Ecological and evolutionary dynamics of a model facultative pathogen: Agrobacterium and crown gall disease of plants

Summary

Many important pathogens maintain significant populations in highly disparate disease and non-disease environments. The consequences of this environmental heterogeneity in shaping the ecological and evolutionary dynamics of these facultative pathogens are incompletely understood. Agrobacterium tumefaciens, the causative agent for crown gall disease of plants has proven a productive model for many aspects of interactions between pathogens and their hosts and with other microbes. In this review, we highlight how this past work provides valuable context for the use of this system to examine how heterogeneity and transitions between disease and non-disease environments influence the ecology and evolution of facultative pathogens. We focus on several features common among facultative pathogens, such as the physiological remodeling required to colonize hosts from environmental reservoirs and the consequences of competition with host and non-host associated microbiota. In addition, we discuss how the life history of facultative pathogens likely often results in ecological tradeoffs associated with performance in disease and non-disease environments. These pathogens may therefore have different competitive dynamics in disease and non-disease environments and are subject to shifting selective pressures that can result in pathoadaptation or the within-host spread of avirulent phenotypes.

Introduction

The spread of pathogens through host populations reflects the interplay of many factors including host-microbe and microbe-microbe interactions as well as broader epidemiological and evolutionary dynamics (Johnson et al., 2015; Parratt et al., 2016). Pathogen life histories vary dramatically and accordingly help determine which factors are most important to the ecological and evolutionary dynamics of a particular infectious disease. Environmentally acquired infectious agents contend with the challenge of transitioning between highly disparate disease and non-disease environments (Sokurenko et al., 2006; Brown et al., 2012). Environmentally transmitted pathogens include those that reside within environmental reservoirs in dormant states such as spores or virions as well as facultative pathogens that are metabolically active in both disease and non-disease environments.

The environmental heterogeneity that these pathogens must navigate can be extreme. Host responses help shape the within-host success of microbes and each environment presents different nutrient availabilities and microbial communities. Transitioning between these environments typically requires that these pathogens remodel their physiology and behavior and involves a shift in evolutionary selective pressures acting on the population. This establishes opportunities for ecological tradeoffs in which traits that are beneficial in one environment may be detrimental in the other environment (Sokurenko et al., 2006; Brown et al., 2012; Mikonranta et al., 2012). As consequence, facultative pathogens may share a number of general features associated with the similarities in their life histories. The importance of the transitions between environmental and host environments largely depends on tight regulatory control of key attributes including virulence functions, motility and surface colonization phenotypes and is shaped by interactions with host defenses and features of the host environment, including the host’s microbiome.

The generalist plant pathogen Agrobacterium tumefaciens, the causative agent of the neoplastic plant disease crown gall, has been an important model bacterium yielding key insights into host-microbe signaling (Venturi and Fuqua, 2013), bacterial cell-to-cell communication (Lang and Faure, 2014) and virulence mechanisms (Christie et al., 2005; Nester, 2015). Further, A. tumefaciens is a prominent biotechnology tool because of its ability to drive horizontal gene transfer to plants (Hwang et al., 2015b) and an economically important pathogen of several plant varieties (Pulawska, 2010). A. tumefaciens pathogenesis and intraspecific interactions have been studied extensively (Platt et al., 2014) and the molecular processes underlying crown gall disease are established (Figure 1). In this review, we examine how this foundational knowledge facilitates the use of A. tumefaciens as a valuable model system for studying the ecological and evolutionary dynamics of diseases caused by facultative pathogens. We will focus our discussion on factors related to the transition between disease and non-disease environments as this defines a key challenge shared by all facultative pathogens. In particular, we will use agrobacterial pathogens as case studies to discuss the importance and consequences of motility and surface colonization phenotypes, interactions with hosts and within host environments, and the consequences of ecological tradeoffs associated with the facultative pathogen life history.

Figure 1.

Facultative pathogens experience disparate ecological factors and evolutionary pressures associated with disease and non-disease environments. Transitions between these environments (double-headed curved arrows) are critically important to the ecological and evolutionary dynamics of these pathogens. In the case of Agrobacterium tumefaciens (white ovals), non-disease environments include soil reservoirs and uninfected host rhizospheres, both of which vary dramatically from the environment of diseased plant rhizospheres or actively infected galls (light brown mass in upper root system). Each environment is associated with a unique microbiome (multicolored ovals) allowing for key differences in the positive and negative microbial interactions that A. tumefaciens experiences in these environments. In addition, the need to colonize plant surfaces (e.g. via the unipolar polysaccharide or UPP, red in inset) and antagonism from plant defenses are key features of host plant environments. Within host environments, important host-microbe interactions also include the utilization of rhizosphere exudates, including those (filled yellow circles) whose catabolism (curved blue arrow pointing to split yellow circles that represent catabolized rhizosphere exudates) is conferred by the At plasmid (blue open circle), and pathogenesis conferred by the Ti plasmid (open circle with purple, orange, and gray regions). The T-DNA (purple region) found on the Ti plasmid is delivered via a type IV secretion system (gray cylinder) encoded by the vir-region (gray region) of the Ti plasmid into the host plant cell (arrow through gray cylinder). This allows for T-DNA insertion into the plant genome stimulating tumorigenesis via expression of oncogene products (filled yellow triangles) and opine (filled purple circles) production (curved arrows originating at integrated T-DNA). Opine uptake and catabolism functions (curved orange arrow pointing to split purple circles that represent catabolized opines) are conferred by the opine catabolic region (orange) of the Ti plasmid. Within the disease environment, the pathogen also faces potential exploitative competition over opines and other plant exudates from other microbes able to use these resources (e.g. arrow representing opine uptake by blue genotype) as well as potential interference competition (dashed, blunt line between white and blue genotypes).

