Variations in glutamine deamidation for a Châtelperronian bone assemblage as measured by peptide mass fingerprinting of collagen

Abstract Peptide mass fingerprinting of bone collagen (ZooMS) has previously been proposed as a method to calculate the extent of the non-enzymatic degradation of glutamine into glutamic acid (deamidation). Temporal and spatial variation of glutamine deamidation at a single site, however, has not been investigated. Here we apply ZooMS screening of Châtelperronian and Early Holocene bone specimens from Quinçay, France, to explore temporal and spatial variation in glutamine deamidation. Our results indicate that chronological resolution is low, while spatial variation is high. Nevertheless, our analysis allows the identification of bone specimens that have undergone diagenetic histories remarkably different (either in length or in type) from spatially related bone specimens. Therefore, ZooMS ammonium-bicarbonate screening is capable of testing bone assemblage homogeneity, which could guide subsequent analysis and interpretation. GRAPHICAL ABSTRACT


Introduction
The presence of intrusive elements of younger or older age in archaeological bone assemblages is a key issue in Palaeolithic archaeology. Examples include a number of burials in Upper Palaeolithic contexts that were demonstrated to be Holocene in age by direct radiocarbon dating (Hoffmann et al., 2011;Conard et al., 2004;Terberger et al., 2001;Benazzi et al., 2014) or reworking of archaeological deposits by (micro)mammals (Enloe, 2012;Discamps et al., 2012). The unrecognized presence of intrusive elements could greatly influence the behavioral or ecological interpretation of a bone assemblage, and various approaches have therefore been proposed to assess the homogeneity of faunal assemblages (among others : Lyman, 1994;Behrensmeyer, 1978;Rendu et al., 2014;Blumenschine et al., 1996;Kuhn et al., 2010). These approaches either focus on preservation coherence, aim to identify the main accumulator(s), or study intrinsic bone properties such as surface modifications and bone density.
ZooMS (Zooarchaeology by Mass Spectrometry; Buckley et al., 2009) is used to establish the taxonomic composition of fragmented bone assemblages (Welker et al., 2015a;Brown et al., 2016;Evans et al., 2016;Charlton et al., 2016;Welker et al., 2016) and the taxonomic origin of worked artefacts (Fiddyment et al., 2015;von Holstein et al., 2014;Ives et al., 2014). A further potential application would be to assess faunal assemblage homogeneity as ZooMS is well suited for studying large numbers of bone and tooth specimens and requires small amounts of material (commonly <30 mg). In this context, glutamine deamidation analysis has previously been suggested as a potential method of studying collagen preservation variability and for the detection of potential outliers Simpson et al., 2016;Schroeter & Cleland, 2016). Glutamine deamidation ratios can be obtained during routine ZooMS screening of faunal samples when adequate extraction protocols are followed van Doorn et al., 2011). Up to now, however, spatial or chronological variability of glutamine deamidation at a single site has not been studied. Therefore, in this study we perform ammonium-bicarbonate ZooMS screening (van Doorn et al., 2011) on a total of 543 bone and dental specimens from Quinçay, France to address this lack of data.
The site of Quinçay is ideal for exploring deamidation variation as it contains an archaeological sequence with scattered Holocene bone specimens lying on top of several in situ Châtelperronian units, likely older than 40ka cal BP, with no indication of elements of intermediate chronological age Roussel & Soressi, 2010;Roussel, 2013;Roussel et al., 2016). Some of the bone specimens found at the surface might even be very recent. Furthermore, the Châtelperronian units at Quinçay contain six pierced animal teeth that are interpreted as beads or personal ornaments (Granger & Lévêque, 1997). Objects such as these, thought to have had "symbolic" meaning, play a pivotal role in discussions on late Neanderthal behavior and the biological attribution of the Châtelperronian (Caron et al., 2011;Higham et al., 2010;Hublin et al., 2012;Soressi & Roussel, 2014;Welker et al., 2016). In this study, we describe variations in glutamine deamidation between archaeological units as well as spatial variation within archaeological units. Next, we discuss the possible interpretational value of these results in relation to specific sedimentary features or artefacts present at the site.

