We performed shotgun metagenome (assessing functional diversity) and 16S rRNA gene V4 amplicon (assessing taxonomic diversity) sequencing of time-series samples from the closed laboratory mesocosm chambers with oil addition (oiled) or without (control) to test whether or not the specialization disturbance hypothesis could explain the microbial community succession patterns (response). Additionally, metagenomic datasets from the Pensacola Beach field study [4] were included for comparison. The latter datasets represented beach sands before the oil had reached the coast (Pre-Spill), while the beach was contaminated (Spill-Oiled and Weathered), and after the oil concentrations in beach sands had reached undetectable levels (Recovered) (Fig. 1). The detailed description of the sample processing, sequencing, and bioinformatic analyses can be found in the Supplementary Online material (Figs. S3 and S4). Nonpareil, a tool that estimates what fraction of the microbial community is represented in a metagenome (i.e., the coverage) by examining the level of redundancy among the metagenomic reads [13], showed that coverage of the sampled microbial communities by sequencing was adequate for comparison [14], with 60–75% sample coverage for oiled mesocosm and 45–70% for control sample. In addition, Nonpareil sequence diversity (Nd), an estimate of the total diversity in sequence space harbored by a microbial community, and other diversity metrics showed that control samples (no oil added) harbored higher diversity. Applying the commonly used pipeline of assigning 16S rRNA gene fragments recovered in the metagenomes against the SILVA database release 132 (ref. [15]) using VSEARCH in QIIME2 [16] and 97% nucleotide identity for a match (closed OTU picking) resulted in 11% fewer reads assigned for control vs. oiled mesocosm metagenomes and 27% fewer reads assigned for clean vs. oiled Pensacola metagenomes.
These results indicated that the control samples potentially harbored more novel (uncharacterized) taxa that could confound taxonomic comparisons due to the comparatively lower number of taxonomically identified sequences. To account for this effect, we employed a manual pipeline with BLASTn [17], and a lower cut-off (90% nucleotide identity) for read assignment to the database (Method 2). Additionally, we performed our analysis based on both 16S rRNA gene amplicon sequences as well as 16S-carrying metagenomic reads, and employed Hill numbers, represented as qD, a group of diversity indices that take into account species abundance and richness to compute the equivalent number of species at an order q, where q adjusts the sensitivity to rare species (see also [18] and references therein). Our results, after rarefying the 16S rRNA gene fragment OTU abundance to the metagenomic dataset with the lowest coverage [18], showed that the inverse Simpson index (2D) was lower in oiled chambers with a mean of 274 (SD = 146) than in control chambers with a mean of 896 (SD = 86; Table 1). The Welch’s t-test revealed a significant difference at alpha 0.05 (p value = 8.89e−4). Amplicon data from the same mesocosm samples (Fig. 1) or analysis at the sequence variant level (ASVs; Fig. S5) showed similar results (Fig. 1). See supplementary results and discussion for further details (Fig S6).
Functional diversity was analyzed based on the number of metagenomic reads matching molecular function gene ontology (GO) terms [19] as previously described [7]. Our analysis showed that functional diversity (1D) was higher in oiled chambers with a mean of 193 equivalent GO terms (SD = 18) compared to control chambers with a mean of 105 equivalent GO terms (SD = 31; Table 1; Chao-Shen Entropy Estimator; p value = 1.14e-05, two-tailed Welch’s t-test).
Collectively, the results presented here from closed system mesocosms, which were designed to limit fluctuations in environmental conditions, stochasticity, and dispersal, showed that the specialization disturbance hypothesis can explain microbial succession patterns following crude oil perturbations in coastal beach sand environments. Recent incubation experiments of microbial communities from sandy soils have also provided evidence in support of the specialization-disturbance hypothesis (preprint available at the time of writing [6]), and the close agreement of these results with those of previous field observations (e.g., Table 1, Fig. 1) [7] suggested that this underlying explanation/mechanism is robust even in light of environmental variation and drift. The high similarity in taxonomic composition between our mesocosms and our previous field data also suggested that the key oil degraders were present in the clean sands at the time of sampling for establishing the mesocosms, 7 years after the DWH oil spill. The survival strategy of oil degraders in the clean sand remains an interesting question, and has implications for whether or not their niche breadth includes uncontaminated sandy sediments. It would be interesting to test whether similar patterns are observed in other habitats (e.g., beach sand from an alternative source lacking a history of oil exposure) and other types of perturbation in order to test the universal applicability of the results reported here. With sufficient background data available on the unperturbed ecosystem, we believe that the approach outlined here based on specialist vs. generalist taxa should be able to elucidate whether or not the specialization disturbance can explain microbial responses to other types of perturbations and/or identify recovered ecosystems.
Source: Ecology - nature.com
