Full text of DOI:10.1359/jbmr.2002.17.6.1118

JOURNAL OF BONE AND MINERAL RESEARCH
Volume 17, Number 6, 2002
© 2002 American Society for Bone and Mineral Research
Mineralization of Developing Mouse Calvaria as Revealed
by Raman Microspectroscopy
CATHERINE P. TARNOWSKI,1 MICHAEL A. IGNELZI JR.,2 and MICHAEL D. MORRIS1
ABSTRACT
Raman microspectroscopy is a nondestructive vibrational spectroscopic technique that permits the study of
organic and mineral species at micron resolution, offers the ability to work with hydrated and dehydrated
specimens in vivo or in vitro, and requires minimal specimen preparation. We used Raman microspectroscopy
to determine the composition of the mineral environments present in mouse calvaria, the flat bones that
comprise the top of the skull. We have acquired Raman transects (lines of point spectra) from mouse calvaria
during a developmental time course ranging from embryonic day 13.5 (E13.5; 6 days before birth) to 6 months
of age. Exploratory factor analysis (FA) reveals the presence of a variety of apatitic mineral environments
throughout the tissue series. The earliest mineral is observed in the fetal day 15.5 (F15.5) mice and is identified
as a carbonated apatite. The presence of a heterogeneous mineralized tissue in the postnatal specimens
suggests that ionic incorporation and crystal perfection in the lattice vary as the mouse develops. This
variation is indicative of the presence of both recently deposited mineral and more matured remodeled
mineral. Band area ratios reveal that the mineral/matrix ratio initially increases, reaches a plateau, and then
increases again. The carbonate/phosphate band area ratio remains constant from F18.5 to postnatal day 3
(PN3) and then increases with age. Insights into the chemical species, the degree of mineralization, and the
multiple mineral environments that are present in normal calvarial tissue will enable us to better understand
both normal and abnormal mineralization processes. (J Bone Miner Res 2002;17:1118–1126)
Key words: Raman spectroscopy, mineralization, bone, mouse, calvaria
osteoblast-derived cells rather than the cartilage-mediated
deposition that characterizes endochondral ossification.
These osteoblast-derived cells sequester the calcium and
phosphate needed to form the initial mineral lattice, which
then is deposited onto a collagen matrix. Needle-like crys-
talline structures have been observed in these osteoblastic
cells both in cell cultures(1) and in sectioned rat calvaria.(2)
Stanford and colleagues found that the initial mineral crys-
tals formed in the osteoblastic cell cultures have X-ray
diffraction (XRD) patterns that resemble those of substi-
tuted apatites found in mature bone tissue.(1) Recently,
Kuhn and coworkers also have studied the mineral crystal-
lites found in embryonic chick cell cultures using Fourier-
INTRODUCTION
dentifying the composition of the mineral and matrix
environments in early mineralized tissue is crucial to
I
understanding the initial mineralization process. It is well
known that the mineral lattice formed during early stages of
development of the calvaria (the flat bones that comprise the
skull), as in the mandible and other skeletal elements, oc-
curs through intramembranous ossification (IO); however,
the composition of the initial mineral deposited in vivo is
unknown. It is recognized that IO is mediated through
The authors have no conflict of interest.
1Department of Chemistry, University of Michigan, Ann Arbor, Michigan, USA.
2Department of Orthodontics and Pediatric Dentistry, University of Michigan School of Dentistry, Ann Arbor, Michigan, USA.
1118

DEVELOPING MOUSE CALVARIA/RAMAN SPECTROSCOPY
1119
TABLE 1. BAND ASSIGNMENTS FOR RAMAN SPECTRA OF
TISSUE
The PO34Ϫ symmetric stretch (␯1), the most prominent
vibrational band in the Raman spectrum of mineralized
tissue, is critical to our understanding of the mineral envi-
ronments present in the tissue. Advantageously, both the
PO34Ϫ symmetric stretch (␯1) frequency and the band shape
depend on the local mineral environment and therefore
change with ionic incorporation and crystallinity. Many
studies on apatitic tissue and synthetic apatitic lattices have
correlated the PO34Ϫ symmetric stretch (␯1) band frequency
Raman frequency
shift (cmϪ1
)
Assignment (intensity)
PO34Ϫ ␯4, (m)a
580
855
Proline ring (w)
876
950–964
Hydroxyproline ring (w)
PO34Ϫ ␯ (POO symmetric stretch; vs)a
1
1001–1003
1030
1065–1070
1245, 1270
1445
Ring breathing mode of phenlyalanine (m) with the mineral environment composition. In particular,
PO34Ϫ ␯ (POO asymmetric stretch; m)a
hydroxyapatite, carbonate-substituted apatite, amorphous
calcium phosphate, octacalcium phosphate, and dicalcium
3
CO23Ϫ ␯ (COO in plane stretch; m)a
1
phosphate dihydrate all have characteristic PO43Ϫ
␯ fre-
Amide III, CONOH stretch (m)
CH2 wag (m)
1
quencies.(12,16) In mineralized tissue Raman spectroscopy, it
is useful to classify the apatitic environments using three
broad categories with three different band frequencies.
