Calcium Phosphate to Hydroxyapatite from Raman ... - Springer Link

Calcif Tissue Int (2006) 79:354 359 DOI: 10.1007/s00223-006-0011-9

Complementary Information on In Vitro Conversion of Amorphous (Precursor) Calcium Phosphate to Hydroxyapatite from Raman Microspectroscopy and Wide-Angle X-Ray Scattering M. Kazanci,1 P. Fratzl,1 K. Klaushofer,2 E. P. Paschalis2 1 2

Department of Biomaterials, Max Planck Institute of Colloids and Interfaces, 14424 Potsdam, Germany Ludwig Boltzmann Institute of Osteology at the Hanusch Hospital, 4th Medical Department, Hanusch Hospital, A-1140 Vienna, Austria

Received: 10 January 2006 / Accepted: 26 July 2006 / Online publication: 17 November 2006

Abstract. In addition to mechanical functions, bones have an essential role in metabolic activity as mineral reservoirs that are able to absorb and release ions. Bioapatite, considered the major component in the mineralized part of mammalian bones, is a calcium phosphate mineral with a structure that closely resembles hydroxyapatite (HA, Ca10[PO4]6[OH]2) with variable chemical substitutions. It is important to note that it continues to be chemically active long after it has been initially deposited. Detailed understanding of changes in the mineral phase as HA matures is essential for understanding how normal bone achieves its remarkable mechanical performance, how it is altered in disease, as well as the effects of therapeutic interventions. A model system for investigation of the in vivo maturation of HA is available, namely, the in vitro conversion of amorphous calcium phosphate (ACP) to HA in a supersaturated solution of calcium and phosphate ions. In the present study, this system was employed to correlate with the changes in chemistry and poorly crystalline HAP crystal size, shape, and habit. The results of the Xray diffraction as well as Raman analyses showed that as the crystallites mature in the 002 and 310 directions both the full width at half-height and wavelength at maximum of the Raman peaks change as a function of reaction extent and crystallite maturation, size, and shape. Moreover, such analyses can be performed in intact bone specimens through Raman microspectroscopic and imaging analyses with a spatial resolution of 0.6 1 l, by far superior to the one offered by other microspectroscopic techniques, thus potentially yielding important new information on the organization and mineral quality of normal and fragile bone. Key words: Amorphous calcium phosphate — Hydroxyapatite — Raman microspectroscopy — Wideangle X-ray scattering

Mineralized tissues consist mainly of poorly crystalline hydroxyapatite (HA, Ca10[PO4]6[OH]2). It is important to note that it continues to be chemically active long Correspondence to: P. Fratzl; E-mail: [email protected]

after it has been initially deposited [1]. Consequently, a heterogeneous distribution of mineral crystallites (in terms of size, shape, and maturity) is expected to be encountered in normal bone, dependent on remodeling rates. This has been shown in microspectroscopic studies of thin tissue sections, employing Fourier transform infrared (FTIR) microspectroscopic and imaging analyses with an optimal spatial resolution of 10 6.3 lm [2 4]. Fairly recently, another type of microspectroscopic analysis has been employed to study the spatial and temporal variation of bone mineral maturity and crystallinity, namely Raman microspectroscopy [5 11]. Despite the fact that infrared spectra have a superior signal-to-noise ratio, Raman analysis offers definite advantages, such as that tissues may be analyzed without the need of prior dehydration, embedding, and cutting of thin sections. It also offers superior spatial resolution (0.6 1 lm) compared to the existing infrared techniques (5 10 lm), enabling analysis of biologically important localities such as individual cement lines, individual lamellae, boundaries around microcracks, human dentin tubules, etc. [12 14]. Raman spectroscopy has been employed to study a wide variety of tissues to determine the composition, relative amounts, locations of the tissue matrix, mineral components, degree of collagen mineralization, mineral crystallinity, and presence of nonstoichiometric carbonate and hydroxide substitutions in the mineral [8 11, 15, 16]. Detailed understanding of changes in the mineral phase as HA matures is essential for understanding how normal bone achieves its remarkable mechanical performance, how it is altered in disease, as well as the effects of therapeutic interventions. A model system for investigation of the in vivo maturation of HA is available, namely the in vitro conversion of amorphous (precursor) calcium phosphate (ACP) to HA in a supersaturated solution of calcium ions with phosphate ions [3, 4]. It has been previously used for the training of infrared microspectroscopic techniques, yielding useful spectral

