The effects of age, sex, weight, and breed on canid methylomes

Abstract

Unlike genomes, which are static throughout the lifespan of an organism, DNA methylomes are dynamic. To study these dynamics, we developed quantitative models that measure the effect of multiple factors on DNA methylomes including, age, sex, weight, and genetics. We conducted our study in canids, which prove to be an ideal species to assess epigenetic moderators due to their extreme variability in size and well-characterized genetic structure. We collected buccal swabs from 217 canids (207 domestic dogs and 10 grey wolves) and used targeted bisulphite sequencing to measure methylomes. We also measured genotypes at over one thousand single nucleotide polymorphisms (SNPs). As expected, we found that DNA methylomes are strongly associated with age, enabling the construction of epigenetic clocks. However, we also identify novel associations between methylomes and sex, weight, and sterilization status, leading to accurate models that predict these factors. Methylomes are also affected by genetics, and we observe multiple associations between SNP loci and methylated CpGs. Finally, we show that several factors moderate the relationship between epigenetic ages and real ages, such as body weight, which increases epigenetic ageing. In conclusion, we demonstrate that the plasticity of DNA methylomes is impacted by myriad genetics and physiological factors, and that DNA methylation biomarkers are accurate predictors of age, sex and sterilization status.

Keywords: DNA methylation; canids; dogs; epigenetic clock.

Figure 1.

Epigenetic age and state of dogs. (a) Models generated using least angle regression (LARS). (b) LARS model of the square root of age. (c) Epigenetic Pacemaker was also used to predict epigenetic states, and the trend line was fit using three functions – linear, logarithmic, and square root.

Figure 2.

Epigenetics of sex, weight and spayed status. (a) Predicted sex values (labelled as Epigenetic Sex on the y-axis) were coloured based on the sample’s sterilization status into five categories: blue for not spayed, Orange for spayed, green for not neutered, red for neutered, and purple for missing values. A two-tailed t-test was performed separately for each sex, on the predicted values of fixed and intact samples. The hermaphrodite (intersex) wolf was not included in the t-tests. (b) The confusion matrix displays the number of correctly and incorrectly predicted spayed status for the female samples during the LOOCV process. The rows are the true sterilization status, while the columns are the predicted sterilization status by the SVM model. (c) The lasso model from sklearn is used to predict the weight of dogs.

Figure 3.

Hierarchical tree and weight GWAS. Different breeds of dogs are shown on a hierarchical clustering tree based on their genotypes and each breed is assigned with a unique colour.

Figure 4.

Association of SNPs and methylation values in cis and trans. (a) SNPs were filtered for sites with variance greater than 0.3 and methylation sites were filtered for sites with variance greater than 0.02, resulting in 264 methylation sites and 118 SNPs. We identified 5 significantly correlated pairs of SNP and methylation sites using Pearson correlation and the Benjamini-Hochberg procedure with a false discovery rate of 0.1, only one of which is a cis pair. One SNP, at location chr30:39,263,693 is significantly correlated with three different methylation sites, all of which are trans. (b) Left-most column: histograms of genotypes across all samples for the two SNPs where significant trans correlations were found. Right 3 columns: violin plots of genotype vs methylation for the four significant trans correlations. Each row corresponds with one of the two SNPs that were found to have significant trans correlations. (c) Violin plot of genotype vs methylation values for the significant cis pair, which had a Pearson correlation coefficient of 0.485. The nearest gene to this cis pair is HPGD.

Figure 5.

Weight moderates epigenetic age. (a) Weight values were sorted into the two categories, low-to-medium weight (blue) and medium-to-high weight (Orange), by using the median weight(42.3 lbs) as the cut-off, and the slopes of epigenetic age vs. actual age for the two weight groups are shown. (b) Age vs epigenetic state plots split by value of the 3 significant traits: PC5, PC7, and heterozygosity. The purple points are samples with values higher than the median and the green points are samples with values lower than the median.