Acta Biológica Colombiana

The ABC of plant immunity

Since plants are continuously threatened by different kind of pathogens, it is imperative to develop new strategies in order to control plant diseases. The most environmental friendly strategy is to exploit natural mechanisms that plants have evolved to control invading pathogens. The activation of an effective immune response depends on the ability of plants to recognize pathogens. Based on the knowledge on molecules produced by pathogens and their recognition by plants hosts, the so called zig-zag model has outlined how to describe immunity systems in plants (Jones and Dangl, 2006). This model states that plants have evolved immune receptors that are able to recognize pathogen-associated molecular patterns (PAMPs) or specialized effector proteins that are present in particular races/strains of pathogens. The recognition of PAMPs depends on the pattern recognition receptors (PRRs) that constitute the first line of molecular defense, known as PAMP triggered immunity (PTI) (Zipfel, 2014). Adapted races or strains of pathogens translocate effector proteins into plant cells to manipulate host components or suppress PTI (Cui et al., 2015). Plants can recognize pathogen effectors through R proteins and the immunity they activate is known as effector triggered immunity or ETI (Chisholm et al., 2006). Although this resistance is most of the time total, it is race specific and can be easily overcome by point mutation in effectors that escape plant recognition (Houterman et al., 2009). ETI is the molecular explanation of the gene-by-gene model proposed by Flor. According to this model, a plant is resistant when the interaction with the pathogen is incompatible. On the other hand, when the plant is susceptible, the interaction is compatible. In this case, there are only two possible phenotypes, resistant or susceptible, and the intermediates are not considered.

The zig-zag model does not include intermediate phenotypes and, consequently, is somewhat partial. A novel "invasion model" was recently proposed in which the classification of the plant immunity is based on the pathogen invasion patterns (IPs) (Cook et al., 2015). These IPs are a large spectrum ofmolecules that are produced and/ or released during invasion and are perceived by invasion pattern receptors (IPRs). The function of IPs can vary from microbiological physiology to host defense suppression and, consequently, they trigger a wide range of continuous defense responses. Plant responses can be either symbiotic or not, depending on the ability of the plant to recognize the IPs with IPRs and activate an IPTR (IP-triggered response). This new model also takes into account the fact that there is a complex interaction of multiple receptors and ligands at the same time and that the outcome of the interaction result from the combined result of all of them. The invasion model was developed as an alternative to the adopted classifications that separate PTI from ETI and in which PAMPs are defined from the host perspective while the effectors are considered from the pathogen perspective (Cook et al., 2015). In its application, the invasion model emphasizes the identification and understanding of molecules produced by the pathogen. Therefore, it is necessary to develop a model that uses the idea of defense as a continuum of responses, with a synergy and interaction between components from the invasion model, but that also shows the plant perspective of the model and its application.

Quantitative resistance enters into the game

In plant populations, when the response to a pathogen is a continuous phenotypic value, varying from highly susceptible to highly resistant individuals, it is considered as quantitative resistance (Huard-Chauveau et al., 2013). QDR is controlled by several genes, each one contributing, to a different degree, to the reduction of the disease (St. Clair, 2010). The term polygenic, or oligogenic, resistance is frequently associated with QDR because of its inherent genetic architecture (Mackay et al., 2009; Niks et al., 2015). Although the concept of QDR is well-defined, it is widely used and sometimes misunderstood or misused. Partial resistance is the most widely used concept in literature to describe an intermediate phenotype when resistance is not total (Niks et al., 2015). It is important to emphasize that this definition must be used to just describe a phenotype and not to define the genetic basis underlying the response (Niks et al., 2015). QDR is often called field resistance because it has been evaluated frequently in polycyclic field conditions (Niks et al., 2015); however, QDR should not been used as a synonym of field resistance since it has also been observed and assessed under greenhouses and controlled conditions.

