Supplementary Material for - Science

www.sciencemag.org/content/347/6223/746/suppl/DC1

Supplementary Material for Amorphous intergranular phases control the properties of rodent tooth enamel Lyle M. Gordon, Michael J. Cohen, Keith W. MacRenaris, Jill D. Pasteris, Takele Seda, Derk Joester* *Corresponding author. E-mail: [email protected] Published 13 February 2015, Science 347, 746 (2015) DOI: 10.1126/science.1253950 This PDF file includes: Materials and Methods Figs. S1 to S14 Tables S1 to S3 Full Reference List

Materials and Methods Consumables Unless otherwise specified, the following chemicals were used without further purification: potassium hydroxide (KOH), sodium hydroxide (NaOH), acetone [(CH3)2CO] (VWR, West Chester, PA); bromoform (CHBr3), Fe(III) nitrate nonahydrate [Fe(NO3)3∙9H2O], Fe(III) phosphate dihydrate (FePO4∙2H2O), sodium fluoride (NaF), calcium nitrate tetrahydrate [Ca(NO3)2∙4H2O], magnesium nitrate hexahydrate [Mg(NO3)2∙6H2O], ammonium phosphate dibasic [(NH4)2HPO4], sodium phosphate (Na3PO4), lactic acid (C3H6O3), magnesium oxide (MgO) (Sigma-Aldrich, St. Louis, MO); Epo-Tek 301 epoxy (Epoxy Technology, Billerica, MA); CarbiMet II SiC grinding paper, Microcut SiC grinding paper, Metadi supreme polycrystalline aqueous diamond polishing suspension, Masterprep Alumina suspension, Trident polishing cloth, Chemomet polishing cloth (Buehler, Lake Bluff, IL); Conductive Liquid Silver Paint (Ted Pella, Redding, CA); Super Glue Cyanoacrylate Adhesive (3M, St. Paul, MN); Ultrapure Water (ρ = 18.2 MΩ·cm) was prepared with a Barnstead Nanopure UF+UV ultrapure water purification system (Thermo-Fisher Scientific, Waltham, MA). Reference Materials Unless otherwise stated, phase purity of reference materials was verified by comparing powder X-ray diffraction (XRD) patterns to literature and simulated patterns. Where required, chemical composition was confirmed by inductively coupled plasma mass-spectroscopy (ICP-MS). Mg-substituted ACP (Mg-ACP) was synthesized based on a modification of the protocols given by Holt et al. and Kibalczyc et al. (35, 36). Briefly, a solution of Na3PO4 (328 mg, 20 mmol) in water (100 mL) was rapidly mixed with a solution of Ca(NO3)2∙4H2O (6.85 g, 29 mmol) and Mg(NO3)2∙6H2O (3.08 g, 12 mmol) in water (100 mL). The resulting white precipitate was immediately collected by vacuum filtration, washed with water, and dried under vacuum to give a white powder. By ICP-MS, the Mg/Ca ratio in Mg-ACP as synthesized was 0.3 (6 wt% Mg). Whitlockite was prepared following the protocol by Fadeev et. Al (37). Briefly, a solution of Ca(NO3)2∙4H2O (23.6 g, 100 mmol) and Mg(NO3)2∙6H2O (28.5 g, 111 mmol) in water (100 mL) was stirred at 25˚C. A solution of (NH4)2HPO4 (8.85 g, 67 mmol) in water (100 mL) was added dropwise, using an addition funnel over the course of ~10 minutes, during which a white precipitate formed. The resulting suspension was stirred for an additional 4 h at 25 ˚C. The precipitate was collected by vacuum filtration, washed with water, and dried overnight at 100 °C in air to give a white powder. The powder was subsequently heated to 700 °C for 3 h in air to give whitlockite. By ICP-MS, the Mg/Ca ratio in whitlockite as synthesized was 0.075 (1.6 wt% Mg). “2-line” ferrihydrite was synthesized based on a modification of the protocol given by Cornell and Schwertman (38). A solution of Fe(NO3)3∙9H2O (8.08 g, 20 mmol) in water (100 mL) at 25˚C was rapidly titrated to pH 7 with 1M aqueous KOH. The resulting red-brown precipitate was collected by vacuum filtration, washed extensively with water, frozen in liquid nitrogen, and 2

