Abstract
Objectives: This article presents estimates of narrow-sense heritability and bivariate genetic correlation for a series of morphological crown variants of the anterior dentition. These results provide insight into the value of dental phenotypes as evolutionary proxies, as well as the development of tooth crowns as integrated or modular structures. Materials and Methods: African American dental casts from the Menegaz-Bock collection were scored for a standard set of dental morphological variables using the Arizona State Dental Anthropology System. Estimates of narrow-sense heritability and genetic correlations were generated using SOLAR v. 8.1.1, controlling for the covariates of age, sex, and birth year. Analyses were run using ordinal/continuous scale variables that were then dichotomized at various breakpoints, consistent with standard practices in dental anthropology. Results: Heritability estimates were low to moderate for most traits, and lower in magnitude than those reported for odontometric data from the same study sample. Only winging, canine shoveling, and canine double shoveling returned narrow-sense heritabilities that did not differ significantly from zero. Genetic correlations were high among antimeres and metameres and low for different traits scored on the same tooth crown. These results affirm standard data cleaning practices in dental biodistance. Double shoveling was atypical in returning strong negative correlations with other traits, shoveling in particular. Conclusions: Additive genetic variation contributes to dental morphological variation, although the estimates are uniformly lower than those observed for odontometrics. Patterns of genetic correlation affirm most standard practices in dental biodistance. Patterns of negative pleiotropy involving lingual and labial crown features suggest a genetic architecture and developmental complex that differentially constrain morphological variation of distinct surfaces of the same tooth crown. These patterns warrant greater consideration and cross-population validation.
Original language | English (US) |
---|---|
Pages (from-to) | 124-143 |
Number of pages | 20 |
Journal | American journal of physical anthropology |
Volume | 167 |
Issue number | 1 |
DOIs | |
State | Published - Sep 2018 |
Keywords
- biodistance
- dental morphology
- heritability
- pleiotropy
- quantitative genetics
ASJC Scopus subject areas
- Anatomy
- Anthropology
Fingerprint
Dive into the research topics of 'Heritability and genetic integration of anterior tooth crown variants in the South Carolina Gullah'. Together they form a unique fingerprint.Cite this
- APA
- Standard
- Harvard
- Vancouver
- Author
- BIBTEX
- RIS
In: American journal of physical anthropology, Vol. 167, No. 1, 09.2018, p. 124-143.
Research output: Contribution to journal › Article › peer-review
}
TY - JOUR
T1 - Heritability and genetic integration of anterior tooth crown variants in the South Carolina Gullah
AU - Stojanowski, Christopher M.
AU - Paul, Kathleen S.
AU - Seidel, Andrew C.
AU - Duncan, William N.
AU - Guatelli-Steinberg, Debbie
N1 - Funding Information: Narrow-sense heritability estimates and significance tests for covariate screening are presented in Table. For most variables, multiple estimates are provided. The first estimate treats the ordinal scale data as continuous (which maximizes variation), with subsequent estimates dichotomized into a binary variable at multiple breakpoints. For some traits, dichotomization above a certain breakpoint resulted in a very low sample frequency. This invariably resulted in kurtosis estimates that were too high and heritability estimates that were spurious. In general, binary variables with sample frequencies less than ∼18% failed to produce results. For the purposes of this discussion, “a trait” refers to the expression of a feature on a specific tooth (i.e., I1 and C1 shoveling are considered different “traits”). Raw scores are presented in Supporting Information Table S1. aL = left; R = right; I = incisor; C = canine; P = premolar. Maxillary and mandibular traits indicated by superscript and subscript, respectively. WING = winging; LCRV = labial curvature; SHOV = shoveling; DSHOV = double shoveling; TD = tuberculum dentale; MR = mesial ridge; DAR = distal accessory ridge; PEG = peg form; CONT = trait treated as a continuous variable; BP = dichotomization breakpoint when trait treated as a binary variable. Dashes are associated with incalculable parameter estimates. While premolars are in the postcanine region of the dentition, premolar double shoveling is included here, because this trait is homologous to double shoveling on the anterior teeth. Binary trait frequencies are listed within parentheses and italicized; these are the sample frequencies when trait expression is dichotomized at indicated breakpoints. See Supporting Information Table S1 for raw scores. bN = sample size for heritability estimation. cK = model kurtosis value. Asterisks indicate a violation of kurtosis assumptions. These model results should be interpreted with caution. dh2 = total heritability estimate (MLE). eAll significant heritability estimates (p value < 0.050) and associated probability value estimates are bolded. fSE = standard error estimate (MLE). gc2 = total covariate estimate (MLE). hAll significant probability value estimates (p value < 0.100) for the covariates of age, sex, age/sex interaction, and birth year are bolded. Covariate screening indicated that age and birth year were not significant covariates for most variables. Although several analyses returned significant p values at the 0.05 and 0.10 level, there was no patterning to these results that suggests a real biological effect is evident, and the significant p values likely result from family-wise error. Sex was another matter. Numerous traits returned significant p values for sex as a covariate (Table). Some of these are also likely the result of family-wise error, for example when only a single test is significant across both sides and all breakpoint values. This suggests a spurious result. There was some evidence for a significant sex effect for shoveling and double shoveling, but this was rarely consistent across sides or breakpoints. The exception was for double shoveling of the premolars. Further, tuberculum dentale returned significant sex covariate p values for the I1, I2, and C1, but only when the traits were dichotomized. This might indicate that the highest degrees of expression were found exclusively (or predominantly) in males, and the effect was only pronounced enough in binary scale. Canine mesial ridge demonstrated a consistent sex effect, as did the distal accessory ridge (particularly when expressed in the mandible). Why the maxillary and mandibular results differ for this trait is unclear. Despite specific exceptions, the overall trend in the morphological dataset contrasts with that observed in the odontometric data previously reported (Stojanowski et al.,), for which sex was a significant covariate in nearly every model. These results, therefore, confirm previous research suggesting sexual dimorphism is limited for morphological data and restricted to specific crown traits, here marginal and cingulum ridge developments of the canine. All but three traits exhibited significantly positive narrow-sense heritability. Winging, C1 shoveling, and C1 double shoveling presented no significant estimates regardless of data scale or breakpoint. Mandibular canine distal accessory ridge exhibited significant heritability for some trait breakpoints, but only for the left side; these results are internally inconsistent and also suggest limited heritability for this trait. We note that the frequency of winging was 3% for this sample, which