Quinçay
Excavated between 1968 and 1990 by François Lévêque, the Grande Roche de La Plématrie at Quinçay (France) is a limestone cave facing the south/south-east in a dry tributary to the river Auxance (Lévêque & Miskovsky, 1983). The small cave is broadly divided into a front area, where limestone blocks (numbered with Roman numerals I to V) cover part of the stratigraphy, and a back area, where such limestone blocks are absent (Roussel et al., 2016). The naming of sedimentary/archaeological units generally follows this division between the front area (units named 'Ensemble', or starting with 'E') and the back area (units named 'Série', or starting with 'S'). During the excavation, lithics and faunal specimens were collected and recorded by stratigraphic units, with subunits usually 5 or 10 cm thick.
The youngest material was found at the back of the cave on top of the stratigraphic sequence, where Lévêque et al. (1997) collected seven Mesolithic lithics on the surface. An additional 10 ceramic fragments attributed to the early Neolithic were found on the surface in the same area. The lithics found in the frontal part of the cave have been described by Lévêque (1979) and have recently been studied by M. Roussel (Roussel et al., 2016). The units Ej, Em and En have been attributed to the Châtelperronian, while the bottom unit Eg has recently been reattributed to the Mousterian of Acheulian Tradition Type B (Roussel & Soressi, 2010). According to unpublished stratigraphic documentation by F. Lévêque, a correlation can be established between the units from the front part of the cave and the units found in the back of the cave ( Table 1). The chronological position of units So and Sb is less clear. In addition, a small quantity of finds was discovered in reworked sediments present in square 4L. These specimens have been assigned by F. Lévêque to a unit called 'remanié' (French: 'reworked'), with excavation notes describing the extent of this feature within the square. From these notes, however, the age of the disturbance or of the material contained within it is unknown.
Faunal remains are preserved mostly within the upper layers (Ej+Sj and Em+Sm), and a paleontological analysis of identified specimens has been published (Lavaud-Girard, 1986). Five out of six pierced teeth come from the front area of the cave, while the sixth originates from the back area. They were all found in the upper layers of the site; five come from layer Em and one from the bottom of layer Ej. They have previously been described (Granger & Lévêque, 1997).
In this study, we measure glutamine deamidation for a large sample of previously unidentified faunal specimens. We compare the glutamine deamidation values from the front area (Ensemble) with the values obtained in the back area (Série) for the various units for which material was studied, including the 'remanié', in both temporal and spatial frameworks. From the notebooks and unpublished documents, the expectation is that all the faunal remains should be of Châtelperronian age, except maybe some specimens found within the units So, Sb and the 'remanié'. We will also briefly discuss the homogeneity of the deamidation values obtained on faunal specimens excavated in close proximity to the pierced teeth.

ZooMS COL1 extraction
Faunal samples were taken from 475 morphologically undiagnostic cortical bone specimens (246 for units Ej+Sj, 222 for units Em+Sm, 28 for units En+Sfi) by random selection. Although Lavaud-Girard (1986) mentions the presence of three identified bone specimens from unit Eg, no bone specimens of this unit were included in the present study. Additionally, we identified and sampled bone and dental specimens from above unit Ej+Sj (n=18 for samples assigned to a combined Sb/Sj and n=3 for unit Sb). Sample sizes for these two units are small due to their limited spatial extent. Further, we sampled 26 specimens from the 'remanié' located in square 4L, to test whether unusual deamidation values were present in this feature compared to surrounding squares. Combined, the total number of studied specimens is 543. Collagen (type I, COL1) was extracted following the protocol outlined by van Doorn et al. (2011). This protocol not only allows calculation of deamidation values without inducing deamidation during protein extraction, but it also facilitates subsequent study of the sample itself by further analyses that would require demineralization. This protocol thereby reduces the need to re-sample the original bone specimen. Such subsequent analyses could include additional ZooMS verification using bone demineralization (Welker et al., 2015a), proteomic LC-MS/MS analysis (Welker et al., 2015b) or ancient DNA analysis (von Holstein et al., 2014).