Ecology shapes pathogen diversification

Agrobacterial diversity

Pathogenic agrobacteria are a highly diverse polyphyletic group that includes the A. tumefaciens species complex (formerly biovar 1), A. rhizogenes (formerly biovar 2), and A. vitis (formerly biovar 3) (Otten et al., 2008). In each of these groups, a virulence plasmid horizontally transferable to other bacteria confers most pathogenesis functions (Platt et al., 2014). Because this complicates taxonomic classification, the pathogenic agrobacteria and related bacteria have been the subject of ongoing taxonomic revision (Farrand et al., 2003; Young et al., 2003; Ormeno-Orrillo et al., 2015). For convenience, we will use a commonly used classification system based on how these microbes interact with plant hosts. Each group of pathogenic agrobacteria either elicit a different pathogenesis phenotype in plants or vary in their plant host range. A. tumefaciens causes crown gall disease and can infect most dicotyledonous plants. A. vitis pathogenesis also causes gall formation but its host range is restricted to grapes and it additionally triggers necrosis of grape roots. A. rhizogenes does not cause gall formation, instead causing hairy root disease (Escobar and Dandekar, 2003).

Of these, the majority of research has centered on A. tumefaciens and thus the discussion in this review will focus on this taxon, often incorporating discussion of relevant information about other agrobacteria. A. tumefaciens possesses a multi-partite genome, generally consisting of a circular chromosome (> 2 Mb) and another large (> 2 Mb) chromid replicon (Slater et al., 2009; Harrison et al., 2010). The chromid of strains belonging to the A. tumefaciens species complex are linear molecules, while the chromids of A. rhizogenes and A. vitis are circular molecules (Slater et al., 2013; Ramirez-Bahena et al., 2014). The genome may also include one or more plasmids (Goodner et al., 2001; Wood et al., 2001; Huang et al., 2015). The tumor-inducing or Ti plasmid (approximately 200 kb and 4% of the genome) encodes most virulence functions and imparts the ability to initiate crown gall disease. The similarly sized root-inducing or Ri plasmids found in A. rhizogenes strains correspondingly encode most functions associated with hairy root disease. The Ti and Ri plasmids are similar in many aspects of their plasmid biology and the mechanisms through which they confer the ability to infect host plants (Suzuki et al., 2009; Platt et al., 2014). The regulation and mechanisms of agrobacteria pathogenesis are well studied and have been the topic of several recent review articles (Gelvin, 2012; Subramoni et al., 2014).

Ti and Ri plasmids are each highly diverse with a mosaic structure containing several clusters of high conservation interspersed with more divergent sequence, reflecting an evolutionary history marked by significant horizontal gene transfer, deletions, and other large-scale genetic events (Otten et al., 2008; Suzuki et al., 2009). Reflecting this plasmid diversity, pathogenic agrobacteria are organized into groups based on variations associated with their Ti or Ri plasmids. Agrobacterial infection causes plants to produce a specific set of opines that varies depending on the particular Ti or Ri plasmid (Figure 1). Accordingly, Ti plasmids have been organized into at least 10 groups based on the sets of opines they stimulate plants to produce during infection (Dessaux et al., 1998). Further illustrating the diversity of Ti and Ri plasmids is the observation that they belong to at least four incompatibility groups—IncRh1, IncRh2, IncRh3, and IncRh4 (Otten et al., 2008). Plasmid incompatibility groups are defined as sets of plasmids that are not stably inherited together when they co-occur within the same genome. In the case of Ti and Ri plasmids, incompatibility reflects significant similarities in the repABC region of the plasmid which encodes functions associated with the replication and partitioning of the plasmid (Pinto et al., 2012). Remarkably, one extreme example of incompatibility is a Ti plasmid that has two repABC regions and is incompatible with both IncRh1 and IncRh2 plasmids (Yamamoto et al., 2017).

