genetyka, geografia i zmiany na kościach.pdf

(217 KB) Pobierz
Genetic, geographic, and environmental correlates of human temporal bone variation
AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 134:312–322 (2007)
Genetic, Geographic, and Environmental Correlates
of Human Temporal Bone Variation
Heather F. Smith, 1 * y Claire E. Terhune, 1 * y and Charles A. Lockwood 2
1 School of Human Evolution and Social Change, Arizona State University, Tempe, AZ 85287-2402
2 Department of Anthropology, University College London, London WC1E 6BT, UK
KEY WORDS geometric morphometrics; molecular distance; cranial morphology
ABSTRACT Temporal bone shape has been shown to
reflect molecular phylogenetic relationships among homi-
noids and offers significant morphological detail for distin-
guishing taxa. Although it is generally accepted that tem-
poral bone shape, like other aspects of morphology, has an
underlying genetic component, the relative influence of
genetic and environmental factors is unclear. To determine
the impact of genetic differentiation and environmental
variation on temporal bone morphology, we used three-
dimensional geometric morphometric techniques to evalu-
ate temporal bone variation in 11 modern human popula-
tions. Population differences were investigated by discrim-
inant function analysis, and the strength of the relation-
ships between morphology, neutral molecular distance,
geographic distribution, and environmental variables were
assessed by matrix correlation comparisons. Significant
differences were found in temporal bone shape among all
populations, and classification rates using cross-validation
were relatively high. Comparisons of morphological dis-
tances to molecular distances based on short tandem
repeats (STRs) revealed a significant correlation between
temporal bone shape and neutral molecular distance
among Old World populations, but not when Native Amer-
icans were included. Further analyses suggested a similar
pattern for morphological variation and geographic distri-
bution. No significant correlations were found between
temporal bone shape and environmental variables: tem-
perature, annual rainfall, latitude, or altitude. Significant
correlations were found between temporal bone size and
both temperature and latitude, presumably reflecting
Bergmann’s rule. Thus, temporal bone morphology ap-
pears to partially follow an isolation by distance model of
evolution among human populations, although levels of
correlation show that a substantial component of variation
is unexplained by factors considered here. Am J Phys
Anthropol 134:312–322, 2007.
C 2007 Wiley-Liss, Inc.
Like other aspects of phenotype, cranial morphology
reflects a combination of genetic and environmental
influences (Moss, 1962, 1972). Within this framework,
some authors have suggested that particular portions of
the cranium may be less prone to variation due to envi-
ronmental variables, and therefore more phylogenetically
informative (Olson, 1981; Strait et al., 1997; Lieberman
et al., 2000a; Harvati, 2001; Wood and Lieberman, 2001;
Harvati and Weaver, 2006a,b). For hominins, traits asso-
ciated with heavy chewing have been argued to be homo-
plastic and consequently unreliable indicators of phylog-
eny (Walker et al., 1986; Wood, 1988; Skelton and
McHenry, 1992; Turner and Wood, 1993; McHenry, 1994,
1996; Lieberman et al., 1996; but see Strait et al., 1997;
Asfaw et al., 1999; Collard and Wood, 2001). The mor-
phology of the cranial base has especially been regarded
as a reliable reflection of genetic relationships, as it
forms very early during ontogeny and ossifies endochon-
drally (Moore and Lavelle, 1974; Olson, 1981; MacPhee
and Cartmill, 1986; Lieberman et al., 2000a,b). The cra-
nial base also mirrors the shape of the developing brain,
which is relatively constrained (Houghton, 1996). Basi-
cranial characters may therefore be less influenced by
epigenetic forces than are the externally sensitive intra-
membraneous bones of the facial skeleton.
The morphology of the temporal bone, as part of the
cranial base, may also reflect neutral molecular distan-
ces within species and phylogenetic relationships among
species. However, the temporal bone also serves a vari-
ety of functional roles, such as posture, hearing, balance,
mastication, and formation of the braincase. Conse-
quently, this element can serve as a test case of the
ways in which cranial morphology covaries with molecu-
lar distances and environmental factors and a test of the
hypothesis that cranial base elements have a strong
genetic component.