Traditionally, QDR has been associated with two important concepts: broad spectrum and durable resistance; however, it is important to stress that these two concepts are not strictly exclusive to QDR. Durable resistance is a concept that was defined by Johnson in 1981 to refer to the resistance that retains its effectiveness in crops that are widely cultivated in an environment that is favorable to the pathogen, but this does not mean that it has to be permanent (Johnson, 1983). QDR has been considered as more durable than qualitative resistance and, consequently, more reliable. However, experimental evidence for this assumption is scarce (St. Clair, 2010). On the other hand, broad spectrum resistance occurs when the defense is effective against two or more types of pathogen species or to several strains or isolates of the same species of pathogen (Kou and Wang, 2010). The term is often used as a synonym for QDR, but it is important to note that there are several examples of strain-specific quantitative resistance loci (QRLs) (Poland et al., 2009). On the other hand, PRRs can confer a broad spectrum defense (Zipfel, 2014) and single R genes can also mediate broad-spectrum resistance (Xiao et al., 2001; Zhao et al., 2004; Narusaka et al., 2009) or can be engineered to achieve this type of resistance (Segretin et al., 2014; Giannakopoulou et al., 2015).

How to study complex traits and QDRs

Continuous traits do not exhibit single phenotype inheritance, such as those occuring in progenies that segregate according to the Mendelian rules (e.g. 3:1 or 15:1). However, this is not a consequence of distinct genetic mechanisms per se. Indeed, the variation in the phenotype observed for a quantitative trait is the result of a multiple genotypic expression of segregating alleles. In addition, it is highly influenced by the environment. As a consequence, it is not possible to clearly address a given phenotype to a particular genotype. For this reason, the Mendelian mechanisms in the study of quantitative traits is masked despite the fact that each gene could be segregated in a Mendelian mode (Griffiths, 2005).

For the above reasons, how to study quantitative traits has represented a new challenge to classical geneticists. The first studies on complex quantitative traits in plants included the association between the size of the seed (a quantitative trait) and the seed coat color (a qualitative trait) (Sax, 1923). The initial idea of quantitative loci mapping was first proposed by Thoday (1961), based on the observation that segregating single gene markers could be linked with loci associated with complex traits. Later, the term quantitative trait loci (QTL) was used for the first time to name the different loci that determine several quantitative traits in tomato (Tanksley et al., 1982).

The logic behind QTL detection is to determinate the relationship between DNA variation, as captured by DNA-based markers, and the observed phenotypic variation (Mackay et al., 2009). The identification of QTLs starts with the construction of a genetic map, where a large group of molecular markers are positioned in linkage groups (chromosomes) based on recombination frequencies. Once obtained, these markers are associated with the phenotypic trait of interest, using the fact that genes responsible for a particular trait and linked molecular markers cosegregate via chromosomal recombination during meiosis (Collard et al., 2005).

In the past, the challenge was to increase the number of molecular markers present in genetic maps for QTL mapping purposes. During the 90s, the development of DNA-based markers revolutionized the ability to detect DNA variations (Phillips and Vasil, 2013). Molecular markers, such as RFLP (Restriction Fragment Length Polymorphism), RAPD (Random Amplified Polymorphic DNA), AFLP (Amplified fragment length polymorphism) and SSR (Simple Sequence Repeat), contributed significantly to the development of high-dense genetic maps, allowing the dissection of qualitative and quantitative traits. Despite significant efforts, the genetic maps obtained through the use of these markers were generally low- density maps because of the lack of markers representing the complete set of recombination events. Thus, the first versions of maize and tomato maps contained only 50 RFLP markers each (Helentjaris et al., 1988) and the first potato map had 135 RFLP markers (Bonierbale et al., 1988). Moreover, at that time, the QTL intervals were large, usually ranging from 10 to 30 cM (Glazier et al., 2002). These limitations were overcome by massive sequencing technologies (Ansorge, 2009). These new genotyping technologies allowed for the high throughput identification of SNPs (Single Nucleotide Polymorphism), which have become the most widely used molecular marker. Nowadays, hundreds or thousands of widely distributed SNPs and their positions in the genome (provide the reference genome is accessible) can be identified in a relatively short period of time and at a low cost. Thereby, in recent years, the number of molecular markers and, therefore, map resolutions have increased, which ultimately leads to the reduction of QTL interval lengths to a few cM (Gautami et al., 2012; Stephens et al., 2014; Soto et al., 2015).

