This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/authorsrights Author's personal copy FT-Raman spectroscopy as a method for screening collagen diagenesis in bone Christine A.M. France a,*, Daniel B. Thomas a,b, Charlotte R. Doney a, Odile Madden a a Smithsonian Museum Conservation Institute, 4210 Silver Hill Rd., Suitland, MD 20746, USA bDepartment of Vertebrate Zoology, National Museum of Natural History, Smithsonian Institution, P.O. Box 37012, Washington, DC 20013, USA a r t i c l e i n f o Article history: Received 28 August 2013 Received in revised form 5 November 2013 Accepted 19 November 2013 Keywords: Raman Bone Collagen Bioapatite Diagenesis a b s t r a c t This study examines Fourier transform (FT) Raman spectroscopy as a non-destructive screeningmethod to determine collagen quality in archaeological and paleontological bones. Bone samples were characterized for collagen quality using well-established elemental abundance analyses (i.e., percentage nitrogen and C:N) as the primary criteria for classification. FT-Raman spectra were collected from outer surfaces and freshly cut cross-sections of both well-preserved and poorly-preserved historic mammal bones. Peak heights and peak areas were studied visually and with bivariate and multivariate statistics. Raman spectra from cross-sections provided the most accurate determination of collagen quality. A ratio of the 960 cm1 and 1636 cm1 peak heights provided the most unambiguous distinction between bones with well- preserved and poorly-preserved collagen. The 960 cm1 and 1636 cm1 peaks represent phosphate anion stretching in the bonemineral and amide carbonyl stretching in the collagen, respectively. FT-Raman spectra fromboneswithwell-preserved collagenproduced a 960 cm1:1636 cm1 ratio of 19.4 or less (after peaks were baseline corrected). This mineral to collagen ratio was typically greater in poorly-preserved samples as organic material tends to be more susceptible to early stages of diagenesis. These criteria nowcan be used to accurately determine collagen quality in bones before sacrificing samples to the lengthy and destructive chemical extractions necessary for carbon-14 dating, stable isotope analyses, proteomic analyses, and other techniques of archaeological or paleontological interest. Published by Elsevier Ltd. 1. Introduction 1.1. Collagen In recent years numerous applications for examining bone collagen in paleontological and archaeological specimens have emerged. Stable isotope analyses are ubiquitous in studies of past diets, ecosystem reconstruction, past climates, and migrations. Carbon-14 dating often is performed on extracted collagen to determine age of specimens. Additionally amino acid sequencing and structural analyses of larger proteins provide insight into evolutionary trends and alternative dating methods. Inherent in the use of bone collagen for all such analyses is the preservation of intact unaltered protein. Collagen is a strong abundant protein comprising approximately 10e25% of total boneweight,most ofwhich is type I collagen (Fratzl, 2008; Kucharz,1992; Nimni,1988; Ramachandran and Reddi,1976). Previous analyses of type I collagen to evaluate degradation have relied primarily on the characteristic elemental abundance and amino acid profile of collagen compared to other vertebrateproteins and organicmaterial found in burial environments (Ajie et al., 1991; DeNiro and Weiner, 1988; Hare, 1980; Ho, 1965; Kucharz, 1992; Liden et al., 1995; Nimni, 1988; Ostrom et al., 1994). Commonly used indicators of collagen quality include w31e37 mol% glycine, w8e14 mol% proline,w7e11 mol% hydroxyproline,w11e16% total nitrogen content,w30e45% total carbon content, and a C:N ratio of 2.8e3.6 (Ajie et al.,1991;Ambrose,1990; Bocherens et al.,1991,1994, 1996, 1997; Coltrain et al., 2004; DeNiro, 1985; Drucker et al., 2001, 2003; Jorkov et al., 2007; McNulty et al., 2002). Limited data is available from spectroscopic examination of collagen quality (Chadefaux et al., 2009; Lebon et al., 2011; Turner-Walker and Syversen, 2002). However, spectroscopic, amino acid, and elemental abundance analyses require considerable time, destructive sam- pling, exposure to embedding resins or chemical extractions, and expensive mass spectrometry which may ultimately indicate a poorly-preserved and unusable specimen. Analysis of bone mineral (i.e., bioapatite) has provided an alternate approach for examining collagen degradation. Bioapatite * Corresponding author. Tel.: þ1 301 238 1261. E-mail address: francec@si.edu (C.A.M. France). Contents lists available at ScienceDirect Journal of Archaeological Science journal homepage: http: / /www.elsevier .com/locate/ jas 0305-4403/$ e see front matter Published by Elsevier Ltd. http://dx.doi.org/10.1016/j.jas.2013.11.020 Journal of Archaeological Science 42 (2014) 346e355 Author's personal copy grains are nested within the collagen matrix and diagenesis of the former may imply degradation of the latter (Nelson et al., 1986; Person et al., 1996; Tütken et al., 2008; Veis, 2003). Several methods for assessing bone mineral alteration have been proposed (Iacumin et al., 1996; Kohn et al., 1999; Michel et al., 1995; Person et al., 1995, 1996; Shemesh, 1990; Tuross et al., 1989; Zazzo et al., 2004), but many are destructive and arguably are supported less well than the direct elemental and amino acid abundancemeasures of collagen itself. In regards to spectroscopic methods, destructive mid-infrared spectroscopy has received more attention for tracking inorganic bone diagenesis (Barrick and Showers, 1994; Lee-Thorp and Sponheimer, 2003) than non-destructive Raman spectros- copy (McLaughlin and Lednev, 2011; Thomas et al., 2011). 1.2. Raman spectroscopy of bone Raman spectroscopy is a non-destructive analytical method that describes functional groups in molecules and polyatomic ions in minerals (Smith and Dent, 2005). The technique uses a laser to scatter light from a sample which allows the analysis to be per- formed at a distance with little to no sample preparation. For this reason, the technique is considered non-invasive and non-contact. We assessed whether Raman spectroscopy could provide a rapid and non-destructive method of evaluating collagen preservation with particular interest in identifying Raman spectral regions that change as collagen degrades. Peaks in Raman spectra are counts of scattered photons, and peak intensities can be used to compare the concentrations of specific chemical bonds among similar samples. Characteristic bone collagen peaks will be larger relative to bioapatite peaks when there is more collagen present and vice versa. One limitation of Raman spectros- copy is interference from sample fluorescence which creates a broader strong signal that can overwhelm sharper