Expression profiling across wild and cultivated tomatoes supports the relevance of early miR482/2118 suppression for Phytophthora resistance.

Plants possess a battery of specific pathogen resistance (R-)genes. Precise R-gene regulation is important in the presence and absence of a pathogen. Recently, a microRNA family, miR482/2118, was shown to regulate the expression of a major class of R-genes, nucleotide-binding site leucine-rich repeats (NBS-LRRs). Furthermore, RNA silencing suppressor proteins, secreted by pathogens, prevent the accumulation of miR482/2118, leading to an upregulation of R-genes. Despite this transcriptional release of R-genes, RNA silencing suppressors positively contribute to the virulence of some pathogens. To investigate this paradox, we analysed how the regulation of NBS-LRRs by miR482/2118 has been shaped by the coevolution between Phytophthora infestans and cultivated and wild tomatoes. We used degradome analyses and qRT-PCR to evaluate and quantify the co-expression of miR482/2118 and their NBS-LRR targets. Our data show that miR482/2118-mediated targeting contributes to the regulation of NBS-LRRs in Solanum lycopersicum. Based on miR482/2118 expression profiling in two additional tomato species-with different coevolutionary histories with P. infestans-we hypothesize that pathogen-mediated RNA silencing suppression is most effective in the interaction between S. lycopersicum and P. infestans Furthermore, an upregulation of miR482/2118 early in the infection may increase susceptibility to P. infestans.

Introduction

Resistance proteins (R-proteins) are fundamentally important in plant pathogen interactions. They recognize pathogen molecules, called effectors, which are secreted by pathogens to hijack plant immune responses [1,2]. Upon recognition, R-proteins trigger a pathogen-specific immune response [3,4]. Such immune responses include the hypersensitive response (HR), resulting in the release of reactive oxygen species, and can ultimately lead to cell death.

Misregulation of R-genes carries high fitness costs. Over-expression of R-genes in the absence of a pathogen can severely decrease fitness [5,6]. By contrast, insufficient R-gene expression during pathogen attack can allow for pathogen infection [7]. While several regulatory mechanisms are at play for different R-genes and R-proteins [8,9], negative regulation via small RNAs was proposed to globally buffer R-gene expression to avoid misregulation [10].

One example of negative regulation of R-genes, specifically of nucleotide-binding site leucine-rich repeats (NBS-LRRs), is suppression by the microRNA (miRNA) family miR482/2118 [11-13]. Targeting of miR482/2118 leads either to the degradation of NBS-LRR mRNA or to an inhibition of the translation of the corresponding mRNAs [14]. The family is one of the most labile miRNA families, displaying low sequence conservation, even between closely related species, despite its widespread presence in the plant kingdom [11,13,15]. Its diversity is in part a consequence of the amino acid variability of its target sequence [16].

A major pathogen of cultivated tomato (Solanum lycopersicum) is Phytophthora infestans. However, P. infestans not only infects crops but also their wild relatives [17-22]. Populations of wild tomato species, given that they are not subjected to breeding, have experienced different evolutionary histories with P. infestans compared to the cultivated tomato. In fact, wild tomatoes harbour R-genes that effectively contribute to the resistance to this pathogen [19,23].

Phytophthora infestans's vast effector repertoire is likely the result of a constant adaptation to its diverse hosts [24]. Among P. infestans's effectors, two were recently identified, which suppress the host's RNA silencing machinery [25-26]. Suppression of the plant RNA silencing pathways would release miRNA targets, including NBS-LRRs, from their miRNA-mediated suppression. Therefore, it has been hypothesized that the regulation of R-genes by miR482/2118 may have evolved into a pathogen detection mechanism, i.e. a counter-defence mechanism by which pathogen-mediated RNA silencing suppression activates the plant immune system [13]. This is at odds with the observed positive influence on pathogen virulence by these effectors [25] and suggests a complex network of NBS-LRR regulation during the infection of plants by their pathogens.

