic andmineral matrix depended on sample orientation (1271, 1669cm�1 and 1046 cm�1, respectively). In both precise configu-rations (zz and xx), Raman bands remained constant, i.e., 960cm�1 for the mineral22,24 and 1244 cm�1 for the organic matrix.

Raman Acquisitions as a Function of Angular Variationsof the Sample. The Raman band intensity variations wereconsistent with results obtained from previous polarizedRaman spectroscopy studies.22,24 Intensity ratios were studiedas a function of sample orientation from 08 to 1808 by steps of108 (Fig. 5). The study focused on the intensity ratios 1271/1243 and 1669/1243 corresponding to the organic matrix, and1046/960 corresponding to the mineral component. The 1243and 960 cm�1 bands were used as references because of their

FIG. 3. Raman spectra of sheep tendon as a function of polarization orientation; xx¼ transversal orientation, zz¼ longitudinal orientation.

FIG. 4. Raman spectra of human femur as a function of polarization orientation; xx¼ transversal orientation, zz¼ longitudinal orientation.

APPLIED SPECTROSCOPY 777

low sensitivity to sample orientation (Fig. 4). Intensity valueswere distributed symmetrically around the 908 angle. The 08 (or1808) orientation corresponded to the Raman spectrum of thefemur in the zz polarization configuration. The 908 orientation

corresponded to the Raman spectrum of the femur in the xx

polarization configuration (Fig. 4). The intensity ratios 1271/

1243 and 1669/1243 (organic matrix) increased until the 908

acquisition. The 1046/960 ratio (mineral crystallites) exhibited

the opposite behavior in the same angular range (Fig. 5). A

reduction in signal intensity ratio between the two principal

orientations was observed in the 0.2–0.3 range.

Intensity Ratio Variations as a Function of HaversianDistance. Intensity ratios were measured as a function of

distance from the Haversian center to the rim of the osteon.

Figures 6a and 6b are representative of intensity ratio variations

among ten different osteons. Analyses were performed along a

length of 35 lm, which covered an average of four lamellae of

the osteon. Along the 35 lm scanning line organic ratios

dropped to their minimum value, whereas mineral ratios

attained their maximum value. This sequence was observed

periodically at intervals of 6–8 lm.

Correlation between intensity ratio and the distance from the

Haversian canal was determined using linear regression.

Regressions were performed on each lamellae after data were

pooled over ten osteons (Table I). The lamellas named No. 1

and No. 4 represent the innermost and the outermost lamellae.

FIG. 5. Intensity ratios of Raman bands 1271/1243, 1669/1243, and 1046/960as a function of sample orientation with respect to the longitudinal polarizationdirection of the incident light.

FIG. 6. (a) Optical image of the innermost lamellae from the Haversian canal; (b) representative Raman acquisitions performed every micrometer along a 35 lmline; intensity ratios of Raman bands 1271/1243, 1669/1243, and 1046/960 obtained from linear scanning. The last intensity ratio is multiplied by 5 for easiercomparison. (c) Schematic representation of the oscillations in intra-lamellar collagen-fibril and mineral-crystal orientations.

778 Volume 64, Number 7, 2010

Table I shows a decrease of coefficient correlation between theinnermost and the outermost lamellae.

DISCUSSION

Bone and osteon structure have been the subject ofnumerous studies. It has long been known that bone tissueconsists of collagen fibrils reinforced with nano-sized mineralparticles. These crystallites have been described as beingstrongly oriented in the direction of the collagen fibrils.29,30 X-ray diffraction experiments on corresponding specimenpositions have revealed that the particle orientations detectedby SAXS coincide with the long axis of the hydroxyapatitecrystals.7,8,31,32 In the case of longitudinal samples, crystallitesand collagen fibrils were oriented in a direction parallel to theHaversian canal.33,34 The results of the present work areconsistent with a twisted plywood organization in theHaversian bone structure at the lamellar level.

Wagermaier et al.35 have shown that osteons, at themicrostructural level, are adapted to the in vivo mechanicalstresses in bone. By application of scanning X-ray diffractionwith a micrometer-sized synchrotron beam, they reconstructedthe three-dimensional orientation of the mineralized fibrilswithin a single osteon lamella at micrometer scale. Measure-ments revealed that the orientation of the hydroxyapatitecrystals varied from 108 to 608 in the innermost lamella, andthis was repeated from one lamella to another. Thisarrangement becomes gradually less ordered as the distancefrom the Haversian canal increases.

In polarized Raman spectroscopy, polarized Raman scatter-ing is observed as a result of interference with the polarizedlight by the vibration of molecules. More precisely, theintensity of a Raman vibrational mode of a molecule dependson the differential of the polarizability tensor and on thegeometrical properties of the molecules constituting thesample.24 In highly oriented samples, such as crystallinesamples, the spectral intensity depends on the orientation of thecrystal axis with respect to the polarization of the incident light.