suppressed the heritability estimate because phenotypic variation in the sample was low. In addition, canine shoveling and double shoveling are not normally used in dental biodistance (but see Nichol,); both traits exhibit reduced ranges of variation in this sample and reflect a limited amount of phenotypic variation. Importantly, kurtosis estimates were not within an acceptable range (even after inorm transformation) for winging and C1 shoveling, indicating poor model fit. Results should be interpreted cautiously for all estimates in Table for which model kurtosis was beyond acceptable limits. For approximately half of the statistically significant estimates, heritabilities were highest when quantifying the variable on a continuous scale. This pattern is best demonstrated for I1 labial curvature where there is a linear correlation between breakpoint, heritability estimate, and trait frequency. For mandibular tooth traits, binary estimates based on breakpoints often produced the higher estimates. For mandibular incisor shoveling this could reflect the fact that shoveling is weakly developed in the mandible and there is less overall variation to begin with, which is also subject to higher observer error due to detailed gradations present on the scoring plaque. Dichotomization ameliorates this issue. The other traits that returned the same pattern (where the binary heritability was higher than the continuous) are subject to the same scoring limitations. Distal accessory ridge, shoveling, and double shoveling are difficult to score at low grades of expression; however, presence/absence distinctions are more apparent. Average heritability across all continuous scale traits was 0.248 and 0.262 for the left and right sides, respectively. Comparable estimates using only significant (nonzero) heritabilities are 0.342 and 0.334 for the left and right sides, respectively. Comparisons of maxillary and mandibular traits produced similar results. Maxillary traits returned higher heritabilities (left = 0.339, right = 0.338) than mandibular traits (left = 0.222, right = 0.186) when averaged over all traits scored in each arcade (the number of mandibular traits was too small for significance testing). On a trait-by-trait basis the same was true (Table). Maxillary incisor shoveling returned a higher heritability than mandibular shoveling, and maxillary canine distal accessory ridge returned a much higher heritability than mandibular distal accessory ridge (for which many of the binary estimates did not significantly differ from 0.0). Estimates of genetic correlation (pleiotropy) allow us to ask different questions of the morphological dataset. Genetic, environmental, and phenotypic correlations are presented by side in Table, by tooth position in Table, and by trait in Table. Among antimeres, the genetic correlations are uniformly high, with many values estimated at 1.0 (Table). Most traits demonstrate evidence of complete antimeric pleiotropy with the exception of maxillary premolar double shoveling, which demonstrates evidence for incomplete pleiotropy and the lowest antimeric correlations. Whether this is spurious or not requires validation in additional samples. Phenotypic correlations among antimeres are also large and positive. Together, these results justify the practice of using the individual count method (i.e., quantifying sample size based on individuals instead of elements, and using only a single antimere per element/trait to represent an individual) in biodistance analysis. When compared to previous assessments of odontometric data (Stojanowski et al.,), results for morphological variables indicate slightly lower antimeric correlation (morphology = 0.708, odontometric = 0.792); however, the difference is not statistically significant (p = 0.08). These data reflect less symmetry in morphological trait expression than in tooth size and, when combined with the lower narrow-sense heritability estimates, may reflect the effects of increased stress in this population, which elevated environmental variance that affected the expression of morphological traits. aL = left; R = right; I = incisor; C = canine; P = premolar. Maxillary and mandibular traits indicated by superscript and subscript, respectively. WING = winging; LCRV = labial curvature; SHOV = shoveling; DSHOV = double shoveling; MR = mesial ridge; TD = tuberculum dentale; PEG = peg form; DAR = distal accessory ridge. Dashes are associated with incalculable parameter estimates. While premolars are in the postcanine region of the dentition, premolar double shoveling is included here, because this trait is considered homologous to double shoveling on the anterior teeth. Breakpoints for traits are uniform across antimeres and were chosen based on highest univariate heritability estimates across the valid antimeric models with the fewest possible distributional assumption violations. bN = correlation analysis sample size; Cov = significant covariates (fixed in the correlation analyses); S = sex; K = kurtosis violation, results should be interpreted with caution. cMaximum-likelihood estimate of genetic correlation. Cases of incomplete pleiotropy indicated by a single asterisks. Cases of complete pleiotropy indicated by two asterisks. dProbability of hypothesis (as indicated in parentheses) being true given pedigree structure with values p < 0.050 bolded. eMaximum-likelihood estimate of environmental correlation. fMaximum-likelihood estimate of derived phenotypic correlation. aL = left; R = right; I = incisor; C = canine. Maxillary and mandibular traits indicated by superscript and subscript, respectively. WING = winging; LCRV = labial curvature; SHOV = shoveling; DSHOV = double shoveling; TD = tuberculum dentale; PEG = peg form; DAR = distal accessory ridge; MR = mesial ridge. Dashes are associated with incalculable parameter estimates. Breakpoints were chosen based on highest univariate heritability estimates across valid models free of distributional assumption violations. bMaximum-likelihood estimate of genetic correlation. Cases of incomplete pleiotropy indicated by a single asterisks. Cases of complete pleiotropy indicated by two asterisks. cProbability of hypothesis (as indicated in parentheses) being true given pedigree structure with values p < 0.050 bolded. dMaximum-likelihood estimate of environmental correlation. eMaximum-likelihood estimate of derived phenotypic correlation. fSex covariate fixed in model. aL = left; R = right; I = incisor; C = canine; P= premolar. Maxillary and mandibular traits indicated by superscript and subscript, respectively. Dashes are associated with incalculable parameter estimates. Breakpoints were chosen based on highest univariate heritability estimates across valid models free of distributional assumption violations. bMaximum-likelihood estimate of genetic correlation. Cases of incomplete pleiotropy indicated by a single asterisks. Cases of complete pleiotropy indicated by two asterisks. cProbability of hypothesis (as indicated in parentheses) being true given pedigree structure with values p < 0.050 bolded. dMaximum-likelihood estimate of environmental correlation. eMaximum-likelihood estimate of derived phenotypic correlation. fSex covariate fixed in model. In Table, we present genetic correlations by tooth position. For upper central incisors the following traits were compared: winging, shoveling, labial curvature, double shoveling, and tuberculum dentale. There is evidence for incomplete or complete pleiotropy for shoveling and labial curvature, shoveling and tuberculum dentale, and labial curvature and tuberculum dentale. This pattern is fairly consistent among antimeres suggesting a non-spurious result. Most correlations were positive with the exception of correlations involving