B-type carbonate-substituted apatite (carbonate substituted
for phosphate in the hydroxyapatite lattice) produces a
Raman-shifted frequency in the range of 955 cmϪ1–959
1665
Amide I, COCON stretch (m)
a Denotes a mineral band.
vs, very strong; m, medium; w, weak.
transform infrared (FTIR) spectroscopy.(3) The earliest min- cmϪ1. In a crystalline nonsubstituted hydroxyapatite, the
eral crystallite, found in 8-day-old cultures, consisted of a PO34Ϫ ␯1 Raman-shifted frequency is found in the region of
carbonated apatite with the presence of monohydrogen 962 cmϪ1–964 cmϪ1. Finally, a band with a Raman-shifted
phosphate (HPO24Ϫ) ions. Kuhn and colleagues also found frequency in the 945 cmϪ1–950 cmϪ1 is indicative of a
that the carbonate content in 60-day-old cell culture apatitic disordered phosphate apatitic lattice. Whether this band
mineral increased from the amount found in the 8-day-old frequency arises from A-type carbonate substitution (car-
culture. Although both of these studies have characterized bonate for hydroxide) or from an amorphous calcium phos-
the earliest mineral environments found in their respective phate is not clear; therefore, we prefer to use the term
systems, both are studies of osteoblastic cell cultures rather “disordered phosphate” to describe this mineral environ-
than of mineralized tissue in vivo.
ment. Most bone tissue contains some amount of each of
Alizarin red/Alcian blue staining has shown that the ini- these three mineral environments; consequently, the result-
tial mineral deposition in mouse calvaria occurs at fetal day ing bone tissue phosphate ␯ band is a sum of all three
1
15.5 (F15.5; 4 days before birth).(4) The shortcoming of bands. Typically, the most prominent mineral environment
alizarin red/Alcian blue staining is that it can determine only identified throughout bone tissue is the carbonated apatitic
the presence of mineral; it is unable to characterize the phosphate. Hence, in a bone spectrum, the phosphate ␯
1
species that constitute the deposited mineral lattice. Raman band peak frequency usually is in the 955- to 959-cmϪ1
spectroscopy overcomes this limitation by using informa- range. The phosphate ␯ band often is asymmetric because
1
tion from the vibrational spectra of the ions within a mineral of the addition of varying contributions from the disordered
lattice to identify the mineral formed in these early stages of phosphate and the more crystalline nonsubstituted phos-
bone development.
phate to the left and right sides of the main band, respec-
Raman vibrational spectroscopy has been used to study tively. Variations in the phosphate ␯ band shape and peak
1
mineralized tissues and cell cultures.(5–13) Raman spectros- frequency can reveal changes in the mineral environment
copy offers several technical advantages. It has excellent because of ionic incorporation or crystallinity at different
spatial resolution (capable of 0.5–1 ␮m) and, in general, the stages of development and are important to our understand-
vibrational bands are narrow so that small frequency shifts ing of the mineralization processes.
and band shape changes often are observed easily. These
We are able to monitor and resolve changes in the phos-
narrow vibrational bands allow one to distinguish between phate ␯1 band shape and frequency as well as the changes in
various chemical species. When Raman spectroscopy is the relative amounts of the rest of the tissue components
used, specimen preparation is simple; specimens need not using a multivariate analysis technique, factor analysis
be transparent or fixed. Hydrated or dehydrated specimens (FA). Although univariate analyses such as band area or
may be subjected to Raman spectroscopy and because Ra- intensity ratios may be more familiar and used more often in
man spectroscopy is nondestructive, the same developing spectroscopic tissue analyses, multivariate analyses are well
tissue can be evaluated over a period of time.
known in the fluorescence imaging and magnetic resonance
In biological tissues, Raman spectroscopy probes the mo- imaging (MRI) communities(17,18) as well as in the chem-
lecular and ionic vibrations of mineralized species such as istry community.(19,20) The advantages of using FA as an
phosphate, carbonate, and monohydrogen phosphate, as well analysis tool is that it has an inherent signal-averaging effect
as the many vibrations that arise from the proteinaceous ma- and has the ability to resolve spectral components that are
trix. In Raman spectra, the symmetrical vibrations of a mole- present in different proportions in a series of spectra. This is
cule or ion are the most intense vibrational bands ob- especially important for bone tissue, which can be very
served.(14,15) The major mineral and matrix bands in the heterogeneous, resulting in considerable variation in its
Raman spectrum of bone tissue are summarized in Table 1.
Raman spectra.

1120
Essentially, a complete FA results in factors and scores;
TARNOWSKI ET AL.
the greater the number of tissue factors, the greater the
number of mineral and matrix environments. In our work, a
factor is a descriptor of a tissue or background component
and resembles a Raman spectrum. Each factor has a corre-
sponding score; this score describes the relative amount of
that factor (tissue component) present in each original spec-
trum of the transect. If two tissue components are found in
different ratios throughout the specimen being probed, their
spectral signatures will be separated into two factors. Their
corresponding scores will reflect the variation throughout
the specimen. For example, if the mineral and the collagen
present in the tissue have a constant ratio throughout that
tissue, FA will produce a factor in which the spectral
signatures of the mineral and the collagen are combined.
However, if the ratio of mineral to collagen varies through-
out the tissue probed, the FA will separate the spectral
signatures of the two species into two separate factors.
These two factors also will have separate scores. Their
individual scores will reveal where the relative amounts of
mineral and collagen differ throughout the tissue probed.
