Plant-pathogen interactions can be considered as a two-way communication processes in which not only the plant is able to recognize a foreign organism and defend itself from it, but the pathogen must also be able to manipulate the biology of the plant to create an optimal environment for its own growth and development avoiding the plant response (Pritchard and Birch, 2011; Smale, 2012; Boyd et al., 2013; Kushalappa et al., 2016).

Plants, unlike animals, lack a defined immune system. They rely on the innate immunity of each cell and the systemic signals occurring at infection sites (Schulze-Lefert and Panstruga, 2011; Bonardi et al., 2012; Lapin and Van den Ackerveken, 2013). However, with the particular characteristics of the plant defense system, the molecular mechanisms used by these organisms are similar to animals (Zipfel, 2014; Chiang and Coaker, 2015; Keller et al., 2016).

Plant pathogens, in general, are divided into biotrophic and necrotrophic pathogens, although there is a third group, called hemibiotrophic (Spoel and Dong, 2012; Okmen and Doehlemann, 2014). The difference between these groups lies in their lifestyle. Biotrophic pathogens obtain their nutrients from living host tissue, while necrotrophic pathogens obtain their nutrients from dead host tissue, and for this reason the mechanisms used by each kind of pathogens to infect a plant is different (Koeck et al., 2011; Lai and Mengiste, 2013).

Hemibiotrophic pathogens combine two strategies; they have an initial phase -biotrophic- in which the pathogen must evade the recognition from the host. This phase is followed by a necrotrophic stage in which toxins are secreted by the pathogen to induce host cell death; typically the visual symptoms start in this phase. Because of the presence of the asymptomatic biotrophic phase, the infection process with hemibiotrophic pathogens is difficult to understand and describe (Lee and Rose, 2010; Koeck et al., 2011; Vleeshouwers and Oliver, 2014).

The genetic basis of plant resistance to pathogens is divided into qualitative resistance (monogenic resistance) and quantitative resistance (polygenic resistance) (Lopez, 2011). Qualitative resistance can be explained by the gene-for-gene model proposed in the fifties by Flor (1971), who determined the basis of inheritance of resistance to flax rust in flax cultivars. According to the type of interaction with the pathogen, plant responses at the molecular level are divided into two types: i) non-host response and ii) host response. These two types of responses which differ mainly by the molecule type (both from the plant and the pathogen), involved in the process (Maekawa et al., 2011; Schmidt and Panstruga, 2011; Li et al., 2013) will be explained in the next two sections.

Non-host resistance, is defined as the event where a plant species in particular is resistant to different kind of pathogens (either bacteria, fungi, oomycetes, or viruses) but, these same pathogens can infect other plant species (Bent and Mackey, 2007; Fan and Doerner, 2012; Bellincampi et al., 2014).

This response spectrum is caused by specific recognition processes between pathogen molecules called MAMPs (microbe-associated molecular patterns), which are recognized by PRR-type (pattern recognition receptors) membrane receptor proteins. PRR structures are types of receptor-like kinases (RLKs) that have functional modular domains (Beck et al., 2012; Monaghan and Zipfel, 2012; Wu et al., 2014a). However, MAMPs are molecules involved not only in the pathogenesis process; most of them, in fact, have been described as essential components of the cell, such as flg22. There are other types of MAMPs such as the HAMPs (herbivore-associated molecular patterns) and the DAMPs (damage-associated molecular patterns; originally called endogenous elicitors), but in general most literature talk about MAMPs to refer to this kind of molecules (Conrath et al., 2015).

The molecular response triggered by the recognition of MAMPs is known as PTI (PAMP-triggered immunity) in which some molecular mechanisms associated with PTI include: production of ROS, Ca⁺² cascades, and the activation of MAPK (mitogen-activated protein kinases) cascades, involving Ca⁺²-dependent proteins that ultimately lead to transcriptional reprograming (Bernoux et al., 2011; Uma et al., 2011; Yoshioka et al., 2011; Bigeard et al., 2015; Trapet et al., 2015). Well-known case-studies include the following:

Bacterial PAMPs

FLS2 receptor (receptor-like kinase flagellin sensing 2) of Arabidopsis thaliana that interacts specifically with the oligopeptide flg22 of Gram-negative bacteria (Lu et al., 2010; Albrecht et al., 2012; Wang, 2012).

EFR receptor (Ef-Tu receptor) recognizes the oligopeptide elf18. A particular characteristic that has been described only for plant species of the Brassicaceae family (Beck et al., 2012; Wang, 2012).