lyophilized to give a brown powder. The lyophilized powder was ground with an agate mortar and pestle. Goethite was synthesized based on the protocol given by Cornell and Schwertman (38). Briefly, a solution of KOH (1.91 g, 34 mmol) in water (68 mL) was rapidly added to a stirred solution of Fe(NO3)3∙9H2O (15.35 g, 38 mmol) in water (38 mL) at 25 ˚C in a polyethylene container. Water was added to the mixture to a final volume of 750 mL. The container was then sealed, and heated at 70 °C for 5 days. The resulting yellow-brown precipitate was collected by vacuum filtration, washed extensively with water, and dried under vacuum. Dry powder was ground in an agate mortar and pestle. Geological reference minerals originated from the following localities: Dolomite: York, PA; Mg/Ca = 1.0 (13.5 wt% Mg). Huntite: Tea Tree Gully, Australia; Mg/Ca = 3.5 (22.8 wt% Mg). High-purity synthetic hydroxylapatite was purchased from Sigma-Aldrich, product number 57479, lot number MKBD1322, purity ≥ 99.995%; Certificate of analysis indicates 2.9 ppm Mg. Amorphous iron (III) phosphate dihydrate and magnesium oxide were purchased from SigmaAldrich. Rodent Incisors Incisors were sourced from Dr. C. Whyne and Dr. M. Akens, Sunnybrook Health Sciences Centre, Toronto, Ontario (Rat, Rattus norvegicus); Dr. A. Deymier-Black, Washington University, St Louis, MO. (Mouse, Mus musculus); Dr. C. Newcomb and Dr. S. Sur, Northwestern University, Chicago, IL. (Mouse, Mus musculus); Furries Leather Shoppe, Okanogan, WA. (Rabbit, Oryctolagus cuniculus); Kevin Thon, Wykoff, MN. (North American Beaver, Castor canadensis). For a comparison of the enamel structure of these species see Figure S1 for lactic acid etched sections and Table S3 for characteristic dimensions. Preparation of enamel samples for further analysis APT and TEM sample preparation and SEM imaging was performed on epoxy embedded and polished cross-sections of mandibular incisor. 29 APT data sets from 11 mandibular incisors of 6 mice and 3 data sets from 1 mandibular incisor of 1 rat were produced. For quantitative etching experiments, a beaver incisor was cut in a plane orthogonal to the cervical-incisal direction into nine sections ~5 mm in width. From six of these, pigmented enamel was mechanically removed by mechanical grinding with a Dremel rotary tool (Robert Bosch Tool Corporation, Mount Prospect, Illinois) equipped with a tungsten-carbide burr. Three of the beaver samples from which the pigmented enamel had been removed were exposed to fluoride as described above. A rabbit incisor was similar cut into three sections ~5 mm in width. For Fe-XAS, the enamel from the tips of two mandibular incisors from one beaver was mechanically removed as single pieces approximately 2 cm in length. 3