Therefore, FA is an excellent choice for detecting variations
in the mineral and matrix environments throughout the
tissue, because these variations are reflected in the original
spectra as small changes in either the shape of the spectro-
scopic bands or in the band ratios of different tissue com-
ponents. It is important to note that as the variation in the
mineral and matrix environments in the tissue increases,
there will be an increase in the number of tissue factors.
FIG. 1. Schematic diagram of the endocranial surface of the mouse
calvaria: (F) frontal bone, (P) parietal bone, (O) occipital bone. Bold
lines indicate the three general areas where Raman transects were
acquired in each mouse.
MATERIALS AND METHODS
objective and focused into an NIR-optimized spectrograph
(HoloSpec f/1.8I; Kaiser Optical Systems, Inc., Ann Arbor,
Specimen preparation
B6CBA F1/J wild-type mice (The Jackson Laboratory, MI, USA). The spectrograph dispersed the Raman scatter
Bar Harbor, ME, USA) were harvested at nine different onto a liquid-nitrogen cooled camera (MASD; Roper Sci-
developmental time points: embryonic days 13.5 (E13.5) entific, San Diego, CA, USA). A 25-␮m entrance slit was
and E14.5; F15.5 and F18.5; postnatal days 1 (PN1), PN3, used to set the spectral resolution of the system to 3–4
PN7, and PN14; and postnatal 6 months. The flat bones of cmϪ1. A motorized translation stage (New England Affili-
the top-most portion of the skull, the calvaria, were removed ated Technologies, Danaher Precision Systems, Salem, NH,
intact and rinsed of debris using phosphate-buffered saline USA) moved the specimen at 2-␮m increments under the
(PBS) at pH 7.2. Then, the tissue was fixed in 100% ethanol focused laser spot to provide a 4-␮m spatial resolution.
for 30–60 minutes, soaked in 95% ethanol for 20 minutes,
Six Raman transects (lines of point spectra) were ac-
and transferred to 70% ethanol for storage at 4°C. Before quired for each developmental time point. For F18.5 and
the spectroscopic experiments, the specimens were soaked each postnatal time point, two mice were examined in three
in PBS to avoid dehydration during spectral acquisition. different areas on the endocranial surface of the parietal
The calvarial specimens were placed between a quartz slide bone. One transect was acquired along the edge nearest to
and coverslip (Esco Products, Inc., Oak Ridge, NJ, USA) the occipital bone, one transect was acquired laterally, and
for the duration of the experiment.
one transect was acquired adjacent to the sagittal suture
(Fig. 1, bold lines). Because there was significantly less
mineral in the F15.5 mice, only one area on the endocranial
surface of the parietal bone was examined per mouse; six
Raman spectroscopy
The method by which the Raman spectra were obtained mice were examined for this time point. All Raman
and the instrument design have been described previously.(8) transects were acquired with a 4-␮m spatial resolution and
Briefly, a 785-nm diode laser (80 mW at specimen; SDL, ranged in length from 200 to 250 ␮m.
Inc., San Jose, CA, USA) was focused onto the specimens
using a 10X/0.5 NA objective (Fluar Series; Zeiss, Thorn-
wood, NY, USA) mounted on a BH-2 microscope frame
Data analysis
(Olympus, Inc., Melville, NY, USA). In the epi-illumination
All data analysis was performed using MatLab 5.3 (Math-
configuration, the Raman scatter was collected by the same Works, Inc., Natick, MA, USA) using vendor-supplied and

DEVELOPING MOUSE CALVARIA/RAMAN SPECTROSCOPY
1121
locally written scripts. All spectra were preprocessed by
removing detector-generated artifacts (spikes) and subtract-
ing the detector dark current.
FA identifies tissue components, as does the examination
of raw spectra, but FA also reveals how various tissue
components change over time with relation to one another.
FA was performed on the data by calculating the covariance
matrix (the data set matrix multiplied by its transpose) of the
transect spectra data set(20) and then carrying out a principal
component analysis, which enhances any small changes in
the Raman spectra throughout the data set. This principal
component analysis results in set of eigenvectors that de-
scribes the original data set. Most of these eigenvectors
describe noise in the system; these noise-describing eigen-
vectors are discarded. The rest of the nonnoise eigenvectors
are rotated manually (added and subtracted from one another),
resulting in “factors.” These factors describe the original
spectral signatures of the various tissue components.
