DISCUSSION

Given that pathogens are prevalent across all ecosystems, an individual’s reproductive success and survival will depend on its ability to resist infection. Natural host populations have been shown to support considerable diversity in resistance (Laine et al., 2011; Salvaudon et al., 2008), and theory predicts that this variation is maintained by pathogen-imposed selection. However, empirical support for the role of selection in generating resistance diversity is to date still scarce. With recent advances uncovering the molecular underpinnings of resistance, it is becoming increasingly feasible to study resistance also in non-model systems.
Here, we established a high-quality de novo transcriptome assembly of P. lanceolata to investigate the gene expression and processes activated in different plant genotypes in response to inoculation of the same pathogen strain. In our study, all five plant genotypes showed unique gene expression patterns. This was clearly demonstrated in the principal component analysis where the genotypes separate, while the inoculated and mock-inoculated replicates of the same genotype cluster together. Significant variation in gene expression patterns among plant genotypes has also been discovered in other studies (Burghardt et al., 2017; Muller, Kersten, Fladung, & Schroeder, 2019). In the redundancy analysis (RDA), inoculation explains only 4 % of the total variation, while genotype-by-inoculation interactions contributes 46 %, suggesting that the genotypes have highly unique responses to pathogen attack. While such genotype specificity may be expected between susceptible and resistant genotypes, the split to resistant versus susceptible phenotypes explains only 9 % of the variance, with considerable transcript expression pattern differences between phenotypes. Overall, the plant genotypes also differ by the number, fold change and the function of the transcripts which are differentially expressed in response to the pathogen. Furthermore, even though the gene expression shows the known induction of JA, SA and ABA signaling pathways, they show highly varying activation patterns with JA and ABA significantly contributing to the genotype by experiment differences in RDA. This suggests that different plant genotypes use different strategies in response to the same pathogen, which could be an important mechanism generating phenotypic resistance diversity. One possible explanation for the diverse responses is the extremely high genetic variation within the species; overall, the transcriptome had very high Watterson θ=0.068 and nucleotide diversity π= 0.077, suggesting effective population sizes in the order of millions. The high genetic diversity where, on average, eight nucleotides out of 100 differ between any two individuals, is likely manifested also in the diverse responses. The experimental take home message of genotype-specific expression pathways to treatment is that including multiple genotypes in experiments and avoiding pooling for RNA-Seq is essential to uncover variation relevant for phenotypic differentiation.
The pathways induced by the pathogen were also visible in the gene expression data. The common responses included the induction of specific nitrate transport genes in susceptible phenotypes as well as elevated expression of photosynthesis-associated genes and related biological processes taking part in chloroplast in all genotypes. The powdery mildew fungi have a contracted carbohydrate metabolism, for example they are not able to degrade pectin, an essential component of plant cell walls (Liang et al., 2018), whereas the lipid metabolism is intact, suggesting that their main source of energy is from lipids. Chlorosis is another hallmark sign of a successful pathogen attack. In our results, the fact that specific nitrate transporters show elevated expression levels as well as chloroplast processes in general, suggests elevated chlorophyll biosynthesis. Together, this suggests that, at least at the early stage of infection, P. plantaginis may target the chloroplast lipids of Plantago to obtain its nutrients, most likely the lipids involved in chlorophyll.
NLRs play an important role in pathogen recognition and downstream defense responses, defense signaling, as well as activation of a hyper sensitive response (Monteiro & Nishimura, 2018). In our study, we discovered 543 NLR isoforms from a transcriptome that is a combination of the five different Plantago genotype NLR repertoires. Majority of these transcripts were expressed at least at some level in all five plant genotypes. Presence-absence polymorphism in a subset of NLRs has been demonstrated across Arabidopsis accessions (Van de Weyer et al., 2019), and hence it could contribute to the slight differences in the numbers of NLRs detected in the genotypes. Overall, the putative number of NLR genes in P. lanceolata, based on the identified 543 NLR isoforms, seems to be quite high compared to the range reported in other species (Baggs et al., 2017). The high count could result from preferential retention of NLR genes after Plantaginaceae specific whole genome duplication event 46-49 million years ago (Li, Zhang, et al., 2019), since it has been proposed that numerous and highly diversified defense genes are particularly favored in long-lived plants to avoid infection by their rapidly evolving pathogens (Plomion et al., 2018).
