|
Post by Admin on Apr 10, 2023 18:35:49 GMT
Data recording Sex and age-at-death estimates are fundamental to any osteological analysis. Sex estimation was based on pelvic and cranial morphology using standard osteological techniques (Buikstra & Ubelaker, 1994). Age-at-death estimates were based on dental development, fusion of epi- and apophyses, cranial and palatal suture closure, and morphological changes of the pubic symphyseal and auricular surfaces (Buikstra & Ubelaker, 1994). For our biological distance analysis we collected dental metric and dental nonmetric trait data. The dental metric dataset consists of 32 mesiodistal (MD) and buccolingual (BL) crown and cervical diameters of the permanent teeth recorded for each individual. Only polar teeth were recorded (UI1, UC, UP3, UM1, LI2, LC, LP3, LM1) in order to reduce genetic covariation between traits, and to minimize potential effects of fluctuating asymmetry and ontogenetic plasticity on adult tooth size (Butler, 1939; Dahlberg, 1945; Sciulli & Cook, 2016; see also Stojanowski, 2003, 2004; Thompson, Hedman, & Slater, 2015). Only left teeth were measured, but when a left tooth was missing, damaged, or affected by wear or pathology, the corresponding right antimere was measured. Maximum MD and BL crown diameters were recorded according to the procedures detailed in Mayhall (1992) and Buikstra and Ubelaker (1994). MD and BL dimensions of the cervix at the cement-enamel junction were recorded according to the procedures detailed in Hillson, Fitzgerald, and Flinn (2005). All measurements were taken using a Mitutoyo pointed blade digital sliding caliper accurate to 0.01 mm. A table listing the summary statistics of the dental metric dataset is provided in Appendix S1, Table S1. The dental nonmetric trait dataset consists of observations for 34 morphological variables in the permanent dentition of each individual. All traits were recorded according to the reference standards of the Arizona State University Dental Anthropology System (ASUDAS) described in Turner, Nichol, and Scott (1991). This system comprises a set of dental casts illustrating expression levels for various traits and specific instructions to ensure a standardized scoring procedure that minimizes observer error. Only traits on key teeth were recorded (Scott, Maier, & Heim, 2016). Scoring followed the individual count method, where a trait was counted only once per dentition, regardless of whether or not the trait appeared bilaterally. In cases where a trait was expressed asymmetrically, the side with the highest expression level was scored (Delgado et al., 2019; Edgar, 2007; Irish & Konigsberg, 2007; Scott, 1980; Sutter & Verano, 2007; Turner, 1985; Turner & Scott, 1977). To ensure accuracy, any observation that was potentially affected by dental wear, caries, or calculus was treated as missing data. We followed the standard procedure and dichotomized the ordinal-scaled dental trait scores into binary categories of absence (0) or presence (1) in order to reduce observer error and simplify data analysis. The applied dichotomization thresholds follow established breakpoints (Irish, 2016; Turner, 1987) that have proven useful for effectively capturing variation across human populations in the Mediterranean Basin (Coppa, Cucina, Lucci, Mancinelli, & Vargiu, 2007; Irish, Lillios, Waterman, & Silva, 2017; McIlvaine, Schepartz, Larsen, & Sciulli, 2014; Parras, 2004; Rathmann, Saltini Semerari, & Harvati, 2017; Ullinger, Sheridan, Hawkey, Turner, & Cooley, 2005). A table listing the summary statistics of the dental nonmetric dataset is provided in Appendix S1, Table S2. All osteological data were collected by the lead author (H.R.). The dataset is publicly available via an online data repository (https://github.com/HannesRathmann/GCSI). The data sheet provides individual-level information about sex, age-at-death, dental metrics, dental nonmetric traits scores, and dichotomized dental nonmetric traits. Data preprocessing A number of data preprocessing