Several recent studies of variation in the temporal
bone have demonstrated this region’s utility in distin-
guishing among species and subspecies of extant great
apes, and for quantifying levels of variation within and
between taxa (Harvati, 2001, 2003; Lockwood et al.,
2002, 2004, 2005; Terhune et al., 2007). In particular,
Lockwood et al. (2004) demonstrated that, using shape
distributions of coordinate data from modern humans,
orangutans, gorillas, chimpanzees, and bonobos, the re-
sultant phylogenetic tree of these taxa was identical to
the molecular phylogeny of these species. Similarly, sev-
Grant sponsors: AMNH Collections Study Grant and ASU Depart-
ment of Anthropology; Grant number: NSF BCS-9982022.
*Correspondence to: Heather F. Smith or Claire E. Terhune,
School of Human Evolution and Social Change, Arizona State Uni-
versity, Box 872402, Tempe, AZ 85287-2402, USA.
E-mail: heather.f.smith@asu.edu or claire.terhune@asu.edu
y These authors contributed equally to this work.
Received 12 December 2006; accepted 8 May 2007
DOI 10.1002/ajpa.20671
Published online 13 July 2007 in Wiley InterScience
(www.interscience.wiley.com).
C
2007 WILEY-LISS, INC.
48985284.013.png 48985284.014.png 48985284.015.png 48985284.016.png 48985284.001.png 48985284.002.png 48985284.003.png 48985284.004.png 48985284.005.png 48985284.006.png 48985284.007.png
TEMPORAL BONE VARIATION IN MODERN HUMANS
313
Fig. 1. Map of the world
showing the approximate lo-
cations of populations used
in the morphological analy-
sis (triangles), populations
used in the molecular analy-
sis (circles), and waypoints
(squares). Lines link the mor-
phological populations and
their genetic representatives.
eral recent studies (Harvati, 2001, 2003; Terhune et al.,
2007) have used the morphology of the temporal bone to
test hypothesized taxonomic divisions among fossil taxa.
Given this background, we sought to investigate the
association between temporal bone morphology and mo-
lecular distance among human populations, together
with geographic distance and external factors such as
environmental variables. Some recent studies have ex-
plicitly evaluated these influences on cranial anatomy
(Relethford, 1994, 1998, 2001, 2002; Gonzales-Jose et al.,
2004; Roseman, 2004). Linear dimensions of the skull
have been shown to reflect genetic relationships of
human populations, such that closely related populations
tend to be more similar in overall cranial form (Rele-
thford, 2001, 2002; Gonzales-Jose et al., 2004; Roseman,
2004). However, selective pressures acting on the skull
of certain human populations have also been identified
and can have a significant impact on cranial morphology
of populations living in regions with extreme tempera-
tures, such as Siberia (Roseman, 2004). Diversifying re-
gional selection due to climate also affects the cranial
morphology of several other human populations (Carey
and Steegmann, 1981; Franciscus and Long, 1991; Rose-
man, 2004).
Harvati and Weaver (2006a,b) analyzed the correlation
between human morphological variation in three cranial
regions – the temporal bone, cranial vault, and facial
skeleton – with molecular distances and environmental
variables. They concluded that the morphology of the
temporal bone and cranial vault are correlated with mo-
lecular distance in human populations, while facial mor-
phology covaries more reliably with environment. The
correlation between temporal bone shape and molecular
distance was significant but low, suggesting that other
factors also play a significant role in patterns of tempo-
ral bone morphology in humans. In addition, temporal
bone centroid size was found to be correlated with tem-
perature, a finding that is consistent with environmental
variation in body size as first outlined by Bergmann (1847).
Our goal is to use an independent dataset and an
expanded set of landmarks on the temporal bone to
replicate part of the study of Harvati and Weaver
(2006a). We also include additional environmental varia-
bles such as rainfall and altitude, and explore the rela-
tionship between morphology and geographic distance.
In general, we are testing the hypothesis that the tem-
poral bone follows an isolation by distance model of evo-
lution in human populations (Wright, 1943). More specif-
ically, three research questions were investigated:
Q1. Are modern human populations significantly differ-
ent in temporal bone morphology?
Q2. What is the strength of the correlation between tem-
poral bone morphology and molecular distance
among populations of modern humans?
Q3. How do external variables such as environmental
differences or geographic distance covary with pat-
terns of temporal bone morphology in humans?