M. Kazanci et al.: In Vitro Conversion of ACP to HA

355

Table 1. 002 and 310 line broadening analysis of ACP Time(minutes)

W(real)(rad)(002)

W(real)(rad)(310)

L002 (nm)

L310 (nm)

10 20 30 60 90 120 150 240 960 1,440 2,880

) ) )

) ) ) ) ) 28·10)3 32·10)3 31·10)3 28·10)3 29·10)3 31·10)3

) ) ) 17.56 17.17 20.25 21.07 23.24 25.9 27.43 29.26

) ) ) ) ) 5.86 5.125 5.29 5.86 5.66 5.29

9·10)3 9.2·10)3 7.8·10)3 7.5·10)3 6.8·10)3 6.1·10)3 5.8·10)3 5.4·10)3

parameters that describe the variation of mineral maturity and, by correlation, crystallinity. In the present study, this in vitro model system was employed to expand on the Raman spectral parameters reported to date [4], which correlate with the changes in chemistry and poorly crystalline HAP crystal size, shape, and habit.

(Bruker, Karlsruhe, Germany) with h/h geometry, 2 mm beam size, and a two-dimensional detector (Bruker) with 1,024 · 1,024 pixels. Frames of data were collected for 20 minutes each, integrated, and merged to produce a diffraction pattern from 20 to 70 2h. The sample-to-detector distance was 1.3 m. DIFFRACplus (Bruker) EVA 7.0 software was employed for data analysis. Data Analysis

Materials and Methods Synthetic Apatites Details on the preparation of ACP, which converts to poorly crystalline HA, have been published elsewhere [3]. The sample size was 10, and the average values for each time point were utilized. Briefly, 1,400 mL of 0.022 M CaCl2 and 600 mL of 0.036 M Na2HPO4, both prepared in 0.15 M Tris buffer, were mixed with constant stirring at room temperature. The reaction was carried out for a maximum of 2 days, at room temperature and in open air (as a result, carbonate was incorporated into the crystals) [3, 4]. Precipitation occurred upon mixing. The solution was stirred constantly, and aliquots were withdrawn at 10, 20, 30, 60, 90, 120, 150, 240, 960, and 2,880 minutes. Precipitates were filtered through a mediumporosity fritted glass filter, washed with ammoniated (pH 10) water to remove unreacted salts while preventing dissolution and reprecipitation, and lyophilized for 24 hours. The powder samples were analyzed by means of X-ray diffraction (XRD) and Raman spectroscopic methods. Raman Spectroscopic Measurements Raman spectra of ACP were taken at room temperature using a confocal Raman microscope (CRM200; WITec, Ulm, Germany) equipped with a piezo scanner (P-500; Physic Instrumente, Karlsruhe/Palmbach, Germany) and high numerical apertures (NA) microscope objectives (·100, NA = 0.90; Nikon, Tokyo, Japan). In a typical experiment, diode-pumped green laser, k = 532 nm (CrystaLaser, Reno, NV, USA) was focused on the material with a diffraction-limited spot size (diameter approximately k/2). The spectra were taken using an air-cooled charge-coupled device (PI-MAX; Princeton Instruments, Evry, France) behind a grating (1,800 g Æ mm)1) spectrograph (Acton) with a resolution of 3 Æ cm)1. X-Ray Analysis Powder XRD was done in microdiffraction mode with CuKa radiation (k = 1.5418 A˚) on a DISCOVER D8 diffractometer

The Raman and XRD results were averaged, baseline-corrected for each band, and smoothed (using Sav-Golay function); in order to analyze the m3PO4, HPO4, and m1CO3 band positions, the broad band in the same region was divided into subbands using GRAMS32 software (Galactic Industries) and underline peaks were fitted. The ratio of the areas was calculated. Determination of Particle Size Particle size was calculated by Scherrer’s equation [17, 18]. The details are given elsewhere [18]. Based on Scherrer’s equation, the ACP average particle size using the (002) reflection peak from the XRD pattern was calculated to monitor the change in the c-crystallographic axis as a function of time (extent of conversion). Results