The genomes of A. tumefaciens often include another megaplasmid designated the At plasmid (often over 500 kb; ~10% of the genome). The functions conferred by these plasmids are less well defined, but they are known to impart a number of metabolic activities associated with rhizosphere environments thereby facilitating plant colonization (Baek et al., 2005; Chai et al., 2007; Haudecoeur et al., 2009; Morton et al., 2014). Consistent with this, pathogenic agrobacteria bearing both the Ti and At plasmids outcompete cells bearing only the Ti plasmid in the rhizosphere of actively infected hosts; in contrast, they decline in frequency following plant senescence (Morton et al., 2014). Like the Ti plasmid, the At plasmid also exhibits dramatic variation across A. tumefaciens strains.

Agrobacterial diversity is, in part, shaped by selective pressures imposed by host plant environments. Reflecting this, some A. tumefaciens and A. rhizogenes strains exhibit relatively narrow host ranges and harbor Ti plasmids larger than and lacking T-DNA region homology with the majority of described Ti plasmids (Unger et al., 1985). Examination of different strains revealed that large scale genomic events (e.g. deletions associated with transposition events) can result in dramatic narrowing of Ti plasmid plant host range, allowing for the rapid emergence of variants with limited host range (Paulus et al., 1991; van Nuenen et al., 1993). Another example illustrating the impact of adaptation to host environments on agrobacterial diversity is found with many strains of A. vitis harboring a separate tartrate utilization plasmid that confers a within host benefit stemming from the ability to catabolize tartrate, a nutrient present at high concentrations on their grapevine hosts (Szegedi et al., 1992; Salomone et al., 1998; Burr and Otten, 1999). Further, strains belonging to the genomovar G8 within the A. tumefaciens species complex carry a number of genes specific to this group. Strain C58 is within genomovar G8, and genome analysis reveals that these specific genes are located within seven genomic islands and include functions related to interaction with plants such as the ability to catabolize plant exudates and plant degradation products (Lassalle et al., 2011).

Transitions between disease and non-disease environments

Motility, chemotaxis, and vir-induction

Like other facultative pathogens, agrobacteria experience a wide range of environments. These include soil reservoirs, the rhizosphere of plants not manifesting crown gall disease, and the rhizosphere of plants with crown gall disease. For simplicity, we will refer to bulk soil and uninfected plant rhizosphere environments as non-disease environments and the rhizospheres of infected plants as disease environments. Motility and chemotaxis via flagellar locomotion is important for the early steps of host colonization and virulence in A. tumefaciens (Shaw et al., 1988; Chesnokova et al., 1997). A. tumefaciens encodes chemotaxis receptors for a variety of monosaccharides and sugar acids, many of which that are exuded from plant host cells (Wright et al., 1998). Several Ti plasmids, and the pAtK84b opine catabolic plasmid carried by the avirulent A. rhizogenes strain K84, encode functions for chemotaxis toward some of the opines that they confer the ability to catabolize (Kim and Farrand, 1998). Plant produced phenolics (e.g. acetosyringone and catechol) activate Ti plasmid virulence (vir) gene expression through the VirA-VirG two component system, and also stimulates chemotaxis towards plant tissues (Bolton et al., 1986; Shaw et al., 1988). A periplasmic binding protein, ChvE, interacts with plant-released sugars and sugar acids and potentiates vir-gene expression through VirA-VirG (He et al., 2009; Hu et al., 2013). Acidic conditions also enhance vir-gene expression partly through another periplasmic regulator, ExoR, which is proteolytically destabilized at low pH, thereby releasing the ChvG-ChvI two component system to activate virG expression (Heckel et al., 2014). Via these mechanisms, multiple signaling pathways orchestrate virulence initiation through finely coordinated virulence gene expression upon recognition of host-associated signals.

Attachment and biofilm formation

Following motility and chemotaxis-mediated association with host tissues, many pathogens adhere to and form biofilms on host tissues in order to initiate virulence functions. Initial surface contact can be mediated by pili as found in a broad range of pathogens (Strom and Lory, 1993; Craig et al., 2004; Craig and Li, 2008). A. tumefaciens contains a gene cluster encoding Type IVb pili (Ctp pili) that promote weak, reversible surface interactions (Wang et al., 2014). Homologous loci found in other bacteria are also thought to encode pili that drive initial, non-specific attachment to surfaces (Kachlany et al., 2000; Goodner et al., 2001; Tomich et al., 2007). Reversible surface association via the Ctp pili in A. tumefaciens likely facilitates the transition into permanent attachment and biofilm maturation through production of exopolysaccharides, most notably a polar adhesin called the unipolar polysaccharide (UPP), which is deployed upon contact with biotic and abiotic surfaces (Figure 1; Tomlinson and Fuqua, 2009; Li et al., 2012; Xu et al., 2012). A. tumefaciens produces a number of other exopolysaccharides, but only the UPP and cellulose have been proposed to play a direct role in adhesion (Matthysse, 1983; Matthysse et al., 2005). Other polysaccharides, including succinoglycan, cyclic β-1,2-glucans, and curdlan have variable and inconsistent roles in attachment and biofilm formation (Heindl et al., 2014). Regulation of attachment in A. tumefaciens is affected by environmental conditions such phosphorus limitation, oxygen availability, low pH, and divalent cations such as Fe and Mn (Danhorn et al., 2004; Ramey et al., 2004; Tomlinson et al., 2010; Xu et al., 2012; Xu et al., 2013; Heckel et al., 2014; Heindl et al., 2014; Feirer et al., 2015). Many of these signals appear to modulate UPP and cellulose production via effects on the bacterial second messenger cyclic diguanylate monophosphate (cdGMP) (Amikam and Benziman, 1989; Barnhart et al., 2013; Xu et al., 2013; Barnhart et al., 2014; Feirer et al., 2015). To what extent these factors modulate attachment during pathogenesis of the host remains poorly understood.