double shoveling, which returned negative correlations with other traits. For upper lateral incisors the following traits were compared: shoveling, double shoveling, tuberculum dentale, and peg-form. There were few significant results although most correlations were positive. Complete pleiotropy is inferred for shoveling and peg-form, while double shoveling produced negative correlations when compared with shoveling (as with the I1). Finally, for maxillary canines the following traits were compared: canine mesial ridge, shoveling, double shoveling, tuberculum dentale, and distal accessory ridge. Unfortunately, many of these tests failed to converge. Successful tests do suggest genetic independence among most of these features, however. Larger samples sizes are needed to resolve the pattern of statistical significance. We note, however, that double shoveling returns negative correlations for many of the canine comparisons, as well. In Table we present genetic correlations for four traits that are scored on multiple teeth: shoveling, double shoveling, distal accessory ridge, and tuberculum dentale. Shoveling demonstrated positive genetic correlations for most comparisons with evidence of complete pleiotropy between LI1 and LI2, LI2 and LC1, RI1 and RI2, and RI2 and RC1. In addition, there was evidence for incomplete pleiotropy between LI1 and LI2, LI2 and LI2, and RI1 and RI2. Correlations were higher for within-arcade comparisons and highest for contiguous teeth regardless of tooth class. This might suggest a spatial field effect. Double shoveling also demonstrated positive correlations for most comparisons, although many failed to reach statistical significance. The pattern of correlations was similar to that documented for shoveling with strong genetic integration of I1, I2, and C1. The inclusion of premolars complicates the interpretation. The P2 presented evidence of shared genetic effects with the incisors and canines, but this was less evident for the P1, and the P1 and P2 do not appear to be genetically integrated. Distal accessory ridge was genetically correlated between isomeres, as expected, but complete pleiotropy was only indicated for the right side. Finally, tuberculum dentale was positively genetically correlated in the maxillary incisors and canines. However, in contrast to shoveling and double shoveling, for this trait the incisors indicated the highest correlation. Tuberculum dentale expression on the canines was indicated to be, in most cases, genetically independent of incisor expression, suggesting that tooth type does affect the manifestation of this trait. Dental morphological data were scored for anterior dental traits following the Arizona State University Dental Anthropology System (ASUDAS) (Turner et al.,). The standards as described by Turner et al. () were followed with the exception of winging for which the absence of winging was scored as a zero, and each tooth was scored for presence of winging as a one. All data were recorded by WND, whose documented scoring precision is within acceptable limits. Previous intra-observer error tests indicated 20% of comparisons varied by one grade on the ASUDAS plaques and 4% varied by more than one grade (Duncan,). In a more recent, unpublished study, Duncan found that 15% differed by one grade and 5% differed by more than one grade. These results align with those of Scott (), whose intra-observer error equaled one grade in 14% of cases and exceeded a single grade in only 2% of cases. They also corroborate Marado's () ASUDAS intra-observer study that yielded a mean overall precision value exceeding 92% when considering variation within a single scoring grade. Following our previous study of Gullah crown dimensions (Stojanowski et al.,), narrow-sense heritability estimates were generated using maximum likelihood variance components analysis in the program SOLAR v. 8.1.1 (Almasy & Blangero,; Blangero et al.,). These analyses are useful in modeling phenotypic variance/covariance and can accommodate pedigree data (i.e., genetic relatedness) from complex and/or unbalanced genealogies, similar to those represented in the Gullah sample (Amos,). Maximum likelihood estimation assesses model fit through serial log-likelihood testing, ultimately converging upon optimized parameter estimates (Carson,; Konigsberg,; Lynch & Walsh,; Shaw,). SOLAR employs the phenotypic covariance expression: Ω=2Φ σ2G+Iσ2E, where Ω represents a covariance matrix, Φ represents a kinship coefficient matrix, σ2G signifies additive genetic variance, I denotes an identity matrix, and σ2E represents random (environmental) variance (Almasy & Blangero,). SOLAR reports a univariate heritability estimate defined as the ratio of quantifiable phenotypic variance attributable to additive genetic variance (h2=σ2G/σ2P) (Almasy & Blangero,; Hlusko et al.,). Four covariates were included in each variance components model: sex, age, sex*age interaction, and birth year. Any covariate whose mean effect was deemed significant (p < 0.10) was subsequently fixed in bivariate genetic correlation analyses. We next generated a suite of bivariate models in SOLAR, employing a multivariate derivation of the phenotypic variance/covariance expression presented above (see Boehnke, Moll, Kottke, & Weidman,; Falconer,; Hopper & Mathews,; Lange, Boehnke, & Opitz,). These models were expressed as Ω=G ϒ 2Φ +E ϒ I, where Ω denotes a phenotypic covariance matrix, G represents a genetic variance-covariance matrix, Υ denotes a Kronecker product operator, Φ signifies a kinship coefficient matrix, E represents an environmental variance-covariance matrix, and I denotes an identity matrix (as applied in Almasy, Dyer, & Blangero,; Blangero, Konigsberg, & Vogler,; Hlusko et al.,; Mahaney et al.,; Williams-Blangero & Blangero,; Williams-Blangero, Blangero, & Beall,). Additive genetic correlations (ρG) and environmental correlations (ρE) between paired characters were estimated with reference to the G and E matrices. These values were also employed in the estimation of bivariate phenotypic correlations (ρP), following ρP=h12 h22 ρG+(1−h12) 1−h22ρE. The significance of parameter estimates was assessed using likelihood ratio tests, comparing a restricted model (parameter of interest held constant) to an unrestricted model (all parameter values estimated) (Almasy & Blangero,; following Hlusko & Mahaney,; Hlusko et al.,). Significance tests for bivariate genetic correlations involved comparing likelihoods between the unrestricted model and restricted model A (genetic correlation fixed at 0.0) and restricted model B (genetic correlation fixed at 1.0) (Almasy & Blangero,). If model likelihoods significantly differed between the unrestricted model and restricted model A, results were interpreted as complete pleiotropy; in other words, all of the trait pair's additive genetic variance can be attributed to the effects of the same genes or overlapping set of genes. If model likelihoods significantly differed between the unrestricted model and restricted models A and B, results were interpreted as incomplete pleiotropy; that is, some the trait pair's additive genetic variance can be attributed to the effects of the same genes or overlapping sets of genes. Correlations that did not differ significantly from 0.0 or 1.0 were under powered and were not interpreted in any specific way (no result). Heritability and genetic correlation estimates were generated based on expression scored from both left and right antimeres. Because SOLAR accommodates only continuous and binary variables, the data were analyzed in two ways. First, the ordinal data were treated as continuous, appealing to the widely held assumption that nonmetric traits of the dentition and skull are characterized by underlying