In this study, we analysed how coevolution of tomatoes and their pathogen P. infestans has shaped miRNA-mediated NBS-LRR regulation and how this regulatory network contributes to resistance in tomato. We first identified miR482/2118 targets associated with P. infestans defence in S. lycopersicum. Next, we studied the expression of SlmiR482/2118 and a set of 12 NBS-LRRs in S. lycopersicum during infection by P. infestans. Although the expression of NBS-LRRs is undoubtedly regulated by multiple mechanisms in addition to negative regulation via miRNAs, we observe examples of strong co-regulation between members of miR482/2118 and their targets. Combining comparative expression analyses of members of the miR482/2118 family in three closely related tomato species (S. lycopersicum, Solanum pimpinellifolium and Solanum arcanum) and analyses of host resistance led to two observations: (i) the least resistant tomato, S. lycopersicum, showed downregulation of several miRNAs from 24 to 96 hours post-inoculation (hpi) relative to the mock control, while its more resistant wild relatives did not and (ii) downregulation of miR482a and miR482f during early time-points of infection (6 hpi) correlated with resistance to P. infestans. Based on these observations, we hypothesize that global pathogen-mediated RNA silencing suppression is more effective in cultivated tomato than in its wild relatives.

Material and methods

Plant material and Phytophthora infestans inoculation

Seeds of S. arcanum were surface sterilized using approximately 5% NaOCl (30 s), washed 3×3 min in sterile H2O, plated on 1.2% H2O agar and incubated in dark for 3 days (16 h/8 h with 18°C/15°C). Afterwards, the seeds were transferred to a 16 L (166 ± 17 µmol quanta m−2 s−1) : 8 D regime. Nine days post sterilization (dps), seedlings were transferred to 0.5% Murashige & Skoog medium [27] with 1% sucrose.

The isolate, IPO-C, of P. infestans was grown on rye-sucrose-agar plates (with 100 µg ml−1 ampicillin, 10 µg ml−1 amphotericin B and 20 µg ml−1 vancomycin; [28]) at 18°C in the dark. Zoospores were isolated and leaflets of S. arcanum were inoculated at 28 dps as described in de Vries et al. [29]. Three biological replicates (three to four seedlings each) were sampled per treatment and time-point (0 hpi, 6 hpi, 24 hpi, 48 hpi, 72 hpi and 96 hpi).

RNA extraction, mRNA purification and cDNA synthesis

Total RNA of S. arcanum was isolated using the Universal RNA/miRNA Purification Kit (Roboklon, Berlin, Germany). RNA from S. lycopersicum and S. pimpinellifolium was used from de Vries et al. [29]. mRNA was purified using the Dynabeads mRNA Purification Kit (Thermo Scientific, Massachusetts, USA).

cDNA libraries for mature miR482/2118 expression analyses were created using miScript Plant RT Kit (Qiagen, Hilden, Germany) using 250 ng total RNA and diluted 1 : 10 with nuclease-free H2O. cDNA libraries for all other expression analyses were created with the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, Vilnius, Lithuania) using 1000 ng total RNA and random hexamer primers and libraries were diluted 1 : 1 with nuclease-free H2O.

cDNA libraries for the modified 5'RNA ligase-mediated rapid amplification of cDNA ends (5′RLM-RACE) were created using the GeneRacer Kit (Invitrogen, California, USA) using 50–100 ng mRNA from infections (24 and 48 hpi) and mock (48 hpi). To identify miRNA cleavage sites, the protocol was modified to omit the enzymatic digest of the cap and proceed directly to the ligation of the 5′ GeneRacer RNA oligo adapter. The SuperScript III RT Module (Invitrogen, California, USA) with the GeneRacer Oligo dT Primer was used for reverse transcription.

5′RLM-RACE

Amplification of 5′RLM-RACE products was performed (1× High Fidelity PCR buffer, 0.6 µM GeneRacer 5′ primer, 0.2 µM of the gene specific primer (electronic supplementary material, table S1), 200 µM dNTPs, 1 mM MgSO4, 3% DMSO and 0.5U Platinum Taq DNA Polymerase High Fidelity) followed by a nested PCR, using 1 µl of the PCR product in a 50 µl reaction (1× High Fidelity PCR buffer, 0.2 µM GeneRacer 5′ nested primer, 0.2 µM of the nested gene specific primer (electronic supplementary material, table S1), 200 µM dNTPs, 1 mM MgSO4 and 0.5U Platinum Taq DNA Polymerase High Fidelity).

