The tomato Slmlo1 line, carrying a loss-of-function mutation in the SlMLO1 susceptibility gene (Bai et al., 2008; Zheng et al., 2016), and the tomato cultivar (cv) Moneymaker (MM) were used in the disease tests as resistant and susceptible control, respectively. The tomato Slmlo1 line was also used as background genotype for RNAi experiments.

The Wageningen isolate of the tomato powdery mildew fungus Oidium neolycopersici and the Wag04 isolate of the barley powdery mildew fungus Bgh, maintained on the susceptible cvs. MM and Manchuria, respectively, were used for inoculation. Infected tomato and barley plants were kept in a greenhouse compartment at 20 ± 2°C with 70 ± 15% relative humidity at the Unifarm of Wageningen University & Research, Netherlands.

The Arabidopsis syntaxin AtPEN1 amino acid sequence (GenBank ID: NP_187788.1; gene AT3G11820) was used as query for a tBLASTn search against the Tomato Genome SL2.50 CDS of Sol Genomics Network (SGN), using default settings. The resulting 21 hits were used for a ClustalW alignment of 41 sequences, also including the complete Arabidopsis Qa SNARE family, grapevine VvPEN1, and barley HvROR2. The gap open cost and the gap extension cost were set equal to 10 and 4, respectively. Afterward, an Unweighted Pair Group Method with Arithmetic Mean (UPGMA) phylogenetic tree was built setting the bootstrap value equal to 100. All the bioinformatic analyses were performed using the CLC sequence viewer software¹.

To generate RNAi constructs, the two tomato cDNA sequences showing the highest similarity to AtPEN1, Solyc10g081850.1.1 (named SlPEN1a) and Solyc01g006950.2.1 (named SlPEN1b), were used to design primer pairs for Gateway cloning (Invitrogen), using the Primer3 software (Untergasser et al., 2012). These primers, named SlPEN1a_gate__Fw/Rev (5′-caccCTGGTTGTGGACCTGGAAGT-3′/5′-TGTCCTCTTCCTTGCTCCTG-3′) and SlPEN1b_gate_Fw/Rev (5′-caccACGAGCTGAAAAACCTCGAA-3′/5′-ACA ACAGACGTCCTCGTCCT-3′), allowed the amplification of 259 and 250 bp PCR products for SlPEN1a and SlPEN1b, respectively. Amplification was conducted using PfuUltraII Fusion HS DNA Polymerase (Agilent Technologies). PCR fragments were first cloned into the Gateway-compatible vector pENTR D-TOPO (Invitrogen) and then transferred to Escherichia coli competent cells (XL10-Gold Ultracompetent Cells, Agilent Technologies). The presence of the inserts was assessed by colony PCR, restriction enzyme digestion and sequencing using the universal M13 primers. Positive clones were used in a LR reaction, which allowed the inserts to be transferred to the pHELLSGATE12 vector (Helliwell and Waterhouse, 2003), harboring the 35S Cauliflower Mosaic Virus (CaMV) promoter for constitutive expression and the marker gene NPTII for kanamycin resistance selection. Plasmids were cloned into E. coli competent cells DH5α and positive colonies were screened by colony PCR and sequencing as described before. Recombinant vectors were finally extracted and transferred to the AGL1+virG strain of Agrobacterium tumefaciens by electroporation.

The tomato Slmlo1 line was used for transformation, that was carried out as described by Appiano et al. (2015). A total of 15 T₁ plants were generated and, after self-pollination, T₂ families were obtained.

To select transgenic plants within each segregating family, the DNA of the T₂ plants was amplified with primer pairs NPTII_Fw/ NPTII_Rev (5′-TCGGCTATGACTGGGCACAAC-3′)/5′-AAGAAGGCGATAGAAGGCGA-3′), and 35S-Fw/Rev (5′-GCTCCTACAAATGCCATCA-3′)/(5′-GATAGTGG GATTG TGCGTCA-3′). The expression of the target genes in selected T₂ families was assessed by real-time qPCR using the primer pairs qPEN1a_Fw/Rev (5′-CGAGATGCTTTGTGCATCAG-3′/5′-CAGTCTCCTTCAGCTCCATTTC-3′) and qPEN1b_Fw/Rev (5′-TGGTTTAGTTGTTGATGGACCTC-3′/5′-ACCCCCATCCAACTTACTTACTTCTC-3′).