but less intense Raman peaks. Fourier transform Raman (FT-Raman), used in this study, uses a near infrared excitation laser which induces less fluo- rescence in teeth and bone (Edwards et al., 2005). A Raman spectrum of bone (Fig. 1A) contains chemical and structural information about collagen and bioapatite that is consistent across animal species (Frushour and Koenig, 1975; Morris and Mandair, 2011). Two prominent spectral bands, identi- fied as amide I and amide III, inform about the collagen peptide backbone (Barth and Zscherp, 2002). Amide I (1560 and 1725 cm1) is due mostly to carbonyl stretching. Amide III (1210 and 1350 cm1) describes two vibrational modes: stretching between carbon and nitrogen atoms and bending of a secondary amine. Bending of methylene groups (d[CH2]) and stretching between carbon and hydrogen atoms (n [CeH]) are vibrational modes common to several peptides and manifest as peaks at 1365e 1500 cm1 and 2820e3035 cm1. Benzene ring-breathing is spe- cific to phenylalanine (n[Phe]) and occurs at w1003 cm1 (Morris and Mandair, 2011). Bioapatite anions also contribute bands to a Raman spectrum of bone (Fig. 1A). The most prominent is a sharp peak attributed to symmetric stretching of phosphate (n1-[PO43]). Four smaller bands describe stretching and bending modes of phosphate (n2, n3, n4) and carbonate (n1-[CO32]) (Awonusi et al., 2007; de Aza et al., 1997; Kravitz et al., 1968). Because organic Fig. 1. Raman spectrum fromspecimen558, a historic humanmetatarsal and anexample of a bonewith ‘well-preserved’ collagen. A) Peaks in a Raman spectrumof bone identifyanions in bioapatite and functional groups in collagen. B) Fluorescence is a source of spectral noise and can be mitigated with baseline correction. Peak intensities used for bivariate ratios are indicated with double-headed arrows. Hatched and shaded regions show peak areas used for bivariate ratio calculations (further detail in the Electronic Supplementary Material). C.A.M. France et al. / Journal of Archaeological Science 42 (2014) 346e355 347 Author's personal copy matter tends to degrademore rapidly than bioapatite inmost burial conditions, we anticipated that bands related to structural features of collagen (i.e., amide I and amide III) would be particularly informative about collagen preservation. The aim of this study was to develop a simple protocol that accurately identifies well-preserved collagen using FT-Raman spectroscopy. Modern and historic bone specimens for which the collagen quality is knownwere analyzed, and the spectra evaluated visually and quantitatively to provide a distinct numeric value characteristic of well-preserved specimens. 2. Materials and methods 2.1. Bone specimens We studied 44 modern and historic (19th century) bone speci- mens from humans and other mammals (Table 1). Thirty-nine historic specimens were excavated from cemeteries and battlefields across disparate locations in the United States (n ¼ 36) and Africa (n ¼ 3) representing a range of taphonomic environ- ments. Five modern specimens were not exposed to burial condi- tions and served as controlled examples of well-preserved collagen. A solid piece was removed from each bone for elemental analysis. For FT-Raman analysis, samples were prepared to expose the outer surface and a fresh cross-section using a chisel or bone saw. Clinging sediment was removed mechanically; no chemical treat- ments were applied. Because bone pieces for elemental analysis were removed prior to Raman analysis, the latter could not be performed in the exact location fromwhich collagen was extracted and was performed in an area immediately adjacent. 2.2. Collagen yield and elemental abundances Collagen quality parameters for each bone specimen were analyzed, including gravimetric measurements of extracted collagen yield, and elemental abundance mass spectrometric Table 1 Sample list and elemental abundance data.a Sample Museum designationb Description Burial location Bone % Collagen Wt% N Wt% C C:N Conditionc Modern BB Cow Unknown Femur 25.9 14.7 36.8 2.9 Good DDB Dog Unknown Femur 10.5 15.0 43.5 3.4 Good DBIII Deer Virginia, U.S. Mandible 31.0 10.8 27.0 2.9 Good WR Whale Unknown Rib 13.6 13.5 40.6 3.5 Good SB Cow Unknown Femur 3.2 15.5 44.0 3.3 Good Historic 281 29LA1091-BOR-42 Human Ft. Craig Cemetery, NM Metatarsal 4.9 14.2 40.4 3.3 Good 343 51RICHARDS-CC-07 Human Congressional Cemetery, DC Metacarpal 12.1 14.8 41.8 3.3 Good 345 GLO-099-2A Human Glorieta Pass Battlefield, NM Femur 13.3 14.2 39.7 3.3 Good 350 GLO-099-2C Human Glorieta Pass Battlefield, NM Femur 6.3 14.2 40.0 3.3 Good 394 6CT58-5-AMM03 Human Walton Family cemetery plot, CT Femur 21.2 15.0 41.7 3.2 Good 396 6CT58-5-AMM05 Human Walton Family cemetery plot, CT Femur 17.6 14.6 40.3 3.2 Good 400 6CT58-5-AMM11 Human Walton Family cemetery plot, CT Femur 19.0 14.5 40.2 3.2 Good 417 51KEYWORTH-CC-04 Human Congressional Cemetery, DC Radius 14.9 14.7 41.0 3.3 Good 420 51KEYWORTH-CC-07 Human Congressional Cemetery, DC Metatarsal 21.5 14.3 40.4 3.3 Good 426 51WHITE-CC-02 Human Congressional Cemetery, DC Fibula 16.4 14.5 40.8 3.3 Good 442 51CAUSTEN-CC-11 Human Congressional Cemetery, DC Metacarpal 23.5 14.2 40.4 3.3 Good 443 51CAUSTEN-CC-13 Human Congressional Cemetery, DC Metacarpal 17.5 14.9 41.8 3.3 Good 454 18PR224-06-300 Human Family tomb, MD Femur 9.7 14.3 40.3 3.3 Good 455 18PR224-99-400 Human Family tomb, MD Tibia 7.3 12.4 35.3 3.3 Good 466 51KEYWORTH-CC-08 Human Congressional Cemetery, DC Tibia 16.8 12.3 34.9 3.3 Good 471 31FOSCUE-ECU-1 Human Foscue Plantation family plot, NC Ulna 7.2 14.1 40.7 3.4 Good 508 ELMINA-A-L49B Human Elmina Settlement, Ghana, Africa Tibia 5.0 12.1 34.6 3.3 Good 525 ELMINA-A-Y51L Human Elmina Settlement, Ghana, Africa Femur 8.0 10.2 28.8 3.3 Good 556 7NCE98A-WOODVILLE-04 Human Woodville Cemetery, DE Fibula 5.1 12.3 36.5 3.5 Good 558 7NCE98A-WOODVILLE-10 Human Woodville Cemetery, DE Metatarsal 16.1 13.6 38.5 3.3 Good 560 7NCE98A-WOODVILLE-SLOPEB Human Woodville Cemetery, DE Ulna 29.7 13.8 39.1 3.3 Good 566 FABC-08-107a Human First African Baptist Church Cemetery, PA Metacarpal 15.3 14.5 40.8 3.3 Good 598 44PWKINCHELOE-SI9114-C Human Kincheloe Plantation family plot, NC Temporal 3.6 12.9 38.8 3.5 Good 255 GETTYS-NPS-965 Human Gettysburg Battlefield, PA Temporal 0.6 8.2 27.9 4.0 Bad 266 29LA1091-BOR-23B Human Ft. Craig Cemetery, NM Talus 1.0 6.5 34.0 6.1 Bad 267 29LA1091-BOR-23C Human Ft. Craig Cemetery, NM Talus 0.5 4.3 25.9 7.0 Bad 268 29LA1091-BOR-23D Human Ft. Craig Cemetery, NM Talus 0.2 5.3 33.9 7.5 Bad 297 TRINITY-EAST 06 Human Trinity Church Cemetery, DC Temporal 0.4 1.9 8.3 5.0 Bad 303 TRINITY-EAST-14 Human Trinity Church Cemetery, DC Temporal <0.2d 7.7 35.2 5.4 Bad 305 TRINITY-EAST-22 Human Trinity Church Cemetery, DC Temporal 0.8 9.7 34.8 4.2 Bad 