The results of XRD analyses are summarized in Table 1 and Figure 1. The average sizes (based on 002 reflection at 26 and the 310 reflection at 40) are listed as a log function of reaction extent. The sizes became longer in the 002 direction throughout the experiment, while they were almost constant in the 310 direction (Fig. 1a). Although Raman values were obtained starting from 10 minutes, 002 diffraction lines were observed right after the band sharpening began at 60 minutes at the nucleation stage (before 60 minutes, the structure was completely amorphous), and 310 diffraction lines were observed after 120 minutes (in the crystalline region). It is easer to detect the sharp peaks than the broader ones. The full widths at half height (FWHH) and band shifts of the m1CO3, m2PO4, m4PO4, m1PO4, m3PO4, HPO4 and shoulder of m1PO4 peaks (nonlinear left portion of m1PO4 band) were monitored as a function of time, and the results are summarized in Table 2. As may be seen,

356


30

these changes are directly (+) or inversely ()) proportional are shown. The stronger correlations (arbitrarily chosen for r2 ‡ 0.5) are listed in bold. Very strong correlations were observed between 002 diffraction peak and Raman shifts and FWHH of bands. The 310 diffraction peak is almost constant at the crystalline stage; therefore, it correlates with Raman bands that show a stable trend in the same region.

25

(a) 002

Nucleation stage

20 Crystalline stage

15 10

(b) 310

Amorphous precursor phase stage

5

Discussion

0 10

100

1000

log(time(min)) Fig. 1. Evolution of ACP particles: (a) length of crystals from 002 diffraction peaks and (b) width of crystal from 310 diffraction peaks as a function of reaction time.

all FWHH values decreased within the first 100 minutes of reaction, after which they remained constant. m1PO4, m4PO4, HPO4, and shoulder of m1PO4 band mass centers shifted to higher energy levels, whereas m3PO4 and m1CO3 band mass centers shifted to lower energy levels and the m2PO4 band mass center remained almost constant throughout the reaction. Figure 2 lists two typical Raman spectra at the beginning (10 minutes) and end (2,880 minutes) of the reaction. As may be seen, the spectral baseline noise is reduced as a function of reaction extent, accompanied by a sharpening of all peaks. The precise wavelength at maximum height of the bands was also recorded as a log function of time, and the results are presented in Figure 3. In this case, the m1PO4 and m4PO4 bands were sharply shifted between 60 and 90 minutes of reaction, which corresponds to the nucleation stages at the X-ray crystallography, after which they assumed relatively constant values. The HPO4 and m2PO4 bands remained constant throughout the experiments. Figure 4 shows the band ratio of different components (a, HPO4 to different PO4 lines and, b, CO3 to different PO4 lines). The maturation of the particles causes a substantial increase in the CO3 to HPO4 ratio and a decrease in the HPO4 to PO4 ratio. The results of XRD and Raman analyses showed that as the crystallites matured, both the FWHH and wavelength at maximum of the Raman peaks shifted as a function of reaction extent. XRD peaks and Raman bands displayed a very strong correlation (linear regression). The correlation between these changes was recorded and the changes in Raman peak FWHH and wavelength at maximum vs. changes in average crystallinity (based on 002 and 310 reflection) are listed in Table 3. More specifically, the r2 values and whether

Normal bone is a hierarchical structure, exhibiting remarkable mechanical behavior. Important information may be obtained from all hierarchical levels. Nowadays, it is widely accepted that both bone quantity and quality contribute to bone strength [19]. Although bone quantity has been measurable for quite some time now (dual-energy X-ray absorptiometry), determination of bone quality is still rather challenging. The major bone constituents are mineral and collagen. One of the obstacles to be circumvented when assessing mineral and matrix tissue properties is tissue heterogeneity at the microscopic level, due to bone remodeling [20]. In addition, both components undergo extensive chemical changes after their initial synthesis and deposition [21]. Mineral crystallites (poorly crystalline HA) keep changing their chemistry (maturity) and size (crystallinity). The importance of mineral crystallinity in determining bone strength is evident in the case of fluoride-treated bone and the resulting fragility despite increased bone mineral density [22]. The rather recent availability of instruments such as micro-computed tomography, small angle X-ray scattering (SAXS), FTIR, and Raman microspectroscopic instruments allows the determination of both structural and material properties of bone at the ultrastructural level [19]. Raman microspectroscopy and imaging analysis offers information on the mineral and organic matrix spatial distribution with a spatial resolution of 1 lm [8, 14]. In the present study, we employed Raman analysis of the well-established chemical conversion of ACP to HAP, a chemical process that results in the in vitro production of mineral crystallites similar in chemistry and size distribution to the ones present in bone [4]. Precipitation in vitro often involves the initial formation of a disordered, more soluble phase that subsequently transforms into a less soluble and usually more ordered form. Termine and Posner [23] proposed that ACP may be a precursor phase in bone formation. In 1972, Brecevic and Furedi-Milhofer [24] showed that in vitro the first-formed phase is ACP. This subsequently transforms into octacalcium phosphate (OCP) and finally into carbonate apatite. Crane et al. [25] observed a m1PO4 955 Æ cm)1 band that was attributable to an OCP-like mineral.