steps were used to ensure that patterns of dental phenotypic variation most closely approximate underlying genotypic variation. First, H.R. quantified his level of intra-observer error by re-measuring a subsample of individuals from Maria D'Anglona (n = 30) in two sessions separated by an interval of 1 week. Dental measurements from the two sessions were compared using Student's t tests performed with the t.test function in R (R Core Team, 2016). None of the comparisons exhibited a significant difference with p < 0.05 (Appendix S1, Table S3). Dichotomized dental nonmetric traits from the two sessions were compared using Cohen's Kappa tests using the cohen.kappa function from the psych package in R (Revelle, 2017). The resulting coefficients ranged from 0.621 to 1.000 with a significance of p < 0.05 for all comparisons (Appendix S1, Table S4). All comparisons indicated that intra-observer error was negligible. Second, sexual dimorphism on dental characters was analyzed using Student's t tests for metric variables and Fisher's exact tests for dichotomized nonmetric variables using the t-test and fisher.test functions in R. For comparison, we only included individuals with secure sex determinations (n = 61). We found that 53% of the metric variables (15 of 28) showed significant differences between sexes with p < 0.05 (Appendix S1, Table S5). For the dichotomized nonmetric variables we found that 4% (1 of 25) exhibited a significant difference between the sexes with p < 0.05 (Appendix S1, Table S6). Hence, all further analyses have to correct for sexual dimorphism on metric features (see Methods section below), while levels of sexual dimorphism in nonmetric features are within an acceptable range as detailed by previous ASUDAS studies (Irish, 2016; Matsumura & Oxenham, 2014; Reyes-Centeno, Rathmann, Hanihara, & Harvati, 2017; Scott & Turner, 1997). Third, inter-trait correlations between dental metric and nonmetric traits were analyzed using the mixed.cor function from the psych package in R. The mixed.cor function computes a heterogeneous correlation matrix consisting of Pearson correlations for metric variables, tetrachoric correlations for dichotomized variables, and biserial correlations for mixed variables. Correlations were generated using all observations with valid data for a pair of variables. The resulting inter-trait correlation matrix is provided in Appendix S2, Sheet 1. The correlation matrix was further visualized using the corrplot function from the corrplot package in R (Wei & Simko, 2017) (Appendix S1, Figure S1). In summary, we found a high amount of integration among dental metric variables (with 71% of all 496 pairwise comparisons exceeding r > 0.5), but general independence among dichotomized traits (with only 7% of all 465 pairwise comparisons exhibiting correlations of r > 0.5 or r < −0.5) as well as general independence among dichotomized and metric features (with only 5% of all 942 pairwise comparisons exhibiting correlations of r > 0.5 or r < −0.5). All subsequent biodistance analyses require independence among dental features. The high amount of integration among dental metric variables was removed using data reduction techniques (see Methods section below). The few instances of inter-correlation between dichotomized descriptors and between dichotomized and metric features were removed by dropping variables from analysis that had the fewest observations overall or that were correlated with multiple variables. Consequently, 18 variables were removed prior to biodistance analysis: Tuberculum Dentale UI1, Tuberculum Dentale UC, Distal Accessory Ridge UC, Distal Accessory Ridge LC, Odontome P1-P2, Mesial and Distal Accessory Cusps UP1, Peg-shaped UI2, Cusp 6 LM1, Cusp 7 LM1, Enamel Extension LM1, Root Number LM2, UP1-MD-CROWN, LP1-MD-CROWN, LM1-MD-CROWN, LI2-BL-CROWN, LI2-MD-CROWN, LI2-BL-CERVIX, and LI2-MD-CERVIX. This reduction procedure resulted in a dataset of 475 individuals and 48 variables (25 metric and 23 nonmetric traits).