307 TRINITY-EAST-25 Human Trinity Church Cemetery, DC Mandible 0.5 7.1 25.9 4.3 Bad 309 TRINITY-EAST-04C Human Trinity Church Cemetery, DC Metacarpal <0.2 5.2 47.2 10.5 Bad 329 7BAYVISTA-DHCA-116X Human Ground burial, DE Metacarpal <0.2 6.3 26.6 5.0 Bad 330 7BAYVISTA-DHCA-118A Human Ground burial, DE Femur 0.5 7.0 22.0 3.7 Bad 401 6CT58-5-AMM13 Human Walton Family cemetery plot, CT Femur 0.5 12.9 23.3 2.1 Bad 501 ELMINA-A-E51B Human Elmina Settlement, Ghana, Africa Femur 4.4 10.0 32.5 3.8 Bad 555 7NCE98A-WOODVILLE-02 Human Woodville Cemetery, DE Metatarsal 1.6 6.4 27.8 5.0 Bad 557 7NCE98A-WOODVILLE-06 Human Woodville Cemetery, DE Femur 0.8 6.5 32.7 5.9 Bad 571 18ST1-103-2426W Human Church cemetery, MD Humerus 0.9 7.5 27.7 4.3 Bad a Elemental abundance data frommost of the historic specimens has been published previously in France and Owsley (in press) and France et al. (2014) and reappears here. b Historic specimens are accessioned objects in the collection of the Smithsonian National Museum of Natural History. c Conditions of ‘good’ and ‘bad’ refer to well-preserved and poorly-preserved collagen, respectively. d Yieldsw0% neared the limit of resolution of our scale and are noted as <0.2% to encompass the inherent error. C.A.M. France et al. / Journal of Archaeological Science 42 (2014) 346e355348 Author's personal copy assessments of C:N ratio, percentage nitrogen, and percentage carbon. Collagen was extracted from solid bone pieces (w500 mg) by modified methods of Longin (1971). Solid samples were decal- cified in 0.6 M HCl at 4 C for several days and rinsed to neutrality. Humic and fulvic acidic contaminants were removed by soaking in 0.125 M NaOH overnight. Samples were rinsed and heated in 0.03MHCl at 95 C overnight to denature and separate the collagen strands. The resulting supernatant was freeze-dried and, in well- preserved specimens, the organic extract contained denatured collagen. The collagen yield was recorded as a percentage of the original bone weight. A portion of each organic extract was weighed (w0.5 mg) and packed into tin cups. Samples were combusted on a Costech 4010 Elemental Analyzer (EA) coupled to a Thermo Delta V Advantage mass spectrometer via a Conflo IV interface. Weight percent nitro- gen andweight percent carbonyieldswere calculated by calibration of peak area with a homogeneous acetanilide standard (associated error is 0.5%). Weight percent yield values were converted to moles andused to calculate atomic C:N ratios. Boneswere assigned a designation of ‘well-preserved’ or ‘poorly-preserved’ based on the previously well-established elemental abundance and C:N ratio parameters indicating collagen quality as outlined in Section 1.1. 2.3. Collection and processing of Raman spectra One Raman spectrum was collected from the outer surface, and one from a freshly cut cross-section, for each specimen. An addi- tional two spectra were collected from the cross sections of five specimens with relatively small proportions of collagen: 266, 309, 401, 455, 508. These additional spectra were collected from different places and were used to assess compositional heteroge- neity across the bone surface. A further two spectra (one from the outer surface and one from the cross-section surface) were collected from specimens 266, 281, 303, 309, 329, 471 and 508 at a lower spectral resolution. Raman spectra were collected with an NXR FT-Raman module coupled to a 6700 Fourier transform infrared spectrometer (Thermo Electron Corporation, Madison, WI, USA). The Raman module was equipped with a continuous wave near infrared Nd:YVO4 excitation laser (1064 nm), CaF2 beam splitter, and thermoelectrically-cooled InGaAs detector. Spectra were collected using a 50 mm laser spot and 1.0 OD neutral density filter that limited laser power to a maximum estimated 0.13 W. Laser power was chosen empirically to maximize signal-to-noise (SNR) ratio without burning the sample. Initially, spectra were a co-addition of 1024 scans across 100e3701 cm1 (4 cm1 resolution). As the study progressed it became clear that spectral noise potentially was confounding correct assignment of some specimens, particularly those with low collagen concentrations (i.e., the collagen peaks were small). To mitigate this problem, some spectra were collected again from a different spot free of accretions or surface staining, and/or with more co-added scans (i.e., 2048 or 4096 scans). Spectra for a subset of 7 samples (listed above) were collected at 8 cm1. Noise reduction through post-processing was explored using three or nine point Savitzky Golay filtering performedwith the ‘sgolayfilt’ function from the signal 0.7e2 package (Signal developers, 2011) in R 2.15.2 (R Core Team, 2012). 2.4. Data analysis Raman spectra were evaluated visually and quantitatively to determine which approach and combination of spectral features produced the best indicator of collagen quality. Spectra collected from outer surfaces and freshly cut cross-sections were considered as two separate datasets for all subsequent analyses. 2.4.1. Visual features of Raman spectra of bone collagen Spectral characteristics of visual interest were peaks typical of bone, the shape of those peaks, fluorescence, spectral noise, and spectral artefacts caused by heating of the sample by the laser. If a spectrum appeared noisy or the baseline was distorted by sample heating, a new spectrumwas collected with adjusted parameters or at a new spot to determine the ideal parameters for collection. 2.4.2. Bivariate data analysis: peak ratios A desired outcome of this research was an unambiguous numeric value to distinguish collagen quality. An ideal statistical indicator would show a range of values that corresponds to well- preserved collagen (as determined by the elemental abundance indicators) that is distinct from the range for poorly-preserved collagen, with no overlap. We calculated simple ratios for all possible pairwise combinations of Raman peaks in the bone spec- trum: The peak ratios grouped into three ranges: 1) ratios from well- preserved specimens only, 2) ratios from poorly-preserved speci- mens only, and 3) ratios from both well- and poorly-preserved specimens. Ratios were scored according to how well they segre- gated well-preserved from poorly-preserved specimens, where spectra in groups 1) and 2) were considered correctly assigned: Ratios were calculated for all 78 pairwise combinations of 13 peak heights, and for all 28 pairwise combinations of eight peak areas (Table 2). Ratios were calculated from the Raman peak in- tensity relative to a modeled baseline; the baseline was modeled in five distinct regions to mitigate the distortion introduced by fluo- rescence and sample heating (Fig. 1, Table 2). The terminal points of each regionwere designated as anchor points for a straight baseline for that region (Fig. 1B). Baseline correctionwas accomplished with Peak height ratio ¼ Peak height at wavenumber AOPeak height at wavenumber B Peak area ratio ¼ Peak area at wavenumber AOPeak area at wavenumber B Classification rate ¼ ð# correctly assigned spectraOtotal number of spectraÞ  100 C.A.M. France et al. / Journal of Archaeological Science 42 (2014) 346e355 349 Author's personal copy the ‘baseline.fillPeaks’ function from the baseline 1.01 package (Liland and Mevik, 2012) in R 2.15.2. For each of the seven samples collected at 8 cm1 resolution, collagen quality was predicted by substituting a lower resolution spectrum for the corresponding 4 cm1 resolution spectrum, and the classification rates were recalculated. 