Effects of nutrient input on functional gene abundance.

Once annotation groups were defined, we calculated the frequency of each group at each site across enriched and reference creeks and tested our hypothesis that nutrient loading increased their frequencies in the enriched creek using nonparametric permutational multivariate analysis of variance (MANOVA) models (27). The models included both creek (reference or enriched) and bulk sediment carbon percentages (Fig. 1C) as factors and were run over Bray-Curtis similarity matrices for each annotation subgroup. These analyses were conducted in the R statistical programming language (v. 3.1.2) using the adonis implementation in the vegan community ecology library (28), using 10,000 permutations within each factor. These analyses provided a robust method to compare the relative contributions of nutrient input in driving differences in annotation frequencies between the creeks relative to that of natural variation between the creeks.