continuous, normal distributions (Carson,; Cheverud, Buikstra, & Twichell,; Corruccini,; Harris,; Konigsberg, Kohn, & Cheverud,; Sofaer, Niswander, MacLean, & Workman,). This also follows the methodology employed by Hlusko and Mahaney () in their analysis of cingular remnant variation in savannah baboons. For these analyses, data were transformed using SOLAR's inorm function to avoid potential error stemming from the violation of distributional assumptions, especially kurtosis. Next, each ordinal variable was decomposed to generate a series of binary variables, using expression grades as dichotomization breakpoints. For example, canine distal accessory ridge is an ordinal trait with six possible expression grades (0 to 5) (Turner et al.,). We dichotomized this variable at each grade, thus using data collected for a single trait to generate five new binary variables (i.e., 1+=present; 2+=present; 3+=present; 4+=present; 5 = present). Heritability estimates were then generated for each of these new binary variables. The casts used in this study were collected by Rene Menegaz-Bock as part of her dissertation work on variation in the anterior dentition. These materials are currently part of the collections of the Department of Anthropology at The Ohio State University (Edgar & Nardin,), and this research was conducted with the approval of their IRB (2012B0529). The Gullah sample includes dental casts from an African American population that was living on James Island, South Carolina during the early to mid-twentieth century (Menegaz-Bock,; Pollitzer,; Pollitzer, Menegaz-Bock, Ceppellini, & Dunn,) and includes 469 casts (of ∼3,400 residents) representing 76 families. While many of the reconstructed genealogies were small, comprising only two generations and a pair of casted individuals, the largest genealogy spans seven generations and includes more than 100 casted individuals. Genealogical reconstruction also resulted in the identification of 22 bilineal families, the largest of which incorporates 313 casted individuals (see also Stojanowski et al.,). The complex nature of the reconstructed genealogies is consistent with ethnographic data that suggests Gullah society engaged in broadly inclusive relationships, emphasizing fairly stable marriages, shared familial ties, and household and/or patrilineal exogamy (Pollitzer,; Twining & Baird,). The casts have previously contributed to analyses of fluctuating dental asymmetry (Guatelli-Steinberg, Sciulli, & Edgar,) and studies of dental morphology (Edgar & Sciulli,) and population history (Edgar,). The Gullah trace their origins to west and central Africa (Edgar,; Parra et al., 2001; Pollitzer,; Rogers,) and are one of the least acculturated African American populations in the United States (Herskovits,; Pollitzer,; Twining & Baird,) with low rates of European and Native American admixture (Benn Torres, Doura, Keita, & Kittles,; McLean, Argyropoulos, Page, Shriver, & Garvey, 2005; McLean et al.,; Parra et al.,; Pollitzer,). Dental morphological data were scored for anterior dental traits following the Arizona State University Dental Anthropology System (ASUDAS) (Turner et al.,). The standards as described by Turner et al. () were followed with the exception of winging for which the absence of winging was scored as a zero, and each tooth was scored for presence of winging as a one. All data were recorded by WND, whose documented scoring precision is within acceptable limits. Previous intra-observer error tests indicated 20% of comparisons varied by one grade on the ASUDAS plaques and 4% varied by more than one grade (Duncan,). In a more recent, unpublished study, Duncan found that 15% differed by one grade and 5% differed by more than one grade. These results align with those of Scott (), whose intra-observer error equaled one grade in 14% of cases and exceeded a single grade in only 2% of cases. They also corroborate Marado's () ASUDAS intra-observer study that yielded a mean overall precision value exceeding 92% when considering variation within a single scoring grade. Following our previous study of Gullah crown dimensions (Stojanowski et al.,), narrow-sense heritability estimates were generated using maximum likelihood variance components analysis in the program SOLAR v. 8.1.1 (Almasy & Blangero,; Blangero et al.,). These analyses are useful in modeling phenotypic variance/covariance and can accommodate pedigree data (i.e., genetic relatedness) from complex and/or unbalanced genealogies, similar to those represented in the Gullah sample (Amos,). Maximum likelihood estimation assesses model fit through serial log-likelihood testing, ultimately converging upon optimized parameter estimates (Carson,; Konigsberg,; Lynch & Walsh,; Shaw,). SOLAR employs the phenotypic covariance expression: Ω=2Φ σ2G+Iσ2E, where Ω represents a covariance matrix, Φ represents a kinship coefficient matrix, σ2G signifies additive genetic variance, I denotes an identity matrix, and σ2E represents random (environmental) variance (Almasy & Blangero,). SOLAR reports a univariate heritability estimate defined as the ratio of quantifiable phenotypic variance attributable to additive genetic variance (h2=σ2G/σ2P) (Almasy & Blangero,; Hlusko et al.,). Four covariates were included in each variance components model: sex, age, sex*age interaction, and birth year. Any covariate whose mean effect was deemed significant (p < 0.10) was subsequently fixed in bivariate genetic correlation analyses. We next generated a suite of bivariate models in SOLAR, employing a multivariate derivation of the phenotypic variance/covariance expression presented above (see Boehnke, Moll, Kottke, & Weidman,; Falconer,; Hopper & Mathews,; Lange, Boehnke, & Opitz,). These models were expressed as Ω=G ϒ 2Φ +E ϒ I, where Ω denotes a phenotypic covariance matrix, G represents a genetic variance-covariance matrix, Υ denotes a Kronecker product operator, Φ signifies a kinship coefficient matrix, E represents an environmental variance-covariance matrix, and I denotes an identity matrix (as applied in Almasy, Dyer, & Blangero,; Blangero, Konigsberg, & Vogler,; Hlusko et al.,; Mahaney et al.,; Williams-Blangero & Blangero,; Williams-Blangero, Blangero, & Beall,). Additive genetic correlations (ρG) and environmental correlations (ρE) between paired characters were estimated with reference to the G and E matrices. These values were also employed in the estimation of bivariate phenotypic correlations (ρP), following ρP=h12 h22 ρG+(1−h12) 1−h22ρE. The significance of parameter estimates was assessed using likelihood ratio tests, comparing a restricted model (parameter of interest held constant) to an unrestricted model (all parameter values estimated) (Almasy & Blangero,; following Hlusko & Mahaney,; Hlusko et al.,). Significance tests for bivariate genetic correlations involved comparing likelihoods between the unrestricted model and restricted model A (genetic correlation fixed at 0.0) and restricted model B (genetic correlation fixed at 1.0) (Almasy & Blangero,). If model likelihoods significantly differed between the unrestricted model and restricted model A, results were interpreted as complete pleiotropy; in other words, all of the trait pair's additive genetic variance can be attributed to the effects of the same genes or overlapping set of genes. If model likelihoods significantly differed between the unrestricted model and restricted models A and B, results were interpreted as incomplete pleiotropy; that is, some the trait pair's additive genetic variance can be attributed to the effects of the same genes or overlapping sets of genes. Correlations that did not differ significantly from 0.0 or 1.0 were under powered and were not interpreted in any specific way (no result). Heritability