Nonnegativity and band shapes were used as rotational
constraints.(21)
Before any univariate analysis, any spectrum without a
signal-to-noise ratio Ͼ3 in the PO34Ϫ ␯1 envelope (920–980
cmϪ1) was identified as a nonmineral spectrum and thus
discarded. Using this criterion, some spectra were discarded
only from F15.5 and 18.5 transects—the transects with the
least mineral deposition. Before band area calculations, a
linear baseline was subtracted from each band envelope and
the band was then offset to zero. The procedures used to
obtain the carbonate/phosphate and mineral/matrix ratios
are similar to those used by Camacho et al. in their analysis
of FTIR human bone tissue spectra.(22)
The carbonate/phosphate ratios for each transect were
calculated from an average of the carbonate/phosphate band
area ratios from three spectra. Only three spectra were
chosen for carbonate/phosphate analysis because in some of
the transect spectra, there was less mineral signal present
and as a result, the signal to noise ratio in the carbonate
envelope was not large enough to determine reliably the
carbonate band area. The three spectra chosen for carbonate/
phosphate analysis contained a strong mineral signal (de-
termined by the mineral scores from the FA) and were
spaced widely throughout the data set so as to include any
variation in the tissue’s mineral environments. Two or three
bands (ϳ1030 cmϪ1 PO34Ϫ ␯3, ϳ1045 cmϪ1 PO43Ϫ ␯3, and
ϳ1070 cmϪ1 CO32Ϫ ␯1) were fit to the baseline-corrected
PO43Ϫ ␯3, CO23Ϫ ␯1 envelope (1015–1090 cmϪ1). The area of
the 1070-cmϪ1 band only was used as a metric for the
carbonate content. The integrated area of the baseline-
corrected PO34Ϫ ␯1 envelope (920–980 cmϪ1) was used as a
FIG. 2. Raman spectra reveal first signs of mineral at F15.5. Selected
Raman spectra from three specimens: E14.5, F15.5, and PN14. Spectra
are offset for clarity. The 785-nm excitation (80 mW) focused onto
specimens through a 10ϫ/0.50 NA objective. There is no detectable
phosphate ␯ band (957 cmϪ1) in the E14.5 spectrum, although it is
1
prominent in the F15.5 and PN14 spectra. The band area and intensity
ratios of the phosphate (957 cmϪ1) to the collagen-dominated protein
(1003 cmϪ1 or 1665 cmϪ1), a reflection of mineral/matrix ratio, in-
crease with age from F15.5 to PN14.
mean mineral/matrix ratio for each transect was used to find
the global average mineral/matrix ratio for each time point.
RESULTS
Raman spectra
Selected Raman spectra from E14.5, F15.5, and PN14
measure of the phosphate content. A global average transects are shown in Fig. 2. The embryonic spectrum
carbonate/phosphate ratio for each time point was obtained (E14.5) contains only a collagen-dominated protein signa-
by averaging the mean carbonate/phosphate ratio from each ture, suggesting that no mineral has been formed at this
of the six transects for each time point.
stage in the areas probed by Raman spectroscopy. This first
envelope sign of mineral appears at F15.5 in the form of an apatitic
The area of the baseline-corrected PO43Ϫ
␯
1
(920–980 cmϪ1) discussed previously was used as a mea- phosphate (956–957 cmϪ1). This same apatitic phosphate
sure of the mineral content in the mineral/matrix ratio. The band is seen at PN14. Both F15.5 and PN14 also contain the
area of the baseline-corrected amide I band envelope protein signature similar to the one found at E14.5. The
(1600–1720 cmϪ1) was calculated and used as a measure of relative intensity ratio of the phosphate ␯1 (956–957 cmϪ1
)
the matrix content. The mineral/matrix ratios were calcu- band to that of the primarily collagen-dominated matrix
lated for each spectrum of a transect and averaged. This bands (either 1003 cmϪ1 or the 1665-cmϪ1 band) in the two

1122
TARNOWSKI ET AL.
FIG. 3. FA reveals the presence of only one resolvable mineral species at F15.5. Factors describe specific chemical species in the tissue; scores
describe the relative amount of each factor at a given location in the tissue probed. Areas where one score varies from another denote variation
in the tissue environment. The 785-nm excitation (80 mW) focused onto specimens through a 10ϫ/0.50 NA objective, spectra acquired at 2-␮m
increments. Data were acquired from the endocranial side of the parietal bone. Mineral subregion factors and scores from an F15.5 specimen.
Factor A, collagen-dominated protein; factor B, carbonated apatitic mineral (PO34Ϫ ␯1, 957 cmϪ1; CO23Ϫ ␯1, 1066 cmϪ1). Four nontissue factors
(two featureless backgrounds, buffer, and quartz) have been omitted for clarity.
spectra with mineral increases dramatically with age (Fig. 2, ences in their scores indicate areas in the tissue where their
F15.5 and PN14), indicating that an increased amount of the
mineral is found in the older tissue.
relative amounts differ.
The FA results of the F15.5 specimens show one of the
greatest advantages of FA—its signal-averaging effect. On
visual examination of the Raman spectrum from the F15.5
specimen with the most mineral signal (Fig. 2, F15.5), there
is no easily discernible carbonate band (CO23Ϫ ␯1, 1065–
1070 cmϪ1). However, the FA shows that the apatitic min-
eral actually is carbonated (Fig. 3, factor B, CO23Ϫ ␯1, 1066
cmϪ1).
Multivariate data analysis
FA: The FA results from the individual transects can be
used to sort the different aged mice into four distinct groups.
The first group contains tissue with no mineral (E13.5 and
E14.5). The second group contains tissue with only one
mineral factor (F15.5 and F18.5). The third (PN1 and PN3)
and fourth (PN7 and PN14 and postnatal 6 months) groups
contained tissues with three and two resolvable mineral
factors, respectively. As mentioned previously, the greater
the number of resolvable tissue factors (descriptors), the
greater the variation throughout the tissue.
PN1 and PN3. At least two mineral environments were
found in each of the transects in the PN1 and PN3 speci-
mens. A set of mineral factors from a PN3 specimen with
three resolvable mineral environments is shown in Fig. 4.
