The structure of temperate phage–bacteria infection networks changes with the phylogenetic distance of the host bacteria

Abstract

With their ability to integrate into the bacterial chromosome and thereby transfer virulence or drug-resistance genes across bacterial species, temperate phage play a key role in bacterial evolution. Thus, it is paramount to understand who infects whom to be able to predict the movement of DNA across the prokaryotic world and ultimately the emergence of novel (drug-resistant) pathogens. We empirically investigated lytic infection patterns among Vibrio spp. from distinct phylogenetic clades and their derived temperate phage. We found that across distantly related clades, infections occur preferentially within modules of the same clade. However, when the genetic distance of the host bacteria decreases, these clade-specific infections disappear. This indicates that the structure of temperate phage–bacteria infection networks changes with the phylogenetic distance of the host bacteria.

1. Introduction

Bacteriophages are shaping bacterial populations and community structures. Although lytic phage are parasitic, temperate phage can be either mutualistic or parasitic [1]. Upon infection they either lyse the cell, or integrate into the bacterial chromosome as prophage. In particular, temperate phage can horizontally transfer genes encoding for antibiotic resistance or virulence factors among bacteria, which critically alters their hosts' phenotype [2].

While we have insights into infection patterns between bacteria and lytic phage [3,4], our understanding of infection patterns between bacteria and temperate phage is still incomplete. Quantifying which temperate phage infect which bacteria is fundamental to understand how prophage-mediated horizontal gene transfer scales up to influence bacterial communities.

The family Vibrionaceae contains the greatest number of reported phage–host systems for the marine environment [5], with an estimated prevalence of 100% for temperate phage [6]. Thus, the genus Vibrio, which consists of 14 defined clades covering 58 species [7], is ideal to investigate infection patterns between temperate phage and bacteria across different phylogenetic scales.

In a previous phage–bacteria infection network (PBIN), we observed that temperate phage isolated from Vibrio alginolyticus could only infect bacteria from the same clade and not from a foreign clade [6]. However, that PBIN defined a pattern of relatively low taxonomic scales as it encompassed only two clades, which differed significantly in their sample sizes. If genetic differences between distinct phylogenetic groups of Vibrio limit the exchange of phage, we would expect a modular network, with nested patterns within each phylogenetic group once the number of different clades is increased [3,4]. We therefore hypothesized that lytic infections by temperate vibriophage occur preferentially within but not across clades and tested this hypothesis on two different phylogenetic scales by empirically generating two PBINs that differ in the genetic distance of the tested bacteria.

2. Methods

(a). Study organisms

We generated two threefold replicated PBINs using bacteria of the genus Vibrio and their derived temperate phage (electronic supplementary material, tables S1 and S2). One network contained strains from low-related Vibrio clades, with a maximum genetic distance of 0.04 (figure 1a) and their derived temperate phage and is called the low-relatedness network (LRN). By contrast, the genetic distance of the second network is about one order of magnitude shorter (genetic distance = 0.005, figure 1c). This network contained three different members of the splendidus-like clade and is called the high-relatedness network (HRN). Isolation and genotyping of all strains has been described elsewhere [8–11].

Figure 1.

Bayesian phylogenetic trees of the LRN (a) and the HRN (c). Modular and nested display of the LRN (b) and HRN (d). Rows and columns represent bacteria and phage. Only strains that are susceptible to at least one phage infection and phage that could infect at least one bacterium are shown. Coloured squares indicate infection success. Each colour corresponds to the clade of the hosts that are infected by a particular phage: alginolyticus clade, blue; fischeri clade, orange; splendidus-like clades in different shades of green, V. splendidus, dark green; V. crassostreae, mossy green; V. cyclitrophicus, neon green; other clades with fewer than three strains per clade, grey.

(b). Bacterial resistance/phage infectivity

Resistance/infectivity was measured by determining the reduction in bacterial growth rate (RBG) imposed by the phage, adapted from [12]; for details, see electronic supplementary material S1 Methods.

(c). Statistical analysis

Phylogenetic analyses were performed as described in [6]; for details, see electronic supplementary material S1 Methods.

(i). Network analysis

Infection data were processed in the form of binary matrices. The same analysis pipeline was used for both networks. Statistics were performed in R version 3.1.2. We first confirmed that the three technical replicates were not significantly different using a Mantel test. Subsequent nestedness and modularity analysis was performed on a consensus matrix (positive infection: infection occurs in at least two replicates) using the packages bipartite, Falcon and lpbrim [13,14].

For each phage, we performed a correlation between the genetic distance of its original hosts and the genetic distances of the hosts it could infect; for details, see electronic supplementary material S4.

(ii). Predicted probabilities

Based on each network, we estimated the probability for every phage to infect a strain from the same clade as the phage's host bacterium or a foreign clade, using a logistic regression with infection success as dependent variable and clade as independent variable. The complete model was analysed using a generalized linear model and an analysis of deviance for which we assumed deviance change to be approximately χ² distributed. Null results (i.e. bacteria not infected by any phage, and phage not able to infect any bacterium) were included in the analysis. Based on these obtained predicted probabilities (PP), we calculated the probability to infect a bacterium from the same relative to a foreign clade, according to the following formula: (PP(Same) − PP(Foreign))/PP(Foreign)_. We used the same analysis to predict the probability for each bacterium to be infected by a phage, from the same relative to a foreign clade.