The association mapping (AM) approach emerged at the turn of the 21^st century as an alternative to the linkage mapping approach for QTL identification. In this case, the analysis is based on the phenomenon of linkage disequilibrium (LD) and the availability of historical recombination events at the population level (Zhu et al., 2008). AM take advantage of the explosion of new genome-scale data, allowing a higher resolution in loci localization assignment, compared with linkage mapping (Zhu et al., 2008). In this case, two strategies for the dissection of complex traits can be followed. The first one is the candidate gene approach and the second one is genome wide association studies (GWAS) (Brachi et al., 2011). While in the QTL linkage mapping approach, quantitative candidate genes are located within an interval; in AM, a direct association between complex traits sand the polymorphic markers, usually SNPs, is achieved (Rafalski, 2002). However, both approaches seem to be complementary in the sense that their ultimate goal is the detection of the genes underlying the quantitative complex trait for further cloning. Several examples of the use of the association mapping approach in QDR studies for the more limiting diseases can be found in recent literature (Benson et al., 2015; Gutiérrez et al., 2015; Arruda et al., 2016; Iquira et al., 2015; Olukolu et al., 2016; Turuspekov et al., 2016). However, despite the broad use of this approach, no genes detected by AM for plant QDR have been cloned so far.

Recently, two approaches had been proposed for studies on quantitative traits. First, there is extreme-phenotype GWAS (XP-GWAS), a new approach combining bulk segregant analysis (BSA) and GWAS (Yang et al., 2015). The second approach takes advantage of the use of clustered regularly interspaced short palindromic repeats (CRISPR-Cas9) and allow to obtain targeted mitotic recombination events without needing to develop directed crosses (Sadhu et al., 2016).

Through this approach, high frequency double strand breaks (DSB) are induced in regions of interest in mitotic cells. Then, the intrinsic cell reparation by homologous recombination (HR), generates recombination events that lead to the formation of a recombinant. Thus, the high efficiency of CRISPR-Cas9 mediate recombination events within 20 kb of the targeted site has been demonstrated. Comparing this rate of recombination with that obtained by random meiotic segregation, the later would require more than seven thousand individuals (Sadhu et al., 2016). The application of XP-GWASandCRISPR-Cas9approaches and their potential use in QDR characterization is highly promising.

In recent years, the concept of the set of all the information supported experimentally irrespective of the followed methodology, about QTLs and their allelic variation for a trait in one species, has received the name QTLome (Salvi and Tuberosa, 2015). Beyond constructing a QTLome, it is necessary to find a way to integrate and give a global meaning to the whole QTL information. This is the challenge of statistical QTL meta-analyses. The detection of common QTLs and the identification of co-locating resistance candidate genes from different experiments and populations have been recently achieved using QTL meta-analyses in maize to find resistance genes for virus diseases (Wang et al., 2016), leaf rust in wheat (Soriano and Royo, 2015) and verticillium wilt in cotton (Zhang et al., 2015).

A new era for QDR studies: phenotyping has the last word

The greatest aim for QDR studies in the past century was to increase molecular markers in mapping populations to capture all the allelic variants of genes that govern complex traits. Advances in high-throughput sequencing technologies have overcome this limitation, at least partially. However, the current challenge is to produce quality phenotype data, increasing molecular information and representing the bedrock of a new era of plant quantitative trait studies that will contribute to a better understanding of QDR (Basu et al., 2015).

Advances in automated precision phenotyping or high-throughput phenotyping apply technologies principally based on image, thermal, spectra and digital sensors, from which quantitative phenotypic information can be generated (Araus and Cairns, 2014). There are several advantages of these approaches. First, the reduction of subjectivity in the determination of disease incidence and symptoms during a particular plant-pathogen interaction. Second, the increase in the number of plants that can be evaluated. Finally, the increase in the reproducibility and the possibility to collect data at numerous time points (Mutka and Bart, 2015).

Some of the more sophisticated technologies for high-throughput phenotyping applied in QDR studies are hyper-spectral imaging, chlorophyll fluorescence imaging and thermal imaging (Mutka and Bart, 2015). Plant diseases produce different spectral reflectance patterns and plants suffering biotic stresses display changes in chlorophyll fluorescence emission (Baker, 2008). It has been shown that pathogens can change plant tissue temperature during the infection process. With these recent technologies, all of these parameters can be measured, even in the early plant phenological stages (Mutka and Bart, 2015). Wheat and sugarcane are some crops where these technologies have been used for detection and study of QDR (Mahlein et al., 2012; Bauriegel and Herppich, 2014; Mutka et al., 2016). Despite the fact that these techniques require a large number of previous evaluations in order to set the parameters for each disease, their potential in phenotyping plant disease is undeniable.