PCR products were amplified with a Phusion High-Fidelity DNA Polymerase (New England Biolabs, Massachusetts, USA) and cloned using a Zero Blunt TOPO PCR Cloning Kit (Invitrogen, California, USA).

Confirmation of infection and infection progress

To confirm successful infection and study the disease progression, leaflets of S. arcanum seedlings were analysed microscopically. The relative necrotic area and pathogen structures were determined according to [29]. Statistical differences in necrotic area over time and between mock and infections were estimated using a Kruskal–Wallis test [30] with a Tukey and Kramer post hoc test, using a Tukey distance approximation [31]. For comparisons of the relative necrotic area between species, normal distribution of the data was evaluated using a Shapiro–Wilk test [32] and then tested for significant differences using a Mann–Whitney U test [33] in R v. 3.2.1. To determine the abundance and life cycle progression of P. infestans at the molecular level, expression of three biotrophic, two necrotrophic and one biomass marker gene were analysed according to [29].

Identification of miR482/2118 family members

Members of miR482/2118 from S. lycopersicum and S. pimpinellifolium have been previously identified in de Vries et al. [15]. Members of miR482/2118 from S. arcanum were identified via a BLASTn against the S. arcanum genome using miR482/2118 precursor sequences of S. lycopersicum as query. The best hits in S. arcanum were aligned to the SlmiR482/2118 precursor sequences and the mature miR482/2118 sequences were determined. Folding of S. arcanum miR482/2118 precursors into hairpins was predicted using RNAfold [34] (electronic supplementary material, figure S1).

Selection of R-genes

We chose R-genes that were (i) predicted to be targeted by one or more members of miR482/2118 and (ii) associated with resistance to P. infestans. The 52 potential miR482/2118 target genes [15] were used as queries for a BLASTn-search against the NCBI nr/nt database limited to S. lycopersicum. The best functional annotated BLAST hit (e.g. excluding hits to entire chromosomes) was recorded. Hits with an e-value of 0, query coverage greater than 90% and an identity greater than 85% to an R-gene associated with resistance against P. infestans in S. lycopersicum or the resistance gene analogues (RGA) complex were determined as likely to be associated with resistance to P. infestans.

qRT-PCR

qRT-PCR was performed using the miScript SYBR Green PCR (Qiagen, Hilden, Germany). miR482/2118 forward primers were designed based on the mature miR482/2118 sequences. miR482/2118 primer specificity was tested by creating a qRT-PCR product for each primer. These qRT-PCR products were purified, and each primer was tested with each qRT-PCR product to determine if and at what annealing temperatures the primers would bind to other miR482/2118 paralogues. For all miR482/2118 primers a binding-specific annealing temperature was determined (electronic supplementary material, table S1). The only exceptions were the primers for SlmiR482h and SpmiR482h, which annealed to miR482h as well as miR482 at all annealing temperatures. SamiR482h was specific because of its slightly different mature miRNA sequence (electronic supplementary material, table S1). As a control, the expression of mature SlmiR156a/b/c, SlmiR166a/b, SlmiR168a/b and SlmiR172a/b was determined. miR390a was used as a reference due to its constant expression across treatments and time-points according to BestKeeper v.1 [35].

Expression of NBS-LRRs in S. lycopersicum was determined using the SsoAdvanced Universal SYBR Green Supermix (electronic supplementary material, table S1; Bio-Rad, California, USA). As reference genes, we used SAND [15], TIP41 [15] and Translation Initiation Factor 3 subunit H (TIF3H; [29]).

Relative abundance and progression of P. infestans were measured using Histone2a (PiH2a). Expression of PiH2a at time-points 24 to 96 hpi was set relative to its expression at 24 hpi. The data were normalized with the plant reference genes (SAND, TIP41 and TIF3H).

