Analysis of covariance.

Since MANOVA revealed that a negative relationship between biological composition (the RNA/DNA ratio and the Symbiodinium genome copy proportion [GCP]) and gene expression significantly distinguished outliers from non-outliers (Fig 4C), multivariate analysis of covariance (MANCOVA) and univariate ANCOVA were performed on the multivariate means (excluding size data) and individual gene expression means, respectively. Only genes for which statistically significant interaction effects were documented between a biological composition parameter and outlier status (analyzed as a categorical variable: outlier [yes] vs. non-outlier [no]) have been presented. For MANCOVA, Wilks’ lambda values are shown for comparisons between a continuous variable and a categorical one, while Exact F values are shown between two continuous variables. For the multivariate data, individual correlations were tested between canonical scores (first axis only) and 1) the Symbiodinium GCP and 2) the RNA/DNA ratio. t = linear regression test statistic. *p<0.05. **p<0.01. ***p<0.0001.