3. Results

(a). Network structure

Genetic distances and infection patterns differed significantly between both networks. The LRN has a modular structure (Qbib = 0.62), with four distinct modules, out of which three could be assigned to either the alginolyticus, splendidus or fischeri clade (figure 1a,b). In support of this, the genetic distance of a phage's host bacterium and the bacteria that this phage can infect are significantly correlated (r² = 0.73, t_df=196 = 15.02, p < 0.001, electronic supplementary material, figure S4a). This indicates that phage that were derived from a specific clade are more likely to infect bacteria from the same than from distantly related clades. Module 4 contained strains from different clades that are susceptible to various phage irrespective of their clade origin, indicating that infections of distantly related Vibrios are possible but not very common.

The HRN shows neither a clade-specific modularity (Qbip = 0.33, electronic supplementary material, figure S3) nor a significant correlation between the genetic distance of the phage's original and the infected hosts (r² = 0.11, t_df=189 = 1.5, p = 0.13, electronic supplementary material, figure S4b). In contrast to the LRN, the HRN is significantly nested (table 1 and figure 1d). Similarly, nested structures appeared at smaller scales, i.e. at the level of each clade-specific module, within the LRN (table 1).

Table 1.

Nestedness statistics for the LRN, HRN and the three clade-specific modules of the LRN. NODF, nestedness measure based on overlap and decreasing fill.

network	matrix connectance	matrix fill	NODF nestedness score	p-value	normalized temperature
HRNa	0.064	337	35.82	0.001	2.02
LRN	0.004	124	18.38	0.053	1.11
alginolyticus modulea	0.68	54	85.7	0.001	1.38
splendidus module	0.54	42	42.16	0.91	0.83
fischeri module	0.43	26	49.17	0.08	1.22

^aSignificant networks.

(b). Relative infection/resistance probability

Across distinct phylogenetic clades, i.e. the LRN, the probability for a bacterium to be infected by a phage is higher when this bacterium belongs to the same phylogenetic clade as the phage's original host (figure 2). Similarly, phage are more likely to infect bacteria from the same clade as their original host, relative to bacteria from distantly related clades. However, at smaller phylogenetic scales (i.e. in the HRN) this pattern disappears.

Figure 2.

Probability of a bacterium being infected by a phage from the same relative to a phage from a distantly related clade (left). Probability of a phage infecting a bacterium from the same relative to a bacterium from a distantly related clade (right). Values of rP > 0: more infections by same. Values of rP < 0: more infections by distantly related. Light grey, LRN; dark grey, HRN.

4. Discussion

Based on two PBINs, which differ in the genetic distance of the original hosts, we tested the hypothesis that lytic infections by temperate vibriophage occur preferentially within rather than across phylogenetic clades. While we observed clade-specific modularity in the LRN, the structure of the HRN is significantly nested. This suggests that the structure of PBINs changes with the phylogenetic distance of the host bacteria and confirms our hypothesis that temperate phage are more likely to infect bacteria from the same clade as their original host, relative to bacteria from distantly related clades.

Clade-specific modularity suggests that either non-host resistance [15] or diversifying coevolutionary induced selection [4] resulted in largely independent Vibrio/phage communities across the Vibrio phylogeny, where genetic differences may limit the exchange of temperate phage between clades. By contrast, nested structures observed within the alginolyticus [6] and splendidus-like clades (present study) are considered to be a result of gene for gene coevolution [16]. Under this scenario, bacteria gain resistance to recently evolved phage by new mutations, thereby maintaining resistance to ancestral phage. Similarly, new mutations lead to phage-host range expansion without loss of the ability to infect ancestral bacteria.

Although temperate phage preferentially infect bacteria that belong to the same clade as their original host, we also observed that infections of distantly related bacteria are possible but not very common. This pattern was mainly driven by four strains (figure 1b). These strains were highly susceptible to most of the phage, irrespective of which original host the phage were isolated from. By contrast, we could not identify a single phage with an equally broad host range. We thus speculate that the observed resistance/susceptibility pattern is driven by divergent bacterial resistance mechanisms rather than by phage-host range. Future studies that identify resistance as well as infectivity mechanisms are needed.

Phage are not only important in shaping bacterial community composition, they also play a pivotal role in bacterial evolution through the movement of important genes among bacteria [2]. It becomes increasingly clear that this phage-mediated DNA transfer can occur across distantly related bacteria and even the apparently highly conserved 16S rRNA gene was found within the genome of a broad host range transducing phage [17]. It is thus paramount to better understand phage-host range and bacterial resistance to predict possible phage-driven horizontal gene transfer events across bacterial species, and thus the emergence of new (drug-resistant) pathogens.

Data accessibility

Data are available from PANGAEA (https://doi.org/10.1594/PANGAEA.889653).

Authors' contribution

C.C.W. and O.R. gained funding and interpreted the data; C.C.W. and H.G. conducted the experiments; C.C.W. devised the study, analysed the data and drafted the manuscript; all authors contributed to the manuscript, approved the final version of the manuscript and agree to be held accountable for the content therein.

Competing interests

We have no competing interests.

Funding

This study was supported by two grants of the Deutsche Forschungsgemeinschaft: WE 5822/1-1 and the Cluster of Excellence 80 ‘The Future Ocean’ given to C.C.W. and O.R.