Relative expression was calculated according to [36]. Data were tested for normality using a Shapiro–Wilk test [32] and equal variance using R v. 3.2.1. Comparisons between infections and mock control were tested using either a two-sample t-test or a Welch two-sample t-test for normally distributed data or a Mann–Whitney U-test [33] for non-normally distributed data.

Results and discussion

One-third of potential nucleotide-binding site leucine-rich repeats targets have high identity to Phytophthora infestans-associated resistance genes

We screened for NBS-LRRs that are potential targets of miR482/2118 and classified as R-genes for P. infestans (electronic supplementary material, table S2). Of the 52 predicted NBS-LRR targets [15], we identified 20 which were annotated as a P. infestans-associated R-gene or the RGA complex, members of which are associated with resistance to the pathogen [37-38]. Of these 20, 17 matched a P. infestans-associated R-gene with an e-value of 0, a query coverage of greater than 90% and an identity of greater than 85% (electronic supplementary material, table S2). Therefore, approximately 33% of the predicted direct NBS-LRR targets of miR482/2118 are associated with resistance to P. infestans.

Nucleotide-binding site leucine-rich repeats are targeted by miR482/2118 in Solanum lycopersicum during infection by Phytophthora infestans infection

Previous studies have used 5′RLM-RACE to test whether the expression of NBS-LRRs is regulated by members of the miR482/2118 gene family [13-14]. Targeting by SlmiR482f of Solyc08g075630.2.1 and Solyc08g076000.2.1, which are associated with P. infestans defence responses (electronic supplementary material, table S2), was only shown in overexpression lines of Nicotiana benthamiana [14]. To test whether these NBS-LRRs are targeted by miR482/2118 in S. lycopersicum, we created 5′RLM-RACE libraries from S. lycopersicum infected with the pathogen and mock-treated (figure 1). In addition, we tested Solyc02g036270.2.1, because it is a functional miR482/2118 target [13] that is not associated with P. infestans resistance (figure 1; electronic supplementary material, table S2).

Figure 1.

Targeting of NBS-LRRs by miR482/2118 family members in S. lycopersicum. In vitro confirmation of NBS-LRR targeting by SlmiR482/2118 using 5′RLM-RACE for Solyc02g036270.2.1 targeted by SlmiR482a (a), Solyc08g075630.2.1 targeted by SlmiR482f (b) and Solyc08g076000.2.1 targeted by SlmiR482f (c). A schematic of the target gene (blue) is on the left. The predicted binding site (P-loop, orange) and its sequence is shown below. The arrows indicate the validated degradation sites. The number of clones supporting the site and the total number of clones sequenced are given above the arrows. Upper numbers indicate clones from the mock controls and lower numbers indicate those from infections. The corresponding PCR products of the 5′RLM-RACE are shown on the right.

A cleavage site is determined by an enrichment of a specific degradation product in the 5′RLM-RACE library. This is established by cloning the degradation products of the gene of interest from the library and analysing how often a specific degradation product was cloned. If the gene of interest has a miRNA cleavage site, the majority of the clones should contain a product cut at the predicted cleavage site. All three tested genes revealed a cleavage site in the region complementary to the miR482/2118 sequences. Moreover, these cleavage products were observed in both mock-treated and P. infestans infected leaflets of S. lycopersicum. Based on clone analyses of the 5′RLM-RACE library of Solyc02g036270.2.1, 15 out of 18 clones were cleaved between nucleotide positions 11 and 12 of the miRNA binding site (figure 1a). For Solyc08g075630.2.1, all clones (24/24) were cleaved between nucleotide positions 12 and 13 of the miRNA binding site (figure 1b). For Solyc08g076000.2.1, 13/17 clones had a cleavage site between nucleotide positions 12 and 13 of the miRNA binding region (figure 1c). Some alternative cleavage products were observed for Solyc02g036270.2.1 and Solyc08g076000.2.1 (figure 1). This is in agreement with Ouyang et al. [14], who also observed an alternative cleavage site for Solyc08g076000.2.1. In summary, we demonstrate that targeting of NBS-LRRs by miR482/2118 is effective in pathogen-challenged and unchallenged plants.