The main mineral factor in all of the specimens was a
carbonated apatitic mineral (PO34Ϫ ␯ ϳ956 cmϪ1; Fig. 4,
1
E13.5 and E14.5. Transects from E13.5 and E14.5 cal-
varia were found to contain only a collagen-dominated
protein. There was no detectable presence of any phosphate
mineral.
factor A). At least one additional resolvable mineral factor
was present in all of transects; two additional mineral fac-
tors were resolvable in 33% and in 50% of the PN1 and PN3
specimens, respectively. A nonsubstituted, more crystalline
apatitic lattice (962–964 cmϪ1)–resolvable factor was
present more often then the disordered phosphate factor
(945–950 cmϪ1).
PN7 and PN14 and postnatal 6 months. In the PN7,
PN14, and 6-month specimens, we were able to resolve only
two separable mineral environments using FA; a set of
mineral factors and scores from one of the PN14 transects is
shown in Fig. 5. All spectra were dominated by the substi-
tuted phosphate mineral environment at ϳ956 cmϪ1 (with
B-type carbonate present at ϳ1069 cmϪ1; Fig. 5, factor A).
A resolvable nonsubstituted apatitic band (ϳ964 cmϪ1; Fig.
5, factor B) was present also in three of the PN7 specimens,
five of the PN14 specimens, and four of the 6-month-old
F15.5 and F18.5. F15.5 and F18.5 calvarial transects all
contained only one mineral factor, an apatitic phosphate
lattice with a small amount of B-type carbonation (PO34Ϫ ␯1,
957 cmϪ1; CO32Ϫ ␯1, 1066 cmϪ1). Figure 3 presents the
factors and scores from an F15.5 transect (four nontissue
factors were omitted for clarity). Each factor describes a
tissue component found in the specimen. Its corresponding
score describes the relative amount of that component found
in the original data set. For instance, there is more signal in
the original data set from the apatitic mineral (factor B) at
50 ␮m and 140 ␮m through the transect than at 20 ␮m or 80
␮m through the transect. Note that the collagen (factor A)
and mineral (factor B) components are not colocalized
throughout the tissue probed; hence, the FA separates these
two tissue components into two separate factors. Differ- specimens. However, in the presence of the more fully

DEVELOPING MOUSE CALVARIA/RAMAN SPECTROSCOPY
1123
FIG. 4. FA reveals the presence
of three resolvable mineral envi-
ronments (959 cmϪ1, 946 cmϪ1
,
and 963 cmϪ1) at PN3. Similar
results were observed at PN1. For
explanation of factors and scores
and for experimental conditions
see legend for Fig. 3. Factor A,
carbonated apatite (PO34Ϫ ␯1, 959
cmϪ1; CO23Ϫ ␯1, 1069 cmϪ1); fac-
tor
B
disordered phosphate
(PO34Ϫ ␯1, 946 cmϪ1); factor C, a
more crystalline, nonsubstituted
apatitic mineral (PO34Ϫ ␯1, 963
cmϪ1). Background factors have
been omitted for clarity.
FIG. 5. Two resolvable mineral
environments (956 cmϪ1 and 964
cmϪ1) found in PN14; the resolv-
able disordered phosphate min-
eral environment is not present.
Similar results were observed at
PN7 and at 6 months. For expla-
nation of factor and scores and
for experimental conditions see
legend for Fig. 3. Factor A, car-
bonated apatite (PO34Ϫ ␯1, 956
cmϪ1; CO32Ϫ ␯1, 1069 cmϪ1); B,
a nonsubstituted apatitic mineral
(PO34Ϫ ␯1, 964 cmϪ1). Back-
ground factors have been omitted
for clarity.
mineralized tissue, the disordered phosphate factor (945– ing that the mineralization was irregular and that the mineral
950 cmϪ1) was separable in only one of the PN14 transects was present in a different proportion than the collagen. (Fig.
and was not found in any of the PN7 or 6-month-old 3) As mineralization progresses, the mineral and collagen
specimens.
are not always resolved, meaning that the mineral is depos-
In the last two groups (the PN1 and PN3 group and the ited evenly on the collagen matrix; in the PN3, PN7, and
PN7, PN14, and 6-month group), more than one spatially PN14 specimens, the mineral and collagen are resolved in
resolvable mineral environment was present. The scores for one-half of the transects, and none are resolved in the
the multiple mineral environments were different from one 6-month-old mice. Hence, as the mouse ages and mineral
another (e.g., Fig. 4, in the 40- to 80- and 120- to 160-␮m deposition is more complete, the mineral is distributed
regions and Fig. 5 in the 40- to 60-␮m region), but varied evenly across the collagen over the distances and the spatial
only a small amount—meaning that the mineral environ- resolution of our measurements.
ments (lattice substitution and crystallite perfection) change
only slightly throughout the distance of the tissue probed.
FA is especially useful in examining tissue in which the
Univariate data analysis
mineral deposition is incomplete. In the younger tissue
Band area ratios: Carbonate/phosphate ratios. Although
(F15.5, F18.5, and PN1), the FA in all of the transects the FA of the F15.5 calvaria revealed the presence of a
produced resolvable collagen and mineral factors, indicat- carbonate-substituted phosphate, the amount was so low

1124
TARNOWSKI ET AL.
