Dental enamel is arguably the toughest and most resilient biological tissue1. Despite the extraordinary mechanical properties that enable withstanding of fatigue and wear, dental enamel has limited ability for self-repair or renewal, unlike bone2. Though considered biologically produced analogues of hydroxy(l)apatite [Ca5(PO4)3OH], the apatites of bone and dental enamel are remarkably dissimilar3. In enamel, through pH-dependent supramolecular self-assembly, a protein called amelogenin plays a key role in guiding the elongated and highly oriented growth of apatite in enamel4,5. As a direct consequence of chemical gradients, the distinct core–shell structure and the residual stresses thus arising significantly impact the dissolution behaviour of human enamel crystallites6. In rodents, the continuously growing incisor teeth7 are protected by Fe-containing pigmented enamel while molar teeth are covered with Mg-rich unpigmented enamel8.

Raman spectroscopy can distinguish between different mineralised biological tissues, including enamel, dentine, cementum, and bone, and reveal vital information about the biological processes underpinning their formation and assembly9. Certain aspects of mineral composition, e.g., carbonate ion (CO32−) incorporation, provide insights into the pathways of bioapatite formation10. The degree of carbonation (i.e., CO32− content), ultimately, affects the long-range order or mineral crystallinity11. Additionally, factors such as local pH influence the availability and incorporation of acid phosphate (HPO42−) into the apatite lattice—i.e., with greater HPO42− being introduced under acidic conditions12. HPO42− containing phases such as octacalcium phosphate are frequently encountered in the mineralised dental biofilm13. And although human and bovine enamel are believed to contain about 5 wt% HPO42−14, Raman studies of human premolar teeth have not been able to detect non-apatitic environments15,16.

Chemical and structural characterisation of biological tissues using Raman spectroscopy is often plagued by the intrinsic autofluorescence17. This process originates from various organic moieties18, but can be suppressed by methods such as deproteinisation with sodium hypochlorite (NaOCl)19. This work uses high-resolution Raman spectroscopy together with scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy (EDX) to probe the major compositional differences (particularly the HPO42−, CO32−, and OH environments) between two distinct types of enamel—pigmented incisor enamel (PIE) and unpigmented molar enamel (UME) in the rat (Rattus norvegicus), with and without deproteinisation. Furthermore, the Raman spectra are compared with geologic hydroxy(l)apatites from the RRUFF™ database20, such as the well-characterised and highly crystalline Holly Springs hydroxy(l)apatite21,22, which are typical reference materials in crystallographic studies of biogenic calcium phosphates23.


The major X-ray emission lines Ca Kα (3.692 keV), Ca Kβ (4.013 keV), P Kα (2.014 keV), Fe Kα (6.403 keV), Fe Kβ (7.058 keV), Fe Lα (0.705 keV), and Mg Kα (1.254 keV) confirmed the differences in Ca/P ratio, Fe content, and Mg content between PIE and UME. The Ca/P ratio of UME (1.30 ± 0.02 at.%) is higher (p < 0.05) than PIE (1.19 ± 0.01 at.%). Similarly, the Mg/Ca ratio of UME (0.011 ± 0.001 at.%) is higher (p < 0.05) than PIE (0.002 ± 0.001 at.%). Whereas the Fe content (Fe/Ca ratio) of UME is negligible, PIE shows significant Fe enrichment (~ 0.15 ± 0.01 at.%), which results in higher (Ca + Mg + Fe)/P ratio (~ 4.2 ± 0.9 at.%), indicating the incorporation of iron at the expense of calcium (Fig. 1). Small amounts of Cl are also detected, which may occur by partial substitution of OH24. Scanning electron microscopy (SEM) of the enamel surface reveals that UME is smooth while PIE is grainy in comparison. UME etches homogeneously with orthophosphoric acid (H3PO4) but PIE is mildly resistant to acid attack, as evident from the islands of an incompletely removed surface layer. When visualised after acid etching, bundles of hydroxy(l)apatite nanowires comprise the enamel prisms, where PIE prisms are wider and longer than UME prisms.