Phenotype has also been studied from an "omics" view (Salvi and Tuberosa, 2015). This new phenotyping method includes transcripts, proteins and metabolites, such as elements directly related to the phenotype, which have led to approaches such as expression-QTLs (eQTLs), protein-QTLs (pQTLs) and metabolite-QTLs (mQTLs), respectively. eQTLs and mQTLsare the more used given their progress in collection, automation and analysis of data (Salvi and Tuberosa, 2015).

From theory to practice: QDR in breeding

The complexity of QDR represents a challenge and an opportunity for plant breeding. A breeding scheme focused on obtaining qualitative disease resistance is relatively simple. It would be enough to introduce a single R gene into a susceptible plant background to confer resistance. On the other hand, in the case of QDR, the introduction of a gene from a QTL can confer a reduction, but not absence of disease. Thus, it would not be possible to get complete resistance until all of the resistance responsible loci are identified. In addition, in contrast to the current relative large repertoire of isolated R genes, the isolation of genes governing QDR for future use in breeding programs has not been an easy task.

For decades, marker-assisted selection (MAS) (Xu and Crouch, 2008) and gene pyramiding (Brun et al., 2010) efforts have been directed toward the use of QTLs with major effects, explaining more than 20 % of phenotypic variance, for introduction into plant resistance breeding programs (Collard et al., 2005). Some examples with great success in achieving high levels of resistance are found in rice (Bustamam et al., 2002), common bean (Miklas et al., 2006) and pearl millet (Sehgal, 2016), but unfortunately this has not been the case for most crops, including staple crops, such as cassava.

In plant breeding programs focusing on QDR, one of the limitations to be considered is the genetic linkage between the genes conferring resistance and closely linked undesirable genes, a phenomenon called linkage drag (Summers and Brown, 2013). Undesirable genes may affect the commercially accepted gene pool and, therefore, modify the quality and crop yield. If linkage drag is not eliminated or decreased, the use of the QTL in the program will be impractical. The MAS strategy has counteracted this phenomenon. Through high throughput genotyping and the use of haplotype analysis of the introgressed region (QTL), linkage drag in seedlings can be detected and tracked in order to subsequently backcross these individuals to resistant varieties lacking drag. This strategy was applied to detect and remove the linkage drag around the Rpv12 gene and confer resistance to powdery mildew in wine grapes (Vitis vinifera L.) (Venuti et al., 2013); alternatively, the marker-assisted recurrent selection (MARS), combined with genomic selection (GS) (Heffner et al., 2009), can also contribute to solving the linkage drag problem for QDR (Summers and Brown, 2013). GS selects plant material carrying whole genome molecular marker that are associated with resistance to a specific pathogen through the prediction of the phenotype using breeding values (BV) (Falconer and Mackay, 1996). These BVs are obtained by the compilation of molecular marker scores, phenotypic data evaluation of several germplasm and populations under a range of environmental conditions and (if it is available) pedigree information. Thus, MARS would increase the frequency of insertion of the desirable gene, decreasing the incorporation of undesirable ones and speeding up the detection of resistance loci with GS.

Another limiting factor in exploiting QTL with the aim of generating new varieties is the effect of the environment on the QTL. Multi-environment analyses in QDR studies offer the opportunity to detect the QTL x environment interaction (Q x E), conditional QTLs (El-Soda et al., 2014) and QTL stability during seasons and crop cycles. Special attention should be given to an eventual QxE interaction in plant quantitative resistance that is widely influenced by the environment and in which heritability values are usually low (Ntare and Williams, 1998).On the other hand, the functional validation of candidate generation an important part of QDR studies, which can be carried out by overexpressing or down regulating the candidate gene by applying genetic engineering (Mittler and Blumwald, 2010) or exploiting the mutant collections (Cavanagh et al., 2008).

In a large number of QDR studies the phenotypic evaluation (host response to the pathogen) is done after an artificial inoculation, employing a particular strain or a group of strains, allowing for the detection ofQTL associated with these strains and leaving aside other genomic regions involved in resistance to other strains. When these QTLs are introgressed in particular varieties and evaluated under naturally diseased fields, where different pathogen strains or races can be found, it is possible to obtain unsatisfactory results. For this reason, it is mandatory that a breeding program starts with the knowledge on the dynamics and diversity of the pathogen populations.

