Comparative analysis of amplicon and metagenomic sequencing methods reveals key features in the evolution of animal metaorganisms

Abstract

Background: The interplay between hosts and their associated microbiome is now recognized as a fundamental basis of the ecology, evolution, and development of both players. These interdependencies inspired a new view of multicellular organisms as "metaorganisms." The goal of the Collaborative Research Center "Origin and Function of Metaorganisms" is to understand why and how microbial communities form long-term associations with hosts from diverse taxonomic groups, ranging from sponges to humans in addition to plants.

Methods: In order to optimize the choice of analysis procedures, which may differ according to the host organism and question at hand, we systematically compared the two main technical approaches for profiling microbial communities, 16S rRNA gene amplicon and metagenomic shotgun sequencing across our panel of ten host taxa. This includes two commonly used 16S rRNA gene regions and two amplification procedures, thus totaling five different microbial profiles per host sample.

Conclusion: While 16S rRNA gene-based analyses are subject to much skepticism, we demonstrate that many aspects of bacterial community characterization are consistent across methods. The resulting insight facilitates the selection of appropriate methods across a wide range of host taxa. Overall, we recommend single- over multi-step amplification procedures, and although exceptions and trade-offs exist, the V3 V4 over the V1 V2 region of the 16S rRNA gene. Finally, by contrasting taxonomic and functional profiles and performing phylogenetic analysis, we provide important and novel insight into broad evolutionary patterns among metaorganisms, whereby the transition of animals from an aquatic to a terrestrial habitat marks a major event in the evolution of host-associated microbial composition.

Keywords: Animal microbiome; Evolution; Holobiont; Metaorganism; Phylosymbiosis.

Conflict of interest statement

The authors declare that they have no competing interests.

Figures

Fig. 1

Average community composition of bacteria (a) and fungi (b) in the mock community samples sequenced via metagenomic shotgun and 16S rRNA gene amplicon techniques (amplicon: V1 V2, V3 V4, one-step, two-step; shotgun: MEGAN based classification (short reads)). c Bacterial genus-level alpha diversity estimates in comparison to the expected community value. d Principle coordinate analysis of the Bray-Curtis distance between methods and the expected community. Ellipses represent standard deviations of points within the respective groups. Sample sizes for the different approaches are Nshotgun = 4, NV1V2-one-step = 3, NV1V2-two-step = 3, NV3V4-one-step = 3, and NV1V2-two-step = 3

Fig. 2

Comparison of bacterial genus richness (a) and Shannon H (b) based on 16S rRNA gene amplicon and shotgun derived genus profiles based on MEGAN highlighting the differences between variable regions, amplification methods, and metagenomic classifier, as well as between the different host organisms. Colors show significance of amplification methods (a, c) or pairwise comparisons of methods (b, d) based on pairwise t tests with Hommel P value adjustment (a, b), and approximate Wilcoxon test for the comparison between environmental categories (c, d). Mean values are shown in gray symbols in plots a and b. Sample sizes are indicated below the samples

Fig. 3

Non-metric multidimensional scaling of Bray-Curtis distances based on genus abundance profiles derived from the different 16S rRNA gene amplicon methods (V1 V2/V3 V4, one-step/two-step) and shotgun-derived genus profiles highlighting a host differences and b differences between host environments (terrestrial/aquatic; see Table 2). Non-metric multidimensional scaling of Jaccard distances based on genus presence/absence profiles derived from the different 16S rRNA gene amplicon methods and shotgun-derived genus profiles highlighting c host taxon differences and d differences between host environments (terrestrial/aquatic; see Table 2). Both panels show a separation based on host organisms and environments and not by method. Large symbols indicate the centroid of the respective host groups and vertical lines help to determine their position in space. Sample sizes are equal to Fig. 2 (see also Additional file 2: Table S1)

Fig. 4

Functional diversities were derived from the number and abundances of MEGAN-based EggNOG annotations. Functional richness between a host organisms and b host environmental groups based is displayed, as well as functional differences between hosts (c) and environmental groups (d). Non-metric multidimensional scaling is based on Bray-Curtis distances on the differences in functional composition between the host organisms displayed (c, d; see Table 3). Large symbols indicate the centroid of the respective groups. Functional variation of communities based on pairwise Bray-Curtis distances within host organism groups and environmental groups. Sample sizes for the host taxa are N = 5, except for D. melanogaster gut tissue (N = 10; see Additional file 2: Table S1)

Fig. 5

a Differences in the Nearest Sequenced Taxon Index (imputation success) between variable regions (average: Z = 0.3869, P = 0.7017, approximate Wilcoxon test; probability: odds ratio = 1.5941, P = 0.1402, Fisher’s test) and amplification method (Z = 0.0667, P = 0.9472, approximate Wilcoxon test; probability: odds ratio = 1.5511, P = 0.1436, Fisher’s test). b Procrustes correlation of imputed- and shotgun-based COG categories among different techniques, with significantly higher correspondence between imputed and measured functional profiles in the V1 V2 compared to the V3 V4 region (F1,18 = 7.8537, P = 0.0118, ANOVA). c Non-metric multidimensional scaling displays Bray-Curtis distances based on functional category abundances (COG categories) derived from PICRUSt (V1 V2/V3 V4, one-step/two-step) and shotgun-based approaches (MEGAN, single assembly). Ellipses represent standard deviations of points within the respective groups

Fig. 6

Multivariate correlation (Procrustes analyses) of phylogenetic distance among host organisms and community distances based on (a) shared presence among samples or (b) differences in abundance in 16S rRNA gene amplicon or shotgun-derived community profiles at different taxonomic cutoffs (Phylum to Genus, additional species level OTUs for amplicon-based profiles). Similar results are shown for the correspondence between phylogenetic distances among samples and their distances based on (a) shared presence or (b) abundance differences in their functional composition. The functional composition was derived from COGs and COG categories imputed from PICRUSt, EggNOG-derived COG categories and genes, and CAZY functions. All correlations are significant at P ≤ 0.05 (10,000 permutations)