For Mg-XAS, four mouse mandibular incisors from two animals were ground to a powder with an agate mortar and pestle and pooled. Enamel was isolated by centrifugation in a bromoformacetone mixture with specific gravity of 2.7 (8 vol% acetone). The sinking enamel fraction was separated from the floating fraction (dentin, bone and cementum) by freezing the centrifuge tube containing the bromoform-acetone solution in an acetone-dry ice mixture and subsequently sawing the frozen tube in half. The two fractions were allowed to thaw in separate tubes, collected by centrifugation, washed extensively with acetone, and dried under vacuum. Nanoindentation was performed on longitudinal and transverse cross-sections of two mandibular incisors from one rat and one mandibular incisor from one beaver. Raman spectra were generated from a fragment of pigmented enamel of one mandibular incisor of one beaver. Mossbauer and XRD data was collected on pigmented enamel mechanically removed from eight incisors by grinding with a Dremel rotary tool equipped with a tungsten-carbide burr, and pooled. Embedding, Grinding and Polishing Lower (mandibular) incisors were dried in air following extraction and embedded in Epo-Tek 301 epoxy and polymerized overnight at 25 °C. Embedded samples were ground using progressively finer grits of Buehler SiC grinding paper (400, 600, 800 & 1200 grit). Ground samples were polished using 3 µm and 1.0 µm polycrystalline aqueous diamond polishing suspensions on a Buehler Trident polishing cloth. After a final polishing step using 0.05 µm Al2O3 suspension on a Buehler Chemomet polishing cloth, samples were rinsed with water and dried under flowing argon gas. Fluoride Exposure Freshly ground and polished mouse incisor cross-sections were submerged in 250 mM aqueous NaF solution and maintained at 37 °C for 24 hr on a rotator. Following fluoride exposure, samples were rinsed with water and dried under flowing argon gas. Beaver incisor samples with exposed outer enamel (pigmented enamel removed) were incubated in an aqueous solution of NaF (250 mM) at 37 °C for 24 hr on a rotator. Qualitative Acid etching Synthetic hydroxylapatite powder (Sigma-Aldrich) was pressed into 6 mm discs and sintered at 1200 °C for 3 hr in alumina crucibles in air. Samples were removed from the furnace and allowed to cool to room temperature before epoxy embedding. Samples were ground and polished as described above.

4

Freshly polished, epoxy-embedded samples (enamel cross sections and sintered OHAp) were etched for 15 seconds (Figure 2) or 1 minute (Figure 1) at 25 °C in 250 mM aqueous lactic acid adjusted to pH 4.0 with NaOH. For a comparison of mouse incisor enamel before and after etching see Figure S12. For a comparison of lactic acid-etched enamel of different species please see Figure S1. Coating For SEM imaging, samples were secured to an aluminum stub with cyanoacrylate adhesive, coated with ~5 nm of Pt for imaging or ~20nm of C (graphite target) for EDS with an Ion Beam Sputter Deposition and Etching System (IBS/e, South Bay Technologies, San Clemente, CA) operating at a base pressure of < 10-4 Pa and working pressure of 10-2 Pa Ar, with two ion guns operating at 8 kV and 3 mA per gun. The coating was grounded to the stub with conductive liquid silver paint. Scanning Electron Microscopy SEM was performed with an FEI Helios Nanolab (Hillsboro, OR) operating at 5 keV with 0.1 – 0.7 nA probe current. EDS measurements were performed at 10-20 keV with 10-20 nA probe current. The SEM was equipped with an energy dispersive silicon drift detector (Bruker, Berlin, Germany) with an ultrathin window (Moxtek AP3.3, Orem, UT). Sections were individually scanned for 5-10 minutes each. Qualitative iron concentration maps were produced based on total detected Fe-Kα fluorescence X-ray counts per pixel. Transmission Electron Microscopy Sample Preparation Transmission electron microscopy (TEM) lamellae were prepared from a polished mouse incisor cross section following established procedures with a DualBeam scanning electron microscope (SEM) and focused ion beam (FIB) instrument (Helios NanoLab, FEI, Hillsboro, OR)(39). A strip of platinum (FIB-Pt) was deposited over a region of interest using the ion beam (30 kV, 93 pA) to locally decompose an organometallic precursor gas,(trimethyl)methylcyclopentadienylplatinum (CH3)3Pt(CpCH3)). A trench was then milled out (30 kV, 6.5 nA) on each side of a 2 µm thick slice of material. The slice of material was cut free (30 kV, 2.8 nA) from the substrate on three sides leaving only a small connecting bridge. An in situ tungsten nanomanipulator probe (Omniprobe) was attached to the free side of the substrate using FIB-Pt (30 kV, 93 pA). The remaining connection to the substrate was milled away (30 kV, 93 pA) and the probe was retracted with the sample. The sample was then welded to a copper TEM half-grid (Omniprobe) using FIB-Pt and the connection to the probe was milled away (30 kV, 93 pA). The lamella was successively thinned to ~100nm at 30 kV (93 pA) at a 1-2º angle grazing incidence milling condition. The majority of the surface amorphization and gallium implantation was removed by low angle milling (~7º) at 5 and 2 kV (28 pA) and the sample was thinned to approximately 6080nm.

