8 Inferring Human Demographic History from Genetic Data
191
population growth. We note in passing though that there are several potential
confounding factors that have not yet been adequately accounted for in studies
of recent population growth, including the effects of purifying and background
selection on the site frequency spectrum and the effects of aggregating different
populations into larger groups such as “Europeans” or “African-Americans.”
8.4
Population Bottlenecks
The early analyses of human mtDNA variation assumed a founder model, whereby
the TMRCA corresponded to the “founding” of a population by a few genetically
similar individuals (e.g., Cann et al. 1987). This in part led to the idea that our
species had undergone a drastic population bottleneck (i.e., a temporary reduction
in effective population size) during the mid- to late-Pleistocene (e.g., Cann et al.
1987; Gibbons 1993). Population genetics theory, though, suggests that the specific
TMRCA has very little correlation with the effective population size at that time.
In particular, the recent TMRCA of mtDNA might be due to random chance or
the action of natural selection. Since demographic events such as bottlenecks are
expected to affect genetic variation across the whole genome, it is straightforward
to analyze nuclear sequence polymorphism data to assess the strength of evidence
for a species-wide bottleneck. Analyses of the HLA region, which has extremely
high levels of diversity due to diversifying selection, show that the human effective
population size has not dropped much below 10,000 for all of our species’
history (Takahata 1993; Ayala 1995), and analyses of unlinked, putatively neutral,
autosomal regions came to similar conclusions (Sjödin et al. 2012).
In contrast, there is strong evidence that at least some human populations have
experienced a recent population bottleneck. Simulations suggest that population
bottlenecks lead to a reduction in levels of nucleotide and haplotype diversity, an
increase in levels of linkage disequilibrium (LD), and a skew in the distribution
of allele frequencies (Fay and Wu 1999; Reich et al. 2001; Wall et al. 2002).
Studies of microsatellite (Tishkoff et al. 1996; Rosenberg et al. 2002), singlenucleotide polymorphism (SNP, e.g., Conrad et al. 2006; Jakobsson et al. 2008),
and sequence (Frisse et al. 2001; Livingston et al. 2004; Voight et al. 2005; 1000
Genomes Project 2010, 2012, 2015; Mallick et al. 2016) data consistently show
that all non-African populations have less variation and more LD than all subSaharan African populations. The simplest explanation for this pattern is that all
non-African populations have experienced at least one population bottleneck in their
recent history. This is consistent with the recent African origin and replacement
model of human evolution (Stringer and Andrews 1988), which posits that modern
humans first evolved in sub-Saharan Africa 150–200 Kya and that modern humans
later expanded and replaced (without admixture) the indigenous “archaic” humans
they encountered in the rest of the world. It is also consistent with the recent African
origin and hybridization model (Brauer 1989), which is identical to the previous
model but allows for a limited amount of hybridization between modern and archaic
humans, but not consistent with some models of modern human evolution such as
191
population growth. We note in passing though that there are several potential
confounding factors that have not yet been adequately accounted for in studies
of recent population growth, including the effects of purifying and background
selection on the site frequency spectrum and the effects of aggregating different
populations into larger groups such as “Europeans” or “African-Americans.”
8.4
Population Bottlenecks
The early analyses of human mtDNA variation assumed a founder model, whereby
the TMRCA corresponded to the “founding” of a population by a few genetically
similar individuals (e.g., Cann et al. 1987). This in part led to the idea that our
species had undergone a drastic population bottleneck (i.e., a temporary reduction
in effective population size) during the mid- to late-Pleistocene (e.g., Cann et al.
1987; Gibbons 1993). Population genetics theory, though, suggests that the specific
TMRCA has very little correlation with the effective population size at that time.
In particular, the recent TMRCA of mtDNA might be due to random chance or
the action of natural selection. Since demographic events such as bottlenecks are
expected to affect genetic variation across the whole genome, it is straightforward
to analyze nuclear sequence polymorphism data to assess the strength of evidence
for a species-wide bottleneck. Analyses of the HLA region, which has extremely
high levels of diversity due to diversifying selection, show that the human effective
population size has not dropped much below 10,000 for all of our species’
history (Takahata 1993; Ayala 1995), and analyses of unlinked, putatively neutral,
autosomal regions came to similar conclusions (Sjödin et al. 2012).
In contrast, there is strong evidence that at least some human populations have
experienced a recent population bottleneck. Simulations suggest that population
bottlenecks lead to a reduction in levels of nucleotide and haplotype diversity, an
increase in levels of linkage disequilibrium (LD), and a skew in the distribution
of allele frequencies (Fay and Wu 1999; Reich et al. 2001; Wall et al. 2002).
Studies of microsatellite (Tishkoff et al. 1996; Rosenberg et al. 2002), singlenucleotide polymorphism (SNP, e.g., Conrad et al. 2006; Jakobsson et al. 2008),
and sequence (Frisse et al. 2001; Livingston et al. 2004; Voight et al. 2005; 1000
Genomes Project 2010, 2012, 2015; Mallick et al. 2016) data consistently show
that all non-African populations have less variation and more LD than all subSaharan African populations. The simplest explanation for this pattern is that all
non-African populations have experienced at least one population bottleneck in their
recent history. This is consistent with the recent African origin and replacement
model of human evolution (Stringer and Andrews 1988), which posits that modern
humans first evolved in sub-Saharan Africa 150–200 Kya and that modern humans
later expanded and replaced (without admixture) the indigenous “archaic” humans
they encountered in the rest of the world. It is also consistent with the recent African
origin and hybridization model (Brauer 1989), which is identical to the previous
model but allows for a limited amount of hybridization between modern and archaic
humans, but not consistent with some models of modern human evolution such as