357

Table 2. Peak positions and FWHH of Raman bands relative to reaction time m1PO4

m2PO4

m3PO4

m4PO4

m1CO3

m1PO4 shoulder

HPO4

Time (minutes)

Shift

FW HH

Shift

FW HH

Shift

FW HH

Shift

FW HH

Shift

FW HH

Shift

FW HH

Shift

FW HH

10 20 30 60 90 120 150 240 960 1,440 2,880

951 951 951 951 955 959 960 961 960 960 960

48 49 49 49 47 45 43 43 42 43 41

430 430 428 429 430 429 430 431 431 431 433

61 62 61 61 55 55 52 52 52 53 52

1,058 1,050 1,048 1,047 1,038 1,045 1,042 1,038 1,044 1,039 1,042

48 62 55 61 44 42 38 34 34 34 35

573 573 570 587 593 590 591 594 592 592 593

90 87 74 78 69 55 53 54 48 49 50

1,084 1,079 1,078 1,081 1,075 1,079 1,078 1,071 1,075 1,071 1,074

46 49 50 50 49 50 49 53 54 52 52

1,016 1,012 1,011 1,013 1,015 1,019 1,019 1,018 1,022 1,019 1,019

42 37 36 29 20 21 18 16 20 17 16

884 880 886 886 891 896 898 894 894 895 894

42 45 40 39 42 47 45 42 42 46 43

PO4 ν1 symmetric stretch PO4 ν2 symmetric bend modes

PO4 ν4 antisymmetric bend modes

PO4 ν3 antisymmetric stretch

CO3 ν1 symmetric stretch

2880 min.

HPO4 symmetric stretch

10 min

400

600

800

1000

Raman shift (cm-1)

Associated with it is a 1,010 Æ cm)1 band that is a P-O stretch of monohydrogen phosphate and is also found in OCP [25]. In our experiment, the precipitation began with ACP (m1PO4 around 950 Æ cm)1), and at the transition point around 90 minutes (at the nucleation stage), we began to observe OCP at m1PO4 955 Æ cm)1. After 90 minutes, the structure completed its maturation and the band position shifted to m1PO4 960 Æ cm)1 (crystalline stage). When the structure evolved from ACP to OCP, the XRD lines became visible. Before this region, the precursor was completely amorphous (Fig. 1). The average size (based on 002 and 310 reflections) was measured by XRD, and the acquired Raman spectra correlated with the changes in mineral size. Raman peaks change as a function of particle size in a dual fashion: shifts in the wavelength at maximum as well as changes in FWHH (Table 2). The strongest correlations

1200

1400

Fig. 2. Raman spectra of the m1, m2, m3 and m4 phosphate, monohydrogen phosphate regions and the m1 carbonate region of ACP. Times (minutes) shown are reaction times.

were observed between changes in size at 002 reflection and m1PO4, m3PO4, m4PO4, and HPO4 peak shifts at maximum as well as changes in FWHH for the m1PO4, m2PO4, m3PO4, m4PO4, and HPO4 peaks and the HPO4/ m1PO4 ratio. In addition, there were strong correlations between particle size at 310 reflection and Raman shifts of m1PO4 and HPO4 and FWHH of m1PO4, m2PO4, m3PO4, m4PO4, and HPO4 bands. Recently, several publications have employed Raman spectroscopy and microspectroscopy to describe variations in mineral crystallinity in both in vitro samples [15] and bone [26], based on changes of the m1PO4 peak. Although changes in this peak are correlated with changes in crystallite maturity/crystallinity, the results of the present study show that it is very sensitive to changes in 002 reflection. Thus, a combinatorial (multipeak) consideration in a series of Raman spectra may