5

Transmission Electron Microscopy TEM was performed with a Hitachi H-7700 (Hitachi High-Technologies Science America, Northridge, CA) operating at 120 kV. Atom Probe Tomography Sample Preparation Samples for APT were prepared using the dual-beam SEM/FIB instrument (Helios Nanolab, FEI, Hillsboro, Oregon) using established protocols(18, 40, 41). A rectangular strap of platinum (FIBPt) was deposited over a region of interest on the polished cross-section using the ion beam (30kV, 93pA) to locally decompose an organometallic precursor gas, (trimethyl)methylcyclopentadienyl-platinum [(CH3)3Pt(CpCH3)], over a region of interest (2 x 25 µm) on a polished cross-section. A wedge of material below the Pt strap was cut out on three sides. The wedge was attached to an in-situ nano-manipulator (Omniprobe, Dallas, TX) using FIB-Pt before cutting the final edge free. Segments 1-2µm wide were cut from the wedge and sequentially affixed to the tops of Si posts in an array (Cameca Scientific Instruments, Madison, WI) with FIB-Pt. Each tip was shaped and sharpened using annular milling patterns of increasingly smaller inner and outer diameters. The majority of the amorphized surface region and implanted gallium in the tip surface was removed by milling at 2 kV, 0.4 nA. Atom Probe Tomography Atom probe tomographic analyses were conducted in a Cameca local-electrode atom-probe tomograph (LEAP 4000XSi, Cameca, Madison, WI) using a pulsed laser (λ = 355 nm, 200-250 kHz, 50-150 pJ per pulse). The DC potential on a microtip during APT was controlled to maintain an evaporation rate of 0.0025 or 0.005 ions per laser pulse. The base temperature of the microtip was maintained at 40 K and the ambient vacuum pressure was below 10-8 Pa. Peak ranges were defined as the entire visible peak and background subtraction was performed using side-band subtraction (42). For spectra see Figure S3. Three-dimensional reconstruction of APT data was performed using the Cameca integrated visualization and analysis software (IVAS) based on published algorithms, assuming a hemispherical tip shape(43, 44). Standard reconstruction parameters, field factor (kf = 3.3) and image compression factor (ξ = 1.33) were used with an electric field-dependent tip radius (r). The average evaporation field (Fe) of the enamel apatite (14 V∙nm-1) was determined from SEM and/or TEM images of microtips after APT analysis. Atomic volumes for the reconstruction were calculated based on the hydroxyapatite crystal structure(26). Validity of reconstruction parameters was confirmed with correlative TEM imaging of a subset of APT samples before and after atom probe analysis (Fig. S11). For an overlay of Mg and F positions in fluoride-treated enamel see Figure S13. Compositions extracted from atom probe data sets are reported as mole fractions, using the unit “at%” (rather than the identical “mol%”) that is commonly used in the atom probe literature.” Compositions are reported in Table S1. The detection limit for Mg is approximately 0.006 at% Mg based on 3 times the standard deviation of the pre-peak background and a Mg2+ peak width (FWHM) of 0.04 m/z. The average