Bone samples weighing between 10-30 mg were immersed in 100 µl ammonium-bicarbonate buffer (50 mM) for one day at room temperature. Next, the samples were centrifuged, the buffer removed and the sample re-immersed in 100 µl ammonium-bicarbonate (50 mM) for 1 hour at 65°C. Afterwards, 50 µl of ammonium-bicarbonate buffer, now containing eluted soluble collagen, was transferred to a new Eppendorf. Protein digestion was accomplished by adding 1 μl of trypsin (Promega) and incubation at 37°C overnight. 1 μl of TFA (5%) was added to quench the enzymatic digestion, and peptides were purified on a C 18 ZipTip (Agilent) with elution in 50 μl of 50% can in 0.5% TFA (Welker et al., 2015a). Species identification was based on peptide markers for relevant species published previously (Buckley et al., 2009;Welker et al., 2016). In case ammoniumbicarbonate buffer extraction failed to provide useful taxonomic information, the same bone sample was demineralized in acid and processed again (following Welker et al., 2015a;Welker et al., 2016). A method blank containing used reagents was processed alongside archaeological samples to check for the presence of COL1 contamination during ZooMS screening, the results of which were negative.

Glutamine deamidation calculation
Glutamine (Q) deamidation was calculated for ammonium-bicarbonate ZooMS spectra for two of the peptides studied by van Doorn et al. (2012;Wilson et al., 2012)). The studied peptides, P1105 (GVQGPPGPAGPR, COL1α1) and P1706 (DGEA-GAQGPPGPAGPAGER, COL1α1), have an identical amino acid sequence in the terrestrial species studied here, thereby removing a potential confounding influence of the primary amino acid sequence on observed deamidation rates (Robinson & Robinson, 2001). They contain only a single possibly deamidating position, and have low deamidation rates compared to other glutamine positions in COL1 , ensuring that saturation is not reached in Late Pleistocene contexts. We report deamidation ratios to range from %Gln=1.0, non-deamidated glutamines, to %Gln=0.0, fully deamidated glutamines. This change can be observed as a shift in the isotope distribution of the peptides (Figure 1). Deamidation ratios were calculated with a signal-to-noise ratio of 4.0, Fitpeaks at 3.0, Fitnoise at 2.0 and with mass steps equal to 0.1 m/z.

Statistical analysis
To distinguish outliers or possible sub-groups among the %Gln values we 1) constructed boxplots with whiskers ranging 1.5 times the interquartile range (IQR), 2) performed a Grubbs test, and 3) calculated density curves. Any %Gln value over 1.5 IQR will be identified by the boxplots as a non-parametric outlier. We tested each of these by performing Grubbs test, designed to parametrically identify outliers when dealing with small sample sizes. We employ 3 versions of Grubbs test, designated to identify a) one outlier on one tail, b) two outliers, each on one tail, and c) two outliers on one tail. Grubbs test identifies significant outliers when p≤0.05. Density curves are constructed with a kernel width of h=0.04. Density curves and boxplots are presented as violin plots, where the boxplot is centrally integrated within the density curves.

ZooMS species identification and species composition
Out of 543 bone and tooth specimens, 11 specimens (2.0%) could not be identified, 31 specimens (5.7%) were attributed to taxonomic groups that could be specified further if missing peptide markers had been present and 501 bone specimens (92.3%) could be identified to family, genus or species level (Table 2). We note that 8 out of 11 unidentified specimens are located in the 'remanié'. To achieve this success rate, 29 bone specimens needed to be demineralized after ammonium-bicarbonate buffer extraction (5.3% of the studied assemblage).
Comparison with two previous studies of the Quinçay fauna assemblage (Lavaud-Girard 1980, 1986) reveals a rather similar species composition between those studies and this study. We note the addition of Sus sp. (Suidae) to unit Ej by ZooMS screening. This taxon was also added by ZooMS to the Châtelperronian at Les Cottés (Welker et al., 2015a). Mustelidae are absent in the ZooMS assemblages studied, while Lavaud-Girard (1986) mentions the presence of weasel (Mustela sp.) and Marten (Martes martes) in units Ej and Em. This could potentially be explained by the absence of sampling complete, identifiable teeth for the ZooMS samples studied here.