Receptor XA21: Identified in rice, associated with specific resistance to various bacterial strains of the Xanthomonas oryzae pv. oryzae species (Lee et al., 2009).

Fungal and oomycetes PAMPs

CeBIP and CERK1 receptor: LysM domains were initially identified as carbohydrate-binding domains in bacteria (Monaghan and Zipfel, 2012). Evidence that these domains (LysM- RLKs) are involved in PTI activity come from their identification in rice (CeBIP) and in Arabidopsis (CERK1) (Miya et al., 2007; Macho and Zipfel, 2014), where these domains bind together and specifically recognize chitin fragments.

EiX1 and EiX2 receptors: EIX (ethylene-inducing xylanase) proteins which induce ethylene synthesis and PR (pathogen-related proteins) gene expression, are plant elicitors identified in tobacco and tomatoes. Their action is associated with a HR response (Ron and Avni, 2004).

Cf-9: identified in tomato was the first protein LRR-RLP and confers resistance to the fungus Cladosporium fulvus).

In the case of oomycetes, the PAMPs and the PRR have not been described yet. There are reports with some approximations like the soluble beta- glucan-binding protein (GBP) from soybean (Glycine max) that recognizes heptaglucosides from the oomycete Phytophthora sojae. Additionally, other PAMPs have been identified that can trigger immune signaling in fungi or oomycetes. For example, plants can recognize fungal ergosterol and other oomycete PAMPs including arachidonic acid, elicitins, the transglutaminase-derived immunogenic epitope Pep13. However, in all of these cases, the PRR proteins have not been identified so far (Zipfel, 2014).

This first line of defense response is usually effective against some pathogens, however, the pathogen has adapted its molecular infection mechanisms to evade or suppress the PTI, secreting proteins (called effectors), that trigger what is known as a host resistance mechanism (Monaghan and Zipfel, 2012; Couto and Zipfel, 2016). This plant-pathogen interaction has been explained by the zig-zag Model proposed by Jones and Dangl (2006) which suggests a co-evolutionary response process involving two plant immunity branches (PTI and ETI) and a transition phase (ETS). The details about this model will be described in the final section.

In addition to the mechanisms of PTI previously described, new evidence suggests that plants also use RNA silencing mechanisms as a defense mechanism. In this case, the plant has specific miRNA sequences that regulate gene expression and defend cells against invasive “nucleic acids”, whether they are transposons, transgenes or viruses (Zvereva and Pooggin, 2012).

The first miRNA identified to be involved in PTI was miR393 in Arabidopsis thaliana in response to Pseudomonas syringae. miR393 is induced by flg22 and then suppresses auxin signaling by negatively regulating mRNAs of auxin receptors, transport inhibitor response 1 (TIR1), AFB2 and AFB3, which allow plants to prioritize defense signaling over plant growth, and trigger a series of defense responses. The role of miRNAs in PTI has also been demonstrated in fungal and oomycete infection, for example, osa-miR7695 was found to accumulate in rice treated with blast fungal mycelia (Fei et al., 2016; Huang, 2016).

Molecular events during ETI (effector-triggered immunity) processes overlap with PTI. This branch of plant immunity occurs within the cell and originates once the host recognizes the effectors secreted by the pathogen, in which plant resistance (R) proteins can perceive these effectors initiating a defense response, including oxidative burst, accumulation of hormones such as salicylic acid (SA) and NOI (nitrogen oxide), MAPK cascades, changes in calcium levels, transcriptional reprogramming and synthesis of antimicrobial compounds, expression of pathogeneses related (PR) genes (Hein et al., 2009; Coll et al., 2011; De Bruyne et al., 2014; Stael et al., 2015).

Depending on the type of pathogen, the plant can induce programmed cell death (PCD) processes, also termed hypersensitive response (HR). These processes seek to block the advance of biotrophic pathogens to avoid an infection in different host tissues. In this process, chloroplasts play a key role in the production of ROS-type molecules and NOI (Robert-Seilaniantz et al., 2007; Lopez, 2011; Presti et al., 2015). For necrotrophic pathogens, the plant cell wall is the first line of defense providing a dynamic interface for interaction with necrotrophic pathogens. The interaction includes serving as a rich source of carbohydrates for the growth of pathogens; acting as a physical barrier to restrict the progression of the pathogens, and playing a role as an integrity sensory system that can activate intracellular signaling cascades in which plant hormones like jasmonic acid (JA) and ethylene (ET) play a major role in the defense response against these pathogens (Vleesschauwer et al., 2014).