6

baseline noise before the Mg2+ peak is 2 counts/bin (0.001 m/z bin width). The highest bin counts in the 24Mg2+ peak for the whole dataset is 350 counts. The detection limit for Fe is approximately 0.05 at% Fe based on 3 times the standard deviation of the pre-peak background and a Fe2+ peak width (FWHM) of 0.04 m/z. The average baseline noise before the Fe2+ peak is 22 counts/bin (0.001 m/z bin width). The highest bin counts in the 56 2+ Fe peak for the whole dataset is 1400 counts. Powder X-ray diffraction For pigmented enamel, 50 mg aliquots were heated for 2 hr at 200, 400, 600, 800 and 1000 °C in alumina crucibles in air. Samples were removed from the furnace and allowed to cool to room temperature. Reference materials were analyzed as is. Samples were suspended in acetone and dispersed on the surface of an off-axis cut quartz single crystal plate (MTI Technologies, Richmond, CA). Remaining acetone was allowed to evaporate. Samples were analyzed with a Scintag diffractometer with an X-ray source equipped with a Cu anode operating at 40 kV and 20 mA. Diffracted intensity was recorded between a 2θ° of 5° and 90° with a step size of 0.05° and a dwell time of 2 seconds per step. An energy dispersive X-ray detector was used to count only diffracted Cu-Kα X-rays and exclude Cu-Kβ X-rays and X-ray fluorescence. Diffractograms are given in Figure S8. Quantitative Acid etching The lingual dentin surfaces and mesial and distal enamel edges of all samples were masked from etching using clear nail varnish (Sally Hansen, COTY Inc, New York, NY). A nail varnishcoated silicon wafer was used as a control. The surface area of enamel not protected by nail varnish was determined using a calibrated stereomicroscope. Each sample was placed in a 2 mL tube containing 1.5 mL aqueous lactic acid (250 mM, pH 4.0). Samples were incubated on a rotator at 37 °C. At the 1, 10 and 100 min time-points, 200 µL aliquots were removed from the etchant solution for analysis by ICP-MS. Mass loss during etching was determined by calculating the total mass of Ca, P, Fe, Mg and Sr in solution and accounting for oxygen by stoichiometry. Trace levels of organics, other metals, water, and carbonate, were not accounted for. The etching rate was determined by fitting a linear model to the cumulative mass loss over time. For plots of mass loss vs. etch times see Figure S14. Microwave Digestion Powdered mineral samples (~10 mg) were added to pre-weighed 10 mL NalgeneTM Oak Ridge high-speed TeflonTM FEP centrifuge tubes (Thermo Fisher Scientific, Waltham, MA, USA). 200 µL of TraceSelect® nitric acid (≥ 69%, Sigma-Aldrich Chemical Co., St. Louis, MO, USA) was then added to the TeflonTM FEP centrifuge tubes and capped. Samples were then placed into TeflonTM Microwave digestion vessels and assembled into the high-pressure rotor (HPR 1000/10) of a Milestone EthosEZ closed microwave digestion system (Milestone, Sorisole, Italy). Samples were treated by microwave irradiation using the following parameters: ramp to 120 °C for 20 min. at 600 W, 120 °C hold for 1 h at 400 W, exhaust for 30 min. Following 7