|
|
|
Post by Admin on Apr 11, 2023 18:20:11 GMT
Analyzing levels of mobility in pre- and post-colonial southern Italy We analyzed mobility under a model of isolation-by-distance (IBD). The IBD model states that individuals that are geographically close to each other tend to be genetically more similar than individuals that are farther apart because of spatially limited gene flow (Wright, 1943). Therefore, a positive correlation is expected between genetic distance and geographic distance. However, the strength of this correlation depends on the level of mobility. If mobility was low, then we would expect genetic distances to be strongly correlated with geographic distances, whereas if mobility was high, we would expect genetic distances to be weakly correlated with geographic distances (Loog et al., 2017). In physical anthropology, IBD models are traditionally used to explain genetic variation using populations as the unit of analysis (e.g., Konigsberg, 1990; Ragsdale & Edgar, 2015; Relethford, 2004). Here, we take a different approach and apply the model to single individuals, a procedure that is well-established and commonly performed in genetics. To analyze mobility through time, we compared the IBD pattern of individuals living in pre-colonial southern Italy (900–700 BC) to the IBD pattern of individuals living in post-colonial southern Italy (700–200 BC). Sample sizes for these tests are detailed in Table 1.
IBD patterns were quantified by estimating the association between inter-individual biological distances (B) and geographic distances (G) using Mantel tests. Mantel tests measure the correlation between two distance matrices against a null model and assess statistical significance via a permutation procedure (Mantel, 1967). We also performed partial Mantel tests to control for the effects of a third distance matrix during our comparisons (Smouse, Long, & Sokal, 1986). This test was used to account for the fact that not all individuals in our analysis are contemporaneous (Konigsberg, 1990; Pinhasi & Cramon-Taubadel, 2009; Reyes-Centeno et al., 2017; Schillaci, Irish, & Wood, 2009). Partial Mantel tests assessed the correlation between biological distances (B) and geographic distances (G), while holding temporal distances (T) constant. Computationally, the partial Mantel test design calculates the correlation of the residuals from the independent regressions B ~ T and G ~ T. Mantel computations were performed using the mantel and mantel.partial functions from the vegan package in R (Oksanen et al., 2016). For all Mantel tests, correlation significance was determined after 1,000 permutations.
Biological distances among individuals were estimated on the basis of dental metric and nonmetric data using the Gower similarity coefficient (Gower, 1971), following the protocol set forth by Paul, Stojanowski, and Butler (2013). Gower coefficients have been extensively used in inter-individual biological distance analyses (e.g., Howell & Kintigh, 1996; Ricaut et al., 2010; Stojanowski & Hubbard, 2017; Stojanowski & Schillaci, 2006) because they can incorporate multiple variable scales (metric and nonmetric traits) and allow for missing data. Nevertheless, the amount of missing data should be reduced as much as possible in order to prevent the comparison of two individuals who share no traits in common. Because not every tooth could be observed in each individual due to poor preservation, wear, or pathology, our dataset comprises large amounts of missing values. We therefore removed the most incomplete variables and individuals from analysis in a systematic stepwise manner to ensure that no more than one-third of the variables were missing for any individual included in the analysis. We removed all MD crown diameters from the analysis because of excessive levels of missing data for these measurements (UI1-MD-CROWN, UC-MD-CROWN, UP1-MD-CROWN, UM1-MD-CROWN, LC-MD-CROWN, LP1-MD-CROWN, and LM1-MD-CROWN). Furthermore, we removed dental nonmetric variables that were monomorphic (Double Shoveling UI1, Mesial Ridge UC, and Root Number LM1) or too data sparse (Anterior Fovea LM1, Deflecting Wrinkle LM1). These reduction procedures left us with a dataset of 91 individuals and 39 variables (21 metric and 18 nonmetric traits), with less than 13% of values missing.
Gower coefficients allow for missing values, thus, in principle, no data imputation is required. Nonetheless, we imputed missing dental measurements because complete metric data rows per individual are necessary for subsequent data processing steps (see below). Missing metric data were imputed following Kenyhercz and Passalacqua (2016) using the k-nearest neighbor (kNN) algorithm using the knn function from the VIM package in R (Kowarik & Templ, 2016). The kNN algorithm searches the entire dataset for cases most similar to the one with missing data and generates a mean to replace the missing value(s). Missing nonmetric trait data were not imputed because they are generally independent (see Data Preprocessing section), making the estimation of missing values difficult or impossible (Stojanowski & Hubbard, 2017).