2.4.3. Multivariate data analysis Raman spectra also were analyzed with multivariate methods, which can be useful when information is dispersed across many re- gions of a spectrum. The wavenumber range of each spectrum was reduced to 3100e360 cm1 and divided into eight regions: 1) 3100e 2800 cm1, 2) 2800e1750 cm1, 3) 1750e1550 cm1, 4) 1550e 1530 cm1, 5) 1530e1150 cm1, 6) 1150e910 cm1, 7) 910e700 cm1, 8) 700e360 cm1. Separate baseline corrections were made for each region using the ‘baseline.fillPeaks’ function from the baseline 1.01 package (LilandandMevik, 2012) inR2.15.2. A regionof the spectrum not informative for collagen or bone mineral subsequently was removed (2800e1750 cm1). Each spectrum was minimume maximum normalized and each dataset was mean-centered (Beebe et al., 1998; Thomas and Chinsamy, 2011). Spectral data were exploredwith principal components analysis (PCA), calculated using the ‘princomp’ function in R 2.15.12 (Fig. 2). The outer surface and cross sectioned datasets were studied separately with a single component partial least squares discrimi- nant analysis (PLSDA). The spectra in each dataset were subdivided intowell-preserved and poorly-preserved groups according to their previously established elemental abundance collagen quality in- dicators. One spectrum was removed from the dataset, and the collagen quality for that spectrum was predicted according to the spectra remaining in each group. A Raman spectrum of a well- preserved specimen predicted to belong in the well-preserved group was considered correctly classified, whereas a well- preserved spectrum predicted to be part of the poorly-preserved group was misclassified (and vice versa for poorly-preserved collagen specimens). Predictions were calculated for each spec- trum (i.e., full cross validation). PLSDA was performed with the ‘plsda’ function from the caret package (Kuhn, 2012) in R 2.15.2. 3. Results 3.1. Visual features of Raman spectra of bone collagen Qualitative assessment of Raman spectra indicates several distinct peaks that visually distinguish collagen quality (Fig. 3). For poorly-preserved bones, the organic peaks show a general trend of decreasing size and shape distortion related to decreasing collagen Fig. 2. Principal component analysis of Raman spectra collected from cross sectioned surfaces of historical bone specimens. A) Clustering of score values along principal component one. B) and C) Principal component one and two loadings identified collagen bands as the two largest sources of spectral variation in the dataset. Table 2 List of peaks in Raman spectra of bone and interval over which corrected baseline was modeled. Peak position (Raman shifted cm 1) Peak area (Raman shifted cm 1) Bond vibration probed Modeled baseline region (Raman shifted cm1) 2978 3035e2820 n-CH 3100e2800 2940 n-CH 2882 n-CH 1671 1725e1560 Amide I 1750e1550 1636 Amide I 1450 1500e1365 d-CH2 1530e1150 1245 1350e1210 Amide III 1070 1140e1055 n1-CO32 1150e910 1045 n4-PO43 1003 n-Phe 960 990e910 n1-PO43 588 635e510 n3-PO43 700e360 430 485e360 n2-PO43 C.A.M. France et al. / Journal of Archaeological Science 42 (2014) 346e355350 Author's personal copy yield. The amide peaks disappear first, followed by shrinkage and often distortion and/or disappearance of the CeH peaks. However, some samples with poorly preserved collagen gave spectra with nicely defined CeH peaks at 2820-3035 cm1. Spectral noise also contributed to decreased peak quality, especially by camouflaging smaller peaks which might otherwise be diagnostic. The collection of good quality spectra required subjective adjustments including choosing clean, lighter colored surfaces for testing, checking multi- ple spots on the same bone, increasing the number of co-added scans, and decreasing the spectral resolution to 8 cm1 in some cases. Spectra from outer surfaces often differed from those of cross- sections. 3.2. Statistical data analysis 3.2.1. Outer surface and cross-section comparison The degree to which Raman spectra of outer bone surfaces differed from cross-sectioned surfaces of the same bone was quantified for each peak ratio. Differences in the peak height ratios from outer surface spectra (Ou) and cross-section spectra (Cs) were recovered as a percentage of the maximum ratio for each pair: Variation between Ou and Cs ¼ ½absðOu CsÞOmaxðOu;CsÞ  100 The average discrepancy between cross-section and outer surface spectra was 31.7% for mineral to collagen peak ratios (n ¼ 39, s ¼ 22.1%). Hence, the outer surfaces of the bones that had been exposed to the burial environment were spectrally different from the freshly cut internal surfaces. Heterogeneity also was observed among spectra collected at different spots along the same cross- sectioned surface. The mineral to collagen ratio showed an average discrepancy of 38.5% (n ¼ 5, s ¼ 24.5%) within the same bone. For samples in which multiple cross-section spectra were collected, a single representative spectrum was chosen for the quality predictions and parameter regressions that follow. These spectra were selected according to two guidelines: 1) the spectrum collected nearest to the collagen extraction sampling site was chosen, or if the extraction sampling site was not apparent, 2) the spectrum with the best SNR was chosen. The full set of spectra is available as electronic supplementary material. 