Navigating the host and its defenses

Host physiology and defenses are key aspects of the within host environment of any pathogen. Plants often activate innate immune responses when their pattern recognition receptors (PRRs) bind microbe- or pathogen associated molecular patterns (MAMPS or PAMPS), which are associated with highly conserved microbial features such as flagellin, elongation factor Tu, peptidoglycan, and lipopolysaccharides (Newman et al., 2013). Though flagellin is one of the most common PAMPs that plants respond to, the Arabidopsis flagellin PRR fails to elicit a response to flagellin from Agrobacterium (Bauer et al., 2001). However Arabidopsis does mount a defense in response to peptidoglycan from Agrobacterium (Erbs et al., 2008) and when its elongation factor PRR detects Agrobacterium elongation factor (Zipfel et al., 2006).

In contrast to many phytopathogens, most plant hosts fail to activate a hypersensitive response (HR) to A. tumefaciens. This is thought to reflect the fact that A. tumefaciens employs several mechanisms to interfere with host defense systems. For example, A. tumefaciens produces a catalase that prevents H₂O₂ accumulation during the early stages of pathogenesis of Arabidopsis thaliana, thereby interfering with HR and other host defensive responses (Xu and Pan, 2000; Lee et al., 2009). The phenolic inducers that activate the VirA-VirG two component system coordinating the expression of pathogenesis functions are themselves toxic and may thereby contribute to the plant’s defense against agrobacteria. However, VirH2 expressed following vir-induction detoxifies most phenolic inducers and even allows for the catabolism of some of these phenolics (Brencic et al., 2004).

Much of the work in this area has focused on interactions between A. thaliana plants and a generally narrow subset of A. tumefaciens taxa. Future work should evaluate how well these interactions represent those occurring on other hosts and with other agrobacteria. For a more thorough discussion of the interactions between A. tumefaciens and plant defenses, as well as how A. tumefaciens impacts hosts tissues during the development of crown gall disease, we recommend two recent review papers (Gohlke and Deeken, 2014; Hwang et al., 2015a).

Consequences of infection—virulence plasmid conjugation

The complex regulatory network underlying the horizontal genetic transfer (HGT) of Ti plasmids involves responses to host cues and quorum sensing signaling (Lang and Faure, 2014). Following transformation of plant cells with T-DNA (Figure 1, Gelvin, 2012), the resulting release of opines from the infected plant tissue is perceived through transcription activators (e.g. AccR, NocR, or OccR) that specifically form complexes with specific opines, and initiate gene expression for catabolism of that specific opine (Figure 1; von Lintig et al., 1991; von Lintig et al., 1994; Kim and Farrand, 1997; Subramoni et al., 2014).

A subset of opines, the so-called conjugative opines, also lead to elevated expression of traR, the primary activator of Ti plasmid conjugative transfer between bacteria (White and Winans, 2007). TraR is a LuxR-type quorum sensing regulator that recognizes the acylated homoserine lactone (AHL) quorum-sensing molecule, N-(3-oxooctanoyl)-L-homoserine lactone (3OC8-HSL), synthesized by another Ti plasmid gene product called TraI, a LuxI-type AHL synthase. TraR-AHL complexes promote transcription of the conjugative transfer genes for the Ti plasmid, tra/trb, including the traI gene, thereby creating a positive feedback induction loop (Piper et al., 1993; Zhang et al., 1993; Fuqua and Winans, 1994; Hwang et al., 1994). TraR-AHL complexes also increase Ti plasmid copy number through activation of the repABC genes (Fuqua and Winans, 1996; Pappas and Winans, 2003; White and Winans, 2007; Pinto et al., 2012). Quorum-dependent regulation of conjugation is further controlled through an antiactivator, TraM, which directly inhibits activated TraR (Hwang et al., 1999). In some cases non-conjugative opines can activate expression of defective, truncated TraR paralogues (called TraS or TrlR) which also can inhibit activation of conjugative transfer genes (Oger et al., 1998; Zhu and Winans, 1998).