FIG. 7. Mineralization begins during fetal period, remains constant
through early postnatal life and increases with age. Mineral/matrix
band area ratios (PO43Ϫ ␯1, 957 cmϪ1; amide I, 1665 cmϪ1) for F15.5
through 6-month-old specimens. The band area ratio increases from age
F15.5 to F18.5, plateaus, and then increases again from the age of PN14
to 6 months old. The values shown are the averaged mineral/matrix
band area ratio for the six transects acquired for each age time point;
the statistical uncertainties are equal to 1 SD.
FIG. 6. Carbonate inclusion in the mineral lattice is constant from
F18.5 to PN3 and then increases with age. Carbonate/phosphate band
area ratios (CO23Ϫ ␯1, 1070 cmϪ1; PO43Ϫ ␯1, 957 cmϪ1) for F18.5
through 6-month-old specimens. The band area ratio is constant from
F18.5 to PN3, and then increases with age. The values shown are the
averaged carbonate/phosphate band area ratio for the six transects
acquired for each age time point; the statistical uncertainties are equal
to 1 SD. The carbonate/phosphate band area ratio for the F15.5 age
specimen is not shown because the carbonate band intensity is too weak
to allow an accurate band area calculation.
vironment detected in the mouse calvaria and continue this
characterization until a more mature mineral environment
was formed. Using Raman spectroscopy, we have shown
that the first appearance of mineral in mouse calvaria was
observed at F15.5, consistent with alzarin red/Alcian blue
staining studies,(4) and that this mineral is a carbonated
apatite, similar to that observed by other initial mineraliza-
tion studies of endochondral ossification(23,24) and cell cul-
tures.(25,26) Although the mineralization of the F15.5 spec-
imens was incomplete, mineral had certainly been deposited
in several areas during the 24 h between E14.5 and F15.5.
These data support previous studies that reported initial
mineral deposition occurs at F15.5.(4) As mineralization
progressed during the postnatal period, FA revealed the
presence of more than one resolvable mineral factor, indi-
cating that the tissue contained multiple mineral environ-
ments. The presence of multiple mineral environments is
indicative of several mineralization processes occurring ei-
ther simultaneously or sequentially.
that the carbonate/phosphate ratios for F15.5 specimens
could not be determined. In the rest of the specimens, the
carbonate/phosphate ratio remained constant from F18.5 to
PN3 and then increased with age (Fig. 6).
Band area ratios: Mineral/matrix ratios. The mineral/
matrix ratio increases rapidly over the first 3 days of min-
eralization (F15.5–F18.5), reaches a plateau, and then in-
creases between the ages of PN14 and the 6-month-old
specimens (Fig. 7). The first and most significant increase in
the mineral/matrix ratio is found between F15.5 and F18.5.
During this period, mineral deposition is occurring rapidly
and as a consequence, the mineral/matrix ratio drastically
increases. The mineral/matrix ratios for the days of F18.5
through PN14 have reached a plateau, because the mineral/
matrix ratio values over this age range are not different.
However, it is evident that mineralization is not complete at
PN14 because the mineral/matrix ratio is less than that of
the 6-month-old specimens.
We have shown that FA is a robust data analysis tool able
to distinguish variations in Raman spectra (and hence in the
original tissue components). In particular, FA enabled us to
examine small variations in the mineral environment. Be-
fore discussing the relevance of each resolvable mineral
DISCUSSION
We have shown that Raman spectroscopy is a powerful environment, it is important to emphasize that the main
technique to study mineralized tissues. Using both univar- mineral environment found throughout all of the tissue is a
iate and multivariate data analysis techniques, we were able carbonated apatite. This substituted apatitic environment
to characterize the composition of the earliest mineral en- has varying levels of carbonate content throughout the de-

DEVELOPING MOUSE CALVARIA/RAMAN SPECTROSCOPY
1125
velopment of the mouse calvaria, but in all of the spectra in incorporation after PN3), an increase in carbonation is ob-
all of the tissue transects, this was the most prominent served with an increase in age. This is consistent with
mineral environment observed. The other two resolvable mineral maturation in fully mineralized tissue.
mineral environments were a nonsubstituted, more crystal-
line apatite and a more disordered phosphate.
We found no evidence in the calvaria for thermodynamic
precursors to apatitic mineral (dicalcium phosphate dihy-
In the F15.5 and F18.5 calvaria, the carbonated apatite drate, octacalcium phosphate, and amorphous calcium phos-
was the only mineral environment observed. This could be phate),(28) postulated to be precursors to biological apatite.
a result of the rapid deposition of a nearly homogeneous It is interesting to note that no other vibrational spectro-
substituted, though not necessarily perfect, crystalline min- scopic studies of cell cultures and developing tissue have
eral. Another more likely possibility is that because there is provided evidence for the proposed hydroxyapatite precur-
less mineral, the signal to noise of the mineral vibrational sors.(1,3,23,25,26,29) It is possible that we did not detect these
bands is decreased. As a result of this decrease in signal to precursors because they were present only in small quanti-
noise, small changes in the shoulders of the PO34Ϫ ␯ band ties and only for a few minutes or hours.