and genetic correlation estimates were generated based on expression scored from both left and right antimeres. Because SOLAR accommodates only continuous and binary variables, the data were analyzed in two ways. First, the ordinal data were treated as continuous, appealing to the widely held assumption that nonmetric traits of the dentition and skull are characterized by underlying continuous, normal distributions (Carson,; Cheverud, Buikstra, & Twichell,; Corruccini,; Harris,; Konigsberg, Kohn, & Cheverud,; Sofaer, Niswander, MacLean, & Workman,). This also follows the methodology employed by Hlusko and Mahaney () in their analysis of cingular remnant variation in savannah baboons. For these analyses, data were transformed using SOLAR's inorm function to avoid potential error stemming from the violation of distributional assumptions, especially kurtosis. Next, each ordinal variable was decomposed to generate a series of binary variables, using expression grades as dichotomization breakpoints. For example, canine distal accessory ridge is an ordinal trait with six possible expression grades (0 to 5) (Turner et al.,). We dichotomized this variable at each grade, thus using data collected for a single trait to generate five new binary variables (i.e., 1+=present; 2+=present; 3+=present; 4+=present; 5 = present). Heritability estimates were then generated for each of these new binary variables. Comparatively few estimates of genetic correlation have been published for dental data. Hlusko and colleagues (2006, 2007, 2011, 2016) presented genetic correlations for a large dataset of baboon and mouse dental metrics and documented low to moderate levels of integration among teeth. Stojanowski et al. () presented genetic correlations for human odontometric data and documented strong integration of mesio-distal crown size across tooth types and classes. In that paper, almost all genetic correlations were positive and significantly different from 0.0, including comparisons across antimeres, isomeres, and metameres. Here, we report the first estimates of genetic correlation for morphological variables that complement decades of work comparing antimeres, isomeres, and metameres using phenotypic correlations (reviewed in Scott et al., a), which do not disassociate environmental and genetic effects (Falconer & MacKay,). The results presented here, based in quantitative genetic theory, more directly assess the assumptions and data cleaning practices commonly used in biodistance analyses (see Scott et al., a). The results for antimeres (Table) were unsurprising and similar to those presented for odontometric data. Estimates were large and positive and, for most comparisons, significantly different from 0.0. This affirms the practice of collapsing left and right side data and indicates strong genetic control of morphological features across the midline. Genetic correlations for metameres also affirm the practice of using a key tooth for each trait to avoid violating the assumption of trait independence (Table). Most cross-tooth correlations for shoveling, double shoveling, distal accessory ridge, and tuberculum dentale were positive, and many were significantly different from 0.0. Finally, analyses of multiple traits scored on the same tooth indicate that most of these characters are genetically independent, as is often assumed (Table). Winging, shoveling, and labial curvature were found to be genetically uncorrelated on the I1 and I2. Canine mesial ridge, shoveling, and distal accessory ridge were found to be genetically uncorrelated on the C1. These traits satisfy the assumption of independence. There were two exceptions, however. First, shoveling and tuberculum dentale on the incisors (but not the canine) were genetically integrated, and significantly so for the I1, which corroborates results presented by Mizoguchi (), Berry (), and Suzuki and Sakai ()—see also Crummett (). Second, shoveling and double shoveling returned consistently negative genetic correlations, positive environmental correlations, and near zero phenotypic correlations. Other negative genetic correlations involving double shoveling were also observed, but these may reflect spurious output due to observational protocol. For example, I1 labial curvature and I2 peg form returned negative genetic, environmental, and phenotypic correlations with double shoveling (Table), which suggests a different interpretation than the results for shoveling. P values for some of these comparisons were not statistically significant, however, these results seem to corroborate multiple studies highlighting the association between variants of the EDAR gene and the expression of shoveling and double shoveling (Kimura et al.,; Park et al.,; Peng et al.,; Tan et al.,). Kimura et al. () also found positive associations between I1 shoveling, I2 tuberculum dentale, and C1 tuberculum dentale and a single nucleotide polymorphism (SNP) in the WNT10A gene. Interestingly, they also noted a negative association between the same SNP and I1 double shoveling, which confirms that some, but not all, genes have the same phenotypic effect for these crown features (incomplete pleiotropy). Here, the low to moderate correlations and low sample sizes result in under-powered statistical tests. Nonetheless, these results are intriguing and by pattern alone suggest a real phenomenon is being captured by the quantitative genetic models that should be evaluated in other populations. Double shoveling is a complex trait with much less research focused on its distribution and inheritance (Dahlberg & Mikkelsen,; DeVoto et al.,; Ohba,; Shirono & Sasaki,; Snyder,; Takagi et al.,). Like shoveling, double shoveling has a dentin component and is thought to serve a functional role similar to that of shoveling—strengthening the tooth to mitigate increased bite forces. Scott et al. () indicate a phenotypic correlation between shoveling and double shoveling of 0.30 and confirm the geographic patterning of the traits is quite similar but not identical (Shirono & Sasaki,). One notable difference is the degree to which double shoveling frequencies vary within populations for which shoveling is nearly fixed (Scott et al., a: Figure 5.4). In the Gullah sample, the phenotypic correlations neared zero, while the genetic correlations were moderately negative in value, which, according to Falconer and MacKay (: 315), suggests that environmental and genetic sources of variation, “affect the characters through different physiological mechanisms.” At this point it is important to stress our assumption that the genetic correlations reflect the effects of pleiotropy and not linkage disequilibrium. It is also important to note that these estimates reflect a balance/average of loci with both positive and negative pleiotropic effects (Cheverud,; MacKay, Stone, & Ayroles,). Negative genetic correlations (antagonistic or negative pleiotropy) are often discussed in the context of aging, life history and fitness trade-offs (Falconer & MacKay,; MacKay et al.,; Roff & Fairbairn,), that is, when the same gene or genes produce opposite phenotypic effects across traits, while maximizing fitness in a stabilized system. Negative pleiotropy can also result from the division of a developmental precursor (Riska,), functional relationships between phenotypes, or the limiting effects of some resource important for morphological development (Norry, Vilardi, & Hasson,). Dental morphological features are often considered selectively neutral; however, this may not be the case with shoveling and double shoveling because of their association with genes related to the morphogenesis of ectodermally derived structures. Variation in the frequency of these traits may, therefore, reflect correlated responses