Glutamine deamidation
Glutamine deamidation ratios for peptide P1105 could be calculated for 457 spectra from ammonium-bicarbonate extracts, of which 16 are dental specimens (see Figure 1 for examples). P1706 was observed in a smaller number of instances (n=266 in total, n=12 dental specimens). We excluded P1706 on the basis of a smaller number of observations for the spatial analysis presented here, and we also excluded dental specimens (unless specified otherwise). Glutamine deamidation values for all included bone specimens are attached in Supplementary Table 1.
Glutamine deamidation ratios are not correlated with bone length (Pearson R=0.051; Figure 2B) or species identification when divided into small, medium, large and megafauna size classes (ANOVA, F (3)=0.374, p=0.77). As we mostly sampled cortical bone specimens, intrinsic bone specimen properties such as bone length, taxonomic origin or bone type therefore do not explain any variation observed in deamidation.
In plan view, values for units Ej, Sj, Em and Sm indicate a relatively heterogeneous %Gln distribution within the cave (Figure 2A). Despite the overall heterogeneous composition, several patterns are apparent when looking at transverse and longitudinal cross-sections of the cave (Figure 3). Transverse deamidation values are lowest in the "I" squares, for both units (c) Boxplots of P1105 deamidation ratios observed for squares located under the limestone blocks, those directly around these blocks and squares located further away (rows 9-11). Statistically significant differences are indicated (p<0.05, *). ( Figure 3B). Although based on a small number of measurements, these squares are closest to the limestone wall of the cave. It is well known that water movement in caves has a significant influence on bone preservation, for example by precipitation along cave walls causing increased mineral dissolution in peripheral cave areas (Bocherens et al., 2008;Salesse et al., 2014). We hypothesize that such processes are indicated by our deamidation data in transverse view for square 1I and 1H. The longitudinal cross-section ( Figure 3A) reveals very similar patterns for units Ej+Sj and Em+Sm. Deamidation values are lowest in squares along row 1, and are constant throughout most of the other rows except for row 7 where there is a dip towards lower values. Low values at the front of the cave (squares 1I, 1J and 1K) can (also) be explained by increased hydrological activity along the rock face containing the cave (eg. an active dripline). The sudden dip in row 7 corresponds with the end of limestone block III. We find significant differences between squares located under the limestone blocks (%Gln P1105 =0.54±0.11) compared to squares directly around the edges of these limestone blocks (0.49±0.11; t (124.01)=3.16, p=0.00). We find no significant differences between squares under the blocks and those located further away from the edges of these blocks in rows 9-11 (0.56±0.11; t(106.07)=-1.60, p=0.11), but do find significant differences between squares along the edges of the blocks and those further away (t (136.27)=−3.79, p=0.00). This suggests that the squares around the blocks are deamidated more than other squares, possibly because the limestone blocks cause water precipitation to run off alongside the edges of these blocks ( Figure 2C). A relatively similar scenario has been proposed to explain areas with differing bone preservation at Hayonim Cave (Stiner et al., 2001). From these observations and comparisons it is clear that deamidation values of individual bone specimens reflect both localized conditions (the presence/absence of limestone blocks) and processes common to caves in general (the presence of a dripline, positioning relative to wet cave walls). No significant differences in deamidation ratios could be observed between the Châtelperronian units (Ej, Sj, Em, Sm, En and Sfi; Figure 4), although the spread in %Gln values is large. Violin plots and the boxplots included within them indicate the presence of non-parametric outliers at the positive end (towards 1.0) for units Sb/Sj ( Figure 1I), Sm ( Figure  1H) and within the 'remanié' (Figure 1J), and at the negative end (towards 0.0) for the 'remanié' ( Figure  1K-L). These will be discussed individually below. The same bone specimens and their associated %Gln P1105 values are also identified as outliers when the data are considered by square, in addition to several outliers in square 9N. These outlying values from square 9N are within the %Gln distribution observed elsewhere, but they are identified as outliers by Grubbs test (Table 3). For square 3J, one specimen had a negative outlying value compared to other values for this square, but was not identified by the Grubbs test as an outlier (Table 3). We do not consider these further.