Molecular explanation of quantitative resistance

Although important progress in understanding and analyzing complex traits has been accomplished in recent years, knowledge on the molecular basis of the QDR is still scarce. Several hypotheses have been generated to explain the function of the genes that control the QDR (Poland et al., 2009). Five genes have been cloned from QTLs, which has enriched the proposed models. The rice Pi21 gene, which encodes for a protein that has a heavy metal-transport/ detoxification domain, confers resistance to several races of Magnaporthe oryzae (Fukuoka et al., 2009). The QDR genes from wheat Yr36 (resistance to Puccinia striiformis f. sp. tritici) and Lr34 (resistance to Puccinia striiformis, P. triticina and to Blumeria graminis) encode for a Kinase-START protein and a pleiotropic drug resistance subfamily of ABC transporters, respectively (Fu et al., 2009; Krattinger et al., 2009). In addition, the RKS1 gene,which encodes for an atypical kinase identified in the model plant Arabidopsis thaliana, confers resistance to most Xanthomonas campestris races, and to the pathovars raphani, incanae or armoraciae (Huard-Chauveau et al., 2013). Finally, the receptor-like protein coded by the ZmWAK gene ofmaize that confers resistance to Sphacelotheca reiliana was cloned recently (Zuo et al., 2015). Considering these discoveries and the gaps in QDR knowledge, different explanations of how it works are plausible.

QDR as a continuous response that depends on gene expression intensity

The expression level of genes involved in plant resistance can play important roles on the intensity of the expressed resistance. Transcriptomic analyses have allowed the identification of global changes in the expression profiles of genes involved plant immunity. Through expression analysis, it was possible to demonstrate that, during incompatible, compatible and non-host interactions, gene expression profiles were almost the same and that differences consisted in the intensity and kinetics of their induction (Tao et al., 2003). Other studies support the overlap between PAMP and ETI at the gene expression level (Navarro et al., 2004; Bozsó et al., 2009; Bozso et al., 2016). For QDR, several studies have reported a direct relationship between the expression level of some genes and the degree of resistance response. The maize ZmWAK gene is induced after pathogen inoculation. This gene is highly expressed in the mesocotyl and, at a lesser extent in the coleoptile of resistant lines, and the expression level of ZmWAKcan be associated with the degree of pathogen growth restriction in mesocotyl and coleoptiles (Zuo et al., 2015). Likewise, the RKS1 gene expression is correlated with the resistance level indifferent Arabidopsis accessions to Xanthomonas campestris pv. campestris (Xcc) strain 568 (Huard-Chauveau et al., 2013). Finally, the induction of the expression of susceptibility increases the susceptibility ofplants, as demonstrated for the susceptibility Pi21 gene from rice. In this case, transgenic plants showing higher expression of this gene were more susceptible to a virulent race of Magnaporthe oryzae (Fukuoka et al., 2009).

QDR can result not only from gene expression level, but also from pattern expression of genes. The wheat Lr34 gene, which was recently cloned from a QTL involved in wheat partial resistance to rust fungi and powdery mildew, is expressed in similar levels in resistant and susceptible plants. However, its expression level significantly varies between wheat seedlings and adult plants, where it is more effective (Krattinger et al., 2009). In this case, the quantitative response depends not on the genetic background of the plant, but on the alteration of defense gene expression by different factors, such as the developmental stage of the plant, tissue or organ location, etc.

The above examples suggest a correlation between the transcription level of QDR genes and the degree of resistance (Fig. 1). For instance, Huard-Chauveau et al. (2013) established, that the expression of RKS1-L in natural accessions was negatively correlated with the disease intensity. A question could be what is the factor that determines the induction level of gene expression? During the expression of plant resistance, it is considered that the induction of defense gene expression result from the activation of a signal pathway, which in turn is dependent on pathogen recognition. According to this, gene expression level could be conditioned depending on the ability, specificity and strength of the interaction between pathogen-derived molecules and plant receptors.

Figure 1: Model explaining the quantitative nature of resistance. As discussed in the text, the expression level of a group of particular genes involved in immunity contributes to the final output of the phenotype. The output intensity is determinate by the number of genes induced and their expression level.