Figure 1
figure 1

Elemental composition and microstructure. (A) PIE is Fe-rich while UME is Mg-rich (EDX). Spectra are normalised to the Ca Kα peak (at ~ 3.692 keV). (BI) PIE surface is grainy (B) and UME surface is smooth (F). Insets: Respective lower magnifications (Scale bars = 25 µm). Enamel prisms in PIE (CE) and UME (GI) after H3PO4 etching. BSE = Backscattered electron mode. SE = Secondary electron mode. Scale bars in (B), (C), (F), and (G) = 10 µm, (D) and (H) = 5 µm, (E) and (I) = 2 µm.

High-resolution Raman spectra were acquired over the 350–1100 cm−1 wavenumber range using the 2400 g mm−1 grating. In enamel and mandibular bone, the major Raman domains are \(\nu\)2 PO43− (symmetric bend) at 370–490 cm−1, \(\nu\)4 PO43− (asymmetric bend) at 545–635 cm−1, \(\nu\)1 PO43− (symmetric stretch) at 960 cm−1, and \(\nu\)3 PO43− (asymmetric stretch) at 1015–1095 cm−125. Estimated from the difference in the background signal over the 350–1100 cm−1 spectral range, UME generates 381 ± 18% higher fluorescence than PIE, which decreases to 262 ± 9% after deproteinisation (Fig. 2). Normalised to the polarisation-insensitive \(\nu\)2 PO43− band 26, the \(\nu\)1 PO43− band (~ 960 cm−1) of enamel is substantially stronger than bone.

Figure 2
figure 2

Background fluorescence and \(\upnu\)1 PO43− intensity. (A) UME generates stronger fluorescence than PIE, both before (NaOCl−) and after (NaOCl+) deproteinisation. Unprocessed spectra without baseline subtraction and cosmic ray removal (averaged Raman spectra). Broken lines indicate the background fluorescence profile. (B) Baseline corrected and normalised spectra. (C) \(\upnu\)1 PO43− band.

The \(\nu\)2 PO43− band consists of sub-components at 428 cm−1 and 450 cm−1, of which the 428 cm−1 sub-component tends to be stronger in carious enamel15 and synthetic hydroxy(l)apatite27 (Fig. 3). The 428/450 cm−1 ratios of PIE (1.05 ± 0.12) and UME (0.95 ± 0.10) are close to unity and comparable (p > 0.15). Upon deproteinisation, the 428/450 cm−1 ratio of UME (0.85 ± 0.05) shows a minor decrease (p = 0.026) and is also lower (p = 0.013) than that of PIE after deproteinisation (1.00 ± 0.11). Bone also shows decreased 428/450 cm−1 ratio upon deproteinisation, from ~ 1.63 to ~ 1.58, and the organic matrix is lost. PIE and UME display comparable \(\nu\) 4 PO43− band profiles. Among the \(\nu\)4 PO43− band sub-components (580 cm−1, 590 cm−1, and 607 cm−1), the 580 cm−1 sub-component in enamel is the strongest while it is significantly weaker for bone and often less intense than the 590 cm−1 sub-component. Furthermore, a shoulder is observed at 621 cm−1 in bone only prior to deproteinisation, and is therefore assigned to the organic matrix25.

Figure 3
figure 3

HPO42−, CO32−, and the organic matrix in PIE, UME, and bone. (A) \(\upnu\)2 PO43− band. Insets: 428/450 cm−1 ratios before (NaOCl−) and after (NaOCl+) deproteinisation. (B) \(\upnu\)4 PO43− band. (C) HPO42−, proline (at 853 cm−1), and hydroxyproline (at 876 cm−1) bands. (D) \(\upnu\)3 PO43−, \(\upnu\)1 HPO42−, \(\upnu\)1 CO32−, and phenylalanine (at 1004 cm−1) bands. Features labelled a, b, d, e, f, g, l, and m represent the organic matrix of bone, c and h indicate HPO42−, i and j are assigned as \(\upnu\)3 PO43−, and k is \(\upnu\)1 CO32−.