1
throughout the transect are not resolved by the FA tech-
Raman microspectroscopy proved to be a valuable char-
nique. As mineralization and growth continue after birth, it acterization tool in the study of early mineralized tissue. It
is possible that both osteoclastic resorption and remodeling offers several advantages over some more familiar bone
of the previously deposited mineral as well as deposition of tissue–characterization technologies such as FTIR, XRD,
new mineral is occurring. If this mineral deposition and and electron microscopy (EM). Of these four technologies,
remodeling is occurring, it could be that the resulting min- Raman spectroscopy is the only nondestructive technique.
eral environment is not homogeneous. The appearance of Raman spectroscopy can be performed on in vitro and
more than one mineral factor in the older tissues is consis- potentially on some in vivo specimens; hydrated specimens
tent with this interpretation.
can be subjected to Raman spectroscopy because the OOH
Besides the most prominent factor (the carbonated apa- stretch does not obscure the tissue spectrum. Raman spec-
tite), FA revealed the presence of a resolvable nonsubsti- troscopy provides spatial information that XRD cannot be-
tuted more crystalline apatite mineral factor in all of the cause of the need to pulverize specimens before XRD
postnatal time points, indicating that either a previously analyses and Raman spectroscopy is capable of better spa-
substituted mineral lattice is being remodeled (to remove tial resolution (resolution down to 0.5 ␮m) than FTIR (10
substitutions and to perfect the lattice) or that a nonsubsti- ␮m). Raman spectroscopy also requires very little specimen
tuted apatitic phosphate lattice is being deposited. The more preparation; the thin sections needed for FTIR analyses are
disordered phosphate environment, present in the PN1 and not necessary for Raman analyses. Like FTIR spectroscopy,
PN3 specimens, suggests that mineral maturation has not Raman spectroscopy also offers molecular speciation infor-
fully occurred and that osteoclastic resorption has not yet mation that XRD and EM cannot. For example, Raman
begun in these areas of the tissues. It could be that the spectroscopy distinguishes a phosphorous atom in a hy-
disordered phosphate shoulder is no longer resolvable from droxyapatite lattice from a phosphorous atom in a phospho-
the substituted apatitic phosphate in the PN7, PN14, and lipid. In this work, we have been able to use Raman mi-
6-month-old calvaria because remodeling is continuing and crospectroscopy to examine mouse calvaria and provide
osteoclastic resorption is beginning, removing some of the detailed compositional information about the tissue envi-
disorder mineral from the lattice. The presence or absence ronment. We believe that Raman microspectroscopy also
of these two additional resolvable mineral factors helps will prove useful in the study of birth defects affecting the
illustrate the complexity of the early mineralization process. skull, including the abnormal mineralization of the calvarial
We propose that carbonate/phosphate ratios that are constant sutures in the disease craniosynostosis.
(F18.5 to PN3) and then increase thereafter are a reflection of
the concentration and availability of carbonate ions in the
mineral environment. We believe that both through remodeling
ACKNOWLEDGMENTS
and new mineral deposition, any ions that are present in the
mineral environment can be incorporated into the apatitic lat-
tice. The work of Boskey et al. supports this hypothesis; using
FTIR spectroscopy, they have shown that patients who have
osteoporosis and are treated with sodium fluoride have in-
creased mineral crystallite sizes in their tissue, indicative
of fluoride ion incorporation.(27) We suggest then, that in
the younger specimens, fewer carbonate ions may be
available for incorporation into the mineral lattice. Birth
may change the ionic equilibrium found in the tissue and
may affect what ions are available for incorporation. The
change in ions integrated into the mineral lattice after birth
may not be measurable immediately; however, with the
continuing, rapid calvarial mineralization, the change in
ionic incorporation may become evident after PN3. As
mineralization continues (growth, remodeling, and ionic
The authors thank Jennifer Dreier (University of Michi-
gan) for her assistance with specimen preparation and Dr.
Jerilyn Timlin (Sandia National Laboratories), Dr. Jeremy
Shaver (Eigenvector Research, Inc.), and Angela Carden
(University of Michigan) for their helpful discussions re-
garding data analysis. This work was supported in part by
the University of Michigan Center for Biomedical Engi-
neering Research (to M.D.M.) and NIH/NIAMS 46024 (to
M.D.M. and M.A.I. Jr.) and NIDCR grant DE 11530 (to
M.A.I. Jr.).
REFERENCES
1. Stanford CM, Jacobson PA, Eanes ED, Lembke LA, Midura
RJ 1995 Rapidly forming apatitic mineral in an osteoblastic
cell line (UMR 106–01 BSP). J Biol Chem 270:9420–9428.

1126
TARNOWSKI ET AL.
2. Hoshi K, Ejiri S, Ozawa H 2001 Localization alterations of 17. Geladi P, Grahn H 1996 Multivariate Image Analysis. John
calcium, phosphorus, and calcification-related organics such as
proteoglycans and alkaline phosphatase during bone calcifica-
tion. J Bone Miner Res 16:289–298.
Wiley & Sons Ltd., West Sussex, UK.
18. Reyment R, Jo¨reskog KG 1996 Applied Factor Analysis in the
Natural Sciences, 2nd ed. Cambridge University Press, Cam-
bridge, UK.
19. Adams MJ, 1995 Chemometrics in Analytical Spectroscopy.
The Royal Society of Chemistry, Cambridge, UK.
20. Otto M, 1999 Chemometrics: Statistics and Computer Appli-
cation in Analytical Chemistry. Wiley-VCH, New York, NY,
USA.