digestion, samples were weighed out to account for sample loss and prepared for ICP-MS analysis. ICP-MS 10-100 µL aliquots of the microwave-digested mineral samples or lactic acid etching solution were placed in metal-free 1.5 mL tubes and diluted up to 10-fold with ultrapure de-ionized water (samples were diluted in order to obtain accurate ppb levels of selected elements). ICP-MS samples were then prepared by adding pre-determined amounts of diluted sample followed by the addition of TraceSelect® nitric acid to a final concentration of 3% v/v, 5 ng/g of a multielement internal standard (CLISS-1 standard containing Bi, Ho, In, 6Li, Sc, Tb, and Y from Spex CertiPrep, Metuchen, NJ, USA), and ultrapure de-ionized water to obtain a final volume of 5 mL. Standards for Na, Mg, P, Ca and Fe (Inorganic Ventures, Christiansburg, VA, USA) were prepared at 1000, 500, 250, 125, 62.5, 31.25, 15.625, 7.8125 and 3.90625 ng/g containing 5 ng/g multi-element internal standard and 3.0% (v/v) TraceSelect® nitric acid. ICP-MS analysis was performed on a computer-controlled Thermo X series II ICP-MS (Thermo Fisher Scientific, Waltham, MA, USA) equipped with a CETAC 260 autosampler (Teledyne CETAC Technologies, Omaha, NE, USA). Each sample was acquired using 1 survey run (10 sweeps) and 3 main peak jumping runs (100 sweeps per run). The isotopes selected for analysis were 23Na (LOD = 1 ppb), 24,25,26Mg (LOD = 0.1 ppb), 31P (LOD = 20 ppb), 44Ca (LOD = 50 ppb), and 57Fe (LOD = 5 ppb) using 45Sc, 89Y, and 115In for internal standardization. Instrument performance was optimized using custom autotune sequence and TuneA solution (Thermo Fisher Scientific, Waltham, MA, USA), followed by a custom performance report. Mg K-edge XAS XAS was performed at the Spherical Grating Monochromator (SGM, 11ID-1) beamline of the Canadian Light Source in Saskatoon, SK. Dry samples were prepared by spreading a small amount of powder on graphite tape (Ted Pella). Samples were scanned relative to the edge Mg K-edge (1303 eV) from -60 to -12 eV in steps of 2 eV, -12 to -8 eV in steps of 0.5 eV, -8 to 30 eV in steps of 0.1 eV, 30 to 190 eV in steps of 0.2 eV, 190 to 300 eV in steps of 0.3 eV and 300 to 400 eV in steps of 0.5 eV with a constant dwell time of 2 seconds/step. Monochromator energy calibration was performed by setting the first absorbance maxima of the MgO reference sample spectra to 1309.5 eV. X-ray fluorescence intensity was measured simultaneously with four solid state silicon drift energy dispersive X-ray detectors (Amptek, Bedford, MA). Incident flux was measured by recording the current from a gold mesh upstream. The exit slit was adjusted and the undulator detuned to reduce flux to prevent saturation of X-ray fluorescence detectors when measuring concentrated reference samples. Between 1 and 7 scans were collected for each sample and averaged. No beam-induced changes were observed when comparing sequential spectra. The Mg X-ray fluorescence intensity was isolated from the total fluorescence intensity containing contributions from X-ray fluorescence from other elements and the scattered incident beam using custom written code in Mathematica (Wolfram Research, Champaign, IL). For Mg K-edge XANES spectra, see Figure S4A.

8

Mg EXAFS Absorption data were normalized, background subtracted using AUTOBK, and converted to kspace using Athena(45). Edge energy (E0) was set to the maximum of the first derivative of the absorption spectra. χ(k) data were weighted by k2 and Fourier transformed over a k range of 2-9.5 Å-1, applying a Hanning window with a sill width of 1 Å-1. Theoretical photoelectron scattering amplitudes and phase shifts based on the crystal structures of dolomite (46), huntite (47), whitlockite (48), and hydroxyapatite (26) were calculated using FEFF6 (49). Shell-by-shell fitting of the EXAFS data was performed in R-space using Artemis (45). An energy shift parameter (E0) was maintained constant for the scattering paths but allowed to vary between samples. The amplitude reduction factor (S02 = 0.8) was determined based on a fit to the dolomite, huntite and whitlockite spectra with coordination numbers constrained based on their respective crystal structures. Multiple scattering in the carbonate reference samples was accounted for following Reeder et al (50). Enamel and ACP EXAFS spectra were fit using a model based on the Ca[II] site of OHAp, consisting of a single Mg-O and two Mg-P scattering paths (25, 51). To minimize the number of fitting parameters, the coordination number and σ2 for the two Mg-P paths were constrained for each sample but allowed to vary between samples. For Mg K-edge EXAFS spectra, see Figure S4B,C. For fitting results see Table S2. Fe K-edge XAS Powdered reference samples that adhered well to Kapton tape (3M 7419, St. Paul MN), were spread uniformly and multiple layers of tape were stacked to optimize X-ray absorption. Poorly adhering powders were diluted with microcrystalline cellulose (Sigma-Aldrich 310697) and pressed into a pellet with a diameter of 13 mm and a thickness of ~200 µm. EXAFS measurements were conducted at beam line 5-BM-D part of the Dupont-Northwestern-Dow collaborative access team at the Advanced Photon Source at Argonne National Laboratory. The energy of the Si(111) two-crystal monochromator was calibrated by assigning an energy of 7112 eV to the first zero-crossing of the second derivative (inflection point) of the absorption edge of an iron foil measured in transmission. The monochromator was detuned to eliminate harmonics. Absorption spectra of reference standards were recorded at the Fe-K edge in transmission by simultaneously measuring the incident and transmitted flux with two ion chambers (FMBOxford IC Spec). An iron foil was placed behind the second ion chamber before the third ion chamber and measured simultaneously with each sample for energy calibration. Due to reduced iron concentration of pigmented enamel compared to reference materials, absorption spectra of these samples were recorded by measuring the total Fe-Kα fluorescence yield with two Vortex ME-4 four element silicon drift detectors (Hitachi High-Technologies Science America, Northridge, CA) placed close to the sample. The whole enamel segments with the dentin removed were placed directly into the beam. Al foil was placed over the detectors to absorb CaK fluorescence. Escape depth of the Fe-Kα fluorescence X-rays is approximately the same as the thickness of the pigmented enamel, 10 μm, such that the majority of the signal originates from the pigmented enamel layer. Background was collected from -150 eV to -20 eV (relative to the edge) with steps of 5 eV; XANES data were collected from -20 eV to +35 eV (k = 3 Å-1) with steps of 0.1 eV. EXAFS data were collected from +35 eV to +750 eV (k = 3 Å-1 – 14 Å-1) with a k1.5 weighted count time 9