MATERIALS AND METHODS
Data collection
A total of 243 individuals from 11 modern human pop-
ulations were included in this study (Fig. 1, Table 1).
Specimens were housed at the American Museum of
Natural History and Arizona State University. Individu-
als with extensive antemortem tooth loss were generally
avoided to minimize the possibility of alveolar resorption
affecting the morphology of the temporomandibular joint
(TMJ). Following Lockwood et al. (2002), 22 landmarks
from the ectocranial surface of the temporal bone were
employed, which describe the morphology of the mandib-
ular fossa, tympanic, mastoid, and petrous portions of
the temporal bone (Fig. 2, Table 2). In comparison, Har-
vati and Weaver (2006a) used 13 landmarks.
An Immersion Microscribe point digitizer was used to
record the three-dimensional coordinates of each land-
mark. These three-dimensional data were then analyzed
using Morphologika (O’Higgins and Jones, 1998). First,
three-dimensional coordinate data were registered
through a generalized Procrustes analysis (GPA) (Gower,
1975; Goodall, 1991; Dryden and Mardia, 1998). Subse-
quently, variation in shape was investigated through
principal components analysis (PCA). Output from these
analyses (Procrustes residuals from the GPA and PC
scores from the PCA) was recorded and copied into other
statistical programs for further analysis. All three-
dimensional data were collected by the second author,
and intraobserver error for a subset of the data set used
here is reported by Terhune et al. (2007).
Data on 783 STRs in matched analogues of nine of the
human populations discussed earlier were used to obtain
American Journal of Physical Anthropology—DOI 10.1002/ajpa
48985284.008.png
314
H.F. SMITH ET AL.
TABLE 1. Modern human populations used in the morphometric analysis
Population a
Total Genetic representative Centroid size Average geographic coordinates
Alaskan Natives
20 None
106.43
68.4N, 166.7W
Australian Aborigines
21 Australians
94.69
34.8S, 138.5E
Hungarians (Medieval)
21 French
98.69
46.6N, 18.4E
Khoisan
19 San
98.21
20.5S, 19.5E
Malaysians
21 Cambodians
100.23
4N, 109.5E
Mongolians
18 Mongolians
103.43
46.9N, 103.8E
Native American (Grand Gulch, Utah)
20 Pima
102.91
37.6N, 109.8W
New Guineans
20 Papua New Guineans
97.52
6.4S, 150.2E
Nubians (Semna South, Sudanese Nubia)
43 Mozabite
98.63
20.0N, 30.1E
Pare (Tanzania)
19 Kenyan Bantu
98.35
4.3S, 38.1E
Southern Indians
21 None
94.64
13N, 77.56E
Total 243
a Specimens were housed at Arizona State University (Nubians) or the American Museum of Natural History (all others).
neutral molecular distances. STRs have been shown to
be particularly useful and appropriate for determining
genetic relationships of populations of Homo sapiens.
These loci are autosomal and evolve neutrally such that
shared mutations are accepted as evidence of common
ancestry. The dataset used here was originally used by
Ramachandran et al. (2005) and Rosenberg et al. (2005)
and consists of the largest and most inclusive STR data-
set published to date. Several of the populations meas-
ured in the craniometric study have not been typed for
STRs, particularly the archaeological samples (the
Nubians and Medieval Hungarians). In these cases, it
was necessary to substitute a representative population
from the same geographic region and/or linguistic group
(Table 1). This practice has been employed in previous
studies of the relationship between morphological and
molecular distances in modern humans (Relethford,
1994; Roseman, 2004; Harvati and Weaver, 2006a,b).
The Alaskan natives and southern India sample had to
be omitted from the molecular analysis as neither they nor
any other comparable population has been typed for a
sufficient number of STR loci. However, these populations
were still included in all other analyses in this study.
Approximate geographic coordinates of population ori-
gins were estimated using an atlas and published infor-
mation for the samples. In the case that a range of coor-
dinates was obtained, an average location was used.
Data were also compiled on environmental variables in
regions from which the populations originated, using
data from nearby weather stations (New et al., 1999,
2000) and almanacs. These included rainfall, tempera-
ture, altitude, and latitude. The link between these envi-
ronmental variables and temporal bone morphology
could stem directly from local adaptations of cranial
shape or indirectly from behaviors mediated by the envi-
ronment, such as diet or activity levels.