3.2.2. Collagen quality predictions: bivariate analysis Raman spectra from cross-sectioned surfaces that were collected and analyzed with the same parameters (1024 co-added scans, 4 cm1 resolution, no spectral smoothing) successfully classified between 38 and 100% of samples using the 960:1636 cm1 peak height ratio (Fig. 4). The classification rate depended on which spectra were selected to represent specimens 266, 309, 401, 455 and 508. Hence, the classification rate was influenced by surface heterogeneity. The highest rate was achieved when spectra had the highest SNR and were collected closest to the collagen sampling site. This ratio cleanly separated the well- and poorly-preserved samples (Fig. 5A). Cross-section spectra from well-preserved samples featured a 960 cm1:1636 cm1 ratio be- tween 9.9 and 19.4 (mean¼ 14.0, 1s¼ 2.6). Cross-section spectra of poorly-preserved specimens featured a 960 cm1:1636 cm1 ratio between 20.4 and 156.6 (mean ¼ 59.3, 1s ¼ 34.9). The classification rate for outer-surface spectra using the 960:1636 cm1 peak height ratio was only 16% because the range of well- and poorly-preserved samples overlaps for all but the highest values (Fig. 5B). The 2978:2882 cm1 ratio was most selective for outer surfaces with 57% correct classifications. Bivariate analyses based on peak area produced a high per- centage of correct classifications. However, ratios most indicative of preservation quality differed for the cross-section and outer surface datasets, and post-processing smoothing, or lack thereof, affected the outcome. Therefore we could not determine a single unam- biguous numeric value indicative of collagen quality for both outer surfaces and cross-sections, which rendered this approach less informative. The highest classification rate was 74% for the ratio of the 485e360 cm1 peak area to the 1500e1365 cm1 peak area, and used unsmoothed 4 cm1 spectra collected from cross sections. The classification rate rose to 100% when the spectral resolution of the cross-section spectrum from specimen 309 was reduced to 8 cm1. For outer surface spectra, the highest classification rate using peak area ratios was 35%. This classification rate was from the ratio of the 635e510 cm1 peak area to the 1500e1365 cm1 peak area and the spectra had been nine-point smoothed. However, in most cases post-processing smoothing did not reliably improve, and often worsened, our predictions overall. 3.2.3. Collagen quality predictions: multivariate analysis Twenty one of 22 well-preserved cross-section spectra and 16 of 17 poorly-preserved cross-section spectra were classified correctly with PLSDA, resulting in a sensitivity of 95%. Two misclassified spectra, specimens 309 and 455, have poorly-preserved and well- preserved collagen, respectively. A principal component (PC) analysis positively loaded collagen bands on the first PC. PC1 score values for specimen 309 clustered among samples with well- Fig. 3. Characteristic spectra of well- and poorly-preserved collagen in bone specimens. C.A.M. France et al. / Journal of Archaeological Science 42 (2014) 346e355 351 Author's personal copy preserved collagen, and the PC1 score value for specimen 455 was intermediate between the cluster of well-preserved and poorly- preserved specimens (Fig. 2). Spectra collected from outer sur- faces were less informative about collagen quality than the freshly cut cross-sections (Fig. 5B). Thirteen external surface spectra were misclassified by PLSDA, resulting in a sensitivity of 68%. While PLSDA identified well-preserved specimens quite successfully, we note that this approach requires significant time and effort to establish an internal laboratory dataset for comparison. 4. Discussion The preservation of collagen in exhumed bone can be evaluated with FT-Raman spectroscopy. Because bone is a composite material of mineral and protein, substances that respond differently to aging and burial conditions, deviations from known proportions are good indicators of preservation. All of our statistical classification methods show promise and improve selectivity over blind selec- tion. The most unambiguous and accurate measure of preservation was achieved through analysis of peak height ratios. Analysis of peak area ratios did not produce a single unambiguous measure of preservation, and PLSDA requires a large internal dataset which is restrictive for smaller laboratories with lower throughput. The bivariate analysis of peak height ratios showed the highest rates of correct classification using the 960 cm1 and 1636 cm1 ratio. A Raman spectrum from a bone with well-preserved collagen will have a 960 cm1:1636 cm1 ratio of 19.4 after baseline correction (value reflects an approximate 95% confidence interval). The 960 cm1 band is attributed to phosphate stretching and is the most distinct mineral band in a spectrum of bone; the 1636 cm1 band is attributed to amide I stretching and provides insight into the peptide backbone of collagen (Barth and Zscherp, 2002; Kravitz et al., 1968). Hence, the peak height ratio of 960 cm1:1636 cm1 informs about collagen fragmentation, an important index of collagen degradation (e.g., Tuross et al., 1988). Despite the CeH stretching region around 2880 cm1 being the most intense collagen region in a Raman spectrum of bone, bands in this region are not accurate predictors of collagen quality. Indeed, aliphatic Ce H functionality occurs in both degraded and intact collagen (Frushour and Koenig, 1975). Bivariate analysis of mineral and collagen peak heights also predicts the amount of intact collagen in bone. Comparison of the 960 cm1:1636 cm1 ratios to the percent collagen extracted from the same bones indicates that decreased collagen quality co-occurs with collagen loss in both outer surfaces and fresh cut cross sec- tions (Fig. 5). This reiterates the information provided by the ratios because a relatively small 1636 cm1 peak indicates that less amide I, a defining component of collagen, is present. Given the variability in collagen content on outer bone surfaces, a ratio in the higher range of 19.4e26.8 still offers some chance of finding intact collagen, but the % collagen yield likely will be low (<5%). Further study of bones in this range might benefit from a larger sample set. The mineral to collagen ratio of a Raman spectrum varied across a bone. Outer bone surfaces tended to have high ratios, indicating Fig. 4. Classification rate based on peak height for collagen quality predictions. Pairwise ratios of baseline-corrected peak heights were calculated for 39 spectra. Fig. 5. Collagen yield compared to peak intensity ratio based on peak height for the 960:1636 cm1 ratio which produced the most accurate prediction of collagen quality in a bivariate analysis. Percent collagen yield indicates the % by weight of organic extract obtained during chemical extractions which is denatured collagen in well- preserved samples and unidentified organic residue in poorly-preserved samples. A) Freshly cut cross-sections. B) Outer surfaces. C.A.M. France et al. / Journal of Archaeological Science 42 (2014) 346e355352 Author's personal copy that collagen has been lost from surfaces directly exposed to burial environments. Well-preserved bones tended to have lower mineral to collagen ratios on freshly cross-sectioned surfaces. Importantly, we observed spectral heterogeneity on cross-sectioned surfaces which indicates heterogeneous preservation within a