Natural and human-driven selection of a single non-coding body size variant in ancient and modern canids

Abstract

Domestic dogs (Canis lupus familiaris) are the most variable-sized mammalian species on Earth, displaying a 40-fold size difference between breeds.¹ Although dogs of variable size are found in the archeological record,^2-4 the most dramatic shifts in body size are the result of selection over the last two centuries, as dog breeders selected and propagated phenotypic extremes within closed breeding populations.⁵ Analyses of over 200 domestic breeds have identified approximately 20 body size genes regulating insulin processing, fatty acid metabolism, TGFβ signaling, and skeletal formation.^6-10 Of these, insulin-like growth factor 1 (IGF1) predominates, controlling approximately 15% of body size variation between breeds.⁸ The identification of a functional mutation associated with IGF1 has thus far proven elusive.^6,10,11 Here, to identify and elucidate the role of an ancestral IGF1 allele in the propagation of modern canids, we analyzed 1,431 genome sequences from 13 species, including both ancient and modern canids, thus allowing us to define the evolutionary history of both ancestral and derived alleles at this locus. We identified a single variant in an antisense long non-coding RNA (IGF1-AS) that interacts with the IGF1 gene, creating a duplex. While the derived mutation predominates in both modern gray wolves and large domestic breeds, the ancestral allele, which predisposes to small size, was common in small-sized breeds and smaller wild canids. Our analyses demonstrate that this major regulator of canid body size nearly vanished in Pleistocene wolves, before its recent resurgence resulting from human-imposed selection for small-sized breed dogs.

Keywords: IGF1; ancient DNA; antisense lncRNA; body size; canid evolution; canine; dog; domestication; long non-coding RNA; wolf.

Figure 1.. Insulin-like growth factor 1 (IGF1) in Canidae.

(A) IGF1-AS variant genotypes and body mass range collected from 1,162 dogs of 230 breeds. Dots represent outliers. Blue diamonds indicate breed body mass averages, boxplots represent interquartile ranges and black horizontal bars show median for each (***P < 0.0001, Mann–Whitney–Wilcoxon tests). (B) Distribution of IGF1-AS alleles in three schnauzer breeds and poodle varieties. Pie chart indicates population proportion based on genotypes. Red=CC, orange=CT, and yellow=TT. (C) Body mass and serum levels of IGF-1 protein (nmol/L) as functions of IGF1-AS genotype. IGF-1 serum protein levels were assayed in 51 dogs, including 13 mixed-breed dogs (*P<0.01, **P<0.001, ***P < 0.0001, Mann–Whitney–Wilcoxon tests); (D) Positive correlation observed between body mass and IGF-1 serum level (Rho Spearman test). Blue line shows the regression line, grey area represents confidence interval. See also Table S1 and Data S1.

Figure 2.. Detection of IGF1-AS variants in ancient and modern genomes.

(A) Map of DNA sampling locations for 35 ancient canids, colored by their genotypes. Circles=dogs; triangles=wolves. Data were merged when several samples were collected from the same site with the same predicted age. Number of samples are indicated between brackets. Ages are given in thousand years before present (k). (B) Genotypes for the IGF1-AS variant in 13 species: 92 whole genome sequences and 58 DNA samples that were Sanger sequenced for the IGF1-AS variant. Map demonstrates a North/South geographic gradient of alleles corresponding to body size. See also Figures S1–S3 and Tables S1–S3.

Figure 3.. Proposed ancestry for Canis lupus lineage based on IGF1-AS allele distribution.

(A) IGF1 locus from UCSC browser showing four IGF1 transcripts (blue) and two IGF1-AS predicted transcripts (CFRNASEQ_AS_00037985, CFRNASEQ_AS_0003798) (red) that overlap IGF1 transcripts by 182 bp. Black arrow indicates the position of IGF1-AS variant (rs22397284). Conservation between dogs, ferret, panda and cats for 40 nucleotides surrounding the IGF1-AS variant (bold) and for the full length IGF1-AS predicted transcript (CFRNASEQ_AS_00037985). The C allele, associated with small sizes in canids, and shared by the four species corresponds to the ancestral allele. (B) Canidae ancestor was likely small and carried the C allele. The large allele arose some time before 53,000 years before present (53,000 ybp) and generated bigger animals (Canis lupus). The ancestral small allele continues to exist in the grey wolf population, albeit at a low frequency. Approximately 15,000 ybp, canine domestication likely began with large wolf-like dogs. Shortly thereafter, human selection of small canids with the ancestral C allele led to preponderance of small modern domestic breeds. Grey arrow reflects actual hybridization observed between coyotes and wolves in East of America. See also Figures S2–S3.

Figure 4.. Relationship between IGF1-AS variant genotype and individual body mass measures in coyotes.

(A) Mean body mass reported by U.S. state for 79 coyotes sampled by universities and museums, as indicated in Methods. Circle size indicates number of individuals. (B) Frequency of the C allele of IGF1-AS variant in 76 samples (distinct from those in A - see Methods) drawn from eight states across the U.S. Both maps illustrate the West-to-East gradient for the coyote population supported by statistical models (C-D). Linear (blue) and quadratic (red) relationships between longitude and, body mass (C), or small allele frequency (D). Lines indicate predicted values from generalized linear models (with binomial error for small allele frequency and Gaussian error for body mass). In both cases, quadratic and linear effects received similar statistical support. West coast coyotes are primarily homozygous (C allele freq = 0.93, mean body mass = 9.18 kg ± 2 SD); East coast coyotes carry all three genotypes (Mean body mass = 16.03 kg ± 3 SD). Mid-U.S. states (Nebraska and Oklahoma) were not included in these estimations. (E) Analysis of 28 coyotes from Pennsylvania demonstrates a significant relationship between IGF1-AS allele status and body mass (*P<0.01, ***P < 0.0001, Mann–Whitney–Wilcoxon tests), but exclude the hypothesis of a local geographic effect on their distributions (at the state scale). See also Tables S1–S2.