and a k-increment of 0.05 Å-1. A base count time of 1 second and a final count time of 10 seconds were used. Absorption data were normalized using Athena (45). Between 2 and 16 scans were collected for each sample and averaged. No beam-induced changes were observed when comparing sequential spectra. For Fe K-edge XANES spectra, see Figure S5A, for pre-edge fit see Figure S9. Fe EXAFS The absorption data were normalized, background subtracted using AUTOBK, and converted to k-space using Athena(45). A constant edge energy (E0) of 7128 eV was used. χ(k) data were weighted by k3 and Fourier transformed over the k range 3-12 Å-1 and applying a Hanning window with a sill width of 3 Å-1. Due to reduced signal to noise ratios in the spectra from pigmented enamel, a reduced k range of 3-9.5 Å-1 was used. Theoretical photoelectron scatting amplitudes and phase shifts based on the crystal structure of goethite(52) were calculated using FEFF6(49). The amplitude reduction factor (S02 = 0.95) was determined based on a fit to the reference goethite spectra with coordination numbers constrained based on the crystal structure(52). Shell-by-shell fitting of the EXAFS data was performed in R-space using Artemis(45). Four paths were fit for each sample, Fe-O, two Fe-Fe paths and an Fe-O-O multiple scattering path. An energy shift parameter (E0) was maintained constant for the scattering paths but allowed to vary between samples. Triangular Fe-O-O multiple scattering paths within the FeO6 octahedra were included following Mikutta(53) with radius, coordination number, and σ2 constrained based on the Fe-O paths, specifically: RFe-O-O = RFe-O(1+ 2/2); CNFe-O-O = 4CNFe-O; σ2Fe-O-O = σ2Fe-O. For Fe K-edge EXAFS spectra, see Figure S5B,C. For fitting results see Table S2. Fe EXAFS linear combination fitting Least squares linear combination fitting(54) was performed with Athena (45) using k3-weighted spectra over a k-range of 2-9.5 Å-1. For linear combination analysis see Figure S10. Raman Spectroscopy Fragments of pigmented enamel were analyzed by Raman spectroscopy using a fiber-optically coupled Raman microprobe (HoloLab Series 5000 Raman Microscope, from Kaiser Optical Systems). Laser excitation at 532 nm was provided through a single-mode optical fiber which was 8 µm in diameter. An 80x ultra-long-working- distance objective (Olympus, Japan) with a numerical aperture of 0.75 was used to focus 1 to 10 mW of laser power into a spot ~1 µm in diameter on the surface of the sample. A volume holographic transmission grating split the collected signal into two beams that were imaged simultaneously onto a CCD array detector (Andor Technology, Ireland) with 2048 channels across the array. This permitted the 100 to 4000 Δcm-1 region studied in this work to be analyzed simultaneously at a spectral resolution of 2.5 cm-1. The absolute wavenumber recorded and the exact position of the laser line were calibrated using Ne and Ar calibration lamps. The relative wavenumber was calibrated by reference to cyclohexane and monitored by daily analysis of a silicon wafer whose recorded peak position is 520.6±0.1 Δcm-1. The intensity scale was calibrated using the white-light spectrum from a temperature-calibrated tungsten lamp. Spectral acquisition times per analysis were typically 64 x 10

