Multi-scale structure and geographic drivers of cross-infection within marine bacteria and phages

Abstract

Bacteriophages are the most abundant biological life forms on Earth. However, relatively little is known regarding which bacteriophages infect and exploit which bacteria. A recent meta-analysis showed that empirically measured phage-bacteria infection networks are often significantly nested, on average, and not modular. A perfectly nested network is one in which phages can be ordered from specialist to generalist such that the host range of a given phage is a subset of the host range of the subsequent phage in the ordering. The same meta-analysis hypothesized that modularity, in which groups of phages specialize on distinct groups of hosts, should emerge at larger geographic and/or taxonomic scales. In this paper, we evaluate the largest known phage-bacteria interaction data set, representing the interaction of 215 phage types with 286 host types sampled from geographically separated sites in the Atlantic Ocean. We find that this interaction network is highly modular. In addition, some of the modules identified in this data set are nested or contain submodules, indicating the presence of multi-scale structure, as hypothesized in the earlier meta-analysis. We examine the role of geography in driving these patterns and find evidence that the host range of phages and the phage permissibility of bacteria is driven, in part, by geographic separation. We conclude by discussing approaches to disentangle the roles of ecology and evolution in driving complex patterns of interaction between phages and bacteria.

Figure 1

Digitized version of the MN matrix with 286 hosts (rows) and 215 phages (columns) in the same orientation as originally published (Moebus and Nattkemper, 1981). The 1332 black cells represent positive interactions between hosts and phages (see Materials and methods). The connectance of the network (interactions/total size) is approximately 0.022≈1332/61490.

Figure 2

Network representation of the study. We observe 38 isolated components. Black nodes represent phages, and white nodes represent hosts. The station IDs of each host and phage are contained in the center of each node.

Figure 3

Modularity sorting of the network. We detect 49 modules (shaded rectangles). The 15 largest modules discussed in the main document begin at the left of the matrix. Black symbols represent those interactions within a module. Gray symbols represent those occurring between modules. The P-value for the observed modularity is smaller than 10⁻⁵.

Figure 4

Modular sort of the internal structure of the 15 largest modules, in the same order as they appear in Figure 3. The significance of modularity is denoted as follows: A/a=statistically modular/antimodular using Bernoulli null model, B/b=statistically modular/antimodular using probabilistic degree null model. X=no significant modular or antimodular.

Figure 5

Nestedness sort of the 15 largest modules. The gray line represents the isocline of the NTC algorithm. A/B=statistically nested using NTC and Bernoulli/probabilistic degree null model, C/D=statistically nested using NODF and Bernoulli/ probabilistic degree null model. X=no significance was found.

Figure 6

Geographical representation of the 15 largest modules. Each module is considered in a separate panel. Large filled circles represent the stations included in the corresponding module; open circles represent the stations not included in the corresponding module. Red and green small circles representing phages and bacteria, respectively, were randomly placed around their corresponding station for improved visibility. A gray line between a red and green circle denotes an interaction between a virus and bacteria.