Analytical methods
The first research question examined the degree to
which the morphology of the temporal bone can discrimi-
nate among populations of Homo sapiens, and was eval-
uated in two ways. First, Procrustes distances between
groups were calculated, and the significance of these
values was assessed via a permutation test (Good, 1993).
This form of significance testing compares the observed
distance (i.e., test statistic) with a distribution of per-
muted distances, where individuals are randomly allo-
cated to each group and a mean distance is calculated.
A test statistic is considered statistically significant
Fig. 2. Twenty-two temporal bone landmarks digitized in
the present study (following Lockwood et al., 2002). Refer to
Table 2 for landmark definitions. Open circles show the relative
positions of landmarks 1 and 18 when these landmarks are not
directly visible.
American Journal of Physical Anthropology—DOI 10.1002/ajpa
48985284.009.png 48985284.010.png
TEMPORAL BONE VARIATION IN MODERN HUMANS
315
TABLE 2. Definitions of landmarks used in the present study
No. Definition
1 Intersection of the infratemporal crest and sphenosquamosal suture
2 Most lateral point on the margin of foramen ovale
3 Most anterior point on the articular surface of the articular eminence
4 Most inferior point on entoglenoid process
5 Most inferior point on the medial margin of the articular surface of the articular eminence
6 Midpoint of the lateral margin of the articular surface of the articular eminence
7 Center of the articular eminence
8 Deepest point within the mandibular fossa
9 Most inferior point on the postglenoid process
10 Anteromedial apex of the petrous part of the temporal bone
11 Most posterolateral point on the margin of the carotid canal entrance
12 Most lateral point on the vagina of the styloid process (whether process is present or absent)
13 Most lateral point on the margin of the stylomastoid foramen
14 Most lateral point on the jugular fossa
15 Center of the inferior tip of the mastoid process
16 Most inferior point on the external acoustic porus
17 Most inferolateral point on the tympanic element of the temporal bone
18 Point of inflection where the braincase curves laterally into the supraglenoid gutter, in coronal plane of the mandibular fossa
19 Point on lateral margin of the zygomatic process of the temporal bone in the coronal plane of the postglenoid process
20 Auriculare
21 Porion
22 Asterion
After Lockwood et al. (2002).
95% of variation). The differentiation
among populations was then assessed using discriminant
analyses with jackknife cross-validation, where prior
probabilities were set equal to group size. Since the Nu-
bian sample was significantly larger than all other sam-
ples used here (n
the values by a pooled within-group covariance matrix,
which assumes that all groups in the analysis have
similar covariance structures (Ackermann, 2002, 2005;
Klingenberg and Monteiro, 2005). This assumption is
tenuous given the sample sizes used here. In contrast,
since Procrustes distances are not scaled by the pooled
within-group covariance matrix, differences in covari-
ance structure between populations should not affect
these distances as drastically as they would affect
Mahalanobis distances. Also, Mahalanobis distances are
affected by uneven sample sizes, while no similar bias
has been noted for Procrustes distances.
The second research question addressed the degree of
concordance between temporal bone shape and genetic
relationships among human populations. This relation-
ship was tested by examining the correlation between
matrices of temporal bone morphology (i.e., size and
shape matrices) and molecular distances. Analogous
studies above the species level have compared phyloge-
netic trees based on morphology with those based on
molecular data (Lockwood et al., 2004; see also Collard
and Wood, 2001; Strait and Grine, 2004; Lycett and Col-
lard, 2005). However, within humans, a tree-like struc-
ture does not apply to population relationships for mor-
phological or molecular information (summarized by
Sherry and Batzer, 1997; Athreya and Glantz, 2003).
The current analysis is therefore restricted to matrix
correlation comparisons.
STR data were analyzed using Arlequin 3.0 (Excoffier
et al., 2005). Data on 783 STRs have been typed for
eight representative populations (Ramachandran et al.,
2005; Rosenberg et al., 2005), and a subset of 404 of the
same STRs has been typed in Native Australians. A ma-
trix of STR population distances was constructed using
Slatkin’s genetic distance, a distance measure analogous
to F ST but specifically designed for microsatellite loci in
assuming a stepwise mutation model (Slatkin, 1995).