Dental measurements were then converted into shape variables by dividing each measurement by the geometric mean for all the measurements in each individual (Jungers, Falsetti, & Wall, 1995). This standardization procedure removes gross size from the data in order to assess differences in the proportionate contribution of individual variables to overall tooth size (Harris & Lease, 2005; Hemphill, 2013, 2016; Irish, Hemphill, de Ruiter, & Berger, 2016; Irish & Kenyhercz, 2013; Paul et al., 2013; Romero, Ramirez-Rozzi, & Pérez-Pérez, 2018; Scherer, 2007). Furthermore, this procedure adjusts for size differences between individuals that may result from sexual dimorphism.
Because the Gower coefficient requires trait independence, we transformed the highly correlated 21 dental metric variables into a smaller subset of uncorrelated factor scores by performing principal component analysis (Pilloud & Kenyhercz, 2016) using the principal function from the psych package in R. Seven unrotated principal components with eigenvalues ≥1 were retained (Kaiser, 1960). Together these components account for 77% of the total variance in the dental measurements. Each component's loadings, eigenvalues, and variance explained are listed in Appendix S1, Table S7. Individual factor scores are provided in Appendix S2, Sheet 2, as well.
Finally, the seven factor scores for individuals were combined with the 18 dental nonmetric trait variables, and we generated a matrix of pairwise Gower distance values among individuals using the daisy function from the cluster package in R (Maechler, Rousseeuw, Struyf, Hubert, & Hornik, 2017). The daisy function estimates Gower distances by converting the Gower similarity coefficient (S ij) into a distance measure by subtracting its value from one (1 − S ij). The Gower distance matrix is presented in Appendix S2, Sheet 3.
Geographic distances among individuals were measured as straight-line distances in meters between the global positioning system (GPS) coordinates for the location of each individual burial. For simplicity, burials from the same necropolis were assigned the same GPS coordinate, taken at the center of the necropolis. Further information about the number of necropoleis belonging to the archaeological sites under investigation are provided in Table 1 and Appendix S1. Geographic distances were generated with the distm function from the geosphere package in R (Hijmans, 2017) and can be found in Appendix S2, Sheet 4.
Temporal distances among individuals were measured as Euclidean distances between the mean age estimates of the date ranges of each individual burial. Many of the burials in our dataset are well-dated with date ranges within a 10–60 year span. In these instances, the mean age estimate may be an adequate approximation of the actual burial's age. However, we note that an equivalent number of burials in our dataset are poorly dated with date ranges within a 200–300 year span. For these burials the mean age estimate may be only a rough approximation of the burial's actual age. The temporal distance matrix is provided in Appendix S2, Sheet 5.
|
|
|
Post by Admin on Apr 12, 2023 18:12:19 GMT
Analyzing differences in group means and variances within and between pre- and post-colonial southern Italy We analyzed differences in group means and variances using a permutational multivariate analysis of variance (PERMANOVA). PERMANOVA compares groups of individuals to test the null hypothesis that the centroids (means) and dispersions (variances) of the groups are equivalent across groups (Anderson, 2001). PERMANOVA shares some resemblance to univariate analysis of variance (ANOVA) in that both partition the sum-of-squares between and within groups and make use of F tests to compare between-group to within-group variance. However, while ANOVA tests the significance of the generated result based on the assumption of normality, PERMANOVA tests the significance directly from the data via a permutation procedure. Moreover, while ANOVA is restricted to univariate datasets with variables measured on a continuous scale, PERMANOVA is based on any inter-individual distance matrix calculated prior to analysis. This allows researchers to choose among a wide range of useful distance measures, including those that have been designed for complex datasets that violate the assumption of normality, that consist of mixed data types, or that contain more variables than individuals. Because of its flexibility, PERMANOVA is widely used in ecology and genetics, but, surprisingly, it is rarely applied in biological anthropology, with a few recent