Mineral crystallinity, i.e., the inverse full-width at half-maximum (FWHM) of the \(\nu\)1 PO43− band, of enamel is higher than bone (Fig. 4). Upon deproteinisation, the FWHM \(\nu\)1 PO43− of PIE (~ 11.65 cm−1) and UME (~ 12.5 cm−1) remains unchanged while that of bone decreases by ~ 3.4% from 15.97 to 15.43 cm−1. Compared to PIE, the \(\nu\)1 PO43− band of UME is shifted to lower wavenumbers. This shift in \(\nu\)1 PO43− position (~ 0.31 ± 0.06 cm−1) becomes particularly evident after deproteinisation (p = 0.0312). Broad features at ~ 878 cm−1 and 1000 cm−1 are attributable to HPO42− groups in enamel28. The HPO42− content (1000/960 cm−1 ratio) of UME is lower (p = 0.0312) than PIE. The \(\nu\) 3 PO43− sub-components at ~ 1027 cm−1 and ~ 1046 cm−1 are observed consistently for enamel and bone. The \(\nu\)1 CO32− band is centred at ~ 1069 cm−1 for enamel and ~ 1071 cm-1 for bone. The CO32− content (\(\nu\)1 CO32−/\(\nu\)1 PO43− intensity ratio) of bone is substantially higher than enamel, increasing from ~ 0.09 to ~ 0.12 upon deproteinisation. Furthermore, the CO32− content of UME is higher than PIE (p = 0.0312).

Figure 4
figure 4

Compositional differences between PIE and UME. Mineral crystallinity, \(\upnu\)1 PO43− peak position, HPO42− content (1000/960 cm−1), and CO32− content (1070/960 cm−1) of PIE and UME.

Extended-range Raman spectra were acquired over the 800–3700 cm−1 wavenumber range using the 1800 g mm−1 grating (Fig. 5). The \(\nu\) OH region (3460–3660 cm−1) shows an asymmetrical band. The OH content, taken as the integral area ratio between \(\nu\) OH (~ 3573 cm−1) and \(\nu\)1 PO43− (930–990 cm−1), of PIE is ~ 148% greater than UME. The feature at ~ 3618 cm−1 is assigned as Ca(OH)2.

Figure 5
figure 5

Hydroxl (OH) content in PIE and UME (1800 g mm−1 grating). (A) Baseline corrected and normalised spectra. (B) \(\upnu\) OH at 3573 cm−1 and Ca(OH)2 at 3618 cm−1. (C) \(\upnu\) OH content (3573/960 cm−1).

Raman spectra of enamel and bone were compared to three geologic hydroxy(l)apatites. Remarkable similarities are noted between enamel (PIE and UME) and hydroxy(l)apatite from the Wessels mine (Northern Cape Province, South Africa. RRUFF: R130713) (Fig. 6). An unidentified feature is seen at ~ 853 cm−1 for R130713, while the characteristic peaks for HPO42− at 878 cm−1 and 1000 cm−1 are absent. On the other hand, hydroxy(l)apatite from the Sapo mine (Minas Gerais, Brazil. RRUFF: R100225) and Holly Springs (Georgia, USA. RRUFF: R060180) more closely resemble bone. Small amounts of amorphous calcium phosphate, indicated by a shoulder at ~ 950 cm−129, are present in all of the geologic hydroxy(l)apatites.

Figure 6
figure 6

Geologic hydroxy(l)apatites. (A) Overview and detail of the 800–1100 cm−1 region of R130713 (Wessels mine, Northern Cape Province, South Africa); R100225 (Sapo mine, Minas Gerais, Brazil); and R060180 (Holly Springs, Georgia, USA). Inset: An unidentified peak (feature labelled x) is noted at ~ 853 cm−1 for RRUFF R130713. HPO42− (features labelled c and h) is absent for all geologic hydroxy(l)apatites. (B) \(\upnu\)1 PO43−, \(\upnu\)2 PO43−, and \(\upnu\)4 PO43− bands.


Better understanding of enamel composition is essential to develop biomimetic and bioinspired strategies for enamel repair30. Some of the recent and highly divergent approaches to repair enamel include protein order/disorder-guided hierarchical mineralised structures31 and epitaxially-grown hydroxy(l)apatite crystals32. In mineralised biological systems, the presence of iron is often associated with high strength—a prime example being the teeth of the common limpet, thought to be the strongest known biomaterial, where iron-containing filamentous crystals of Goethetite [α-FeO(OH)] comprise the reinforcing phase33. Likewise, the presence iron contributes to the overall mechanical properties of rodent pigmented enamel8. In the northern short-tailed shrew (Blarina brevicauda), iron pigmentation is not confined to the incisors but exists as a general feature of high stress areas on most teeth34. Curiously, iron pigmentation of the dental enamel has also been observed in mammalian species as early as the late Cretaceous period35.