These complex gene regulatory mechanisms ensure that Ti plasmid conjugation rates and copy number increase in response to opine availability and the density of the agrobacterial population. This intricate control reflects the complex evolutionary dynamics at work in controlling the Ti plasmid copy number and HGT. Experimental evolution of a strain that constitutively expresses the quorum sensing regulon under lab conditions results in the spread of mutants that minimize the costs of expressing quorum sensing regulated functions (Tannieres et al., 2017). This result illustrates that there are significant costs associated with expression of these systems. The availability of opines likely counterbalances Ti plasmid costs, including those associated with conjugation, and provides a fitness benefit to harboring the otherwise costly plasmid. These conjugation events may result in novel plasmid-host background combinations when the Ti plasmid is delivered to plasmidless agrobacteria. Alternatively, Ti plasmids may be delivered into genotypes that already contain a Ti plasmid. In the case of similar Ti plasmids, a quorum sensing regulated entry exclusion system may reduce the efficiency of these events (Cho et al., 2009). However, when conjugation does result in co-resident Ti plasmids there are several possible outcomes that are potentially significant to the evolution of Ti plasmids. First, the co-resident plasmids may separate due to incompatibility which could result in novel plasmid-host background combinations (Cho et al., 2009). Second, the plasmids may exchange genetic regions or co-integrate via homologous recombination. Such events, coupled with subsequent transposition and large scale deletion events may help account for the chimeric nature and high degree of diversity exhibited by Ti and Ri plasmids (Otten et al., 1992; Otten et al., 2008).

Additionally, other host factors may modulate HGT of virulence plasmid transmission in A. tumefaciens. Plant-produced γ-amino butyric acid (GABA) accumulates in acidic wound conditions (Figure 1) and activates the At plasmid blcABC operon after being taken up through the proline/GABA receptor and ABC-transporter (atu2422 and braE (atu2427), respectively) and converted to semi-salicylic acid (SSA) (Chevrot et al., 2006). Salicylic acid (SA), a plant defense molecule, has also been shown to reduce vir expression and activates expression of blcABC genes (Yuan et al., 2007; Yuan et al., 2008). BlcC is a γ-butyrolactonase that is capable of degrading the 3OC8-HSL and is proposed to be involved in quorum quenching (Khan and Farrand, 2009; Haudecoeur and Faure, 2010; Lang and Faure, 2014). Other components in the pathway from γ-butyrolactone (GBL) to succinate, including GBL, γ-hydroxybutyrate (GHB), and semi-salicylic acid (SSA) also increase blcABC gene expression (Haudecoeur and Faure, 2010; Lang and Faure, 2014; Subramoni et al., 2014). It is possible that quorum quenching in A. tumefaciens may provide a competitive advantage. In addition to enabling metabolism of plant-release GBL for energy, this system would generally reduce the energy expenditure due to quorum signaling by A. tumefaciens and other proximal bacteria, (Chevrot et al., 2006; Chai et al., 2007; Yuan et al., 2007; Subramoni et al., 2014), although there has been some disagreement about on the strength of these effects (Khan and Farrand, 2009).

Consequences of ecological tradeoffs

Eliciting host pathogenesis is a costly behavior that can greatly reduce a pathogen’s cellular growth rates and fitness (Ackermann et al., 2008; Platt et al., 2012a; Morton et al., 2014; Peyraud et al., 2016). Each episode of host infection results in a dramatic shift in the pathogen’s biotic and abiotic environment. This requires a physiological reprogramming, shapes selective pressures acting on pathogen populations, and can alter the microbial community dynamics (Figure 2). As an example of this, the endophytic bacterial community of grapevine galls caused by A. vitis is distinct from the community found within analogous healthy tissues, suggesting that opine availability results in a shift in the bacterial community (Faist et al., 2016). Pathogens must deal with host defenses as well and inter- and intra-specific microbial competition, the burden of which can severely limit metabolic potential. Numerous strategies have evolved to promote the maintenance of burdensome pathogenic functions across the wide variety of niches that facultative pathogens occupy.

Figure 2.

Facultative pathogens often infect hosts from environmental populations or reservoirs (leftmost curved arrow). As a consequence, pathogenesis results in a dramatic shift in the pathogen’s biotic and abiotic environment. In this highly simplified schematic, each color represents a different microbial genotype. These may be resident microbiota, microbes that subsequently colonize the host, or novel genetic variants arising via genetic changes (e.g. mutation or horizontal gene transfer). The space defined by the solid curvy lines represents the within-host environment, while the space outside this represents the pathogen’s environmental reservoir. Dynamic selective pressures stemming from changes in the competitive environment, environmental conditions, and host responses can result in shifts within host-associated microbial populations and communities as the disease progresses (arrows exclusively inside the within-host environment). At any point over the course of the infection there is potential for shedding of the host-associated microbiota into the environmental reservoir (curved arrows exiting the within-host environment).

Tradeoff between growth and pathogenesis

The expression of virulence functions is often very costly leading to reduced vegetative growth rates, thereby establishing a tradeoff associated with pathogenesis (Maharjan et al., 2013; Ferenci, 2016; Peyraud et al., 2016). Many pathogens limit these costs by tightly regulating the expression of pathogenesis functions, thereby minimizing costs and promoting the evolutionary maintenance of these traits. It is common for pathogens to regulate virulence functions in a density-dependent manner, assuring a sufficient number of pathogenic bacteria to cause disease (Gama et al., 2012). The plant pathogen Ralstonia solanacearum coordinates virulence gene expression through a quorum-dependent global regulator, phcA (Peyraud et al., 2016). Expression of phcA has been shown to reduce fitness and metabolic potential, but it is essential for successful invasion of host xylem tissue (Peyraud et al., 2016). Thus for this and other systems there is a selective pressure to curtail expression of costly virulence functions, but through the coordination afforded in a quorum-regulated system, the majority of cells in the population concertedly participate in the behavior.