21. Jestel NL, Shaver JM, Morris MD 1998 Hyperspectral Raman
line imaging on an aluminosilicate glass. Appl Spectrosc 52:
64–69.
22. Camacho NP, Rinnerthaler S, Paschalis EP, Mendelsohn R,
Boskey AL, Fratzl P 1999 Complementary information on
bone ultrastructure from scanning small angle X-ray scattering
and Fourier-transform infrared microspectroscopy. Bone 25:
287–293.
23. Mendelsohn R, Jassankhani A, DiCarlo E, Boskey A 1989
FT-IR microscopy of endochondral ossification at 20 ␮m
spatial resolution. Calcif Tissue Int 44:20–24.
24. Rey C, Hina A, Tofighi A, Glimcher MJ 1995 Maturation of
poorly crystalline apatites: Chemical and structural aspects in
vivo and in vitro. Cell Mater 5:345–356.
25. Rey C, Kim H-M, Gerstenfeld L, Glimcher MJ 1996 Charac-
terization of the apatite crystals of bone and their maturation in
osteoblast cell culture: Comparison with native bone crystals.
Connect Tissue Res 34:343–349, 397–403.
26. Rey C, Kim H-M, Gerstenfeld L, Glimcher MJ 1995 Structural
and chemical characteristics and maturation of the calcium-
phosphate crystals formed during the calcification of the or-
ganic matrix synthesized by chicken osteoblasts in cell culture.
J Bone Miner Res 10:1577–1588.
3. Kuhn LT, Xu YT, Rey C, Gerstenfeld LC, Grynpas MD,
Ackerman JL, Kim HM, Glimcher MJ 2000 Structure, com-
position, and maturation of newly deposited calcium-
phosphate crystals in chicken osteoblast cell cultures. J Bone
Miner Res 15:1301–1309.
4. Johansen VA, Hall SH 1982 Morphogenesis of the mouse
coronal suture. Acta Anat (Basel) 114:58–67.
5. Carden A, Morris MD 2000 Application of vibrational spec-
troscopy to the study of mineralized tissues. J Biomed Opt
5:259–268 (review).
6. Lawson EE, Barry BW, Williams AC, Edwards HGM 1997
Biomedical applications of Raman spectroscopy. J Raman
Spectrosc 28:111–117.
7. Kale S, Biermann S, Edwards C, Tarnowski C, Morris M,
Long MW 2000 Three-dimensional cellular development is
essential for ex vivo formation of human bone. Nat Biotechnol
18:954–958.
8. Timlin JA, Carden A, Morris MD, Rajachar RM, Kohn DH
2000 Raman spectroscopic imaging markers for fatigue-
related microdamage in bovine bone. Anal Chem 72:2229–
2236.
9. Timlin JA, Carden A, Morris MD 1999 Chemical microstruc-
ture of cortical bone probed by Raman transects. Appl Spec-
trosc 53:1429–1435.
10. Timlin JA, Carden A, Morris MD, Bonadio JF, Hoffler CE,
Kozloff KM, Goldstein SA 1999 Spatial distribution of phos-
phate species in mature and newly generated mammalian bone
by hyperspectral Raman imaging. J Biomed Opt 4:28–34.
11. Lin D-L, Tarnowski CP, Zhang J, Dai J, Rohn E, Patel AH,
Morris MD, Keller ET 2001 Bone metastatic LNCaP-
derivative C4–2B prostate cancer cell line mineralizes in vitro.
Prostate 47:212–221.
12. 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.
13. Kirchner MT, Edwards HGM, Lucy D, Pollard AM 1997
Ancient and modern specimens of human teeth: A Fourier
transform Raman spectroscopic study. J Raman Spectrosc
28:171–178.
14. Schrader B, ed. 1995 Infrared and Raman Spectroscopy: Meth-
ods and Applications. VCH Publishers, Inc., New York, NY,
USA.
15. Chalmers JM, Everall NJ 1993 Vibrational spectroscopy. In:
Hunt BJ, James MI (eds.) Polymer Characterisation. Blackie
Academic & Professional, an imprint of Chapman & Hall,
Bishopbriggs, Glasgow, Scotland, pp. 69–114.
27. Boskey AL 1990 Bone mineral and matrix: Are they altered in
osteoporosis? Orthop Clin North Am 21:19–29.
28. Nancollas GH, Lore M, Perez L, Richardson C, Zawacki SJ
1989 Mineral phases of calcium phosphate. Anat Rec 224:
234–241.
29. Boskey AL, Pleshko Camacho N, Mendelsohn R, Doty SB,
Binderman I 1992 FT-IR microscopic mappings of early min-
eralization in chick limb bud mesenchymal cell cultures. Calcif
Tissue Int 51:443–448.
Address reprint requests to:
Michael A. Ignelzi, Jr., D.D.S., Ph.D.
Department of Orthodontics and Pediatric Dentistry
The University of Michigan School of Dentistry
1011 North University Avenue
W. K. Kellogg Building, Room 2532
Ann Arbor, Michigan 48109-1078, USA
16. de Grauw CJ, de Bruijn JD, Otto C, Greve J 1996 Investigation
of bone and calcium phosphate coatings and crystallinity de-
termination using Raman microspectroscopy. Cell Mater Received in original form May 21, 2001; in revised form January
6:57–62. 7, 2002; accepted January 22, 2002.