Transgenic T₂ plants showing high silencing level for each construct were crossed and eight F₁ individuals were obtained. Four-week-old F₁ plants were tested through qPCR for the expression of SlPEN1a and SlPEN1b, using the primer pairs mentioned above.

Three disease tests were performed using the adapted powdery mildew pathogen O. neolycopersici. The first one encompassed 20 plants for each of the 15 T₂ families, seven Slmlo1 plants and five MM plants. The second one included four T₂ families (two carrying the RNAi::SlPEN1a construct and two carrying the RNAi::SlPEN1b construct), each including 15 transgenic plants, five non-transgenic plants and five Slmlo1 plants. The third test included eight F₁ individuals, 14 T₃ transgenic individuals silenced for SlPEN1a, 19 T₃ individuals silenced for SlPEN1b, and 8 Slmlo1 individuals.

In all cases, the inoculation was performed by spraying plants with a suspension of conidiospores obtained from heavily infected leaves of MM plants and adjusted to a final concentration of 2^∗10⁴ spores per milliliter, as described by Pavan et al. (2008). Inoculated plants were grown at 20 ± 2°C with 70 ± 15% relative humidity and day length of 16 h in a greenhouse of Unifarm of Wageningen University & Research, Netherlands.

Disease evaluation was carried out 15 days after inoculation. Powdery mildew symptoms were visually scored on each plant using a scale of symptoms severity ranging from 0 to 3 (Bai et al., 2008), and means and standard deviation were calculated for each T₂ family. In addition, in the case of the second and third disease tests, fungal biomass was quantified by real-time qPCR as reported by Huibers et al. (2013). Briefly, plant and fungal genomic DNAs were isolated from infected tomato leaves using the Qiagen DNeasy Plant Mini Kit and amplified with the primer pairs On-Fw (5′-CGCCAAAGACCTAACCAAAA-3′)/On-Rev (5′-AGCCAAGAGATCCGTTGTTG-3′), designed on O. neolycopersici internal transcribed spacer (ITS) sequences (GenBank accession number EU047564), and Ef-Fw (5′-GGAACTTGAGAAGGAGCCTAAG-3′)/Ef-Rev (5′-CAACACCAACAGCAACAGTCT-3′), designed on the tomato Elongation Factor 1α (Ef1α) reference gene (Løvdal and Lillo, 2009). Relative fold-change of the ratio between fungus and tomato gDNAs was inferred by the 2^-ΔΔC_T method (Livak and Schmittgen, 2001; Pfaffl, 2001) and results were analyzed by the Student’s t-test.

Three transgenic plants from the T₂ families RNAi::SlPEN1a_I and RNAi::SlPEN1b_I were selected by PCR using the 35S and NPTII primer pairs. Around 18 days after sowing, these plants, together with three plants of the Slmlo1 line, were transferred to an infection chamber.

A dry inoculum of Bgh was brushed off heavily infected barley leaves with a paintbrush and applied on the third and fourth leaves of 4-week old tomato plants. At least three samples for each inoculated plant were collected 72 hours post-inoculation (hpi). These samples were stained with trypan blue as described by Freialdenhoven et al. (1996) and mounted on glass slides with a 1:1 (v/v) glycerol:water solution. Observation of the slides was done using a Zeiss Axiophot bright field microscope and pictures were taken with an Axiocam ERc5s. For each sample, more than thirty infection units (one infection unit representing a germinated spore) per slide were observed. The pathogen penetration rate was estimated as the percentage of units displaying hypersensitive response. Statistical analysis was carried out using the Student’s t-test.

To detect conserved motifs in the syntaxin family, the alignment used for phylogenetic analysis was given as input to the BOXSHADE software², setting as 1.0 the fraction of sequences that must agree. Furthermore, with the aim of identifying residues specifically conserved in syntaxins acting in powdery mildew defense, another ClustalW alignment was performed, using proteins of the SYP-1b subclade. This new alignment was also fed to the BOXSHADE software.

Moreover, the same dataset was used for a codon-based evolutionary analysis, based on the difference of nonsynonymous-to-synonymous substitutions per non-synonymous and synonymous sites (dN-dS). The single-likelihood ancestor counting (SLAC) method implemented by the Datamonkey web server ³ was performed. The default p-value of 0.1 was used as threshold for statistical significance to make predictions on the kind of selection pressure (negative, neutral, or positive) acting on each codon.