4 seconds. All Raman analyses were carried out in a laboratory air environment. For Raman spectra see Figure S7. Reduced laser power (1 mW) was necessary when measuring the pigmented enamel sample to prevent changes to the sample due to laser induced heating. Pigmented enamel spectra acquired with high laser power resembled spectra from the mineral Xanthoxenite [Ca4Fe3+2(PO4)4(OH)2·3H2O], RRUFF ID R050234(55). Mössbauer Spectroscopy Pigmented enamel was mechanically removed from 8 beaver incisors and pooled. Mössbauer spectra were recorded at room temperature with a constant-acceleration spectrometer (Wissel GMBH, Germany) in a horizontal transmission mode using a 50 mCi 57Co source. Appropriate amounts of sample was loaded into the sample cell for the measurement. The velocity scale was normalized with respect to metallic iron at room temperature; hence all isomer shifts were recorded relative to metallic iron. Mössbauer spectra were fitted by assuming Lorentzian line shapes using the NORMOS (Wissel GMBH) least-square fitting program. The isomer shifts, quadrupole splittings, peak line widths and peak areas were determined from the fitted subspectra. For Mössbauer spectra see Figure S7. Nanoindentation A Hysitron (Minneapolis, MN) 950 Triboindentor with a diamond Berkovitch indenter was calibrated in air and on a standard calibration quartz crystal. Calibration on quartz was performed before and after sample measurements to ensure that the system was not drifting during experimentation. Arrays of indents were made covering the outermost 50 µm of the enamel, at a distance of ~2.5 µm between indents. Individual indents were made following a standard trapezoidal function with three successive 5-second segments: loading, hold and unloading steps. The peak load force was 1500 µN. Indentation load cycles were processed with standard methods using the Hysitron software. Reported indentation hardness values are averages of N = 3-17 individual indents. For nanoindentation hardness plots see Figure S2.

11

Fig. S1

Enamel comparison: Scanning electron micrographs of lactic acid etched cross-sections of beaver, rabbit, rat, and mouse incisor enamel. Inner, transition, outer, and pigmented enamel regions are labeled. Thickness of enamel layers are summarized in Table S3.

12

Fig. S2

Enamel nanoindentation and relative iron concentration. Nanoindentation was performed to determine indentation hardness of pigmented and non-pigmented enamel in rat (left column) and beaver (right column) enamel in both longitudinal (top row) and transverse (bottom two rows) cross sections. Black arrows on incisor cross-section schematics indicate the approximate area where indentations were performed. The approximate pigmented area (based on SEM imaging of etched microstructure and SEM-EDS normalized Fe concentration measurements) is shaded brown in the schematics and on the plots. Outer (OE) and transition (TE) enamel are indicated on plots. Areas at the far side of the tooth where pigmentation is absent were used as a control (bottom row). Relative iron concentration was measured by SEM-EDS on longitudinal crosssections of rat and beaver enamel. Each data point is an average of N independent indents, where N = 11-17 (A), 10-13 (B), 3-4 (C), 4-8 (D), 3-4 (E), 3-4 (F). Data points where N