to selection of other phenotypes. As such, it seems unlikely that shoveling and double shoveling are the direct targets of fitness trade-offs. However, these data may reflect a trade-off in the sense of evolvability (Klingenberg,). That is, evolution has shaped the human (and broadly mammalian) anterior dentition in such a way that the lingual surfaces are less constrained in their ability to develop evolutionary novelties. The mesial and distal marginal ridges, accessory marginal ridges, lingual fossa, and lingual cingulum display tremendous variability across human populations (and hominin taxa), which contrasts sharply with the labial surfaces of the same teeth. In fact, with the exception of truly anomalous features, double shoveling is the only labial variant consistently scored for the anterior dentition (labial curvature, in actuality, characterizes overall crown shape and not elaboration of the labial surface). These results may reflect the quantitative genetic signature of a highly structured, but rarely discussed, aspect of organismal development. Nevertheless, these observations pertain to one specific population, a population that also has low degrees of expression of both traits. Because genetic correlations are not stable across different environments (Sgró & Hoffman,), and because quantitative genetic analyses are population-specific, further work is needed to confirm the robusticity of these correlations. In comparison to previously published heritabilities for Gullah odontometrics (Stojanowski et al.,), morphological traits produced overall lower estimates. In this article, average heritability across all significant continuous scale traits was 0.342 (left side) and 0.334 (right side), while average heritability for odontometric variables was 0.557 and 0.482 for the left and right sides, respectively (see Stojanowski et al.,). The difference between odontometric and morphological average heritability across all traits is statistically significant (left side: p = 0.001; right side: p = 0.005). The effect of different measurement scales (continuous vs. ordinal/binary) likely has some impact on these comparisons. Consistent with the previously published odontometric results (Stojanowski et al.,), heritability estimates for Gullah dental morphology are lower than most previously reported estimates for morphological variables (but see Ohno, for low concordance rates). For example, in the Gullah sample I1 shoveling heritability was 0.50 as compared with estimates of 0.68, 0.66, 0.75 (average of 0.82, 0.86, 0.84, 0.49 across relative pairings), 0.52, and 0.68 in other populations (Blanco & Chakraborty,; Hanihara et al.,; Mizoguchi, b; Scott et al., a). Mizoguchi (b) estimated heritabilities for incisor lingual fossa depth at 0.22 and 0.75 for the I1 and I2, respectively, and re-calculated heritabilities from Aoyagi (), Hanihara et al. (), and Hanihara et al. () at 0.44, 0.27, and 0.80, respectively. Therefore, while this trait-specific comparison seems to reflect diminished heritability for incisor shoveling in this population, the estimate is not the lowest reported. We note that the average heritability across all traits (∼0.34) is remarkably similar to the average of Nichol's () transmissibility estimates for the full suite of morphological features (∼0.36) or just the anterior tooth features (∼0.34). We have previously outlined reasons why heritability estimates are expected to be lower in the Gullah sample, which include differences in methods and study designs and a complex pedigree structure that includes more distant kin relationships. Unlike previous studies that focus on sibling or parent–offspring correlations, the approach used here should minimize the effects of shared environment and dominance confounders. For this reason we feel the Gullah data are more reflective of the realities of archaeological datasets (see Stojanowski et al.,). However, there is also reason to believe the lower estimates of heritability reflect the specific social-demographic history of the study sample. Heritability will be reduced when additive genetic variance is low or when environmental variance is high. The former is constrained by patterns of gene flow and assortative mating, and there is ethnographic data to suggest the Gullah were somewhat endogamous (Pollitzer,), an interpretation bolstered by genetic studies documenting very low rates of European or Native American admixture in this population (Benn Torres et al.,; McLean et al.,; Parra et al.,; Pollitzer,). In addition, at the time the casts were collected, the James Island Gullah experienced a number of challenges that are expected to increase environmental variance and developmental stress, including low median wages, low life expectancy, high disease load, and limited economic opportunities (Newby, 1973; Pollitzer,). These expectations were born out in Guatelli-Steinberg and colleagues’ (2006) study of the same casts, in which fluctuating and directional odontometric asymmetry was higher than expected, even when compared with Late Prehistoric and Late Archaic period populations. The effects of developmental stress on dental tissues are well-known (Corruccini, Townsend, & Schwerdt,; Fearne & Brook,; Garn, Osborne, & McCabe,; Harila-Kaera, Keikkinen, Alvesalo, & Osborne,; McKee & Lunz,; Riga, Belcastro, & Moggi-Cecchi,; Zameer et al.,). As such, the low to moderate estimates of heritability reported here should not be over-interpreted, and in fact, the pattern of values is more relevant than their absolute values. For example, results presented in Table support standard practices in dental morphological biodistance analysis, especially with respect to the use of key (sensu Scott et al., a) teeth. For incisor shoveling, the heritability ranks were: I1>I2>C1, with the C1 returning estimates of 0.0. This result supports use of the I1 as the key tooth for this trait and affirms the practice of ignoring canine shoveling altogether. Maxillary incisor shoveling heritability estimates were uniformly higher than mandibular estimates. Interestingly, in the mandible the within-class patterning was reversed (I2>I1), also supporting the fact that the mandibular I2 is considered the more stable, polar tooth in that arcade (but see Scott,). For double shoveling, the same pattern was evident: I1>I2>C1, which affirms the use of the I1 as the key tooth for this trait. Double shoveling of the maxillary premolars, however, returned the highest heritability estimates among all tooth classes. This is the one trait that is scored on both anterior teeth and premolars (although rarely, if ever, discussed for premolars), and the unusual pattern of heritabilities might suggest an etiological disconnect for premolar double shoveling. Nichol () noted a similar pattern in his analyses of dental morphology using complex segregation analysis; however, it should also be stressed that the inter-tooth genetic correlations for double shoveling are mainly positive, with pleiotropic effects indicated for a number of trait pairs (Table). Tuberculum dentale produced the following ranks: C1>I1>I2. Of note, the I1 estimate was higher than the I2 estimate, despite the latter typically being used as the key tooth for this trait (Scott et al., a). These data suggest the canine may be the better alternative. Finally, the canine distal accessory ridge was compared between the maxilla and mandible. As with incisor shoveling, the maxillary estimates were much higher than the mandibular; in fact, the mandibular canine results were unexpectedly low for all breakpoints, even when controlling for sex as a covariate. This result is difficult to explain. Both traits are scored similarly in each arcade with well-developed observation plaques (Scott