Co-regulation of members of miR482/2118 and their nucleotide-binding site leucine-rich repeats targets is time-dependent

Given that a third of the miR482/2118 potential targets in S. lycopersicum are associated with disease resistance to P. infestans in S. lycopersicum, we chose a subset of 11 NBS-LRRs associated with P. infestans resistance and Solyc02g036270.2.1 (as a positive control for cleavage, but a negative control in terms of P. infestans resistance) to study the co-regulation of NBS-LRRs and miR482/2118 in this interaction. We quantified the expression of the seven members of miR482/2118 and 12 NBS-LRRs in infected and uninfected plants across five time-points (6 to 96 hpi) (figures 2a and 3).

Figure 2.

Expression of miR482/2118 family members in S. lycopersicum, S. pimpinellifolium and S. arcanum. Relative expression (log2) in infected compared with mock-control plants of S. lycopersicum (a), S. pimpinellifolium (b) and S. arcanum (c) of the seven miR482/2118 family members at 6, 24, 48, 72 and 96 hpi relative to mock control. The bars represent the average relative expression of the mature miRNAs and the error bars indicate the standard error of the mean (SEM). Significant differences of the relative expression of the miRNA in infected versus mock-treated plants at a specific time-point are indicated by *(p-value < 0.05), **(p-value < 0.01), ***(p-value < 0.001) and ns (not significant).

Figure 3.

Expression and co-regulation of SlmiR482/2118 and their NBS-LRR targets. Relative expression (log2) of potential NBS-LRR targets of miR482/2118 in infected compared with mock-control plants of S. lycopersicum. Bars show the mean expression and error bars indicate the SEM. Statistical differences in relative expression in infected versus mock-treated plants at a specific time-point are indicated by *(p-value < 0.05), **(p-value < 0.01), ***(p-value < 0.001) and ns (not significant). Filled circles below each gene corresponds to the miRNA(s) predicted to target each NBS-LRR. Arrow heads indicate significant up or downregulation of the members of SlmiR482/2118 at a given time-point: upward arrow heads indicate significant upregulation and downward arrow heads indicate significant downregulation of the miRNA. The arrow heads are coloured according to their respective miRNA. Vertical lines between miRNA arrow heads and the relative expression of the NBS-LRR highlight significant negative co-regulation between members of the SlmiR482/2118 family and their targets at a specific time-point.

To identify to what degree the miRNAs show similar expression patterns in response to infection, we compared the expression of the individual miRNAs and recorded how often two miRNAs showed the same expression pattern in parallel at a given time-point, to see whether both show (i) significant upregulation, (ii) significant downregulation or (iii) no differential regulation between infection versus mock. Overall, all SlmiR482/2118 miRNAs show similar dynamics in expression, with the same expression pattern of two miRNAs for 3.1 ± 0.9 time-points, on average (figure 2a).

By contrast, two NBS-LRRs show, on average, the same expression pattern at 2.0 ± 1.3 time-points (figure 3). This is a significantly lower co-regulation compared to that observed for miR482/2118 (p-value = 0.0002). Such differences in co-regulation suggest that despite active targeting by miR482/2118 in S. lycopersicum, NBS-LRRs are likely to be regulated by other mechanisms in addition to the regulation by miR482/2118.

Next, we evaluated how often pairs of miR482/2118 and NBS-LRRs are co-regulated and what type of co-regulation they are subjected to (i.e. negative co-regulation, positive co-regulation or no differential regulation of both miRNA and target). In total (over all time-points), we evaluated 95 miR482/2118–NBS-LRR combinations (electronic supplementary material, figure S2). In 45 pairs, the NBS-LRRs are predicted to be post-transcriptionally regulated, while 50 are predicted to be translationally regulated (electronic supplementary material, table S2). If a target is post-transcriptionally regulated, one would predict a negative co-regulation of target and miRNA. This means that if the miRNA is significantly upregulated, the target should be significantly downregulated and vice versa. Nevertheless, positive correlations between miRNA and target mRNA levels have been reported [39-41]. Additionally, positive co-regulation has been observed for miRNAs [39] that suppress their targets translationally [42], suggesting that translational repression can lead to positive co-regulation. If the miRNA is not differentially regulated between infection versus mock treatment, the target should not be either.