Other pathogens, such as Salmonella typhimurium, initiate virulence functions through phenotypic noise, where a small, isogenic population, to their own detriment, stochastically expresses costly, virulence behaviors that benefit a surrounding infecting population (Ackermann et al., 2008). In this way, a subset of the population can incur costs of virulence to the benefit of a faster-growing subpopulation that does not activate virulence (Ackermann et al., 2008; Diard et al., 2013). While some of the individuals incur the often heavy burden of virulence initiation, the larger population is mostly unaffected and can initiate subsequent virulence expression because of retention of unexpressed virulence traits.

In A. tumefaciens, virulence induction significantly reduces fitness by dramatically decreasing population growth rates (Platt et al., 2012b). The VirA-VirG two component system accordingly limits this costly expression to the host environment based on a range of plant rhizosphere cues, including phenolic compounds, specific sugars, acidic conditions, and low phosphorus levels (Venturi and Fuqua, 2013). Further, the initial consequences of virulence induction are seemingly offset by the availability and metabolic access to opines that are produced as a result of successful infection (Guyon et al., 1993; Platt et al., 2012a). However, the heavy cost of vir-induction establishes a strong selective pressure favoring the spread of avirulent mutants (Platt et al., 2012a). This likely accounts for several observations of mutants unable to infect hosts spreading under in vitro vir-inducing conditions (Fortin et al., 1992; Fortin et al., 1993) and in planta (Belanger et al., 1995; Llop et al., 2009). The results of Llop et al. (2009) suggest that these mutants more often colonize the diseased host from environmental reservoirs rather than arising from de novo mutation in the infecting strain.

In addition to the large cost associated with expression of virulence genes, there is also a significant but comparatively smaller carriage cost to harboring a Ti plasmid when carbon or nitrogen are limiting resources (Platt et al., 2012b). Because of this, non-disease environments, wherein opines are absent, are predicted to be a sink due to competition from saprophytic agrobacteria or other competitors (Platt et al., 2012a). For example, Krimi et al. (2002) observed a decrease in the frequency of pathogenic genotypes within the agrobacterial population associated with the absence of infected plants. Experimental support for this idea comes from the observation that pathogenic agrobacteria are rapidly outcompeted by non-pathogenic agrobacteria when coresident in a rhizosphere without opines. Conversely, pathogenic genotypes outcompete the non-pathogenic agrobacteria when competing in a rhizosphere into which opines are excreted (Guyon et al., 1993). The consequences of non-disease environments acting as a sink for pathogenic agrobacteria likely has a profound effect on the epidemiology of crown gall disease.

Competition within plant microbiomes

Once pathogens stably colonize a host, interference competition between the pathogen and members of the microbiome can affect the ability of the pathogen to maintain its host-associated niche (Figures 1 and 2; Ghoul and Mitri, 2016). Interference competition can be mediated by antagonistic behaviors such as Type VI secretion systems (T6SSs) and production of antimicrobials (Massey et al., 2004; Basler et al., 2013; Kapitein and Mogk, 2013; Ghoul and Mitri, 2016). Initiation of competitor killing may be beneficial during initial host colonization to compensate for decreased fitness due to expression of costly virulence functions. It is likely important for A. tumefaciens to stabilize its niche in the face of intense competition among rhizosphere microbiota. Among these mechanisms, A. tumefaciens activates a T6SS under acidic conditions through the ChvG-ChvI two component system (Wu et al., 2012; Heckel et al., 2014). Interestingly, pairwise competition with the opportunistic human pathogen Pseudomonas aeruginosa hinges on the A. tumefaciens T6SS. The A. tumefaciens T6SS invokes lethal retaliation by P. aeruginosa (via its own T6SSs) under laboratory conditions, but results in the dominance of A. tumefaciens over P. aeruginosa in planta (Ma et al., 2014). Other factors also influence this interaction, such as production of diffusible molecules that can inhibit growth or disperse established A. tumefaciens biofilms (Barreteau et al., 2009; Hibbing and Fuqua, 2012; Ghequire et al., 2017).