exceptions, for example, Allen and Cramon-Taubadel (2017). For each time period, we explored diversity through space by comparing group centroids and group dispersions across archaeological sites. Furthermore, we analyzed diversity through time by comparing the group centroid and group dispersion of individuals living in pre-colonial southern Italy (900–700 BC) to the group centroid and group dispersion of individuals living in post-colonial southern Italy (700–200 BC). Sample sizes for these tests are detailed in Table 1. The PERMANOVA analysis was performed using the adonis function implemented in the vegan package in R, based on inter-individual Gower distances of dental metric and nonmetric data (see Method section above). To confirm that significant results in our PERMANOVA analysis reflect differences in group centroids rather than group dispersions, we checked for homogeneity of group dispersions using a multivariate analogue of Levene's test (Anderson, 2006) implemented in the betadisper function from the vegan package in R. For all tests, statistical significance was determined after 1,000 permutations. To further ease the interpretation of inter-individual relationships, we visualized the Gower distance matrix using principal coordinates analysis (PCoA). Group dispersions were visualized using boxplots. All graphics were created in R using functions described above and the ggplot2 package (Wickham, 2009). Identifying individual ancestries in post-colonial southern Italy We identified individual ancestry using naïve Bayes classification based on dental nonmetric traits. Naïve Bayes is a simple yet powerful classification technique based on Bayes' theorem (Cichosz, 2015). Conceptually, this technique classifies an individual of unknown ancestry (i.e., the test individual) based on a single trait into pre-defined ancestry groups (in our case, Italians and Greeks) by calculating its posterior probability of belonging to Italians, P(I), or to Greeks, P(G), and assigning it to the group with the higher posterior probability. The approach can be extended to multiple traits by sequentially applying Bayes' theorem. The method is flexible as it allows for missing variables. If a trait is missing in a test individual, the calculation of the posterior probability for this particular trait is skipped in the chain of sequentially applied Bayes' theorems. Naïve Bayes classification makes two assumptions. First, it assumes that a test individual certainly derives from one of the reference groups. This assumption seems reasonable in our case, however, we will consider this issue further in the discussion below. Second, when applied to multiple traits, it assumes that the traits are independent. This assumption generally holds true when using dental nonmetric trait data as discussed in the previous section (see Appendix S1, Figure S1 and Appendix S2, Sheet 1). Because naïve Bayes classification is conceptually simple, allows for missing variables, and has few assumptions, it is widely used in classification studies based on dental nonmetric trait data (Bailey, Weaver, & Hublin, 2009; Edgar, 2005; Herrmann, Plemons, & Harris, 2016; Scott et al., 2018). We trained our naïve Bayes model with two ancestry reference groups, Italians and Greeks. The Italian reference sample (n = 241) consists of pooled data from two sites in southern Italy (Incoronata and Maria d'Anglona) dating to the pre-colonial period (900–700 BC). The Greek reference sample (n = 116) consists of pooled data from three sites in central and southern Greece (Corinth in Corinthia, Akraiphia in Boeotia, and Karystos in Euboea) dating from the prehistoric to the Classical period (1100–350 BC). The dental nonmetric trait data from Greece were previously gathered by two of us (B.K. and E.N.) (McIlvaine et al., 2014; Nikita, Schrock, Sabetai, & Vlachogianni, 2019). All dental data from Greece were collected from the same key teeth as those utilized in this study (Scott et al., 2016) and were dichotomized using the same criteria detailed above. Prior to analysis we removed traits that showed strong inter-correlations in our southern Italian sample (see Data Preprocessing section above) and we dropped traits that were monomorphic across ancestry reference groups (Double Shoveling UI1). Ultimately, this reduction procedure left us with a training dataset comprising 237 Italians and 100 Greeks, characterized by a battery of 21 dental nonmetric trait variables. None of the three observers whose data contributed to the naive Bayes classification observed the same dentitions; thus, an inter-observer error