Twenty-one putative syntaxins were retrieved in the tomato genome with a BLAST search using the Arabidopsis AtPEN1 amino acid sequence as input. These were used for a phylogenetic study, together with the 18 Arabidopsis syntaxins, barley HvROR2 and grapevine VvPEN1. The resulting phylogenetic tree was composed by five clearly distinct clades (Figure 1). The distribution of Arabidopsis syntaxins within the clades fully matched with their previous assignment to the five subfamilies SYP-1, -2, -3, -4, -8 (Sanderfoot et al., 2000). Within each clade, at least one tomato putative syntaxin was found. The most represented clade (SYP-1), harboring 21 homologs, was further partitioned into four subclades (indicated as SYP-1a to -1d in Figure 1).

Two putative tomato syntaxins, Solyc10g081850 and Solyc01g006950, were assigned with strong bootstrap support to the subclade SYP-1b, containing all the functional homologs shown to be involved in powdery mildew penetration resistance, i.e., AtPEN1, HvROR2, and VvPEN1 (Collins et al., 2003; Feechan et al., 2013). Therefore, Solyc10g081850 and Solyc01g006950 were renamed SlPEN1a and SlPEN1b, respectively (Figure 1). Protein sequences of SlPEN1a and SlPEN1b are highly related with each other (85.3% amino-acid conservation) and with AtPEN1 (72.3 and 71.7% amino-acid conservation, respectively).

To gain insights into the functional role of SlPEN1a and SlPEN1b, we obtained RNAi silencing constructs for each of the two genes and transformed the tomato Slmlo1 line. Eleven RNAi::SlPEN1a and four RNAi::SlPEN1b T₁ plants were obtained and self-pollinated to provide T₂ families.

Two weeks after the inoculation, all the MM plants were covered by the fungal mycelium and had an average disease index (DI) score of 3 (Supplementary Table S1). Individuals of the Slmlo1 line showed no symptoms on the 4th leaf. Since weak mycelium growth was occasionally observed on the first three true leaves, an average DI score of 0.62 was assigned to them.

The 11 T₂ families obtained with the RNAi::SlPEN1a construct showed various levels of symptoms corresponding to average DI scores ranging from 0.47 to 1.49. The four T₂ families obtained with the RNAi::SlPEN1b construct displayed average DI scores ranging from 0.8 to 0.95.

Individuals of five selected T₂ families associated with the highest average DI scores (namely RNAi::PEN1a_I, RNAi::PEN1a_II, RNAi::PEN1a_III, RNAi::PEN1b_I, and RNAi::PEN1b_II), were screened for the presence/absence of the construct. Plants carrying the RNAi constructs were identified and further referred to as RNAi::SlPEN1a(+) and RNAi::SlPEN1b(+). A higher average DI score was assigned to them compared to T₂ plants not harboring the corresponding RNAi constructs (Supplementary Table S1). To verify that higher susceptibility of individuals carrying the RNAi constructs was due to reduced SlPEN1a and SlPEN1b transcripts, the expression level of these genes was evaluated. The SlPEN1a homolog was significantly silenced in transgenic individuals of the families RNAi::SlPEN1a_I(+) and RNAi::SlPEN1a_II(+) compared to Slmlo1 plants. In addition, no unwanted cross-silencing of SlPEN1b was found (Figure 2). Transgenic plants identified in family RNAi::SlPEN1a_III did not show a significant silencing of the target gene. However, this family was the least susceptible of the three T₂ families carrying the RNAi::SlPEN1a construct. The SlPEN1b homolog was strongly silenced in transgenic plants of families RNAi::PEN1b_I and _II compared to Slmlo1 plants, while the expression of SlPEN1a was not significantly reduced.

A total of five non-transgenic individuals segregating from the five T₂ families were grouped to evaluate the expression of SlPEN1a and SlPEN1b. The expression of these genes did not significantly change compared to Slmlo1 plants (Figure 2).

Aimed to confirm our results, the four families showing high silencing level of the target genes (RNAi::PEN1a_I and _II; RNAi::PEN1b_I and _II) were selected for a second disease test (Figure 3). Compared to the Slmlo1 plants, RNAi::SlPEN1a(+) individuals showed once more clear fungal sporulation (Figure 3A) and significantly increased fungal biomass (Figure 3B). Although RNAi::SlPEN1b(+) individuals showed slightly higher DI than Slmlo1 plants, fungal biomass did not significantly increase (Figure 3B). Non-transgenic individuals of the four T₂ families displayed the same DI of Slmlo1 individuals (Figure 3).