In rodents, differences in enamel architecture between unpigmented molar enamel, which forms during embryogenesis, and pigmented incisor enamel, which forms during post-natal life, relate to genetic control of ameloblast differentiation involving distinct mechanisms at these distinct phases of life36. Autophagy related 7 (ATG7) protein is essential for the secretion of iron from ameloblasts37. Moreover, iron deficiency leads to gross loss of pigmentation and enamel hypoplasia/aplasia38. Although amelogenin plays a fundamental role in achieving the precise crystal habit, the enzyme matrix metalloproteinase-20 prevents protein occlusion inside apatite crystals39.

Fe enrichment of pigmented enamel is possible through partial substitution of Ca2+ without major changes in PO43− internal modes40, although a constriction in lattice parameters is expected41. PIE appears to resist acid attack, which has earlier been attributed to the presence of Ca2+ and Mg2+ substituted ferrihydrite8. However, in the present work, micro-Raman spectroscopy has not revealed evidence of iron oxides or oxy(hydroxides) in PIE42. And though it not straightforward to ascertain the oxidation state of Fe (Fe2+ or Fe3+) from EDX, alone, Fe-L2,3 electron energy-loss near-edge structure (ELNES) of pigmented Fe-rich enamel from the rodent Myocaster coypus suggests a predominantly Fe3+ state43. Under the assumption that Fe occupies Ca sites in iron-pigmented enamel, the Fe/Ca ratio of 0.15 equates to ~ 13% Ca substitution and therefore ~ 5.15% mass difference. Ab initio calculations of 42Ca isotopic substitution for 40Ca, which equates to ~ 5% mass difference at the Ca sites, have revealed that the expected Raman shifts for vibrational modes above ~ 600 cm−1 (for example the \(\nu\)1 PO43− band) do not exceed ~ 1 cm−144. Here, high-resolution Raman spectroscopy reveals this very small shift in \(\nu\)1 PO43− peak position for the first time. In unpigmented enamel, Mg2+ accumulates within intergranular regions of amorphous calcium phosphate6,45. Compared to rat molars, as reported here, the Mg content at the surface of human molars is nearly twice as much at the enamel surface and progressively increases towards the dentinoenamel junction46.

Fourier transform infrared spectroscopy studies have suggested the presence of non-apatitic environments (e.g., HPO42− groups) in porcine enamel47. Here, high-resolution Raman spectroscopy confirms the presence of HPO42− in both pigmented and unpigmented rat enamel. HPO42− is thought to be a precursor phosphate source for enamel apatite48. Therefore, detection of higher HPO42− at the surface of PIE (vs. UME) may be a function of tissue age, as has been reported across different developmental stages of porcine enamel49. It has been suggested that acidic conditions favour the fast growth of highly crystalline hydroxy(l)apatite by dissociating calcium phosphate aggregates into Ca2+ and PO43− ions, which would otherwise block crystal growth and lead to lower crystallinity50. If the higher crystallinity and greater HPO42− content of PIE (vs. UME) can be explained by a more acidic environment, it must be determined how this acidic pH is regulated, e.g., if it is biologically driven. Removal of OH from the local environment through incorporation into the apatite lattice, also more abundant in PIE than in UME, further points towards the presence of acidic conditions. Nevertheless, the OH content of PIE is lower than values of human and boar enamel reported by Pasteris and co-workers27. The anticorrelation between CO32− content and crystallinity with little apparent influence of HPO42− warrants further investigation and raises the question whether crystallinity correlates with CO32− only.