In some cases, agrobacterial infection may indirectly result in host environment conditions that are unfavorable to other microbes. For example, tomato expression of the A. rhizogenes oncogene rolB results in both stimulation of meristem formation associated with the production of adventitious, “hairy” roots (Britton et al., 2008) and increased resistance to the foliar fungal pathogens Alternaria solani and Fusarium oxysporum (Arshad et al., 2014). Pathogens may also outcompete rival microbes in host environments via exploitive competition, wherein resource consumption limits the availability of that resource for its competitors (Figures 1 and 2; Ghoul and Mitri, 2016). For example, production of siderophores could increase competitiveness in low iron environments, and increased motility and nutrient sequestration could limit nutrient availability to competitors (Ghoul and Mitri, 2016; Niehus et al., 2017). A. tumefaciens engineers a competitive nutritional advantage during plant infection through the production of opines (Figure 1), the catabolism of which is also predominantly encoded on the Ti plasmid (Kim et al., 2008; Platt et al., 2012a). Strains of a few other species of rhizosphere bacteria are also able to catabolize opines (Guyon et al., 1993; Moore et al., 1997; Brencic and Winans, 2005; Farrand et al., 2007). Consequently, pathogenic agrobacteria face interspecific competition for the utilization of opine resources within the disease environment. In addition, pathogenic agrobacteria compete for access to opines with other agrobacteria that harbor different Ti plasmids or other plasmids conferring the ability to catabolize opines (Lang et al., 2017). The biocontrol agent A. rhizogenes K84 provides a particularly dramatic example. This avirulent, opine catabolic strain produces several antimicrobials that target virulent nopaline type A. tumefaciens and A. rhizogenes strains (Penyalver et al., 2001; Kim et al., 2006; Platt et al., 2014).

A number of other potential biocontrol strains influencing the establishment of crown gall disease have been identified. For example, A. vitis F2/5 can inhibit the establishment of crown gall disease in grapevine hosts, but fails to do so on other host plants such as tobacco (Kaewnum et al., 2013). This strain is non-tumorigenic but causes necrosis of grape tissues (Zheng and Burr, 2016) and, similar to the biocontrol agent A. rhizogenes K84, A. vitis F2/5 carries an opine catabolic plasmid (Szegedi et al., 1999). However unlike K84, F2/5 does not appear to produce a bacteriocin antagonizing the growth of tumorigenic A. vitis strains (Burr et al., 1997). The mechanism of F2/5’s grape tumor inhibition is not fully characterized, however, appears to depend on this strain producing a nonribosomal peptide synthetase product that interferes with host transformation by tumorigenic A. vitis strains (Kaewnum et al., 2013; Zheng and Burr, 2016). In contrast, the mechanism of inhibition by other potential biocontrol agents of agrobacterial pathogens does appear to involve the production of bacteriocins. These include A. vitis E26 (Wei et al., 2009) and A. vitis VAR03-1, both of which produce bacteriocins that inhibit A. tumefaciens, A. rhizogenes, and A. vitis strains (Kawaguchi et al., 2008; Li et al., 2009).

Several non-agrobacterial rhizobacteria are also potential biocontrol agents inhibiting the establishment of crown gall disease. For example, inoculation of roots with Bacillus subtilis BSCH14 prior to planting in the field resulted in significant inhibition of crown gall disease incidence (Rhouma et al., 2008). Further, another study identified that a number of different rhizobacteria, including Pseudomonas putida UW4 and Burkholderia phytofirmans PsJN, that produce the enzyme 1-aminocyclopropane-1-carboxylate deaminase (ACCD) antagonize crown gall disease development on tomato plant hosts by both A. tumefaciens Sh-1 and A. vitis S4 (Toklikishvili et al., 2010).

Pathogenic agrobacteria compete with several other microbes over host produced opines. Other bacterial and even fungal taxa have also evolved the ability to utilize opines in addition to related agrobacteria and rhizobia (Beauchamp et al., 1990; Bell et al., 1990; Bergeron et al., 1990). Therefore, although opines are custom nutrients produced almost exclusively as a result of infection by pathogenic agrobacteria, they are not an exclusive private good. The degree to which opine utilization influences the competitiveness of A. tumefaciens during infection and whether spatial or other factors are important for niche stabilization remains to be experimentally addressed.

Selection for avirulence—resisting invasion by cheaters through HGT

Pathogenesis initiates a shift in the selective pressures acting on pathogens and can influence the ecological dynamics of infecting populations (Figure 2). Especially in cooperative systems, there exist selective pressures that act to disfavor organisms from participating in virulence, resulting in a rise of cheaters with diminished virulence phenotypes that can outcompete the virulent progenitor (Gama et al., 2012). For example, loss of cooperative siderophore production occurring during chronic Pseudomonas infections has been attributed to the competitive social interactions fostered within the host environment (Andersen et al., 2015). Similarly, avirulent variants that arise during S. typhimurium infections have a selective advantage during host colonization due to the loss of costly expression of virulence functions and cooperation (Ackermann et al., 2008; Diard et al., 2013). In the face of these negative selective pressures, cooperative traits must be supported or reinforced to allow for persistence across generations and subsequent infection cycles.

