Modularity and predicted functions of the global sponge-microbiome network

Abstract

Defining the organisation of species interaction networks and unveiling the processes behind their assembly is fundamental to understanding patterns of biodiversity, community stability and ecosystem functioning. Marine sponges host complex communities of microorganisms that contribute to their health and survival, yet the mechanisms behind microbiome assembly are largely unknown. We present the global marine sponge-microbiome network and reveal a modular organisation in both community structure and function. Modules are linked by a few sponge species that share microbes with other species around the world. Further, we provide evidence that abiotic factors influence the structuring of the sponge microbiome when considering all microbes present, but biotic interactions drive the assembly of more intimately associated 'core' microorganisms. These findings suggest that both ecological and evolutionary processes are at play in host-microbe network assembly. We expect mechanisms behind microbiome assembly to be consistent across multicellular hosts throughout the tree of life.

Fig. 1

Drivers of modularity of the global sponge–microbiome network. a Distribution of temperature (in °C) from sponge sampling sites across network modules identified at the whole network level. Thick line in the middle of boxes represents the median values; box limits are lower and upper quartiles; whiskers, 1.5× interquartile range; points, outliers. b Composition of host type (high- vs. low-microbial abundance sponges, HMA and LMA respectively) per module. c Composition of marine ecoregions from where the samples were collected within each module. Labels for the modules on the x axis in a apply to b and c. Data use to generate this figure are available as Supplementary Data 1

Fig. 2

Sponge phylogeny is associated with core network modularity. Sponge species belonging to the same module identified over the sponge core–microbiome network display the same colour. Stars, crosses and filled squares refer to sponge type according to whether they are high- or low-microbial abundance species (HMA and LMA respectively). NA = no information for host type is available. Only 144 out of the 156 host species considered in this study are shown because 10 of the original host species were not resolved to the species level and the sponge species Ircinia fasciculata and Strongylacidon conulosa did not have matches in the Open Tree of Life from which this tree was extracted (see Methods). For a complete list of sponges, their module membership and types see Supplementary Table 4

Fig. 3

Degree–ratio relationships for sponges in the core network. Degree (i.e. number of OTUs found within the host) vs. ratio of within-module links vs. across-module links of the sponge species in the sponge core–microbiome network, based on the core microbiome modularity. Each point corresponds to a sponge species. Different colours represent species memberships in the modular classification. Haliclona species are highlighted (with species name and a triangle) to illustrate the degree of heterogeneity, both in terms of OTUs and the degree of their intra-module connectivity, among low-microbial abundance species from the same genus

Fig. 4

Microbial abundance drives sponges’ network connectivity. Within-module degree (z) vs. participation coefficient (P) for the sponges the core network. Different colour classes show different sponge types, divided into high- and low-microbial abundant species (HMA and LMA, respectively). NA = no microbial abundance classification is available. LMA sponges generally share a greater number of OTUs with species in other modules than HMA ones, i.e. larger P (KS test: D = 0.57601, p-value < 0.001). However, the extent to which sponges tend to harbour a large fraction of OTUs from those found in their module is variable across both host types

Fig. 5

Modules of the core network are functionally distinct. Constrained distance-based redundancy analysis shows a clear separation of the functional profiles of sponges belonging to different network modules. Module was the sole constraining variable for this analysis and hence the second dimension comes from unconstrained ordination. The axis derived from the modular classification of the sponges (dbRDA1) explains 17.64% of the total variation, while the first axis of unconstrained ordination explains 38.78% of the variation. Module 0 (black dots) comprises original modules 1, 2, 5 and 6 in the sponge core–microbiome network, which were not found to be significantly different from each other by pairwise PERMANOVA comparison in terms of predicted metabolic functions