single bone. This has been observed previously in microscale imaging and spectroscopy of bone composition (Chadefaux et al., 2009; Lebon et al., 2011; Turner-Walker and Syversen, 2002), although our Raman technique identifies such patterns non-destructively. The variation in mineral to collagen ratios, and hence the variation in collagen abundance or preservation, encourages collagen sampling and Raman spectral measurement from the same location on the bone. Collagen would ideally be extracted from the same location that a 960 cm1:1636 cm1 peak height ratio of 19.4 has been measured. We note that our analyses were conducted on historic bones buried for a relatively short amount of time. While a 960 cm1:1636 cm1 peak height ratio of 19.4 is still likely to indicate well-preserved collagen in older specimens, more exten- sive exposure to diagenetic elements over the course of thousands of years may create additional heterogeneity within such bones. Spectra from well-preserved and poorly-preserved specimens differ visually. Peaks in the amide III (1245 cm1), amide I (1636 and 1671 cm1), eCH2 (1450 cm1), and eCH (2882, 2940, and 2978 cm1) regions were noticeably reduced in poorly-preserved samples, while peaks in the inorganic regions maintained in- tensity. This most likely is due to a preferential degradation of proteins during diagenesis and has been observed in previous studies of short-term burial (McLaughlin and Lednev, 2011). In particular, the intensity of the amide III peak was visually reduced in poorly-preserved specimens (Fig. 4). The collagen backbone is constructed from amide bonds and the loss of relative intensity at 1245 cm1 likely signifies the fragmentation of collagen due to a bacterial preference for these relatively high energy amide bonds (Balzer et al., 1997; Grupe, 1995; Grupe and Turban-Just, 1998, 2000; Turban-Just and Schramm, 1998). When samples were misclassified, we found that their scores fell near the intersection of well- and poorly preserved values, and that the amount of collagen present was relatively low. One cross-section spectrum (specimen 309) resembled the spectra of well-preserved samples, although the elemental and yield data indicated poor collagenpreservation.Thisdiscrepancymightbeexplainedbysurface heterogeneity or collagen degradation during extraction. Regarding surface heterogeneity, Raman spectroscopy may have identified a localized region ofwell-preserved collagen in an otherwise degraded sample with little intact collagen remaining. Alternatively, Raman spectroscopy may have identified a collagen-like structure or degraded peptide fragment susceptible to the chemical extraction process. The collagen in specimen 309 may be in the final stages of diagenesis where the protein is largely denatured and is more sus- ceptible to removal during the earliest decalcification steps. Some cross-section spectra represented poorly-preserved collagen despite the elemental and yield data indicating good collagen preservation. Here the discrepancies were likely due to the previously discussed surface heterogeneity. The elemental and yield data homogenizes a larger volume of bone than is analyzed by the w50 mm laser spot during Raman analysis. Although a bone may contain well-preserved collagen, these small localized regions with relatively low amounts of collagen can show a ‘poor-preser- vation’ spectrum. This underscores the importance of collecting high quality Raman spectra with good SNR, and, when a bone with well-preserved collagen is indicated, removing the sample from the same site as the Raman spectrum. We recommend the following methods for a quick determina- tion of collagen quality based on Raman spectra. First collect a spectrum from the outer surface to determine whether collagen is preserved there. Select a spot for this initial test that has been brushed clean of surface dirt and accretions, and appears free of staining. A Raman spectrum will record the presence of any ma- terials in the laser beam path, so it is important that the path be free of burial dirt, adhesives, consolidants, or other surface contami- nants. This visual screening is necessarily subjective, and it in- creases the chance of finding well-preserved collagen from the outer surface. This requires no physical damage or alteration of the bone. A majority of samples produced good quality spectra using the co-addition of 1024 scans across 100e3700 cm1. More co- added scans may be needed (i.e., 2048 or 4096) when the amount of intact collagen is low and/or the bone has a darker color. All samples in this study were analyzed initially at 4 cm1 resolu- tion; collecting spectra at lower resolution (i.e., 8 cm1) may reduce the influence of spectral noise on peak heights and improve selectivity for collagen preservation. If the baseline-corrected 960 cm1:1636 cm1 peak height ratio is 26.8, the bone is un- likely to yield intact collagen. For example, a bone in our study set with an outer surface 960 cm1:1636 cm1 peak height ratio<26.8 had a 61% chance of bearing well-preserved collagen; an outer surface ratio of <19.4 increased that likelihood to 67%. These probabilities should allow investigators to quickly isolate an adequate sample set for further analyses such as radiocarbon dating, stable isotope analyses, or proteomics. If more samples are needed and one is willing to devote more analytical time to each sample, additional spectra can be collected from outer surfaces to identify cases where a bone might have heterogeneous preserva- tion across the surface. Alternately, collection of additional spectra from fresh cross-sections with a baseline corrected 960 cm1:1636 cm1 peak height ratio 19.4 increases the likeli- hood of extracting well-preserved collagen to 95%. 5. Conclusions This study examined FT-Raman spectroscopy as a means to screen historic bone samples for the presence of well-preserved collagen. A majority of the samples produced good quality spectra using the co-addition of 1024 scans across 100e3700 cm1 (4 cm1 resolution). Although a few samples required variations on these settings to produce quality spectra, we suggest that these param- eters offer a quick screening method with a high probability of accurate results. It should be noted that Raman as described here is a spot analysis and a spectrum from one location may not be representative of the entire bone specimen. Variation in collagen quality was observed across a bonewith the outer surfaces typically showing greater alteration than internal surfaces. Future work will examine older archaeological and paleontological specimens to determine if this pattern of heterogeneity changes with more exposure to diagenetic influences. Despite this heterogeneity, collagen quality still was predictable with high accuracy. A