R proteins recognize effectors and activate the defense-signaling network (Hogenhout et al., 2009; Song et al., 2009; Gururani et al., 2012). Generally, this type of resistance confers complete and specific resistance; that is why it is also called race-specific resistance (de Jonge et al., 2011; Saintenac et al., 2013). The R proteins are codified by NB-LRR genes, one of the largest and most variable gene families found in plants (Collier and Moffett, 2009).

Most R proteins belong to a subgroup of a family of proteins called STAND (signal transduction ATPase with numerous domains). NBS-LRR (nucleotide-binding site; leucine rich repeats) proteins that are subdivided into two subclasses depending on their N-terminal domain, -TIR- (Toll/Interleukin-1 receptors) domain or -CC- (coiled coil) domain, and are known as NBS-LRR-TIR and NBS-LRR-CC, respectively (Marone et al., 2013; Wu et al., 2014a).

For signaling, the NBS-LRR-CC proteins generally required a GPI anchored protein named non-race specific disease resistance 1, while NBS-LRR-TIR proteins require an enhanced disease susceptibility 1 for signaling. Additionally, the NBS-LRR-CC proteins are found in dicots and monocots whereas NBS-LRR-TIR are restricted to dicots (Chiang and Coaker, 2015; Cui et al., 2015). The mechanism that activates R proteins and the subsequent signaling cascade in ETI is still being debated. Related to recognition, the simplest model is the direct interaction model in which there is a physical interaction between the pathogen effector and the R protein. An example of this mode of interaction occurs between the pita CC-NB-LRR immune receptor in rice and the AvrPita effector of the fungus Magnaporthe grisea (Liu et al., 2011).

There are no particular characteristics that distinguish the way in which these proteins can sense different classes of pathogens (Collier and Moffett, 2009). However, the recognition process could be modeled in a more complex way through an indirect recognition. This form of recognition has led to the development of alternative recognition models:

Guard hypothesis. The guard hypothesis suggests that R proteins can detect changes or alterations caused by the effector to the host “guard” protein. One of the cases reported for this model corresponds to the RIN4 (RPM1 interacting protein 4) protein of A. thaliana, which is associated with two CC-NB-LRR-RMP1 and RPS2-type proteins. RIN4 is the target protein for AvrRpm1 and Avrpt2 effectors which, because of their protease activity, induce cleavage of RIN4, and this cleavage is detected by R proteins (Caplan et al., 2008; Van der Hoorn and Kamoun, 2008).

Decoy hypothesis. The “decoy” protein mimics the pathogen effector target, so the decoy functions mainly to restrict the pathogen but is not involved in the immune response (van der Hoorn and Kamoun, 2008). This model has been discussed mainly from the evolutionary point of view, it is expected that in the presence of the R gene, natural selection favors the decoy protein, but in the absence of the R gene, natural selection will cause the protein to decrease its affinity for the effector (Saintenac et al., 2013; Wu et al., 2015). Van der Hoorn and Kamoun (2008), explain the new model and offer four study cases to explain the model as follows:

As we previously explained, each immunity branch has specific molecules, but, knowing the molecules and the process that produce them is not the only information that we need to understand the plant immunity. The next section provides information about the zig-zag model that until today is still the most accepted model. However, there are updated or new versions of the model.

Zig-zag model. In the most basic interaction, the zig zag model involves an interaction between the pathogen and the host. The interaction can be divided in four phases:

Phase 1: plants detect MAMPs via PRRs to trigger PAMP-triggered immunity (PTI).

This model is being reevaluated, as some authors argue that describing a pathosystem as a model of interaction between molecules is a reductionist view of a process that is clearly highly complex. Other authors express concerns regarding the confusion that could arise from the terms of avirulence genes, virulence genes and effectors (Cook et al., 2014; Pritchard and Birch, 2014). The intent of this debate is not to invalidate any model, but to draw attention to certain issues discussed in the opinion article by Pritchard and Birch (2014), who describe six limitations of the zig-zag Model:

Molecular approach: It does not include DAMP. Therefore, it is suggested that the model is restricted to interactions with biotrophic pathogens.

Authors like Fei and collaborators (2016) have completed the 4 phases of the model introducing the effect that miRNAs have on the response of ITP and TSI. For example, miR393, which targets genes that are involved in auxin signaling (TIR1, AFB2, and AFB3) is induced upon treatment with flg22. The repression of auxin signaling during infection enhances host PTI by hormone crosstalk. Effectors from pathogens can suppress the levels of plant miRNAs, such as miR393, to enhance susceptibility. However, the miR482 family, a negative regulator of plant Resistance (R) genes, can also be repressed upon detection of effectors to enhance effector-triggered immunity (ETI). Thus, although the zig-zag model is still maintained in its most basic sense as far as lines of defense are concerned, nevertheless, it is accepted that depending on the pathosystem, the model could have more components and could be updated if necessary.