358

B

962

Raman shift (cm-1)

Raman shift (cm-1)

A


960 958 956 954 952 950 10

100

1084 1082 1080 1078 1076 1074 1072 1070 10

1000

100

D

1060

Raman shift (cm-1)

Raman shift (cm-1)

C

1055 1050 1045 1040 1035

1100 1000 900 800 700 600 500 400 300

10

100

1000

10

100

log(time(min))

F

600 595 590 585 580 575 570

900 895 890 885

10

100

10

1000

100

1.5

12 Time vs HPO4/ν1 PO4 Time vs HPO4/ν4 PO4

1.0

B

10

Time vs HPO4/ν2 PO4

8 Ratio

A

6 4

0.5

2 0.0 10

100 log (time (min))

1000

log (time(min))

log(time(min))

Ratio

Fig. 3. Raman band shifts of (a) m1PO4, (b) m1CO3, (c) m3PO4, (d) HPO4 and m2PO4, (e) m4PO4, and (f) shoulder of m1PO4 by log of reaction time.

880

565

2.0

1000

log (time (min))

Raman shift (cm-1)

Raman shift (cm-1)

E

1000

log(time(min))

log(time(min))

1000

0 10

yield information not only on changes in mineral crystallinity but, more specifically, on changes in the 002 and/or 310 directions. In conclusion, the present study demonstrates that useful information may be obtained via Raman spectroscopic analysis pertaining to changes in mineral size. Moreover, such analyses can be performed in intact bone specimens through Raman microspectroscopic and imaging analysis with a spatial resolution of 0.6 1 lm, by far superior to the one offered by other microscopic techniques such as FTIR and SAXS, thus potentially

Time vs CO3/ν1 PO4 Time vs CO3/ν2 PO4 Time vs CO3/HPO4

100 log (time (min))

1000

Fig. 4. Changes in (a) HPO4 and (b) CO3 to PO4 bands (area) ratios by log of reaction time.

yielding important new information on the organization and mineral quality of normal and fragile bone. References 1. Rey C, Shimizu M, Collins B, Glimcher MJ (1991) Resolution enhanced Fourier transform infrared spectroscopy study of the environment of phosphate ion in the early deposits of a solid phase calcium phosphate in bone and enamel and their evolution with age: investigation in the m3 PO4 domain. Calcif Tissue Int 49:383 388 2. Pleshko NL, Boskey AL, Mendelsohn R (1991) Novel infrared spectroscopic method for the determination of


359

Table 3. Correlation between Raman bands and particle size

002 vs. m1PO4 shift m2PO4 shift m3PO4 shift m4PO4 shift m1CO3 shift HPO4 shift m1PO4 FWHH m2PO4 FWHH m3PO4 FWHH m4PO4 FWHH m1CO3 FWHH HPO4 FWHH HPO4/m1PO4 310 vs. m1PO4 shift m2PO4 shift m3PO4 shift m4PO4 shift m1CO3 shift HPO4 shift m1PO4 FWHH m2PO4 FWHH m3PO4 FWHH m4PO4 FWHH m1CO3 FWHH HPO4 FWHH

3.

4.

5.

6.

7.

8.

r2

Slope

0.75 0.40 0.70 0.90 0.53 0.62 0.75 0.80 0.60 0.80 0.52 0.90 0.85

+ + ) + ) + ) ) ) ) + ) )

0.90 0.30 0.40 0.50 0.40 0.78 0.87 0.77 0.71 0.86 0.46 0.87

+ ) ) + ) + ) ) + ) + +

crystallinity of hydroxyapatite minerals. Biophys J 60:786 793 Ou-Yang H, Paschalis EP, Boskey AL, Mendelsohn R (1999) Two-dimensional vibrational correlation spectroscopy of in vitro hydroxyapatite maturation. Biopolymers 57:129 139 Gadaleta SJ, Paschalis EP, Betts F, Mendelsohn R, Boskey AL (1996) Fourier transform infrared spectroscopy of the solution mediated conversion of amorphous calcium phosphate to hydroxyapatite: new correlations between Xray diffraction and infrared data. Calcif Tissue Int 58:9 16 Sauer GR, Zunic WB, During JR, Wuthier RE (1994) Fourier transform Raman spectroscopy of synthetic and biological calcium phosphate. Calcif Tissue Int 54:414 420 Penel G, Leroy G, Rey C, Bres E (1998) MicroRaman spectral study of the PO4 and CO3 vibrational modes in synthetic and biological apatites. Calcif Tissue Int 63:475 481 Walters MA, Leung YC, Blumenthal NC, LeGeros RZ, Konsker KA (1990) A Raman and infrared spectroscopic investigation of biological hydroxyapatite. J Inorg Biochem 39:193 200 Tarnowski CP, Ignelzi MA Jr, Wang W, Taboas JM, Goldstein SA, Morris MD (2004) Earliest mineral and matrix changes in force induced musculoskeletal disease as revealed by raman microspectroscopic imaging. J Bone Miner Res 19:64 70