Cooperative pathogenesis genes can be stabilized by the dissemination of pathogenicity genes via HGT (smith, 2001; Dimitriu et al., 2014). There are numerous mechanisms by which bacterial pathogens can horizontally acquire novel genetic material, including conjugation, transformation, transduction, and gene transfer agents (GTAs). Among these mechanisms, conjugation is arguably the most common and a stable mechanism to faithfully regulate and coordinate dissemination of virulence genes (Cabezon et al., 2015; von Wintersdorff et al., 2016). Conjugative elements that contain pathogenicity functions include integrative and conjugative elements (ICEs) and conjugative plasmids. ICEs were initially discovered due to their ability to confer resistance to antibiotics and heavy metals, but pathogenicity and symbiotic functions are also frequently contained within these elements (Johnson and Grossman, 2015). For example, Pseudomonas aeruginosa contains a pathogenicity island, PAPI-1, that also encodes its own conjugative machinery (Carter et al., 2010). Similarly, the symbiotic island, ICEMlSym^R7a allows members of the genus Mesorhizobium to form nodules on Lotus species and is self-conjugative (Ramsay et al., 2006).

More commonly pathogenicity functions are carried on self-transmissible plasmids that contain numerous virulence genes as well as maintenance and replication functions, and these elements can be quite large (Sengupta and Austin, 2011). Very tight regulation of copy number, gene expression, and plasmid maintenance limits the cellular burden of plasmids, as well as promoting proper segregation and dissemination. In several cases, unstable plasmid maintenance may be compensated by increased interspecies HGT, as seen in recent experiments in Pseudomonas (Hall et al., 2016). In A. tumefaciens, coordination of HGT with the availability of opines and the density of the virulent population (via quorum sensing) may provide a means to ensure delivery and stabilization of the plasmid in avirulent cells that colonize or arise via mutation within the disease environment. Dissemination of the Ti plasmid in the presence of opines may also stabilize the Ti plasmid by allowing the plasmid to colonize genetic backgrounds that can outcompete the infecting strain. Opine-derived fitness benefits are thought to be critical for the persistence of cooperation and stabilization of virulence in A. tumefaciens populations (Platt et al., 2012a). However, because of the diversity in plasmid types and the importance of spatial and temporal variation, the potential value of Ti plasmid conjugation in promoting the persistence of pathogenic agrobacteria within natural disease environments remains elusive.

Concluding Remarks

The plant pathogen A. tumefaciens has proven a valuable system for the study of host-microbe and intraspecific microbe-microbe interactions (reviewed by Venturi and Fuqua, 2013; Kado, 2014; Lang and Faure, 2014; Hwang et al., 2015a). Extensive work has also examined key aspects of its virulence plasmid biology (Platt et al., 2014) and colonization of host and non-host surfaces (Heindl et al., 2014). In this review, we argue that the biology of this pathogen and the foundation of mechanistic understanding positions A. tumefaciens well as a model for the ecology and evolution of infectious diseases caused by facultative pathogens. An understanding of the ecology and evolution of pathogens with this life history necessarily requires inclusion of their dynamics within and the linkages between populations residing within these disparate disease and non-disease environments. Towards that goal, future research should address some current knowledge gaps related to how A. tumefaciens behaves in its environmental reservoirs (e.g. soil and commensal states in plant rhizospheres) as well as how it transitions between disease and non-disease environments via shedding and disease transmission. This effort would be well served by a comprehensive sampling of agrobacteria diversity, blind to the historical biases toward pathogenic agrobacteria found associated with crown gall disease tissues. Another opportunity for future research involves determining whether A. tumefaciens undergoes significant pathoadaptation via within-host evolution, as has been documented in a number of pathogens that infect from non-disease environments (Sokurenko et al., 1998; Marvig et al., 2015). Interspecific interactions (e.g. Ma et al., 2014) as well as competitive interactions with avirulent agrobacteria (Kim et al., 2006) can have dramatic consequences on the success of pathogenic agrobacteria on host tissues. Despite this, the consequences of crown gall disease on plant microbiomes has not been well characterized nor is it known whether or not the composition of the host microbiome significantly alters the initiation or progression of crown gall disease.

Originality-significance statement.

A wide range of animal and plant pathogens are capable of infecting hosts but are not obligately host-associated, and can stably multiply in a free-living state – these pathogens are described as facultative. Facultative pathogens are thus confronted by diverse challenges associated with disease and non-disease environments. Despite their importance, the impact of the facultative pathogen life history on the ecology and evolution of infectious diseases has not been well addressed by past work. Understanding the consequences of the environmental heterogeneity these pathogens experience on the ecology and evolution of the diseases they cause is of particular interest. In this minireview, we discuss how the extensive study of interactions between Agrobacterium tumefaciens and their plant hosts, as well as with other microbes, makes this system well poised as a model facultative pathogen. Few model facultative pathogens have been established and those that exist are often experimentally cumbersome. In contrast, A. tumefaciens is a tractable experimental system that itself has an intricate and well characterized biology. We focus our discussion on several features common among facultative pathogens, including the importance of transitions between disease and non-disease environments. We also discuss how ecological tradeoffs associated with performance in disease and non-disease environments may result in environment-specific ecological and selective pressures.