The predicted NLR transcripts resulted in 58 clusters of varying sizes, with 22 transcripts in the largest cluster. We also found considerable variation in the branch lengths among clusters, which could indicate variation among clusters in their likelihood to accumulate mutations (Tucker, Ackerman, Eads, Xu, & Lynch, 2013). Indeed, NLR genes are among the fastest evolving gene families in plants. They often form tandemly arrayed gene clusters, and this is believed to be critical for the fast pace of their structural and functional diversification (Meyers, 2003; Michelmore & Meyers, 1998). Frequent homologous recombination events and errors produced during the process, followed by diversifying selection, may generate the structural diversity needed to match high effector evolution rates characteristics of the microbes they encounter (Jacob, Vernaldi, & Maekawa, 2013; McDowell & Simon, 2006). NLR genes are also under evolutionary pressure resulting from inappropriate activation of cell death. If the plant cannot control NLR-activated cell death, it leads to decreased fitness (Phadnis & Malik, 2014). In particular, we found an increase in the number of homologs of Arabidopsis RPP13 gene, which is involved in defense against downy mildew (Peronospora parasitica ) in Arabidopsis, as well as other defense processes and signaling (Bittner-Eddy, Crute, Holub, & Beynon, 2000; Rentel, Leonelli, Dahlbeck, Zhao, & Staskawicz, 2008). While none of the clusters had significant dN/dS values, we found between one and twelve loci under selective pressure in 23 of the clusters. The cluster with the highest number of loci under selection has been suggested to be involved in defense against viruses in Arabidopsis (Revers et al., 2003). The H statistic analysis confirms the transcripts identified by the dN/dS analysis (10 transcripts), as well as identifies 13 other NLR transcripts that may have been under putative selection pressure, two of which show selection in the LRR domain responsible for detecting the type of pathogen.
We find that the NLR transcripts are differentially expressed in response to the pathogen treatment, and that this response varies according to genotype. Transcripts of many NLR genes are known to accumulate in response to defense induction or related stimuli (Lai & Eulgem, 2018). For example, 75 of the 124 studied Arabidopsis NLR genes were found to exhibit at least two‐fold higher transcript levels in response to one or more of the 15 implemented defence‐related treatments (Mohr et al., 2010). Up‐regulation of NLR transcripts after defence induction has also been observed in other plant species, such asBrassica rapa , soybean and rice (Brechenmacher et al., 2015; Chen, Pang, Chen, Zhang, & Piao, 2015; Ribot et al., 2008). While we were more likely to observe up-regulation of NLR expression levels in response to the pathogen treatment, this was not consistent across transcripts and genotypes. This is in line with recent studies on crop plants testing different genotypes in response to pathogen infection (Cruz-Miralles, Cabedo-Lopez, Perez-Hedo, Flors, & Jaques, 2019; Sari et al., 2018; Sari, Bhadauria, Vandenberg, & Banniza, 2017). Plants have evolved mechanisms to stabilize their basal expression levels, and to reduce the fitness costs of an overexpressed immune response that could have more deleterious effects on plant fitness than the infection (Fei, Xia, & Meyers, 2013). This may explain the down-regulation of some of the NLR transcripts we observe in both susceptible and resistant phenotypes.
Our results here are well in line with the extensive phenotypic variation, and highly strain-specific disease resistance measured inP. lanceolata in earlier studies (Hockerstedt et al., 2018; Jousimo et al., 2014a). High levels of variation in resistance seems to be nearly ubiquitous across natural host populations that experience pathogen-imposed selection without any human interference, in contrast to agricultural systems (Laine et al., 2011; Salvaudon et al., 2008). We show that phenotypic resistance may be generated by different mechanisms. First, we discovered a large repertoire of candidate NLRs inP. lanceolata. We also find evidence of selection generating diversity in a subset of the identified NLRs. Moreover, we discovered that the genotypes have unique expression profiles in response to pathogens, a mechanism which may further contribute to phenotypic variation. Indeed, this high level of genetic and expression profile diversity may be the key to successful defense against pathogens in sessile plants that lack a long lasting immune memory (Hall et al., 2009; Roux & Bergelson, 2016). Finding different mechanisms that contribute to phenotypic resistance is nontrivial, given how effectively this variation may be utilized to predict and control disease epidemics (Mundt, 2002a). Moreover, resistance in agricultural crops is highly prone to breakdown following pathogen adaptation, and for many commercially important pathogens, the known effective resistance genes are becoming limited. Wild plant populations are currently identified as the most promising source of genes required for development of sustainable agriculture (Fu et al., 2019). In conclusion, characterizing the architecture of resistance in natural host populations may yield unprecedented light on the potential of evolution to generate variation, and it can have broad and long-lasting impacts in our food production environments.