Organic contamination of UME to a greater extent than PIE is hardly surprising since the latter is continually lost to wear and replaced by pristine mineral. Change in the 428/450 cm−1 ratio of UME, from ~ 1 (indicating high symmetry of PO43− groups) to 0.85 after deproteinisation, suggests a reduction in symmetry and that UME is more susceptible than PIE to alterations. The detection of Ca(OH)2 points towards the presence of CaO, which readily reacts with atmospheric humidity51. Finally, simultaneous increases in mineral crystallinity and CO32− content of bone upon deproteinisation are artefactual and imply loss of recently deposited extracellular matrix and poorly crystalline mineral at the bone surface52.

In summary, the chemical contrasts between pigmented and unpigmented enamel in rodents, including HPO42− content, CO32− content, mineral crystallinity, reflect ameloblast function and point towards putative differences in the specific local environmental conditions (e.g., the interplay between pH and the HCO3- buffer system53) of the organic extracellular matrix and matrix metalloproteinase-20 activity during enamel biomineralisation. While the precise functional role of iron in tooth development remains unclear, iron accumulation in rodent incisors (and the presence of iron in mature ameloblasts) is related to the continuously erupting nature of this tooth54. This characteristic feature of rodent incisors also serves to explain the higher HPO42− content of PIE (vs. UME). On the other hand, the high CO32− content of UME is attributed to B-type substitution (i.e., CO32− for PO43−) typical of biological apatites55, and contributes to lower crystallinity together with Mg2+56.

Materials and methods

Hemi-mandibles of adult Sprague Dawley rats, obtained as part of an unrelated study, were fixed in 10% neutral buffered formalin, defatted in acetone (~ 30 min), and stored in Hank’s Balanced Salt Solution (Gibco™) at 4 °C (Fig. 7). The organic constituents were removed by exposure to 10% NaOCl (3 h at room temperature). The experiment was approved by the local Animal Ethics Committee at the University of Gothenburg (Dnr 5.8.18-12983/2021) and performed in accordance with relevant guidelines and regulations.

Figure 7
figure 7

Experimental workflow.

Scanning electron microscopy

Backscattered electron (BSE) imaging and energy dispersive X-ray spectroscopy (EDX) were performed in a Quanta 200 environmental SEM (FEI Company, The Netherlands) equipped with an INCA EDX system (Oxford Instruments GmbH, Wiesbaden, Germany) operated at 1 Torr water vapour pressure, 20 kV accelerating voltage, 0–10 keV spectral energy range, and 10 mm working distance (NaOCl+ ; 1 spot per sample, n = 6). To visualise enamel prisms, deproteinised hemi-mandibles were etched with 85% H3PO4 (90 s at room temperature) and Au sputter coated (~ 15 nm thickness). Secondary electron (SE) imaging was performed in an Ultra 55 FEG SEM (Leo Electron Microscopy Ltd, UK) operated at 5 kV accelerating voltage.

Micro-Raman spectroscopy

Micro-Raman spectroscopy was performed using a confocal Raman microscope (Renishaw inVia Qontor) equipped with a 633 nm laser and LiveTrack™ focus-tracking technology57. The laser was focussed down on to PIE (labial aspect), UME (buccal aspect), and bone (mandibular ramus) surface using a × 50 objective. The Raman scattered light was collected using a Peltier cooled CCD deep depletion NIR enhanced detector. Using the 2400 g mm−1 grating (348–1104 cm−1 wavenumber range, step size of 0.75 ± 0.04 cm−1), Raman spectra were obtained from enamel (9 spots per sample) at 8 s (NaOCl−) or 4 s (NaOCl+) integration time and 10 accumulations, and from bone (one spot per sample) at 10 s (NaOCl−) or 5 s (NaOCl+) integration time and 20 accumulations. Using the 1800 g mm−1 grating (800–3700 cm−1 wavenumber range, SynchroScan wide-range scanning mode; step size of 1.0 ± 0.15 cm−1), Raman spectra were obtained from enamel (3 spots per sample, NaOCl+) at ~ 60 s integration time and 10 accumulations. The laser power at the sample was ~ 15 mW. Background subtraction and cosmic ray removal were performed using intelligent spline fitting in Renishaw WiRE 5.4 software.

Statistical analysis

The Wilcoxon signed-rank test was used for statistical analysis. Mean values ± standard deviations are presented and p values < 0.05 were considered statistically significant.