Variation of BMP3 contributes to dog breed skull diversity

Abstract

Since the beginnings of domestication, the craniofacial architecture of the domestic dog has morphed and radiated to human whims. By beginning to define the genetic underpinnings of breed skull shapes, we can elucidate mechanisms of morphological diversification while presenting a framework for understanding human cephalic disorders. Using intrabreed association mapping with museum specimen measurements, we show that skull shape is regulated by at least five quantitative trait loci (QTLs). Our detailed analysis using whole-genome sequencing uncovers a missense mutation in BMP3. Validation studies in zebrafish show that Bmp3 function in cranial development is ancient. Our study reveals the causal variant for a canine QTL contributing to a major morphologic trait.

Figure 1. Quantitative and qualitative assessments of PC1 on canine cranioskeletal shape.

(A) Gray wolf (mesocephalic, ancestor to dogs) (B) Afghan hound (dolichocephalic), (C) Leonberger (mesocephalic), (D) Pug (brachycephalic). (E) Boxplots of PC1 (corresponding breed names are listed in Table S2). (F) Surface scans of a gray wolf skull illustrate morphological changes associated with PC1. Columns (left to right) are dorsal, lateral, and rostral views. Top row: a gray wolf skull morphed by positive PC1. Middle row: a gray wolf skull (no morphing). Bottom row: a gray wolf skull morphed by negative PC1. Pseudocoloring of the gray wolf skull indicates rostrum (ros) and neurocranium (nc). Line indicates width of the zygomatic arches (za).

Figure 2. PC1 GWAS and fine mapping at CFA32.

All GWAS used the mixed-model GEMMA. Chromosomes listed on the x-axis, −log₁₀(P) on the y-axis. SNPs remaining significant following Bonferroni correction are colored blue. Q-Q plots of observed versus expected −log₁₀(P) are depicted on right, with full SNP dataset (black circles), pruned dataset (grey circles), expected values (red lines), and 95% confidence intervals (black lines). Scan results using breed-sex averages of PC1 without (A) and with a breed-sex average size covariate (B). Including a size covariate in the mixed-model overcorrects, leading to loss of associations on CFA 30 and 32.(C) Scan results using PC1 breed-sex averages and breed-sex size covariates. In this scan, only breeds whose neurocranium size ranked within the smallest 50% of our dataset where analyzed. By reducing relatedness disparity in our study population, the association on CFA32 remains significant despite size correction. (D) Scan results using all breed-sex averages of PC1, but excluding extreme brachycephalic breeds (Pug, Pekingese, Boston Terrier, Shih Tzu, Brussels Griffon, French Bulldog, Bulldog, Boxer, Cavalier King Charles Spaniel, Chihuahua). (E) Average log(H_O ratios) or F_ST from ten-SNP sliding windows. (F) Regional H_O or F_ST values, and their respective Lowess best fit curves.

Figure 3. Genetic variation at the CFA32 QTL includes a brachycephaly-associated missense mutation within BMP3.

For display purposes, we set the reference sequence to be the allele most common to Pekingese and Bulldog. Variants located within 8.15–8.27 Mb (A) or the 85 kb critical interval (B) are illustrated (homozygous reference = yellow, heterozygous = orange, homozygous variant = red). (A) Pekingese and Bulldog agree across an 85 kb interval (black bar) including BMP3 (red) and a portion of PRKG2 (aqua). Line graphs below genes plot conservation (phastCons4way) and association (−log₁₀(P)) with respect to variant position (28). (B) Variants of interest met one or more of the following criteria: conserved (phastCons4way score ≥0.7), associated (an association P-value among the smallest 5% of P-values, see Materials and Methods), exonic (untranslated regions and coding), or splice (located within 20 bp of an exon boundary). Forty-eight variants of interest remained after applying filtering criteria, including a F452L mutation in BMP3 at position 8,196,098.

Figure 4. Zebrafish cranioskeletal development requires Bmp3 function.

(A–H) Wholemount RNA in situ hybridization of bmp3 expression at 48 hpf (A,B), 72 hpf (C), and 96 hpf (D) stages. Anterior to the left. (A) Dorsal view, (B–D) lateral view. Pharyngeal arches indicated by brackets, pectoral fins by red arrowheads. Wholemounts (E,H,K) and alcian blue cartilage stains (F,G,I,J,L,M) of 96 hpf embryos from uninjected (E–G) and morpholino-injected embryos (h–j, mild phenotype, n = 72/177; k–m, severe phenotype, n = 83/177). Phenotypic severity is distinguished by tail curling (compare insets). Loss of jaw structures (black arrows) and frontal bossing (white arrowheads) is apparent in both classes of morphants. Cartilage is severely dysmorphic, hypoplastic, or absent following Bmp3 knockdown. Abbreviations correspond to ceratobranchial (cb), ceratohyal (ch), eythmoid plate (ep), hyosymplectic (hs), Meckel's (m), palatoquadrate (pq), and trabeculae (tr) cartilages.