Invasion model.This model was proposed by Cook et al. (2015), the authors took into consideration some limitations of the zig zag model such as: the model is restricted in terms of what microbe-associated molecule patterns (MAMPs) the plants can perceive through pattern recognition receptors (PRRs).

The invasion model has been explained in a similar way than the the zig-zag model, except in the aspects related to the definition of the immunogenic molecules. The authors suggest that these molecules must be represented as a continuum, and they argue that these molecules play other roles beyond the pathogenicity. Thus, the evolution can affect these molecules and have effects in an interaction model. In this sense, if a molecule has a role in a different process some evolutive forces can alter them, producing changes in the interaction process or even in the fitness of the species.

Multicompent model. The model was proposed by Andolfo et al. (2016). Like in the previous model, the authors start showing the disadvantages of the zig-zag model such as the fact that the model only describes two perception layers (PTI and ETI).

The multicomponent model has two components: activation and modulation, and it is divided in three phases as follows:

Interaction: two principal effects are detected: i) modifications of virulence factor targets and ii) specific alterations of primary plant metabolism.

In this case, the authors try to emphasize that the model will be determined specifically by the pathogen´s life cycle. On the other hand, this model is in certain way connected with a new approach it which R genes and effectors are described like independent molecules, in a way that better articulate all the plant-pathogen related information for using molecular techniques to open a new era in crop breeding.

The two last models do not pretend to invalidate the original zig-zag model; the idea with these new perspectives is to bring a new concept in plant immunity with a holistic vision of a process that is not only related to a molecular interaction. Also, the model try to take into account other aspects that can affect the fitness of the pathogen and the host.

The advent of omics technologies has introduced a new and wide range of identification tools not only for one gene but also for areas of the genome that are associated with the pathogen response and the pathogen´s effectors (Brauer et al., 2014; Vayssier-Taussat et al., 2014; Thynne et al., 2015). Genomic studies should focus on identifying these areas of the genome, as they allow for visualizing the plant’s response mechanisms and even making predictions about the evolution of these areas.

In silico approaches allow generating different views, not only at the species level but also at the population level, to describe, based on genome information, the evolution of plants to achieve durable resistance to diseases (Burdon and Thrall, 2014; Karasov et al., 2014; Knief, 2014; Brown, 2015).

As mentioned at the beginning of this review, the growing demand for food production is generating new challenges to maintain crop production rates. Different authors suggest that genome editing technologies will allow to introduce or modify characteristics of interest that ensure high quality productions and quantity of different crops (Andolfo et al., 2016; Baltes and Voytas, 2015; Hendel et al., 2015; Kole et al., 2015).

Three mechanisms for genome editing are recognized: ZFN (Zinc Finger Nucleases), TALEN (TAL effector nucleases) and CRISPR/Cas9. These technologies seek to modify the genome introducing DBS (double strand breaks), under the control of specific nucleases, such action may cause deletions, insertions or modifications in the genome that can result from the DNA repair mechanisms NHEJ (Non-homologous end joining) and HR (Homologous recombination) (Huang et al., 2016; Mahesh, 2016; Nagamangala et al., 2015; Weeks et al., 2016).

CRISPR/Cas9 is currently proposed as a simple and useful alternative for genome editing, since it allows the modification of multiple sites along the genome. Several case studies have been reported in plants using this technology (Fauser et al., 2014; Jia and Nian, 2014; Lowder et al., 2015; Shan et al., 2014). Weeks and collaborators (2014) present a complete review of the different case studies that have been developed in species such as Arabidopsis thaliana, wheat and rice using CRISPR-Cas9 technology, and it is expected that soon more and more advances in the breeding for disease resistance will come from the use of this technology.

The information available on plant-pathogen interactions has raised many questions. Certain questions are focused on the “simplicity” with which these very diverse mechanisms have been described, as we can see in the new models mentioned. The dynamic and comprehensive approach is what is now referred to as biological dynamic networks, where different aspects of the model, even evolutionary aspects, must be taken account (Pritchard and Birch, 2011; Niks et al., 2015).

According to Vleesschauwer et al. (2014), studies in rice are providing new insights, often revealing unique complexities, for this reason the description of a biological process is not only related to the molecules per se, it is necessary to analyze the organism to describe how these plant-pathogen relationships can modulate the interaction in a niche and in a population.