The degree and significance of the correlation between
the distance matrices from the molecular and morpho-
43), a reduced sample of 20 ran-
domly chosen individuals was used for this analysis.
DFAs were conducted using SPSS (version 11.0.1).
Although Procrustes superimposition scales all speci-
mens to the same unit centroid size, size related shape
changes (i.e., allometry) are not removed. Therefore, to
assess the role of allometry, a size matrix (i.e., a matrix
of the absolute differences in centroid size between
groups) was calculated and compared with the Pro-
crustes distance (or shape) matrix using a Mantel test
(Mantel, 1976; Smouse et al., 1986) in PopTools, an add-
on for Microsoft Excel. Additionally, correlations between
centroid size and shape were evaluated by regressing
the principal component axes on centroid size using
Morphologika.
For each analysis, morphological distances (i.e., size or
shape distance matrices) were compared to the variable
of interest (e.g., molecular or environmental distances).
Both Procrustes and Mahalanobis distances were calcu-
lated for all populations used here, and these two dis-
tance measures were found to be significantly correlated
(r 5 0.662; P \ 0.001). Analyses using both of these dis-
tance measures were found to lead to the same general
pattern of results. However, while a number of authors
(Ackermann, 2002; Strand Viðarsd ´ ttir et al., 2002; Har-
vati, 2003; Harvati et al., 2004; McNulty, 2005; Harvati
and Weaver, 2006a,b) have previously used Mahalanobis
distances in analyses such as this, only Procrustes dis-
tances are reported here, as Mahalanobis distances
attempt to account for within group variation by scaling
5
American Journal of Physical Anthropology—DOI 10.1002/ajpa
(P-value 0.05) if it is reached or exceeded in less than
5% of the random permutations. Second, a discriminant
function analysis (DFA) was conducted using the first 40
PC scores from the PCA of Procrustes coordinates (which
accounted for
[
48985284.011.png
316
H.F. SMITH ET AL.
TABLE 3. Structure matrix for the discriminant function analysis (first 20 PCs only) showing the correlations
between each of the PC axes and discriminant functions
Function
1
2
3
4
5
6
7
8
9
10
PC1
0.131
0.439
2 0.088
2 0.105
0.145
2 0.127
0.230
0.034
0.113
2 0.052
PC2
0.140
0.057
0.182
0.246
2 0.033
0.312
2 0.200
2 0.196
0.051
0.056
PC3
0.076
2 0.211
2 0.061
0.063
0.327
2 0.118
0.071
0.122
2 0.124
2 0.169
PC4
2 0.022
2 0.068
0.070
0.070
0.043
0.164
0.316
2 0.005
0.203
2 0.133
PC5
2 0.037
2 0.010
0.124
0.149
2 0.028
2 0.381
0.102
0.037
2 0.052
0.309
PC6
0.189
0.018
2 0.238
0.379
2 0.051
2 0.153
0.031
0.094
2 0.113
0.144
PC7
2 0.069
0.228
2 0.073
0.186
0.142
0.152
2 0.062
0.165
2 0.357
2 0.068
PC8
0.029
0.022
0.141
2 0.136
0.193
0.037
2 0.049
2 0.069
2 0.033
2 0.093
PC9
0.138
2 0.055
0.013
0.060
0.172
0.014
0.175
2 0.137
0.023
0.009
PC10
2 0.173
0.079
0.029
0.287
0.133
2 0.005
0.149
0.038
0.223
0.018
PC11
0.053
0.035
0.149
2 0.091
0.043
0.060
0.130
0.165
2 0.094
0.286
PC12
0.012
0.037
0.037
0.006
0.059
2 0.080
2 0.049
0.141
0.087
0.164
PC13
0.061
2
0.028
2
0.039
2
0.140
2
0.066
0.156
2
0.135
0.445
2
0.081
2
0.062
PC14
0.102
2
0.094
0.208
0.224
2
0.125
2
0.022
2
0.050
0.064
2
0.056
2
0.244
PC15
2
0.105
0.015
0.192
0.121
0.023
2
0.035
2
0.044
0.042
0.071
0.053
PC16
0.020
2
0.030
0.084
2
0.046
0.010
0.211
0.206
0.012
2
0.065
0.242
PC17
2
0.019
0.085
0.032
0.023
0.127
0.072
2
0.143
0.225
0.390
2
0.093
PC18
2
0.011
2
0.098
0.015
2
0.037
0.182
0.123
2
0.142
0.025
0.029
0.250
PC19
0.010
2
0.033
0.196
2
0.001
0.110
2
0.092
2
0.070
0.210
2
0.034
2
0.096
PC20
2
0.049
2
0.073
2
0.018
0.037
0.150
0.015
0.086
2
0.276
2
0.129
0.064
logical analyses was assessed using a Mantel test, again
in PopTools.