9. Timlin JA, Carden A, Morris MD (1999) Chemical microstructure of cortical bone probed by Raman transects. Appl Spectrosc 53:1429 1435 10. Timlin JA, Carden A, Morris MD, Bonadio JF, Hoffler CE II, Kozloff KM, Goldstein SA (1999) Spatial distribution of phosphate species in mature and newly generated mammalian bone by hyperspectral Raman imaging. J Biomed Optics 4:28 34 11. Akkus O, Adar F, Schaffler MB (2004) Age-related changes in physicochemical properties of mineral crystals are related to impaired mechanical function of cortical bone. Bone 34:443 453 12. Tarnowski CP, Ignelzi MA, Morris MD (2002) Mineralization of developing mouse calvaria as revealed by Raman microspectroscopy. J Bone Miner Res 17:1118 1126 13. Tsuda H, Arends J (1994) Orientational micro-Raman spectroscopy on hydroxyapatite single crystals and human enamel crystallites. J Dent Res 73:1703 1710 14. Kazanci M, Roschger P, Paschalis EP, Klaushofer K, Fratzl P (2006) Bone osteonal tissues by Raman spectral mapping: orientation-composition. J Struct Biol (in press) 15. Stewart S, Shea DA, Tarnowski CP, Morris MD, Wang D, Francheschi R, Lin DL, Keller E (2002) Trends in early mineralization of murine calvarial osteoblastic cultures: a Raman microscopic study. J Raman Spectrosc 33:536 543 16. Carden A, Morris MD (2000) Application of vibrational spectroscopy to the study of mineralized tissue. J Biomed Optics 5:259 268 17. Danilchenko SN, Kukharenko OG, Moseke C, Protsenko IY, Sukhodub LF, Sulkio-Cleff B (2002) Determination of the bone mineral crystallite size and lattice strain from diffraction line broadening. Cryst Res Technol 37:1234 1240 18. Rusu VM, Ng C-H, Wilke M, Tiersch B, Fratzl P, Peter MG (2005) Size-controlled hydroxyapatite nanoparticles as self-organized organic-inorganic composite materials. Biomaterials 26:5414 5426 19. Fratzl P, Gupta HS, Paschalis EP, Roschger P (2004) Structure and mechanical quality of the collagen-mineral nano-composite in bone. J Mater Chem 14:2115 2123 20. Urist MR (1980) Fundamental and Clinical Bone Physiology. Lippincott: Philadelphia 21. Currey JD (2002) Bones Structure and Mechanics. Princeton, NJ: Princeton University Press 22. Roschger P, Fratzl P, Klaushofer K, Rodan G (1997) Mineralization of cancellous bone after alendronate and sodium fluoride treatment: a quantitative backscattered electron imaging study on minipig ribs. Bone 20:393 397 23. Termine JD, Posner AS (1967) Amorphous/crystalline interrelationships in bone mineral. Calcif Tissue Res 1:8 23 24. Brecevic LJ, Furedi-Milhofer H (1972) Precipitation of calcium phosphates from electrolyte solutions. II. The formation and transformation of precipitates. Calcif Tissue Res 10:82 90 25. Crane NJ, Popescu V, Morris MD, Steenhuis P, Ignelzi MA (2006) Raman spectroscopic evidence for octacalcium phosphate and other transient mineral species deposited during intramembranous mineralization. Bone 39:434 442 26. Crane NJ, Morris MD, Ignelzi MA, Yu GG (2005) Raman imaging demonstrates FGF2-induced craniosynostosis in mouse calvaria. J Biomed Optics 10: art. no 031119: 1 8