Kazanci et al. employed Raman polarized measurements andmicro-imaging techniques to study the orientation compositionof an osteon from the femoral midshaft.24 They showed thatRaman intensities depend on two factors: (1) the chemicalcomposition of the tissue, and (2) the orientation of themineralized fibrils with respect to the linear polarizationdirection of the exciting beam. Their results showed that m1

PO4/amide I, used for the calculation of material properties,was influenced by the polarization direction of the incidentlight and might lead to erroneous conclusions. They suggestedtaking into consideration other Raman intensity ratios that areless sensitive to the polarization effect. In a more recent study,Kazanci et al. carried out Raman polarized measurements on an

osteon in the longitudinal and transversal directions.25 Thisapproach allowed them to separate out orientation andcomposition. They showed that m1 PO4/amide I mainlydisplayed lamellar orientation, and that m2PO4/amide III andm1CO3/m2PO4 displayed variations in bone composition. Thepresent work focused on the polarization effects on Ramanband ratios from mineral content (m3 PO4/m1 PO4) and organiccontent (amide III (1271 cm�1)/amide III (1243 cm�1) andamide I/amide III (1243 cm�1)). These ratios were found to besimultaneously dependent on the molecular orientation of bothmineral and collagen components within single lamellae.

Figure 5 plots the Raman intensity ratios for mineral (1046/960) and organic (1271/1243 and 1669/1243) content as afunction of sample orientation with respect to the linearpolarization of the incident beam. Raman intensities reachedtheir maximum in the zz orientation, without completeextinction in the xx orientation (20–30 % lower). In otherwords, the molecular orientation was not strictly parallel to thezz direction. Both collagen fibrils and mineral crystallites werenot strictly orientated in the longitudinal bone axis. Theintensity variations revealed the simultaneous tilt of intra-lamellar collagen–fibril and mineral–crystal orientations. Or-ganic and mineral ratios were then calculated as a function ofdistance from the center of the Haversian canal to the rim of theosteon (Fig. 6b). The polarization direction was keptlongitudinal for each measurement. Ratios varied with a periodof 6–8 lm, which corresponded to the thickness of one lamella.Orientation variations were observed in the innermost lamellaeand diminished as the distance from the Haversian canalincreased. This observation is confirmed by statistical analysis.Correlation coefficients between intensity ratio and distancefrom the Haversian canal diminish strongly between theinnermost and the outermost lamellae. A schematic represen-tation of the oscillations in intra-lamellar collagen-fibril andmineral-crystal orientations are presented in Fig. 6.

CONCLUSION

The present study capitalizes on the advantages of Ramanmicrospectroscopy—i.e., high spatial resolution, simultaneouscharacterization of mineral and organic compounds, andinfluence of molecular orientation—to simultaneously observeorientation oscillations in collagen fibrils and mineral crystalsat the lamellar scale. In comparison with previous polarizedRaman studies, this work reveals new characteristic Ramanband ratios of mineral and organic content. This originalapproach brings new information about the relative orientationof mineral and collagen of the femoral cortical bone. The co-alignment of the collagen fibrils and the apatite crystals withinsingle lamellae is no longer respected as the distance from theHaversian canal increases.

ACKNOWLEDGMENTS

The authors would like to thank Emilie De Paepe for her contributionconcerning the figures and Pr. Jean-Paul Francke and the staff of the departmentof anatomy of the Medical School of the University of Lille who kindlyprovided the bone sample.

1. S. Weiner, W. Traub, and H. D. Wagner, J. Struct. Biol. 126, 241 (1999).2. A. Bigi, A. Cacchioli, A. M. Fichera, C. Gabbi, M. H. J. Koch, L.

Ragionieri, A. Ripamonti, and N. Roveri, J. Biomed. Mater. Res. 41, 289(1998).

3. S. Hong, S. Hong, and D. Kohn, J. Mater. Sci.: Mater. Med. 20, 1419(2009).

TABLE I. Correlation coefficient evaluated from linear regression onthe intensity ratio for each lamellae over 10 osteons. Signs (�) and (þ)stand for negative or positive correlation respectively; **¼p , 0.0001; *¼p , 0.05. Lamellae No. 1 and No. 4 correspond to the innermost and theoutermost lamellae from the Haversian canal, respectively.

Ratio Lamellae No. 1 Lamellae No. 2 Lamellae No. 3 Lamellae No. 4

1045/960 (�) 0.538** (�) 0.517** (�) 0.319* (�) 0.212*1271/1243 (þ) 0.549** (þ) 0.479** (þ) 0.149 (þ) 0.274*1669/1243 (þ) 0.406* (þ) 0.379* (þ) 0.275* (þ) 0.248*

APPLIED SPECTROSCOPY 779

4. M. A. Rubin and I. Jasiuk, Micron 36, 653 (2005).5. N. Sasaki, N. Matsushima, T. Ikawa, H. Yamamura, and A. Fukuda, J.