Another aspect related to the transcriptional control of resistance level is the requirement for the presence of alternatively spliced transcripts. The classical example is the N gene from tobacco, initially known as conferring total resistance in tobacco to Tobacco Mosaic Virus (TMV). Through alternative splicing, both short N_S and long N_L transcripts are produced. During the initial phase of infection, the N_L version, encoding the full-length N protein, is more abundant, but, four hours after inoculation, a short truncated version is more abundant. If only one of the two variants is present, complete resistance is lost (Dinesh-Kumar and Baker, 2000). Concerning the specifically QDR related gene, above mentioned RKS1 gene, two transcripts were identified, with differences in length between resistant and susceptible Arabidopsis accessions (Huard-Chauveau et al., 2013). Another example is the Yr36 gene from wheat, which can result in up to six alternative transcript variants; however, only one of them encodes for a protein containing a complete START domain. This transcript is differentially regulated by temperature and is the only one that is up-regulated after inoculation with the fungus Puccinia striiformis f. sp. tritici (Fu et al., 2009).

To summarize, the transcription level differences shown by QDR genes suggest that it is necessary to incorporate the information on gene expression into the DNA variation data in order to achieve a systematic genetic approach and, thus, gain a better understanding of the molecular bases of the quantitative response (Mackay et al., 2009).

R weak alleles

As mentioned before, the first step in the activation of plant immunity is pathogen-derived molecules recognition. In ETI, a specific, strong and direct or indirect interaction between R proteins and the corresponding effector (named Avr) results in the activation of a signaling pathway, leading to immunity, which, in most cases, is associated with an HR response. Such as Avr-R recognition results in only one possible phenotype, i.e. R. Several studies on QTLs have demonstrated the presence within them of typical qualitative R genes, encoding for NB-LRR in QTLs, suggesting that the molecular bases of the pathogen recognition during QDR can exhibited common features with those of ETI. Supporting this hypothesis is the fact that both responses, ETI and QDR, share common molecular components (Roux et al., 2014b). In this model, the cause of the partial nature of QDR is the presence of "weak alleles" of R genes (Roux et al., 2014b). How to explain that an R protein confers only partial resistance? R protein is responsible for the recognition of a specific effector or the activity of it on a pathogenicity target protein. A weak allele of the R protein can correspond to a protein that is able to interact with an effector (or pathogenicity target), but its affinity is not high enough to induce a normal full response.

Allelic variation

The R weak alleles hypothesis can be considered as an extreme case of R gene allelic variation resulting in QDR. However, other QDR genes have shown allelic variation. Indeed there is evidence that QDR is associated with allelic variation of genes that differ in structure from canonical R proteins and that are important for plant defense. For example, the recently cloned QDR gene, ZmWAK, exhibits alleles differing in seven substitutions and a deletion between the resistant and susceptible maize lines. Although, in this case, polymorphisms affecting protein function were not found (Zuo et al., 2015), these polymorphisms could affect the interaction with other molecules or could prevent protein complexes formation. A similar situation was observed for the RKS1 gene, which, even if it is present in both resistant and susceptible accessions, has several SNPs that have been found to be associated with both phenotypes. These SNPs are located within both the coding and the 5' and 3' regulatory regions. In fact, one of the identified in RKS1 susceptible alleles corresponds to a stop codon. The authors suggested that mutations could be associated with susceptibility as a consequence of altered RKS1 long transcript expression (Huard-Chauveau et al., 2013). Polymorphisms have been also found between resistant and susceptible wheat plants, which are located within the Lr34 gene; two of them were located in exons and one in an intron. The polymorphisms located in exons are present in the resistance cultivar and correspond to a 3 bp deletion and an SNP that changes the aminoacid tyrosine for histidine and affect the first transmembrane domain of the ABC transporter (Krattinger et al., 2009). Resistant and susceptible rice cultivars have seven polymorphisms between them, located in the genomic region that harbors the Pi21 gene. Two of these polymorphisms correspond to deletions and were associated with the corresponding phenotype. Polymorphisms in this region, between different cultivated rice accessions, allowed for the identification of 12 haplotypes and revealed the natural variation of QDR genes. Only one haplotype was associated with resistance (Fukuoka et al., 2009). Further studies that include the intermediate phenotypes that are in between the lines or accessions that have already been evaluated, will resolve the role of polymorphisms in QDR. It is important to stress that these studies were conducted on contrasting lines, representing extreme resistant and susceptible phenotypes, and it would be interesting to evaluate the expression levels of QDR genes in individuals showing a gradient of phenotypic response.