To further investigate the functional role of SlPEN1b and its relation with SlPEN1a, we performed an additional experiment using the F₁ progeny obtained by crossing well-silenced RNAi::SlPEN1a_I(+) and RNAi::SlPEN1b_I(+) individuals. Significantly reduced expression level of the target gene was observed in the eight F₁ individuals and the selfing progenies of their parents, which were included in the same experiment (Figure 4). Compared to the control group of eight Slmlo1 plants, the F₁ progeny and RNAi::SlPEN1a(+) plants showed a significantly higher fungal biomass. However, F₁ individuals were not significantly more susceptible than plants carrying the RNAi::SlPEN1a(+) construct (data not shown).

Since it has been shown that PEN genes are involved in resistance to non-adapted powdery mildew species, we set-up an assay to investigate the role of the two tomato syntaxins, SlPEN1a and SlPEN1b, in the interaction with barley powdery mildew fungus Bgh. An artificial inoculum of this pathogen was used to inoculate RNAi::SlPEN1a_I(+) and RNAi::SlPEN1b_I(+) individuals, together with Slmlo1 plants. An histological analysis was performed to observe if pathogen entry rate changed in these transgenic plants. We noticed that the percentage of infection units showing HR increased dramatically from 22 to 72% in RNAi::SlPEN1a-I(+) individuals compared to the Slmlo1 line (Figure 5). A slight increase (2.5%) of HR was also found in RNAi::SlPEN1b_I(+) plants, although this value was not statistically different from the one of the Slmlo1 line (Figure 5). The increased HR rate in RNAi::SlPEN1a_I(+) plants suggests that silencing SlPEN1a favors Bgh penetration. Indeed, HR has been assumed to act as a post-penetration defense mechanism in non-host interactions (Lipka et al., 2005).

Aiming to investigate the magnitude and direction of natural selection acting on syntaxins involved in defense against powdery mildew fungi, we used the nucleotide sequence of SYP-1b homologs for a codon-based SLAC evolutionary analysis. This analysis is based on the dissimilarity level between non-synonymous substitution (dN) and synonymous substitution (dS) values. Evidence for negative selection was found on 77 codons, associated with amino acid residues scattered in various syntaxin domains (Qa-SNARE, membrane-spanning, and the three auto-inhibitory helix domains known as Ha, Hb, Hc) (Figure 6 and Supplementary Table S2). Alignment of the SYP-1b subclade proteins revealed that 65 of these residues are invariable throughout the dataset, suggesting they might play a crucial role for protein function.

We performed an additional alignment with the same set of syntaxins used for phylogenetic analysis. In total, we found that 24 out of 65 residues that are invariable in the SYP-1b subclade are not conserved in other syntaxins, thus pointing to the identification of residues that are, possibly, specifically important for pathogen defense (Supplementary Figure S1). In addition, we noticed two amino acid differences (highlighted in red) between the proteins playing a major role in penetration defense (AtPEN1/SlPEN1a/VvPEN1/HvROR2) and the other two proteins of subclade SYP-1b (AtSYP122/SlPEN1b) (Supplementary Figure S2).

In this study, a genome-wide survey allowed the identification of 21 syntaxins in the tomato genome. The number of tomato homologs is consistent with similar genome-wide studies showing that A. thaliana contains 18 syntaxins, O. sativa 14 and P. trichocarpa 22 (International Rice Genome Sequencing Project, 2005; Tuskan et al., 2006; Lipka et al., 2007). The identification of 10 putative tomato SYP1 members corroborates previous findings indicating that this subfamily, containing homologs localized in the plasma membrane, is more represented than the other subfamilies SYP2, -3, -4, and -8 in plants (Lipka et al., 2007) (Supplementary Table S3).

