Fig 1.
Map of Kenyan honey bee sample locations.
The monticola bee is associated with the isolated highland forests and the scutellata bee occurs in the surrounding lowland savannahs. High mountain peaks where monticola has been found are indicated with black triangles. Mount Kenya and Mau boxes indicate sample locations (grey = monticola; yellow = scutellata). Mt Kenya Forest (average sample elevation 2,300 m above sea level; n = 10), Mt Kenya Savannah (1,100 m asl; n = 9), Mau Forest (2,900 m asl; n = 10) and Mau Savannah (1,900 m asl; n = 10).
Table 1.
Population genetic statistics.
Fig 2.
Population interrelationships.
(A) Neighbor joining-tree inferred from pairwise FST-distances between the four Kenyan population samples and previously sequenced honey bee populations from Africa (inside ellipse), the Middle East (syriaca; anatoliaca) and Europe (ligustica; carnica; iberiensis; mellifera). KE = Kenya. SA = South Africa. The position of the Eastern honey bee A. cerana used as outgroup is indicated with dashed line (not drawn to scale). (B) Interrelationships and divergence between highland and lowland bees as inferred from whole-genome estimation of the average pairwise genetic distances (dXY) between all specimens. (C) The proportion of alternative population interrelationships inferred from 10 kbp non-overlapping segments across the genome. Topology nr. 1 groups populations by location whereas nr. 3 groups them by environment. (D) Results from (C) subdivided by chromosomes. Color codes as in (C).
Fig 3.
Genome-wide and localized divergence between highland and lowland populations.
(A) Genome-wide plot of allele frequency differences (FST) of every nuclear SNP (n = 8,021,515) segregating between highland bees (n = 20) and lowland bees (n = 19). Divergent regions r7 (chromosome 7; blue) and r9 (chromosome 9; green) contains nearly all FST>0.5 SNPs (n = 24,441). Black line indicates overall FST across 10 kbp non-overlapping windows. (B) Number of SNPs subdivided by FST (0.05 bins) for r7, r9 and the rest of the genome. Stars indicate high-FST intervals with unexpectedly large numbers of SNPs. (C) Magnified view of the r7 block (left; including sub-parts i and ii) and r9 (right). Window-based FST as in (A). (D) Interrelationships and divergence between highland and lowland bees after exclusion of r7 and r9 SNPs. NJ-tree to the left based on average pairwise genetic distances (dXY) between all specimens. NJ-tree to the right inferred from FST between the four populations.
Table 2.
Differentiated regions on chromosomes 7 and 9.
Fig 4.
The distribution of highland and lowland haplotypes as inferred from genotypes at divergent SNPs.
The proportion of genotypes where a sample is either homozygous for the honey bee reference sequence allele (0/0), homozygous for the non-reference allele (1/1) or heterozygous (0/1) is shown for every sample at both the r7 (n = 16,001) and r9 blocks (n = 8,440). Bar plots indicate the proportions of each genotype for all divergent SNPs (FST>0.5). Bottom panels indicate altitude, location and classification of all Kenyan samples.
Fig 5.
Haplotype interrelationships and divergence.
(A) NJ-tree showing the interrelationships between Kenyan highland and lowland r7 haplotypes and the haplotype diversity of other honey bee populations. The position of the outgroup A. cerana is indicated with dashed line (not drawn to scale). (B) The corresponding analysis of the r9 region. (C) NJ-tree inferred from pairwise genetic distances (dXY) between highland and lowland r7 haplotypes. Outgroup branch not drawn to scale (dXY = 6.9%). (D) The corresponding analysis of the r9 region.
Table 3.
Divergence and diversity.
Fig 6.
Window-based estimates of divergence and relative genetic diversity between highland and lowland haplotypes.
(A) Divergence and diversity for the r7 region (r7i+r7ii; shaded) on chromosome 7 (10 kbp non-overlapping windows). From the top: i) Average pairwise genetic distances (dXY) between forest highland (r7h) and savannah lowland (r7l) haplotypes; diamond symbols are dXY for genes (centered at gene-body mid-points); yellow diamonds are octopamine receptors; ii) four-way population interrelationships as in Fig 2C and 2D; iii) genetic diversity on r7h relative to r7l; iv) average per-haplotype sample mapping depth for r7h (grey lines) and r7l (yellow bars) with three >10× mapping peaks shared between highland and lowland haplotypes indicated with stars (per-haplotype average coverage 131×, 13× and 13×, respectively); v) the difference in mapping depth after normalizing across the whole chromosome. Scale bar indicate scaffold coordinates and orientation (? = unknown orientation) in the reference genome sequence. (B) The corresponding data for the r9 region. Shared mapping peak at 13× per-haplotype coverage indicated with star.