Results:
First, we evaluated whether there were 2-way interactions between gene
and tissue and between gene and country to determine whether subsequent
analyses should be carried out separately for each gene. There was a
significant interaction between gene and tissue (F4,660= 5.69, p < 0.001). Inclusion of individual ID and country as
random effects also revealed significant repeatability of
among-individual (r = 0.26, 95% CI = 0.23, 0.28) and among-country (r =
0.42, 95% CI = 0.28, 0.59) variation in gene expression.
Next, we were interested in determining which covariates best explained
the among-country variation observed. Our model selection results (Table
2) revealed that the same set of factors were consistent predictors of
expression for all three genes: tissue, population type, temperature
predictability, genetic group, and latitude; urbanization was a
significant predictor for TET2 only. Sex, body mass, altitude, and
precipitation predictability were excluded from all models. Moreover,
the best-fit models were quite effective at explaining variation in
expression of all three genes (conditional R2 range:
0.36 – 0.48; Table 2)
Finally, given that our first analysis revealed significant among
individual and among country repeatability, and our model selection
revealed that the same set of factors explained variation in gene
expression for all three genes, we evaluated correlations in gene
expression among-countries, among-individuals, and within-individuals.
We found strong support for correlations in gene expression among
countries and within individuals, but not among individuals (see Figure
3). Controlling for tissue type, in countries where house sparrows
exhibited high average expression of DNMT1, they also had had average
expression of DNMT3 (r = 0.15, 95% CI = 0.11, 0.35) and TET2 (r = 0.16,
95% CI = 0.11, 0.36), and similarly high expression of DNMT3 was
associated with high expression of TET2 (r = 0.18, 95% CI = 0.11,
0.36). However, when controlling for these among country correlations,
there was no correlation between an individual’s average expression of
DNMT1 and DNMT3 (r = 0.000, 95% CI = -0.018, 0.059), between DNMT1 and
TET2 (r = 0.00, -0.024, 0.031), or between DNMT3 and TET2 (r = 0.00,
95% CI = -0.02, 0.06). There were significant within individual
correlations in gene expression. Higher residual expression of DNMT1 was
positively correlated with expression of DNMT3 (r = 0.23, 95% CI =
0.20, 0.26) and TET2 (r = 0.16, 95% CI = 0.13, 0.19), and higher
expression of DNMT3 was correlated with higher expression of TET2 (r =
0.20, 95% CI = 0.17, 0.23), controlling for tissue type.