et al., a; Turner et al.,). Ranges of score variation are also similar in the Gullah sample with binary frequencies around 70–80% for both traits, thus complicating any explanation of why the arcades differ so markedly. Nonetheless, there was a general trend toward higher heritabilities in the maxillary dentition. This was true when all nonzero heritabilities were compared, and on a trait-by-trait basis for characters scored in both arcades (shoveling, distal accessory ridge). We previously reported a similar pattern for odontometric data in the Gullah sample (Stojanowski et al.,) and had interpreted the results as reflecting greater developmental constraints on maxillary tooth size. Results presented here suggest the morphological variability of specific crown components may be under the same developmental constraints. Comparatively few estimates of genetic correlation have been published for dental data. Hlusko and colleagues (2006, 2007, 2011, 2016) presented genetic correlations for a large dataset of baboon and mouse dental metrics and documented low to moderate levels of integration among teeth. Stojanowski et al. () presented genetic correlations for human odontometric data and documented strong integration of mesio-distal crown size across tooth types and classes. In that paper, almost all genetic correlations were positive and significantly different from 0.0, including comparisons across antimeres, isomeres, and metameres. Here, we report the first estimates of genetic correlation for morphological variables that complement decades of work comparing antimeres, isomeres, and metameres using phenotypic correlations (reviewed in Scott et al., a), which do not disassociate environmental and genetic effects (Falconer & MacKay,). The results presented here, based in quantitative genetic theory, more directly assess the assumptions and data cleaning practices commonly used in biodistance analyses (see Scott et al., a). The results for antimeres (Table) were unsurprising and similar to those presented for odontometric data. Estimates were large and positive and, for most comparisons, significantly different from 0.0. This affirms the practice of collapsing left and right side data and indicates strong genetic control of morphological features across the midline. Genetic correlations for metameres also affirm the practice of using a key tooth for each trait to avoid violating the assumption of trait independence (Table). Most cross-tooth correlations for shoveling, double shoveling, distal accessory ridge, and tuberculum dentale were positive, and many were significantly different from 0.0. Finally, analyses of multiple traits scored on the same tooth indicate that most of these characters are genetically independent, as is often assumed (Table). Winging, shoveling, and labial curvature were found to be genetically uncorrelated on the I1 and I2. Canine mesial ridge, shoveling, and distal accessory ridge were found to be genetically uncorrelated on the C1. These traits satisfy the assumption of independence. There were two exceptions, however. First, shoveling and tuberculum dentale on the incisors (but not the canine) were genetically integrated, and significantly so for the I1, which corroborates results presented by Mizoguchi (), Berry (), and Suzuki and Sakai ()—see also Crummett (). Second, shoveling and double shoveling returned consistently negative genetic correlations, positive environmental correlations, and near zero phenotypic correlations. Other negative genetic correlations involving double shoveling were also observed, but these may reflect spurious output due to observational protocol. For example, I1 labial curvature and I2 peg form returned negative genetic, environmental, and phenotypic correlations with double shoveling (Table), which suggests a different interpretation than the results for shoveling. P values for some of these comparisons were not statistically significant, however, these results seem to corroborate multiple studies highlighting the association between variants of the EDAR gene and the expression of shoveling and double shoveling (Kimura et al.,; Park et al.,; Peng et al.,; Tan et al.,). Kimura et al. () also found positive associations between I1 shoveling, I2 tuberculum dentale, and C1 tuberculum dentale and a single nucleotide polymorphism (SNP) in the WNT10A gene. Interestingly, they also noted a negative association between the same SNP and I1 double shoveling, which confirms that some, but not all, genes have the same phenotypic effect for these crown features (incomplete pleiotropy). Here, the low to moderate correlations and low sample sizes result in under-powered statistical tests. Nonetheless, these results are intriguing and by pattern alone suggest a real phenomenon is being captured by the quantitative genetic models that should be evaluated in other populations. Double shoveling is a complex trait with much less research focused on its distribution and inheritance (Dahlberg & Mikkelsen,; DeVoto et al.,; Ohba,; Shirono & Sasaki,; Snyder,; Takagi et al.,). Like shoveling, double shoveling has a dentin component and is thought to serve a functional role similar to that of shoveling—strengthening the tooth to mitigate increased bite forces. Scott et al. () indicate a phenotypic correlation between shoveling and double shoveling of 0.30 and confirm the geographic patterning of the traits is quite similar but not identical (Shirono & Sasaki,). One notable difference is the degree to which double shoveling frequencies vary within populations for which shoveling is nearly fixed (Scott et al., a: Figure 5.4). In the Gullah sample, the phenotypic correlations neared zero, while the genetic correlations were moderately negative in value, which, according to Falconer and MacKay (: 315), suggests that environmental and genetic sources of variation, “affect the characters through different physiological mechanisms.” At this point it is important to stress our assumption that the genetic correlations reflect the effects of pleiotropy and not linkage disequilibrium. It is also important to note that these estimates reflect a balance/average of loci with both positive and negative pleiotropic effects (Cheverud,; MacKay, Stone, & Ayroles,). Negative genetic correlations (antagonistic or negative pleiotropy) are often discussed in the context of aging, life history and fitness trade-offs (Falconer & MacKay,; MacKay et al.,; Roff & Fairbairn,), that is, when the same gene or genes produce opposite phenotypic effects across traits, while maximizing fitness in a stabilized system. Negative pleiotropy can also result from the division of a developmental precursor (Riska,), functional relationships between phenotypes, or the limiting effects of some resource important for morphological development (Norry, Vilardi, & Hasson,). Dental morphological features are often considered selectively neutral; however, this may not be the case with shoveling and double shoveling because of their association with genes related to the morphogenesis of ectodermally derived structures. Variation in the frequency of these traits may, therefore, reflect correlated responses to selection of other phenotypes. As such, it seems unlikely that shoveling and double shoveling are the direct targets of fitness trade-offs. However, these data may reflect a trade-off in the sense of evolvability (Klingenberg,). That is, evolution has shaped the human (and broadly mammalian) anterior dentition in such a way that the lingual surfaces are less constrained in their ability to develop evolutionary novelties. The mesial and distal marginal ridges, accessory marginal ridges, lingual fossa, and lingual cingulum display tremendous variability across human populations (and hominin taxa), which contrasts sharply with the labial surfaces of the same teeth. In fact, with the exception of truly anomalous features, double shoveling is the only labial variant consistently scored for the anterior dentition (labial curvature, in actuality, characterizes overall crown shape and not elaboration of the labial surface). These results may reflect the quantitative genetic signature of a highly structured, but rarely discussed, aspect of organismal development. Nevertheless, these observations pertain to one specific population, a population that also has low degrees of expression of both traits. Because genetic correlations are not stable across different environments (Sgró & Hoffman,), and because quantitative genetic analyses are population-specific, further work is needed to confirm the robusticity of these correlations. The casts used in this study were collected by Rene Menegaz-Bock as part of her dissertation work on variation in the anterior dentition. These materials are currently part of the collections of the Department of Anthropology at The Ohio State University (Edgar & Nardin,), and this research was conducted with the approval of their IRB (2012B0529). The Gullah sample includes dental casts from an African American population that was living on James Island, South Carolina during the early to mid-twentieth century (Menegaz-Bock,; Pollitzer,; Pollitzer, Menegaz-Bock, Ceppellini, & Dunn,) and includes 469 casts (of ∼3,400 residents) representing 76 families. While many of the reconstructed genealogies were small, comprising only two generations and a pair of casted individuals, the largest genealogy spans seven generations and includes more than 100 casted individuals. Genealogical reconstruction also resulted in the identification of 22 bilineal families, the largest of which incorporates 313 casted individuals (see also Stojanowski et al.,). The complex nature of the reconstructed genealogies is consistent with ethnographic data that suggests Gullah society engaged in broadly inclusive relationships, emphasizing fairly stable marriages, shared familial ties, and household and/or patrilineal exogamy (Pollitzer,; Twining & Baird,). The casts have previously contributed to analyses of fluctuating dental asymmetry (Guatelli-Steinberg, Sciulli, & Edgar,) and studies of dental morphology (Edgar & Sciulli,) and population history (Edgar,). The Gullah trace their origins to west and central Africa (Edgar,; Parra et al., 2001; Pollitzer,; Rogers,) and are one of the least acculturated African American populations in the United States (Herskovits,; Pollitzer,; Twining & Baird,) with low rates of European and Native American admixture (Benn Torres, Doura, Keita, & Kittles,; McLean, Argyropoulos, Page, Shriver, & Garvey, 2005; McLean et al.,; Parra et al.,; Pollitzer,). Funding Information: NSF Research Grant, Grant number: BCS- 1063942; BCS, Grant number: 1750089; Arizona State University College of Liberal Arts and Sciences, Grant number: CL2014.15; IRB at The Ohio State University, Grant number: 2012B0529 Publisher Copyright: © 2018 Wiley Periodicals, Inc.
PY - 2018/9
Y1 - 2018/9
N2 - Objectives: This article presents estimates of narrow-sense heritability and bivariate genetic correlation for a series of morphological crown variants of the anterior dentition. These results provide insight into the value of dental phenotypes as evolutionary proxies, as well as the development of tooth crowns as integrated or modular structures. Materials and Methods: African American dental casts from the Menegaz-Bock collection were scored for a standard set of dental morphological variables using the Arizona State Dental Anthropology System. Estimates of narrow-sense heritability and genetic correlations were generated using SOLAR v. 8.1.1, controlling for the covariates of age, sex, and birth year. Analyses were run using ordinal/continuous scale variables that were then dichotomized at various breakpoints, consistent with standard practices in dental anthropology. Results: Heritability estimates were low to moderate for most traits, and lower in magnitude than those reported for odontometric data from the same study sample. Only winging, canine shoveling, and canine double shoveling returned narrow-sense heritabilities that did not differ significantly from zero. Genetic correlations were high among antimeres and metameres and low for different traits scored on the same tooth crown. These results affirm standard data cleaning practices in dental biodistance. Double shoveling was atypical in returning strong negative correlations with other traits, shoveling in particular. Conclusions: Additive genetic variation contributes to dental morphological variation, although the estimates are uniformly lower than those observed for odontometrics. Patterns of genetic correlation affirm most standard practices in dental biodistance. Patterns of negative pleiotropy involving lingual and labial crown features suggest a genetic architecture and developmental complex that differentially constrain morphological variation of distinct surfaces of the same tooth crown. These patterns warrant greater consideration and cross-population validation.
AB - Objectives: This article presents estimates of narrow-sense heritability and bivariate genetic correlation for a series of morphological crown variants of the anterior dentition. These results provide insight into the value of dental phenotypes as evolutionary proxies, as well as the development of tooth crowns as integrated or modular structures. Materials and Methods: African American dental casts from the Menegaz-Bock collection were scored for a standard set of dental morphological variables using the Arizona State Dental Anthropology System. Estimates of narrow-sense heritability and genetic correlations were generated using SOLAR v. 8.1.1, controlling for the covariates of age, sex, and birth year. Analyses were run using ordinal/continuous scale variables that were then dichotomized at various breakpoints, consistent with standard practices in dental anthropology. Results: Heritability estimates were low to moderate for most traits, and lower in magnitude than those reported for odontometric data from the same study sample. Only winging, canine shoveling, and canine double shoveling returned narrow-sense heritabilities that did not differ significantly from zero. Genetic correlations were high among antimeres and metameres and low for different traits scored on the same tooth crown. These results affirm standard data cleaning practices in dental biodistance. Double shoveling was atypical in returning strong negative correlations with other traits, shoveling in particular. Conclusions: Additive genetic variation contributes to dental morphological variation, although the estimates are uniformly lower than those observed for odontometrics. Patterns of genetic correlation affirm most standard practices in dental biodistance. Patterns of negative pleiotropy involving lingual and labial crown features suggest a genetic architecture and developmental complex that differentially constrain morphological variation of distinct surfaces of the same tooth crown. These patterns warrant greater consideration and cross-population validation.
KW - biodistance
KW - dental morphology
KW - heritability
KW - pleiotropy
KW - quantitative genetics
UR - http://www.scopus.com/inward/record.url?scp=85051106499&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85051106499&partnerID=8YFLogxK
U2 - 10.1002/ajpa.23612
DO - 10.1002/ajpa.23612
M3 - Article
C2 - 30055011
AN - SCOPUS:85051106499
SN - 0002-9483
VL - 167
SP - 124
EP - 143
JO - American journal of physical anthropology
JF - American journal of physical anthropology
IS - 1
ER -