It is known that specific SYP1 homologs in Arabidopsis are involved in other physiological processes besides defense against pathogens. Specifically, SYP111 controls cytokinesis (Lauber et al., 1997), whereas SYP123 and SYP132 mediate root hair tip growth (Ichikawa et al., 2014). Furthermore, the SYP132 ortholog in Medicago truncatula homolog (MtSYP132) and Nicotiana benthamiana (NtSYP132) were found to be involved in the symbiosome formation of Sinorhizobium meliloti and Pseudomonas syringae resistance, respectively (Catalano et al., 2007; Kalde et al., 2007; Limpens et al., 2009). Phylogenetic analysis presented in this study suggests that the tomato syntaxin Solyc06g053760 is an ortholog of Arabidopsis SYP111. Evolutionary relatedness between SYP132 and a couple of tomato syntaxins, Solyc01g056810 and Solyc10g081580, might indicate that these genes are SYP132 co-orthologs due to a recent duplication event after the split of two lineages from an Arabidopsis-tomato common ancestor.

Two tomato syntaxins, SlPEN1a and SlPEN1b, are grouped in the phylogenetic clade also containing homologs previously reported to play a major (AtPEN1, HvROR2, and VvPEN1) or minor (AtSYP122) role in penetration resistance to powdery mildew fungi (Figure 1). Silencing of SlPEN1a compromised penetration resistance to both adapted and non-adapted powdery mildew pathogens in the Slmlo1 tomato line, showing that SlPEN1a is likely the functional ortholog of HvROR2, AtPEN1 and VvPEN1. This information might be useful for tomato breeding and for future studies on plant–pathogen interaction using tomato as model organism.

Similarly to barley, we found that the impairment of the tomato SlPEN1a syntaxin in the Slmlo1 background macroscopically restored the susceptibility to the powdery mildew disease. In contrast, the Atmlo2pen1 double mutant, although being characterized by wild-type levels of fungal penetration, still displayed a macroscopically resistant phenotype (Consonni et al., 2006). AtPEN1 was shown to act as a negative regulator of the hormone salicylic acid, known to enhance post-penetration defense mechanisms against biotrophic (Zhang et al., 2007). Possibly, differently from AtPEN1, SlPEN1a and barley HvROR2 have minor or no role in the regulation of salicylic-acid-mediated defense pathways. However, further testing needs to be done to support this hypothesis.

The Atpen1Atsyp122 double knock-out mutant displays a severely stunted and necrotic phenotype (Assaad et al., 2004). In the present study, transgenic individuals, silencing both SlPEN1a and SlPEN1b homologs, were devoid of markedly visible pleiotropic effects which might be due to residual gene expression. Alternatively, they may not affect the growth and development of the plant as their corresponding Arabidopsis orthologs. Also in this case, further tests are required to prove this hypothesis.

In this work, a codon-based evolutionary analysis allowed us to detect 77 codons under significant negative selection in the SYP-1b subclade, containing the syntaxins associated with defense against powdery mildew fungi. Moreover, alignment of all the 41 syntaxin sequences pointed out the occurrence of 24 residues specifically conserved in the SYP-1b subclade. Three of these residues are located in the Qa-SNARE domain, whereas 10 are found in the auto-inhibitory motifs Ha, Hb or Hc helices of Qa-SNARE proteins. These helices, when folded in a closed conformation, prevent the exposure of the Qa-SNARE domain and thus the formation of SNARE complexes (Collins et al., 2003). In barley ror2 mutant, a deletion of 31 amino acids (S118-E148) of the Hc helix occurred, which might have led to a constitutively open state of the protein resulting in enhanced binding to HvSNAP34 (Collins et al., 2003). Interestingly, the deletion contains four of the 24 conserved residues identified in this study, namely G123, P124, T133 and G138, in HvROR2 sequence, suggesting an important function of these residues.

Besides the above discussed 24 amino acids specifically conserved in the SYP-1b subclade, we pointed out two different amino acids present in AtSYP122 and SlPEN1b but not in any of the four SYP1b syntaxins with well-characterized role in plant–pathogen interactions, namely A111 and I286 (Supplementary Figure S2). In the study of Pajonk et al. (2008), authors obtained a chimeric syntaxin by swapping the first 175 amino acids of the AtPEN1 N-terminal domain (including the residue A111), with the corresponding sequence of AtSYP122. The construct was then used to transform pen1 mutant. The authors did not observe any difference in the level of Bgh entry rate between the obtained transformants and their corresponding genetic background. This observation suggests that N-terminal region of the AtPEN1 protein contains amino acids of critical importance for its functionality, possibly including the A111 residue mentioned above. Future functional analyses, aimed at targeting these amino acids, might unravel their importance toward the functional specialization observed between AtPEN1/ SlPEN1a/VvPEN1/HvROR2 and AtSYP122/SlPEN1b.