Milk digestion, it does a body good

A few weeks ago I presented a tentative model for how lactose tolerance (the ability of adults to digest milk easily and efficiently) spread throughout the world. Here is what I offered:

1) A haplotype block, A, associated with the LCT gene, and a particular SNP on that block, -13910*T, is responsible for Eurasian lactose tolerance (here are some numbers across populations)

2) In African populations which are lactose tolerant this haplotype, A, and the SNP, -13910*T, are not always found.

3) This implies that separate mutational events resulted in alternative genetic strategies which maintained lactose digestion in adults across the Old World.

A new paper adds more nuance, and makes me retract my simple model. From the abstract:

The -13910*T allele occurs at very high frequency in northern Europeans as part of a very long haplotype (known as A), and promotes binding of the transcription factor Oct-1. However, -13910*T is at very low frequency in many African milk drinking pastoralist groups where lactase persistence phenotype has been reported at high frequency. We report here for the first time, a cohort study of lactose digester and non-digester Sudanese volunteers and show there is no association of -13910*T or the A haplotype with lactase persistence. We support this finding with new genotype/phenotype frequency comparisons in pastoralist groups of eastern African and Middle Eastern origin. Resequencing revealed three new SNPs in close proximity to -13910*T, two of which are within the Oct-1 binding site. The most frequent of these (-13915*G) is associated with lactose tolerance in the cohort study, providing evidence for a cis-acting effect. Despite its location, -13915*G abolishes, rather than enhances Oct-1 binding, indicating that this particular interaction is unlikely to be involved in lactase persistence. This study reveals the complexity of this phenotypic polymorphism and highlights the limitations of C-13910T as a diagnostic test for lactase persistence status, at least for people with non-European ancestry.

There is evidence for a disproportionate number of selective sweeps outside of Africa, and I imagined that the mutation which confers function around LCT to have been one of those sweeps. In Europeans the region around LCT is the largest known haplotype block, suggesting powerful selection that homogenized the genome around it and hasn't had time to be disrupted by recombination. I am more skeptical now of a singular Eurasian sweep, the data from the Middle Eastern populations suggest that they too have alternative lactose tolerance strategies. Additionally, I was a bit surprised at the geographic granularity of the ability to digest milk, urban Arab populations are far less tolerant than Bedouins (the difference is even greater from what I can tell than between north and south Indians, and more like the variance between north and south Europeans). This makes us consider the special characteristics of northern European diet over the past 10,000 years, as lactose tolerance is nearly ubiquitous in that region (as is the -13910*T SNP). If Middle Eastern populations have their own genetic kis, that suggests to me that that north Indian populations also have their own novel alleles (one point to consider though is that north Indian populations seem to have a far higher percentage of the A haplotype block which is common in Europe). Yann offered that this might be another trait like skin color, multiple genetic variants in multiple regions converging upon the same phenotypic outcome. Finally, if you read the paper closely you'll note that there are likely gene-gene interaction effects and trans as well as cis acting factors, which might have relevance for dominance vs. additivity across a locus. Things just got a lot more complicated....

Note: If there are different genetic strategies for digesting lactose, does that mean they are all equally fit? No. But, the ability of a mutation to invade and sweep across populations is contingent upon the relative fitness it confers vis-a-vis the population median fitness. One can imagine that even if alleles differ in their fitness despite their ability to confer a common trait (e.g., the European lactose digestion allele being inferior in efficiency to the Middle Eastern one), if the chasm is small enough than stochastic factors might loom large enough to make fixation of the more fit allele far less likely.