Finally, environmental variables and geographic dis-
tances for populations were evaluated to determine how
they covary with temporal bone morphology. Environ-
mental distance matrices were generated for each envi-
ronmental variable: temperature, rainfall, latitude, and
altitude. A single overall environmental distance matrix
(Euclidean distance, incorporating data from all four
environmental variables) was also calculated in Pop-
Tools. To address the possibility that environmental fac-
tors influenced morphological difference, the morphologi-
cal distance matrices were compared to each environ-
mental matrix using a Mantel test.
To test the association between geography and mor-
phology, geographic great circle distances among popula-
tions were calculated. Great circle distances use latitude
and longitude and take into account the fact that these
coordinates are on the circumference of a sphere to cal-
culate distances between two locations. A geographic ma-
trix was generated using great circle distances and
including five waypoints (Fig. 1), geographic locations
through which populations would have had to travel
when migrating between two continents (Relethford,
2004; Ramachandran et al., 2005). This practice takes
into account the conclusion that most human migrations,
until recently, did not usually traverse large bodies of
water (Ramachandran et al., 2005). The inclusion of
waypoints, therefore, permits a more accurate estimate
of the migrational distance among populations, rather
than a line of minimal geographic distance that could
run across an ocean. The pairwise distance between any
two populations was calculated as the sum of the dis-
tance between Population 1 and the waypoint, and
between the waypoint and Population 2, plus any dis-
tances between waypoints if more than one waypoint fell
between the populations. Following Ramachandran et al.
(2005), waypoints included were Anadyr, Russia; Cairo,
Egypt; Istanbul, Turkey; Phnom Penh, Cambodia; and
Prince Rupert, Canada. Geographic distances among
populations on the same continent were calculated as
normal great circle distances. It is probable even within
TABLE 4. Eigenvalues, distribution of variance, and canonical
correlations for the discriminant function analysis
Function Eigenvalue
%of
variance Cumulative %
Canonical
correlation
1
6.39
40.81
40.81
0.93
2
2.78
17.75
58.56
0.86
3
1.44
9.18
67.74
0.77
4
1.29
8.23
75.97
0.75
5
1.21
7.74
83.71
0.74
6
0.91
5.82
89.53
0.69
7
0.58
3.69
93.22
0.61
8
0.45
2.87
96.09
0.56
9
0.33
2.11
98.20
0.50
10
0.28
1.80
100.00
0.47
continents that migrational distances are affected by
geographical barriers and are not simply great circle dis-
tances; this factor is considered later in discussing the
results. The hypothesis that temporal bone morphology
covaries with geographic distance was then assessed by
comparing the geographic matrix with the morphological
matrix using a Mantel test.
For all analyses, alpha was set at 0.05. All correlations
are reported as Pearson product moment correlation
coefficients (r).
RESULTS
In the DFA, the first function is influenced by a vari-
ety of principal components and accounts for just over
40% of variance among populations (Tables 3 and 4). As
expected, contributions of subsequent functions diminish
rapidly (Table 4).
Permutation tests of the Procrustes distances among
populations were all statistically significant with P-val-
ues of less than 0.001 (Table 5). The DFA with crossvali-
dation demonstrates that the populations can be distin-
guished relatively well, with classification rates between
56 and 85% (mean 73%) (Table 6). For 11 populations of
roughly equal sample size, the expected proportion of
correct random classifications is
9%, so these results
indicate high success rates.
American Journal of Physical Anthropology—DOI 10.1002/ajpa
48985284.012.png
Zgłoś jeśli naruszono regulamin