Deamidation values for a single 'remanié'.
A total of 26 bone specimens from square 4L were studied coming from a 'remanié', a sedimentary unit composed of reworked material. Bone specimens designated as coming from this feature span a depth of 40 centimetres, encompassing the full depth of this feature as described in the excavation notebooks. The majority (8/11) of the unidentified specimens come from this feature. Analysis of glutamine deamidation values from the remaining bone specimens indicate both non-deamidated (1.0; Leporidae specimen, Figure 1J) and highly deamidated specimens (0.12/0.13; Bos/Bison specimens, Figure 1K-L). The majority of bone specimens have deamidation values similar to those obtained for surrounding squares for units Ej and Em ( Figure 1A-F, Figure 5). Both the  boxplot and the Grubbs test indicate that the high and the low %Gln values are statistical outliers (Table 3). Therefore, this reworked feature contains some material that has undergone diagenetic processes different (either in nature or in length) from those of neighbouring squares, with the majority of bone specimen %Gln values comparable to neighbouring squares. The outlying values were observed for bone specimens that were excavated at the bottom of the 'remanié' (Figure 5).

Deamidation values for units Sb and Sb/Sj.
At the back of the cave the presence of additional sedimentary units above unit Sj has been noted (Table 1). On top of unit Sb, isolated lithic and ceramic artifacts are present, attributed to Holocene time periods . In addition, detailed drawings of the cave surface indicates the presence of a concentration of artefacts on the surface, as well as a possible shallow pit in the area of squares 11M/11N/12M/12N (Roussel et al., 2016). The extent of this feature is not known.
Three bone specimens from unit Sb were sampled. To these can be added 18 bone specimens associated with a sedimentary unit that potentially includes material from both Sb and Sj (termed unit Sb/Sj). We find that the one %Gln value we obtained for unit Sb is significantly less deamidated compared to the values obtained for units Ej+Sj, Em+Sm and En+Sfi (%Gln=0.88, Figure 4, Figure 1G). This bone specimen originates from square 11K, for which no additional specimens were available.
Importantly, we identify a comparable value in unit Sb/Sj (square 10K, %Gln P1105 = 0.86, Figure 1I), suggesting that material attributed to this sedimentary unit indeed comprises a mixture of material. This value is identified as an outlier compared to other %Gln values obtained for this square by both the non-parametric boxplot and the parametric Grubbs test ( Figure  6; Table 3). These values are approximated by one additional bone specimen in square 10M (%Gln P1105 = 0.86, Figure 1H), which is attributed to unit Sm. Again, this specimen is an outlier in our analysis by square, as well as an outlier for the unit Sm (Figures 4 and 6; Table 3). This specimen potentially  represents intrusive material unrecognized during the excavation of this square, or might represent an exceptionally well-preserved bone specimen (Schroeter and Cleland 2016). Direct radiocarbon dating of this bone specimen could confirm or refute either hypothesis, as we realize that the number of bone specimens securely attributed to unit Sb is unfortunately small.

Discussion
Glutamine deamidation and spatial variation.