We observed that the direction of co-regulation is not static for every miR482/2118–NBS-LRR combination but can shift between time-points. Such rapid shifts in co-regulation may result from switches between translational and post-transcriptional suppression. For example, Solyc08g076000.2.1 shows an alternating pattern of co-regulation with SlmiR482f (figure 3; electronic supplementary material, S2) and is regulated by both modes [14], despite its prediction to be regulated translationally (electronic supplementary material, table S2).

We determined at which time-points co-regulation was most prevalent, suggesting a potential influence of miR482/2118 on NBS-LRR-regulation. The greatest co-regulation occurred at 48 hpi with 10/12 NBS-LRRs showing co-regulation with at least one of their respective SlmiR482/2118 members (electronic supplementary material, figure S2). High co-regulation was also detected at 6 and 72 hpi for 9/12 NBS-LRRs. All three time-points are biologically interesting: 6 hpi is a crucial time-point for infection success, as early HR significantly contributes to resistance against P. infestans [20]. Between 48 and 72 hpi, P. infestans switches from a biotrophic (i.e. requiring nutrients from a living host) to a necrotrophic phase (i.e. inducing host cell death) [29].

Solanum arcanum is less susceptible to Phytophthora infestans than its two relatives

We found that co-regulation of miR482/2118 with their targets was time-dependent, and more prevalent at time-points critical for infection success and transitions in the pathogen's life cycle. To place this in context with resistance, we compared the response of three tomato species, S. lycopersicum, S. pimpinellifolium and S. arcanum, to P. infestans. These host species differ in their evolutionary and ecological histories. S. lycopersicum has long been subjected to artificial selection. Furthermore, high-density monocultures of crop species can allow for higher pathogen loads and potentially higher pathogen diversity in the cultivated species [43].

Solanum pimpinellifolium and S. arcanum have partially overlapping ranges: S. pimpinellifolium's habitat spans from Central Ecuador to Chile, while S. arcanum occurs in Northern Peru [44]. Furthermore, their habitats overlap with that of P. infestans [45-47], allowing for exposure to and coevolution with the pathogen. Indeed, R-genes associated with resistance to P. infestans have been isolated from S. pimpinellifolium [23,48]. In addition, S. pimpinellifolium is facultative self-compatible and S. arcanum is predominantly self-incompatible [44]. Mating system differences can influence the evolutionary history of the hosts and their adaptation potential. We therefore hypothesize that the different hosts will show variation in their resistance to P. infestans because they experienced different evolutionary histories.

In our previous study [29], we evaluated the relationship between pathogen abundance, the presence of pathogen infection structures and disease symptoms in S. lycopersicum and S. pimpinellifolium. Here, we describe our new results on S. arcanum and compare these with the results from S. lycopersicum and S. pimpinellifolium. The relative necrotic area of S. arcanum increased significantly at 48 hpi (figure 4c; electronic supplementary material, S3a). Although the variance of relative necrotic area was higher in 72 and 96 hpi compared with 48 hpi, the relative necrotic area did not increase significantly beyond 48 hpi (electronic supplementary material, figure S3a). The abundance of P. infestans increased significantly from 24 to 48 hpi, and from 48 to 72 hpi (electronic supplementary material, figure S3b). The lack of correlation between relative necrotic area and P. infestans abundance at 72 hpi may stem from a delayed transition to the necrotrophic phase. For S. pimpinellifolium and S. lycopersicum we pinpointed the transition from biotrophy to necrotrophy to a time between 48 and 72 hpi [29]. For S. arcanum, we observed haustoria from 24 hpi onwards, and developing and mature sporangia at 72 and 96 hpi (electronic supplementary material, figure S3c). In agreement with this, most marker genes for biotrophy are expressed throughout the infection, but the sporulation marker Cdc14 was only expressed from 72 hpi onwards (electronic supplementary material, figure S3d,e), suggesting that the transition to necrotrophy occurred between 48, 72 hpi. However, the number of all infection structures was lower in S. arcanum compared with the other two species (figure 4e). As less virulent isolates of P. infestans also show a reduction in haustoria compared with more virulent isolates [49], this suggests that P. infestans is less infective and has a delayed life cycle transition on S. arcanum.