bivariate analysis of the 960 cm1:1636 cm1 (mineral to collagen) peak height ratio in freshly cut cross-sections produced the most unambiguous and accurate index of collagen quality. Samples with a mineral to collagen ratio of 19.4 are considered well-preserved. Samples with a mineral to collagen ratio of19.4 show a systematic decrease in peak height intensities associated with degradation of the collagen peptide backbone while the intensities of peaks associated with the mineral decrease to a lesser degree. This suggests a preferential breakdown of organic material in bone during diagenesis which has implications for the duration of burial beyond which a bone will no longer yield reliable data from its collagen and other organic components. Collagen preservation was predicted better from cross section surfaces than from outer surfaces. Alteration of outer surfaces most often resulted in well-preserved samples being misclassified as C.A.M. France et al. / Journal of Archaeological Science 42 (2014) 346e355 353 Author's personal copy poorly-preserved. This would result in several usable specimens being eliminated from further destructive analyses, but would still allow one to select a subset of well-preserved specimens that is highly likely to produce quality results in further testing. Labora- tories can now use this pre-screening method before outsourcing analyses and spending additional time or money on radiocarbon dating, stable isotope analyses, or protein sequencing. Acknowledgments The authors acknowledge K. Bruwelheide, S. McGuire, S. Mills, W. Miller, D. Owsley, and A. Warmack for assistance with sample procurement and preparation. Financial support for D. Thomas provided by Smithsonian Peter Buck Post-doctoral Fellowship. Financial support for C. Doney provided by National Science Foundation Research Experience for Undergraduates Award #SMA- 1156360. All analyses performed at the Smithsonian OUSS/MCI Stable Isotope Mass Spectrometry Laboratory and the Smithsonian MCI Modern Materials Laboratory at the Smithsonian Museum Conservation Institute, Suitland, MD. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.jas.2013.11.020 References Ajie, H.O., Hauschka, P.V., Kaplan, I.R., Sobel, H., 1991. Comparison of bone collagen and osteocalcin for determination of radiocarbon ages and paleodietary reconstruction. Earth Planet. Sci. Lett. 107, 380e388. Ambrose, S.H., 1990. Preparation and characterization of bone and tooth collagen for isotopic analysis. J. Archaeol. Sci. 17, 431e451. Awonusi, A., Morris, M.D., Tecklenburg, M.M.J., 2007. Carbonate assignment and calibration in the Raman spectrum of apatite. Calcif. Tissue Int. 81, 46e52. Balzer, A., Gleixner, G., Grupe, G., Schmidt, H.L., Schramm, S., Turban-Just, S., 1997. In vitro decomposition of bone collagen by soil bacteria: the implications for stable isotope analysis in archaeometry. Archaeometry 39, 415e429. Barrick, R.E., Showers, W.J., 1994. Thermophysiology of Tyrannosaurus rex: evidence from oxygen isotopes. Science 265, 222e224. Barth, A., Zscherp, C., 2002. What vibrations tell us about proteins. Q. Rev. Biophys. 35, 369e430. Beebe, K.R., Pell, R.J., Seasholtz, M.B., 1998. Chemometrics: a Practical Guide. John Wiley and Sons, New York. Bocherens, H., Billiou, D., Patou-Mathis, M., Bonjean, D., Otte, M., Mariotti, A., 1997. Paleobiological implications of the isotopic signatures (13C, 15N) of fossil mammal collagen in Scladina Cave (Sclayn, Belgium). Quat. Res. 48, 370e380. Bocherens, H., Fizet, M., Mariotti, A., 1994. Diet, physiology and ecology of fossil mam- mals as inferred from stable carbon and nitrogen isotope biogeochemistry: impli- cations forPleistocenebears. Palaeogeogr. Palaeoclimatol. Palaeoecol.107, 213e225. Bocherens, H., Fizet, M., Mariotti, A., Lange-Badre, B., Vandermeersch, B., Borel, J.P., Bellon, G., 1991. Isotopic biogeochemistry 13C, 15N of fossil vertebrate collagen: application to the study of a past food web including Neandertal man. J. Hum. Evol. 20, 481e492. Bocherens, H.M., Pacaud, G., Lazarev, P.A., Mariotti, A., 1996. Stable isotope abun- dances (13C, 15N) in collagen and soft tissues from Pleistocene mammals from Yakutia: implications for the palaeobiology of the Mammoth Steppe. Palae- ogeogr. Palaeoclimatol. Palaeoecol. 126, 31e44. Chadefaux, C., Hô, A.L., Bellot-Gurlet, L., Reiche, I., 2009. Curve-fitting micro-ATR- FTIR studies of the amide I and amide II bands of type I collagen in archaeo- logical bone materials. e-PreservationSci. 6, 129e137. Coltrain, J.B., Harris, J.M., Cerling, T.E., Ehleringer, J.R., Dearing, M., Ward, J., Allen, J., 2004. Rancho La Brea stable isotope biogeochemistry and its implications for the palaeoecology of the late Pleistocene, coastal southern California. Palae- ogeogr. Palaeoclimatol. Palaeoecol. 205, 199e219. de Aza, P.N., Santos, C., Pazo, A., de Aza, S., Cuscó, R., Artús, L., 1997. Vibrational properties of calcium phosphate compounds. 1. Raman spectrum of b-trical- cium phosphate. Chem. Mater. 9, 912e915. DeNiro, M.J., 1985. Postmortem preservation and alteration of in vivo bone collagen isotope ratios in relation to palaeodietary reconstruction. Nature 317, 806e809. DeNiro, M.J., Weiner, S., 1988. Chemical, enzymatic, and spectroscopic character- ization of “collagen” and other organic fractions from prehistoric bones. Geo- chim. Cosmochim. Acta 52, 2197e2206. Drucker, D., Bocherens, H., Bridault, A., Billiou, D., 2003. Carbon and nitrogen iso- topic composition of red deer (Cervus elaphus) collagen as a tool for tracking palaeoenvironmental change during the Late-Glacial and Early Holocene in the northern Jura (France). Palaeogeogr. Palaeoclimatol. Palaeoecol. 195, 375e388. Drucker, D., Bocherens, H., Pike-Tay, A., Mariotti, A., 2001. Isotopic tracking of sea- sonal dietary change in dentine collagen: preliminary data from modern caribou. Earth Planet. Sci. 333, 303e309. Edwards, H.G., Jorge Villar, S.E., Nik Hassan, N.F., Arya, N., O’Connor, S., Charlton, D.M., 2005. Ancient biodeterioration: an FT-Raman spectroscopic study of mammoth and elephant ivory. Anal. Bioanal. Chem. 383, 713e720. France, C.A.M., Owsley, D., Hayek, L.C., 2014. Stable isotope indicators of provenance and demographics in 18th and 19th century North Americans. J. Archaeol. Sci. 42, 356e366. France, C.A.M., Owsley, D., 2013. Stable carbon and oxygen isotope spacing between bone and tooth collagen and hydroxyapatite in human archaeological remains. Int. J. Osteoarchaeol. (in press). http://dx.doi.org/10.1002/oa.2300. Fratzl, P., 2008. Collagen: Structure and Mechanics. Springer, New York. Frushour, B.G., Koenig, J.L., 1975. Raman scattering of collagen, gelatin, and elastin. Biopolymers 14, 379e391. Grupe, G., 1995. Preservation of collagen in bone from dry, sandy