So far, three published studies have reported on glutamine deamidation ratios in archaeological bone collagen measured by MALDI-TOF-MS Simpson et al., 2016;Welker et al., 2016) while several other studies report on ancient bone protein deamidation investigated through LC-MS/ MS analysis. Mikšík et al. (2014) discuss the deamidation of glutamines and asparagines in a Medieval mummy from Verona, Italy. Based on LC-MS/MS measurements, the deamidation was noted as present or absent for individual glutamine or asparagine positions. Overall, 77.9% of the assigned glutamines were deamidated in bone (Mikšík et al., 2014). For both peptides studied here, their study notes the presence of a deamidated glutamine, not unexpected given the environmental conditions at which the mummy was deposited (Mikšík et al., 2014). Also based on LC-MS/MS data, two other studies presented collagen deamidation data for various Pleistocene mammals (Welker et al. 2015b;Orlando et al., 2013). Undated mammoth and horse  bones presented average glutamine deamidation in 45.9% and 42.9% of observations for permafrost samples and 82.2% and 76.1% for temperate samples (Orlando et al., 2013). The deamidation frequency for individual glutamine positions is not reported in this study, however, and so comparison of specific amino acid positions with other studies is difficult. In a context chronologically younger than that of the Middle Pleistocene horse, deamidation ratios were obtained for directly dated Toxodon specimens (59.2±24.5% for P1105), Macrauchenia bone specimens beyond the radiocarbon limit (82.8 ±14.3% for P1105), and an undated but presumably younger horse specimen (18.9±18.4%; all data from (Welker et al., 2015b). For these specimens, the variation between deamidation ratios of a single bone specimen and those between different bone specimens might be a reflection of environmental conditions rather than chronological factors, despite that the age of some of these bone specimens is unknown (Schroeter & Cleland, 2016). It has been demonstrated elsewhere that some protein extraction procedures can induce deamidation during demineralisation and/or once proteins becomes solubilized (Ren et al., 2009;Hao et al., 2011;Simpson et al., 2016), and these extraction procedures generally differ between published sources. All of the aforementioned studies, except for van ), Welker et al. (2016 and some of the data presented in Simpson et al. (2016), employed slightly different extraction conditions and so direct comparisons between datasets are difficult. The extraction procedure used here, in van Doorn et al. (2012), Welker et al. (2016 and in some data reported in Simpson et al. (2016), is identical and optimized to extract already soluble protein from the bone microenvironment (van Doorn et al., 2011;2012). Previously, it has been demonstrated that this extraction protocol does not induce Figure 7 Deamidation context for the pierced teeth found in squares (a) 2K, (b) 3K, (c) 3J, and (d) 5L. Only bone specimens with a spit accuracy of 10 cm or less are included. Note that squares 2K, 3J and 5L are covered by the limestone blocks. Filled symbols: P1105. Open symbols: P1706. Violin plots per peptide are included at the bottom, with shading similar to the points. Shaded regions indicate the approximate depth at which the Quinçay beads were found, which is unknown for the two beads in square 3K. deamidation for the studied peptides when performed on modern bone samples (Welker et al., 2016). Furthermore, we focused our main analysis on one collagen peptide (P1105) that has an identical primary amino acid sequence between the different terrestrial species included in the study, thereby removing a potential confounding effect of the primary amino acid sequence on glutamine deamidation rates (Robinson & Robinson, 2001;Robinson et al., 2004).
The variation in %Gln observed for P1105 in several sites from the van  study and for the Châtelperronian units at Quinçay is large. This variation might be a reflection of localized geological and environmental conditions, rather than chronology, and is in agreement with suggestions made elsewhere (Schroeter & Cleland, 2016). Such variation has not been investigated more specifically previously but could explain part of the %Gln variation of any given site. For Les Cottés, previous deamidation data for peptide P1105 is similar to those obtained here for Quinçay  and at the Grotte du Renne (Welker et al., 2016). At Les Cottés, the stratigraphy includes sterile layers separating some archaeological horizons, including sterile units around the studied Châtelperronian stratigraphic unit 06 . Therefore, the sterile sediments are an argument against the observed variation in %Gln being the result of younger intrusions.
Our workflow enables large-scale explorations of %Gln variation at a site, both spatially and temporally, when a sufficient number of samples are included. From a taphonomy point of view, any outlying bone specimens could subsequently be selected for more expensive and destructive analysis such as radiocarbon dating. This would enable a more directed radiocarbon dating program to be undertaken in order to understand the chronological composition of a bone assemblage, as it has recently been demonstrated that successful ZooMS analysis often also results in successful collagen extraction for AMS radiocarbon dating of the same bone specimen (Harvey et al., 2016). A comparative radiocarbon chronology for Quinçay is currently lacking, and, unfortunately, several of the bone specimens identified as statistical outliers here are too small for direct AMS radiocarbon dating.