Figure 4.

Infection progress in S. arcanum in comparison to S. lycopersicum and S. pimpinellifolium. Necrotic area on the leaflets of S. lycopersicum (a), S. pimpinellifolium (b) and S. arcanum (c) for mock-treated (upper row) and infected (lower row) leaflets. Comparison of the relative necrotic area during P. infestans infection in S. arcanum (blue), S. pimpinellifolium (yellow) and S. lycopersicum (purple) (d). Statistical differences in relative necrotic area for the three species were calculated per time-point and are indicated by different letters above the boxes. The p-value cut-off was 0.05. Comparison of the number of haustoria, developing and mature sporangia of P. infestans after infection of S. arcanum (blue), S. pimpinellifolium (yellow) and S. lycopersicum (purple) (e). All data for S. pimpinellifolium and S. lycopersicum were published previously in de Vries et al. [29].

Across all species, sporangia develop the earliest (48 hpi) in S. lycopersicum (figure 4e). The relative necrotic area 72 and 96 hpi is also the highest in S. lycopersicum (figure 4a–d). Taken together, this suggests that, although all species are susceptible to P. infestans, they are so by a variable degree: S. lycopersicum is likely the most susceptible, followed by S. pimpinellifolium and finally S. arcanum, which is the least susceptible of all three species.

MiR482a and miR482f are candidate miRNAs for defence responses against Phytophthora infestans

We evaluated the miRNA expression between the tomatoes in relation to their resistance phenotype. Compared with S. lycopersicum, expression between pairs of miRNAs was significantly less correlated in the wild tomatoes: in S. pimpinellifolium pairs of miR482/2118 members showed the same expression pattern at 2.2 ± 1.2 time-points (p-value = 0.012; figure 2b) and in S. arcanum at 2.2 ± 0.9 time-points (p-value = 0.005; figure 2c). Lower co-regulation suggests additional gene-specific regulatory mechanisms in the wild tomatoes. By contrast, the cultivated tomato appears to have a more global co-regulation of miR482/2118 expression. These differences in co-regulation between wild and cultivated tomatoes could result from (i) differences in the evolutionary history of these plants (i.e. artificial versus natural selection) that brought about a more streamlined regulation of expression of miR482/2118 in S. lycopersicum or (ii) greater sensitivity to pathogen manipulation of host RNA silencing in S. lycopersicum, for example, due to pathogen-secreted RNA silencing suppressors. The latter is of interest because two RNA silencing suppressors have been previously described in P. infestans [26,50]. Additionally, we observed a substantial downregulation of additional miRNAs in S. lycopersicum that do not target NBS-LRRs (SlmiR156a/b/c, SlmiR166a/b, SlmiR168a/b and SlmiR172a/b) in S. lycopersicum from 24 hpi onwards (electronic supplementary material, figure S4). Of these four, only SlmiR172a/b is implicated to function in P. infestans resistance, albeit by a different mechanism [51].

Next, we examined the relationship between the expression of miR482/2118 miRNAs and the life cycle of P. infestans. We focused on 6, 48 and 72 hpi, because they are critical time-points during infection by P. infestans and they correspond to the time frame when the greatest co-regulation between pairs of SlmiR482/2118 and their targets is detected (figure 3; electronic supplementary material, S2). We observed that six of the seven miRNAs were upregulated at 6 hpi (figures 2a and 3), which should result in enhanced suppression of their NBS-LRR targets. This was indeed true for three of the NBS-LRR targets screened: Solyc02g036270.2.1, Solyc08g075630.2.1 and Solyc08g076000.2.1. The gene Solyc02g036270.2.1 served as a reference NBS-LRR, because it was so far not reported to be associated with resistance to P. infestans.