soil. J. Archaeol. Sci. 22, 193e199. Grupe, G., Turban-Just, S., 1998. Amino acid composition of degraded matric collagen from archaeological human bone. Anthropol. Anz. 56, 213e226. Grupe, G., Balzer, A., Turban-Just, S., 2000. Modeling protein diagenesis in ancient bone: towards a validation of stable isotope data. In: Ambrose, S.H., Katzenberg, M.A. (Eds.), Biogeochemical Approaches to Paleodietary Analysis. Kluwer Academic, New York, pp. 173e187. Hare, P.E., 1980. Organic geochemistry of bone and its relation to the survival of bone in the natural environment. In: Behrensmeyer, A.K., Hill, A.P. (Eds.), Fossils in the Making: Vertebrate Taphonomy and Paleoecology. The University of Chicago Press, Chicago, pp. 208e219. Ho, T., 1965. The amino acid composition of bone and tooth proteins in Late Pleistocene mammals. Proc. Natl. Acad. Sci. 54, 26e31. Iacumin, P., Bocherens, H., Mariotti, A., Longinelli, A., 1996. Oxygen isotope analyses of co-existing carbonate and phosphate in biogenic apatite: a way to monitor diagenetic alteration of bone phosphate? Earth Planet. Sci. Lett. 142, 1e6. Jorkov, M.L.S., Heinemeier, J., Lynnerup, N., 2007. Evaluating bone collagen extrac- tion methods for stable isotope analysis in dietary studies. J. Archaeol. Sci. 34, 1824e1829. Kohn, M.J., Schoeninger, M.J., Berker, W.M., 1999. Altered states: effects of diagenesis on fossil tooth chemistry. Geochim. Cosmochim. Acta 63, 2737e2747. Kravitz, L.C., Kingsley, J.D., Elkin, E.L., 1968. Raman and infrared studies of coupled PO43 vibrations. J. Chem. Phys. 49, 4600e4610. Kucharz, E.J., 1992. The Collagens: Biochemistry and Pathophysiology. Springer- Verlag, Berlin. Kuhn, M., 2012. Caret: Classification and Regression Training, Version. 5.15-045. http://CRAN.R-project.org/package¼caret. Lebon, M., Müller, K., Bahain, J., Fröhlich, F., Falguères, C., Bertrand, L., Sandt, C., Reiche, I., 2011. Imaging fossil bone alterations at the microscale by SR-FTIP microspectroscopy. J. Anal. At. Spectrom. 26, 922e929. Lee-Thorp, J., Sponheimer, M., 2003. Three case studies used to reassess the reli- ability of fossil bone and enamel isotope signals for paleodietary studies. J. Anthropol. Archaeol. 22, 208e216. Liden, K., Takahashi, C., Nelson, D.E., 1995. The effects of lipids in stable carbon isotope analysis and the effects of NaOH treatment on the composition of extract bone collagen. J. Archaeol. Sci. 22, 321e326. Liland, K.H., Mevik, B.H., 2012. Baseline: Baseline Correction of Spectra, Version 1.1- 0. http://CRAN.R-project.org/package¼baseline. Longin, R., . New method of collagen extraction for radiocarbon dating. Nature 230, 241e242. McLaughlin, G., Lednev, I.K., 2011. Potential application of Raman spectroscopy for determining burial duration of skeletal remains. Anal. Bioanal. Chem. 401, 2511e2518. McNulty, T., Calkins, A., Ostrom, P., Ganghi, H., Gottfried, M., Martin, L., Gage, D., . Stable isotope values of bone organic matter: artificial diagenesis experiments and paleoecology of Natural Trap Cave, Wyoming. Palaios 17, 36e49. Michel, V., Ildefonse, P., Morin, G., 1995. Chemical and structural changes in Cervus elaphus tooth enamels during fossilization (Lazaret cave): a combined IR and XRD Rietvold analysis. Appl. Geochem. 10, 145e159. Morris, M.D., Mandair, G.S., 2011. Raman assessment of bone quality. Clin. Orthop. Relat. Res. 469, 2160e2169. Nelson, B.K., DeNiro, M.J., Schoeninger, M.J., DePaolo, D.J., Hare, P.E., 1986. Effects of diagenesis on strontium, carbon, nitrogen, and oxygen concentration and iso- topic composition of bone. Geochim. Cosmochim. Acta 50, 1941e1949. Nimni, M.E., 1988. Collagen. CRC Press Inc., Boca Raton, USA. Ostrom, P.H., Zonneveld, J., Robbins, L.L., 1994. Organic geochemistry of hard parts: assessment of isotopic variability and indigeneity. Palaeogeogr. Palaeoclimatol. Palaeoecol. 107, 201e212. Person, A., Bocherens, H., Mariotti, A., Renard, M., 1996. Diagenetic evolution and experimental heating of bone phosphate. Palaeogeogr. Palaeoclimatol. Palae- oecol. 126, 135e149. Person, A., Bocherens, H., Saliege, J., Paris, F., Zeitoun, V., Gerard, M., 1995. Early diagenetic evolution of bone phosphate: an X-ray diffractometry analysis. J. Archaeol. Sci. 22, 211e221. R Core Team, 2012. R: A Language and Environment for Statistical Computing, Version 2.15.2. http://www.R-project.org/. C.A.M. France et al. / Journal of Archaeological Science 42 (2014) 346e355354 Author's personal copy Ramachandran, G.N., Reddi, A.H., 1976. Biochemistry of Collagen. Plenum Press, New York. Shemesh, A., 1990. Crystallinity and diagenesis of sedimentary apatites. Geochim. Cosmochim. Acta 54, 2433e2438. Signal developers, 2011. Signal: Signal Processing, Version 0.7-2. http://r-forge.r- project.org/projects/signal/. Smith, E., Dent, G., 2005. Modern Raman Spectroscopy e a Practical Approach. John Wiley and Sons, Chichester. Thomas, D.B., Chinsamy, A., 2011. Chemometric analysis of EDXRF measurements from fossil bone. X-ray Spectrom. 40, 441e445. Thomas, D.B., McGoverin, C.M., Fordyce, R.E., Frew, R.D., Gordon, K.C., 2011. Raman spectroscopy of fossil bioapatited a proxy for diagenetic alteration of the oxygen isotope composition. Palaeogeogr. Palaeoclimatol. Palaeoecol. 310, 62e70. Turban-Just, S., Schramm, S., 1998. Stable carbon and nitrogen isotope ratios of individual amino acids give new insight into bone collagen degradation. Bull. Soc. Geol. Fr. 169, 109e114. Turner-Walker, G., Syversen, U., 2002. Quantifying histological changes in archae- ological bones using BSE-SEM image analysis. Archaeometry 44, 461e468. Tuross, N., Behrensmeyer, A.K., Eanes, E.D., Fisher, L.W., Hare, P.E., 1989. Molecular preservation and crystallographic alterations in a weathering sequence of wildebeest bones. Appl. Geochem. 4, 261e270. Tuross, N., Fogel, M., Hare, P.E., 1988. Variability in the preservation of the isotopic composition of collagen from fossil bone. Geochim. Cosmochim. Acta 52, 929e 935. Tütken, T., Vennemann, T.W., Pfretzschner, H.U., 2008. Early diagenesis of bone and tooth apatite in fluvial and marine settings: constraints from combined oxygen isotope, nitrogen and REE analysis. Palaeogeogr. Palaeoclimatol. Palaeoecol. 266, 254e268. Veis, A., 2003. Mineralization in organic matrix frameworks. In: Dove, P.M., DeYoreo, J.J., Weiner, S. (Eds.), Reviews in Mineralogy and Geochemistry, Bio- mineralization, vol. 54. The Mineralogical Society of America, Washington, pp. 249e289. Zazzo, A., Lécuyer, C., Sheppard, S.M.F., Grandjean, P., Mariotti, A., 2004. Diagenesis and the reconstruction of paleoenvironments: a method to restore original d18O values of carbonate and phosphate from fossil tooth enamel. Geochim. Cos- mochim. Acta 68, 2245e2258. C.A.M. France et al. / Journal of Archaeological Science 42 (2014) 346e355 355