From a spatial point of view, ZooMS ammoniumbicarbonate screening could be employed as a powerful method of screening bone preservation between excavation seasons, thereby directing subsequent excavation and bone specimen sampling decisions. For example, ZooMS has previously been combined with other biomolecular analyses such as dietary stable isotope analysis (Vaiglova et al., 2014;Charlton et al., 2016), ancient DNA analysis (Evans et al., 2016;von Holstein et al., 2014;Meiri et al., 2014), or additional palaeoproteomic analysis using LC-MS/MS (Welker et al., 2015b;Welker et al., 2016). The heterogeneity in %Gln values among the 'bulk' %Gln values present in a given square or sedimentary unit might be used to inform bone specimen selection for additional biomolecular analysis (aDNA, palaeoproteomics) towards those specimens with less deamidation. In this regard, future work should focus on exploring a possible correlation between collagen deamidation, DNA deamination and biomolecule fragmentation (for example DNA sequence length or non-enzymatic protein cleavage frequency), as both deamidation and deamination are temperature dependent and pH dependent (Lyndahl & Nyberg, 1974;Scotchler & Robinson, 1974;Catak et al., 2009;Allentoft et al., 2012).
The deamidation context of the Quinçay pierced teeth.
The Quinçay pierced teeth are, together with the Châtelperronian personal ornaments at the Grotte du Renne, Arcy-sur-Cure, unique cultural artefacts attributed to Châtelperronian Neanderthals (Hublin et al., 1996;Granger & Lévêque, 1997;Caron et al., 2011;Welker et al., 2016). The stratigraphic placement of these items is, therefore, of importance in understanding and interpreting late Neanderthal behavior (Roussel et al., 2016). We present glutamine deamidation data spatially related to these items, but could not study the ornaments themselves.
All pierced teeth come from the lower part of unit Ej (square 3J) or the upper part of unit Em (squares 2K, 3K, 5L and 10L). No comparative faunal material has been studied from square 10L, despite attempts to locate such specimens. For the tooth found in this square and for one of the two found in square 3K no Table 4 Deamidation data related to the Quinçay pierced teeth. Grubbs test for 2 outliers on one tail requires sample size to be between 3 and 30 and was therefore not calculated for squares 3J and 3K. precise stratigraphic provenance is available, although both are assigned to unit Em. We tested non-parametrically for the presence of outliers using boxplots (Figure 7), but found no outliers in the tested squares for P1105 above 1.5 IQR. Testing for the presence of one or two outliers, the Grubbs test further indicates the absence of a statistical outlier in the relevant squares (Table 4). Neither do we observe the presence of a relationship between burial depth and deamidation values (Figure 7) or aberrantly high or low %Gln values at the approximate depth the pierced teeth were found.
Taken together, for each square the %Gln for P1105 and P1706 indicate that the sampled bones constitute a single population without values similar to that of modern bone specimens or those obtained for unit Sb. The wide spread in deamidation data obtained here, and elsewhere, precludes any significant statements on the presence of intrusions chronologically close in time to the Châtelperronian units. Based on archaeological observations the Quinçay pierced teeth are of a Châtelperronian origin.

Conclusion
Assessing and understanding bone assemblage homogeneity/heterogeneity is often central to subsequent analysis, such as faunal analysis, palaeoproteomics or aDNA. Here we demonstrate that ZooMS ammonium-bicarbonate screening is capable of highlighting particular patterns in glutamine deamidation, both in spatial and temporal contexts. As demonstrated before, chronological resolution is low or absent. Therefore, our analysis allows identifying bone specimens that have undergone different diagenetic histories (either in length or type). As such elements could represent intrusive material, ZooMS allows such specimens to be selected from a larger sample set for subsequent analysis. In this regard, the identification of bone specimens with glutamine deamidation ratios similar to modern bone specimens in a reworked feature ('remanié') at Quinçay highlights the potential of such ZooMS screening. In addition, future comparisons between glutamine deamidation, DNA deamination and biomolecular fragmentation of DNA and proteins could potentially be used to identify bone specimens that are more suitable for subsequent biomolecular analysis through ZooMS analysis. The results are promising as the analysis is applicable to bone assemblages studied using ZooMS without additional costs.