We compared the expression patterns of SlmiR482/2118 with those in the close relatives of S. lycopersicum. In S. pimpinellifolium, only two SpmiR482/2118 members were significantly upregulated at 6 hpi (figure 2b). In S. arcanum, none of the seven SamiR482/2118 members were upregulated at this time-point (figure 2c). Moreover, four out of seven SamiR482/2118 were significantly downregulated at 6 hpi in S. arcanum (figure 2c), which was the most resistant tomato species. All of these members of miR482/2118 have targets associated with P. infestans defence in S. lycopersicum (electronic supplementary material, table S2). In fact, the R-gene targets of SlmiR482a and SlmiR482f were significantly downregulated at 6 hpi in S. lycopersicum (figure 3). Therefore, the downregulation of SamiR482a and SamiR482f upon infection in S. arcanum might be related to the enhanced resistance observed in this species. This downregulation of SamiR482/2118 in S. arcanum in the presence of the pathogen could allow for an earlier response to the pathogen, because the predicted NBS-LRR targets would not be repressed during the first 6 h, as they are in S. lycopersicum. Taken together, these results point to miR482a and miR482f as potential regulators of P. infestans-associated defence responses.

Given the substantial co-regulation of miRNAs and their targets at 48 and 72 hpi, we evaluated the association of miR482/2118 expression with the life cycle progression of P. infestans on its hosts. In the biotrophic phase (prior to 72 hpi), P. infestans requires a living host. High R-protein activity during this time frame could lead to earlier pathogen perception and activation of HR/cell death, which in turn would limit pathogen spread [20]. In the necrotrophic phase, P. infestans induces host cell death [52-53]. High R-protein activity at this time-point may not be beneficial to the host, but instead benefit the pathogen. An effective plant resistance response during the necrotrophic phase may include the suppression of cell death-inducing proteins, such as R-proteins, perhaps through an upregulation of miR482/2118. By contrast, if pathogen-mediated RNA silencing suppression were effective at these later time-points, one would expect a downregulation of miRNAs, including miR482/2118.

At 48 hpi, four SlmiR482/2118 (SlmiR482f, SlmiR482, SlmiR482/h and SlmiR5300) were downregulated specifically in infected plants (figure 2a). While this does not exclude a plant-mediated downregulation of miR482/2118, the downregulation of the non-NBS-LRR regulating miRNAs (electronic supplementary material, figure S4), indicates that pathogen-mediated RNA silencing suppression may play a role here. In agreement with this, a P. infestans RNA silencing suppressor, potentially involved in silencing the miRNA-mediated silencing pathway, has its highest expression in the main biotrophic phase [26]. In S. arcanum, three SamiR482/2118 members (SamiR482a, SamiR482b and SamiR482h) were downregulated during the infection compared to the control (figure 2c). By contrast, none were downregulated in S. pimpinellifolium (figure 2b).

After the transition to necrotrophy at 72 hpi, the following miRNAs were upregulated: SlmiR482a, SlmiR482b and SlmiR482f in S. lycopersicum, SamiR482a, SamiR482h and SamiR5300 in S. arcanum and SpmiR482a in S. pimpinellifolium (figure 2). None of the control SlmiRNAs were significantly upregulated (electronic supplementary material, figure S4), suggesting a miRNA-specific plant response at this time-point. miR482a is upregulated at 72 hpi in the infections across all three species, despite the small lag in S. arcanum for the transition from biotrophy to necrotrophy. This would suggest that upregulation of miR482a is a consistent phenotype associated with a plant defence response during the necrotrophic phase of P. infestans. This is further supported by the negative co-regulation of SlmiR482a and its target Solyc11g06530.1.1 at this time-point (figure 3; electronic supplementary material, S2).

Conclusion

In this study, we investigated the expression of miR482/2118 during the infection of P. infestans on three different tomato species. We found that co-regulation of mature SlmiR482/2118 and their targets in cultivated tomato was highest during the initial phase of infection and during the life cycle transition of P. infestans from biotrophy to necrotrophy. Across-species comparisons of the gene expression of mature miR482/2118 and of the strength of resistance led to two main conclusions: (i) Co-evolution of P. infestans and S. lycopersicum may have resulted in a more efficient pathogen-mediated RNA silencing suppression compared with its more resistant sister species; and (ii) miR482a and miR482f could be identified as candidate miRNAs for mediating the resistance response of tomatoes to P. infestans.