test could not be performed. However, each observer followed the same ASUDAS definitions to score dental nonmetric traits, which ensures a standardized scoring procedure with minimal observer error. Moreover, all three observers are experienced in collecting data of this kind. In a previous study on inter- and intra-observer reliability, Nichol and Turner (1986) reported that most dental traits can be observed with adequate levels of replicability. In their study, misclassification of ranked traits by more than one grade was low (6–10%) for between-observer comparisons. Our trait dichotomization approach reduces inter-observer error even further by collapsing ranked trait scores into simplified categories of “present” or “absent” in such a way that slight scoring discrepancies are eliminated. Thus, we consider inter-observer error, although potentially present, to be negligible. Classification algorithms are sensitive to unbalanced sample sizes, biasing the prediction model towards the more common reference group. Therefore, prior to analysis we created equal sample sizes for our two reference groups by randomly removing individuals from the Italian sample until both Greek and Italian reference groups had identical sample sizes of 100 individuals. We constructed our naïve Bayes classification model using the naiveBayes function from the e1071 package in R (Meyer, Dimitriadou, Hornik, Weingessel, & Leisch, 2017). The estimated conditional probabilities for the 21 dental nonmetric traits are presented in Appendix S1, Table S8. We then applied the classification model to our test data consisting of 45 individuals from post-colonial southern Italy (Table 1) with no more than one-third of variables missing (i.e., at least 14 of 21 dental nonmetric traits preserved). Validation of the classification model was performed using the leave-one-out cross-validation method (LOOCV), where a single individual in the training data is removed and used to validate the model constructed on all other individuals in the training data. This procedure is repeated such that each individual in the training data is used once for validation. The LOOCV procedure was performed for all individuals that had at least 14 of 21 dental nonmetric traits preserved.
|
|
|
Post by Admin on Apr 13, 2023 18:47:15 GMT
3 RESULTS Levels of mobility in pre- and post-colonial southern Italy Table 2 displays the Mantel test results of IBD for pre- and post-colonial southern Italy. A positive and statistically significant Mantel correlation indicates a presence of IBD caused by limited regional mobility, whereas a negligible or negative and statistically insignificant Mantel correlation signals an absence of IBD caused by higher levels of regional mobility. For the pre-colonial period, the Mantel test revealed presence of IBD as indicated by a positive and statistically significant correlation between inter-individual biological distances and geographic distances. A similar result was obtained when we performed a partial Mantel test to control for temporal variation due to differential burial dates. In contrast, for the post-colonial period, the Mantel test revealed absence of IBD as indicated by a negligible and negative and statistically insignificant correlation between inter-individual biological distances and geographic distances. Results were similar when we performed a partial Mantel test to control for temporal variation in burial dates. Table 2. Mantel tests of isolation-by-distance (IBD) for pre- and post-colonial time periodsa Time period IBD r p Pre-colonial B ~ G 0.126 0.001b B ~ G, T 0.127 0.001b Post-colonial B ~ G −0.020 0.615 B ~ G, T −0.008 0.569 a Simple Mantel tests correlating inter-individual biological distances (B) against geographical distances (G). Partial Mantel tests correlating B and G, while controlling for the effect of temporal distances (T). Shown are Pearson correlation coefficients (r) and probability values (p). b Statistically significant difference at the 0.05 level. Differences in group means and variances within and between pre- and post-colonial southern Italy Table 3 displays the PERMANOVA and Betadisper results of differences in group means and group variances through space (between archaeological sites within a period) and time (between pre- and post-colonial periods). Figure 2 illustrates the PERMANOVA and Betadisper results in PCoA plots and boxplots, respectively. We performed three comparisons. First, we compared individuals in different archaeological sites during the pre-colonial period (i.e., Incoronata and Maria d'Anglona). PERMANOVA analysis