Polymorphisms can be present even in promoter sequences, e.g. in the ZmWAK gene; however, no association with function or phenotype has been studied for these variations (Zuo et al., 2015). Additional studies are required to reveal if there is an association between such polymorphisms and QDR. In addition, polymorphisms located in introns or in promoters can modify transcription factor binding and splicing events, resulting in various quantities of transcripts or differential timing and tissue specificity of gene expression (Mackay et al., 2009).

The studies presented here to exemplify allelic variation are important, not just because they show that polymorphisms result in different alleles, but also because they were found in QDR genes that were cloned and validated, and could result in quantitative resistance. The number and nature of polymorphisms found in a QDR gene or in its genomic region could define the level of the phenotype. Huard-Chauveau et al., 2013 suggested that the QDR phenotype could be due to the additive effect or interaction of SNPs present in the identified haplotypes (Huard-Chauveau et al., 2013). Likewise, we propose that there is a highly resistant phenotype exhibited by a haplotype for a QDR gene or its genomic region and that variation of this haplotype would lead to the quantitative nature of the resistance. Moreover, additional SNPs could be located in other genes or other genomic regions that account for the resistance. Furthermore, susceptible alleles could compete with resistant ones for the interaction with scaffold proteins of molecular signaling complexes (Huard-Chauveau et al., 2013). Complementary studies with accessions that represent whole the range of different resistance levels and the corresponding sequence polymorphisms will help to tell if the polymorphisms present in QDR sequences are associated with the phenotype in order to validate this hypothesis.

Kinases and signaling

Kinases are essential components in plant cell biology and regulate different processes, such as biotic stress (Afzal et al., 2008; Parniske, 2008), hormone signaling (Santner and Estelle, 2009), growth (Hematy and Hofte, 2008), cell differentiation and other physiological processes (De Smet et al., 2009). Serine-threonine kinases are important signaling transduction components of both PTI and ETI (Zipfel, 2014) and MAP kinase cascades regulate downstream defense responses (Schwessinger and Zipfel, 2008). Additionally, pseudokinases are described as being important in signaling network control (Huard-Chauveau et al., 2013). In this context, it is reasonable to consider genes involved in the signaling pathway, including MAP kinases, as putative key elements of QDR.

Several proteins that have been characterized as involved in QDR have proved to be kinases or pseudokinases. RKS1 from Arabidopsis (Huard-Chauveau et al., 2013) is a typical kinase, ZmWAK from maize contains a kinase domain (Zuo et al., 2015) and Yr36 from wheat has a kinase domain similar to Arabidopsis WAK-like kinases (Fu et al., 2009). In addition, through eQTLs in barley, a gene that encodes for a "putative histidin-kinase" was identified as an important component of the resistance to Puccinia graminisf. sp. tritici (Druka et al., 2008). This biochemical characteristic opens the door to possibilities for the role that these proteins may play in QDR, as for example like transmitting molecular signals.

Miscellaneous

One interpretation of how QDR works at the cellular level considers QDR as corresponding weak PTI or ETI (López, 2011; Kushalappa et al., 2016). In this way, resistance known as qualitative could also be polygenic and be achieved if all the components are present and function correctly (Kushalappa et al., 2016). Therefore, the quantitative counterpart may have missing components. The missing concept here not necessarily corresponds to complete absence of a particular component, but to differential quantities of resistance related metabolites, proteins coded by R genes, or PRRs, which in turn are regulated by other genes (Kushalappa et al., 2016). Therefore, the more defense-related components are missing (or diminished), the more resistance is reduced. An alternative, but not excluding, hypothesis states that differences in defense responses are the consequence of the sensitivity of the components to input signals. It was hypothesized that resistant plants display robust responses because they are insensitive to small changes in input signals (Tao et al., 2003); therefore, the remaining range of responses of QDR could be more sensitive to this change. QDR have been recently redefined due to the cloning of some of the corresponding genes and it has been stated that the involved proteins do not belong to a specific group, such as in the case of R genes, but may have several functions (Navabi et al., 2005; Poland et al., 2009; Bryant et al., 2014; Roux et al., 2014a). Thus new molecules, which previously have not been described as important during plant-microbe interactions, could be responsible for resistance (Roux et al., 2014b). Nevertheless, it seems that R genes actually have roles in QDR, but with a weak contribution, as compared to other genes with different structures and functions (Corwin et al., 2016).