In this study, we leveraged brain imaging and genome-wide genotyping from 626 European Americans, as well as skull measurements obtained on an independent set of 1,170 individuals of European ancestry, to test the hypothesis that skull and brain morphology, like genetic background [14
], mirror geography within Europe. We found that skull and brain morphology vary continuously across Europe as evidenced by the weak clustering between and large variation within populations and are geospatially structured. In particular, we observed a significant NW-SE gradient in morphology that is independent of body size and involves predominantly frontotemporal cortical areas.
Sokal and Uytterschaut [17
] and Sokal et al. [18
] reported that craniometric variation is spatially structured across Europe but found no clear continental trend and further could not recapitulate a map of Europe. Moreover, these studies found that cranial variation was associated with geography and language family more than time period (Early and Late Middle Ages and a Recent Period). Therefore, European cranial diversity has retained some of its spatial patterning for at least 1,500 years. However, the data by Sokal and Uytterschaut [17
] and Sokal et al. [18
] were limited to population averages of only 10 different cranial measurements from populations with non-uniform geographic coverage of Europe. In contrast, our finding of significant skull and brain variation along a NW-SE cline is consistent with genetic and archaeological evidence for range expansions and gene flow due to prehistoric population movements along this geographic axis in Europe. In this ‘demic diffusion’ model [31
], human populations gradually expanded into new geographic areas and carried with them novel genetic (and hence phenotypic) variation. If migrating humans had progressive admixture with individuals local to those regions or replaced local individuals and experienced a succession of bottlenecks [32
], then a gradient of gene frequencies should have been formed. For example, the migration of anatomically modern humans out of Africa approximately 40,000 years ago, postglacial re-expansions from refugia in southern Europe, and the introduction of farming from SW Asia within the last 10,000 years could have generated a NW-SE gradient that is manifest in the genetic diversity across populations since these events involved movement of people primarily from southern to northern Europe [9
However, historical population expansions and invasions – e.g. Greek, Jewish, Phoenician [37
], and Viking [38
] – have also contributed to patterns of genetic diversity in Europe [39
] and may have masked genetic clines formed by earlier migrations. Furthermore, in contrast to demic diffusion, cultural diffusion could explain the spread of agricultural technology and cultural adaptations across Europe due to information – not gene – flow, and the pattern of Y chromosome diversity in modern Europeans reinforces this model [36
]. Therefore, the existence of genetic and craniometric clines in modern European populations suggests at least two theories: (1) prehistoric population movements made such a dominant contribution to the structure of genetic variation in Europe that more recent gene flow has not masked it, and (2) local environmental factors and selection generated clinal variation or acted to restore clinal variation after gene flow occurred. One intriguing possibility for such an environmental factor is the cultural conditions associated with possessing agricultural technologies, e.g. sedentarism, altered diet including milk consumption [40
], and new disease exposures [41
]. As these technologies spread progressively from SE to NW Europe over several 1,000 years [33
], natural selection may have acted either directly or indirectly to alter brain morphology, thus creating the clinal variation found in this study.
The genetic basis for this geographic trend in skull and brain variation is strengthened by the observation of the trend in North American individuals of European ancestry. The environment – e.g. nutrition and prenatal healthcare – can influence skull morphometry due to developmental plasticity [12
] and may be correlated with genetic variation across Europe. In contrast, environmental exposures may vary among European Americans, but this variation is less likely to be geographically structured based on individuals’ European ancestry. Furthermore, Ashkenazi Jewish individuals are geographically dispersed in Europe and yet are genetically quite similar and genetically intermediate between SE European and Middle Eastern populations [26
]. This provides further support that the observed NW-SE clinal variation in brain morphology is driven by genetic more than environmental differentiation of these populations.
Brain volume and cortical surface area in humans increased dramatically after our evolutionary divergence from non-human primates, peaking some 35,000 years ago, and has since decreased in conjunction with body size [45
]. Ruff et al. [45
] state that some modern humans may be expressing a genetic potential for increased body size and cranial capacity that we inherited from our prehistoric ancestors, and these genes may also contribute to the craniometric clinal variation found in this study. Craniometric studies have shown that random genetic drift can explain the morphological differences between Neanderthals (who had significantly larger brains) and modern humans [46
] and can account for most differences between human populations worldwide [9
]. Larger brain size in modern humans is thought to be weakly correlated with increased cognitive performance [47
], and therefore could have been subject to mild positive selection. Alternatively, larger skulls may have been selected for their increased globularity – i.e. a greater volume to surface area ratio – and thus reduced heat loss for a given brain size. Consequently, larger skulls may have provided a small selective advantage to humans in colder climates [48
] and indirectly resulted in modestly increased brain volumes in northern European populations. However, whether or not the morphological diversity we have observed across Europe resulted from neutral genetic drift, natural selection, or a combination of both is an open question.
There are limitations to the data sets we analyzed, but we contend that they are sources of noise that would mask and not falsely generate the clinal geographic variation in skull and brain morphology that we have described. First, the collection of European skulls that we analyzed came from individuals living during a broad range of time periods within the past 500 years. Average skull morphology at any particular location in Europe has likely evolved over time as the result of environmental changes and genetic demographic shifts due to regional migrations. However, we found no relationship between the age (medieval or modern) and geographic location of the populations from which the skulls were obtained. Therefore, this sampling of skulls from different time periods had the potential to obscure continental variation in skull morphology. It is striking that, despite this source of noise, we find strong evidence for clinal skull variation along a NW-SE axis.
Second, both skull and brain morphological data sets include individuals with estimated genetic ancestry from all 4 geographic quadrants of Europe, but NW and SE populations were more highly represented than NE and SW populations. In particular, 62% of skull measurements and 88% of brain measurements come from individuals with NW or SE European ancestry. Therefore, we had greater statistical power to detect morphological variation along a NW-SE geographic axis. In addition, greater than 87% of craniometric variation must be attributed to processes other than NW-SE clinal variation.
A third limitation of this study is the use of subjects diagnosed with MCI and AD. Reductions in brain volume, cortical area, and cortical thickness have been consistently observed in AD and, to a lesser extent, in MCI. If subjects with a diagnosis of AD or MCI were more likely to have inferred genetic ancestry from SE Europe, then we might have falsely attributed differences between NW and SE European brain morphology to genetic background and not neurological disease status. However, neither AD nor MCI diagnosis were significantly associated with degree of inferred NW-SE ancestry. Moreover, brain morphology remains significantly associated with NW-SE ancestry if we exclude subjects with AD, and cortical surface area shows trend level association in the small sample of healthy controls.
In summary, we found consistent and highly significant clinal variation in both skull and brain morphology in Europe despite these sources of noise. We predict that this geographic trend and additional subtle trends will be even more apparent in a modern cohort of healthy, non-elderly adults sampled from across the European continent.
It is plausible that genes responsible for cortical expansion during human evolution retain a role in brain development and contribute to normal variation in brain morphology within and between modern human populations [3
]. Identifying these genes could contribute to our understanding of developmental abnormalities associated with neuropsychiatric diseases such as autism and schizophrenia. In this context, admixture mapping may prove to be a powerful strategy for identifying genomic regions responsible for overt brain morphology differences among individuals of European ancestry. Independent of the use of inferred ancestry for identifying genes, our results indicate that studies seeking to identify genes that influence brain morphology should consider genetic background, as it reflects historical mixing and then isolation of populations.