revealed a statistically significant difference in group centroids and/or group dispersions. Betadisper analysis indicated that the significant PERMANOVA result was due to nonhomogeneous group dispersion and estimated the average dispersion for Maria d'Anglona as 0.205, whereas the average dispersion for Incoronata was estimated as 0.168. Thus, inhabitants of Maria d'Anglona were about one and a half times more variable as those living in Incoronata. Nevertheless, the PCoA plot also indicates some degree of biological separation between the centroids of the two sites. Second, we compared individuals in different archaeological sites during the post-colonial period (i.e., Passo di Giacobbe, Metaponto, and Siris). PERMANOVA found no significant difference in group centroids and group dispersions. Note that for this analysis we removed the single individual from Taras because PERMANOVA and Betadisper analyses require that groups consist of at least two entities. Third and last, we compared individuals in pre-colonial southern Italy to individuals in post-colonial southern Italy. PERMANOVA analysis estimated that there was no significant difference in group centroids and group dispersions. Table 3. Analysis of variance tables for permutational multivariate analysis of variance (PERMANOVA) and Betadiper models testing differences in centroid (mean) and dispersion (variance) of inter-individual Gower distances across sites and time periodsa PERMANOVA Betadisper Test df SS MSS F R2 p SS MSS F p Differences across sites in pre-colonial period 1 0.152 0.152 3.868 0.090 0.001b 0.014 0.014 6.097 0.018b Residual 39 1.531 0.039 0.910 0.088 0.002 Differences across sites in post-colonial periodc 2 0.046 0.023 0.541 0.023 0.882 0.002 0.001 0.338 0.715 Residual 46 1.975 0.043 0.977 0.131 0.003 Differences across pre- and post-colonial periods 1 0.072 0.072 1.704 0.019 0.129 0.001 0.001 0.001 0.985 Residual 89 3.747 0.042 0.981 0.220 0.002 a Shown are factor and residual degrees of freedom (Df), sum of squares (SS), mean sum of squares (MSS), F-statistic values (F), proportion of explained variance (R2) and probability values (p). b Statistically significant difference at the 0.05 level. c To assess differences across sites during the post-colonial time period, we removed the single individual from Taras because PERMANOVA and Betadisper analyses require that groups consist of at least two entities.
|
|
|
Post by Admin on Jul 13, 2023 18:37:37 GMT
Archaeological questions and genetic answers: Male paternal kinship in a copper age multiple burial from the eastern Italian Alps Abstract In 2010, a detailed anthropological study regarding two adult individuals and one foetus-newborn buried in a Copper Age (ca. 3000–2700 cal. BC) multiple grave site found in Ora/Auer in the eastern Italian Alps (Alto Adige/ Südtirol) was published. This exceptional archaeological finding provides a rare insight into an inhumation prehistoric ritual discovered within a natural alpine rock shelter. Due to the presence of an infant in the grave, the authors were doubtful of the anthropological male sex assignment given to both adult individuals found there. Additionally, correlation of non-metric traits suggested a possible kinship between them. To determine the biological sex and to investigate genetic relatedness, we performed a paleogenetic investigation (shotgun and enrichment data of ∼2 million SNPs) of the two adults. We successfully analyzed the ancient DNA of these individuals, confirming that biologically, both were males (XY). Moreover, through kinship analyses and data from unilinear-transmitted markers (Y-Chromosome and mitochondrial DNA), we detected a first-degree paternal kinship (most likely father-son). This study underlines the importance of interdisciplinary dialogue between archaeology, anthropology and palaeogenomics, demonstrating how the latter can significantly support the interpretation of funerary contexts. 