Biomech. 22, 157 (1989).6. G. B. Timothy, M. G. Haviva, C. M. Shannon, W. Johanna, B. Alan, and

M. R. Christopher, Anat. Rec. B, New Anat. 274, 157 (2003).7. I. Zizak, O. Paris, P. Roschger, S. Bernstorff, H. Amenitsch, K.

Klaushofer, and P. Fratzl, J. Appl. Crystallogr. 33, 820 (2000).8. I. Zizak, P. Roschger, O. Paris, B. M. Misof, A. Berzlanovich, S.

Bernstorff, H. Amenitsch, K. Klaushofer, and P. Fratzl, J. Struct. Biol. 141,208 (2003).

9. A. C. Neville, Biology of Fibrous Composite: Development Beyond theCell Membrane (Cambridge University Press, Cambridge, 1993).

10. M.-M. Giraud-Guille, L. Besseau, and R. Martin, J. Biomech. 36, 1571(2003).

11. C. Krafft, D. Codrich, G. Pelizzo, and V. Sergo, J. Raman Spectrosc. 40,595 (2009).

12. A. Boskey, E. DiCarlo, E. Paschalis, P. West, and R. Mendelsohn,Osteoporos. Int. 16, 2031 (2005).

13. M. E. Ruppel, D. B. Burr, and L. M. Miller, Bone 39, 318 (2006).14. J. A. Timlin, A. Carden, M. D. Morris, R. M. Rajachar, and D. H. Kohn,

Anal. Chem. 72, 2229 (2000).15. J. A. Timlin, A. Carden, and M. D. Morris, Appl. Spectrosc. 53, 1429

(1999).16. S. Gourion-Arsiquaud, J. C. Burket, L. M. Havill, E. DiCarlo, S. B. Doty,

R. Mendelsohn, M. C. H. van der Meulen, and A. L. Boskey, J. BoneMiner. Res. 24, 1271 (2009).

17. P. Matousek, E. R. C. Draper, A. E. Goodship, I. P. Clark, K. L. Ronayne,and A. W. Parker, Appl. Spectrosc. 60, 758 (2006).

18. G. Penel, C. Delfosse, M. Descamps, and G. Leroy, Bone 36, 893 (2005).19. J. R. Ferraro and K. Nakamoto, Introductory Raman Spectroscopy

(Academic Press, London, 1994).

20. H. Tsuda and J. Arends, J. Dent. Res. 73, 1703 (1994).

21. G. Leroy, N. Leroy, G. Penel, C. Rey, P. Lafforgue, and E. Bres, Appl.

Spectrosc. 54, 1521 (2000).

22. G. Leroy, G. Penel, N. Leroy, and E. Bres, Appl. Spectrosc. 56, 1030

(2002).

23. A. C. Ko, M. Hewko, M. G. Sowa, C. C. Dong, B. Cleghorn, and L.-P. i.

Choo-Smith, Opt. Exp. 16, 6274 (2008).

24. M. Kazanci, P. Roschger, E. P. Paschalis, K. Klaushofer, and P. Fratzl, J.

Struct. Biol. 156, 489 (2006).

25. M. Kazanci, H. D. Wagner, N. I. Manjubala, H. S. Gupta, E. Paschalis, P.

Roschger, and P. Fratzl, Bone 41, 456 (2007).

26. J. Chalmers and P. Griffiths, Handbook of Vibrational Spectroscopy(Wiley, New York, 2002).

27. B. Wopenka, A. Kent, J. D. Pasteris, Y. Yoon, and S. Thomopoulos, Appl.

Spectrosc. 62, 1285 (2008).

28. A. Awonusi, M. D. Morris, and M. M. J. Tecklenburg, Calcif. Tissue Int.

81, 46 (2007).

29. S. Weiner and H. D. Wagner, Annu. Rev. Mater. Sci. 28, 271 (1998).

30. W. J. Landis, K. J. Hodgens, J. Arena, M. J. Song, and B. F. McEwen,

Microsc. Res. Tech. 33, 192 (1996).

31. P. Fratzl, S. Schreiber, and K. Klaushofer, Connect. Tissue Res. 34, 247

(1996).

32. S. Rinnerthaler, P. Roschger, H. F. Jakob, A. Nader, K. Klaushofer, and P.

Fratzl, Calcif. Tissue Int. 64, 422 (1999).

33. M.-G. Ascenzi, A. Ascenzi, A. Benvenuti, M. Burghammer, S. Panzavolta,

and A. Bigi, J. Struct. Biol. 141, 22 (2003).

34. A. Rindby, P. Voglis, and P. Engstrom, Biomaterials 19, 2083 (1998).

35. W. Wagermaier, H. S. Gupta, A. Gourrier, M. Burghammer, P. Roschger,

and P. Fratzl, Biointerphases 1, 1 (2006).

780 Volume 64, Number 7, 2010