1. Introduction In 2010, Rizzi and colleagues published a detailed anthropological and taphonomic study of an exceptional archaeological find in Alto Adige/Südtirol (hereinafter referred to as South Tyrol), located in the eastern part of the Italian Alps (Fig. 1a). Inside a natural niche at the base of a high cliff in the present municipality of Ora/Auer (Fig. 1b) in the Adige Valley, human remains of three individuals were found in primary deposition (Fig. 2a). Two were adults of approximately 30 (Ind. A) and 40+ years of age (Ind. B). The third individual was a foetus-newborn with an estimated gestational age of 38–40 weeks in utero (Ind. C). Osteological samples were radiocarbon (14C) dated to the early, but not initial, Copper Age (Ind. A: LTL-3607A 4279 ± 45 BP, 3021–2705 cal. BC [95.4%]; Ind. B: LTL-3607A 4268 ± 35 BP, 3008–2707 cal. BC [95.4%], Tecchiati, 2013). Among items found in the grave were a red deer antler tool and some bone beads which are postulated to be associated with the burial (Rizzi et al., 2010). Fig. 1. a. Map displaying the location of Ora/Auer in the Adige Valley in Alto Adige/Südtirol. b. Excavation of the natural rock niche where the burial site was discovered in 2007 (photo courtesy of Jasmine Rizzi). Fig. 2. a. The original positions of Ind. A and B found in the natural rock shelter at Ora/Auer (Rizzi et al. 2010, modified). White crosses show where the osteological remains of Ind. C (the foetus-newborn) were found at the time of discovery. b. Ind. A cranium basal view shows the hole on the left PP (arrow) that was sampled for genetic analysis. There are several reasons why this burial site is particularly unique and important. First, findings of inhumed individuals in this territory are extremely rare and only a few cases have been reported in the literature (De Marinis, 2003, Steiner et al., 2017, Tecchiati, 2011). In fact, archaeological evidence dated to the Copper Age and to the Early Bronze Age suggests a complex funerary practice in South Tyrol as well as the Northern Alps that was more ritual than intentional (Baur, 2006, Tecchiati, 2013, Tecchiati et al., 2016). The ritual involved different kinds of body manipulations (e.g., exhumation, burning, fragmentation), which ended with deposition of the burned human remains often together with cultural objects and animal bones (e.g., Barbiano/Barbian Gostner in Coltorti et al., 2010, Dal Ri et al., 2002; Millan Stockner in Tecchiati, 2013). Moreover, the incinerated bones were sometimes surrounded by stone circles like those found at Varna Circonvallazione in Isarco Valley, South Tyrol (Tecchiati, 2014). Second, Copper Age burial sites with clear tomb structures in natural rock shelters, such as the one found in Ora/Auer, are very rare in South Tyrol (Tecchiati, 2013). This is in contrast to the neighboring Trentino region where this funerary practice is found more often (Mottes and Nicolis, 2019). There are some other similarities to the Trentino area, including the discovery of another foetus buried in an Early Bronze Age vase at Ora/Auer that dated to the later phases of usage of the tomb (burial in pithos, Nicolis, 2001). In their study, Rizzi and colleagues (2010) used morphological and morphometric methods to postulate that the two adult individuals (Ind. A and Ind. B) were male. However, they doubted this sex estimation conclusion due to the presence of the foetus-newborn (Ind. C) originally found on the chest of Ind. B, as indicated by the taphonomic investigation. This initial discovery suggests that Ind. B could be a female who was buried together with her child. The burial site at Ora/Auer may therefore have represented a family burial. Additionally, the authors report that both adult individuals exhibited the same cranial and postcranial non-metric traits (i.e., lambdoid ossicles, suprameatal spines and depressions, fusion of the os trigonum with the astragalus as well as the third femoral trochanter, Rizzi et al., 2010), suggesting a possible kinship. In the absence of genetic analyses, classic osteology was used to infer a relationship between individuals through the presence of non-metric or epigenetic traits (Berry and Berry, 1967). The latter are skeletal and dental variants considered to be bio-distance indexes (Lippi, 2009). However, the analysis of non-metric traits present methodological limitations as unrelated individuals could share morphological traits by chance (Vai et al., 2020). In fact, these traits only give some preliminary indications to infer kinship that need to be genetically validated to provide a reliable result (Ricaut et al., 2010). To further explore this rare prehistoric find and clarify ambiguities in the interpretation of anthropological results, we performed a paleogenetic analysis to answer two specific questions: 1) are the two adults biological males?; 2) are they genetically related? Ancient DNA (aDNA) analyses were done on nuclear and unilinear transmitted markers (mitochondrial DNA and Y-Chromosome) of Ind. A and B found at the Ora/Auer site. Unfortunately, a paleogenetic analysis of the foetus-newborn (Ind. C) was not possible due to the paucity and poor preservation of the infant’s immature bones. www.sciencedirect.com/science/article/pii/S2352409X2300278X
|
|