Host targets of effectors of the lettuce downy mildew Bremia lactucae from cDNA-based yeast two-hybrid screening

Plant pathogenic bacteria, fungi and oomycetes secrete effector proteins to manipulate host cell processes to establish a successful infection. Over the last decade the genomes and transcriptomes of many agriculturally important plant pathogens have been sequenced and vast candidate effector repertoires were identified using bioinformatic analyses. Elucidating the contribution of individual effectors to pathogenicity is the next major hurdle. To advance our understanding of the molecular mechanisms underlying lettuce susceptibility to the downy mildew Bremia lactucae, we mapped a network of physical interactions between B. lactucae effectors and lettuce target proteins. Using a lettuce cDNA library-based yeast-two-hybrid system, 61 protein-protein interactions were identified, involving 21 B. lactucae effectors and 46 unique lettuce proteins. The top ten targets based on the number of independent colonies identified in the Y2H and two targets that belong to gene families involved in plant immunity, were further characterized. We determined the subcellular localization of the fluorescently tagged target proteins and their interacting effectors. Importantly, relocalization of effectors or targets to the nucleus was observed for four effector-target pairs upon their co-expression, supporting their interaction in planta.


82
Generation of the Y2H prey library and bait constructs 83 A Lactuca sativa cv. Olof cDNA library was constructed using Invitrogen Custom Services (Invitrogen,   84 Carlsbad, CA). Briefly, RNA was isolated using phenol/chloroform extraction from mock-treated Library screening 120 The mating method was used for library screening [13]. Diploid yeast cells were plated on Sc -Leu -121 Trp -His medium to identify interacting bait-prey combinations. Retransformation of bait and prey 122 candidates was performed to confirm interactions. A detailed description can be found in S1 file.

124
Bioinformatic analysis of effectors and lettuce targets week-old N. benthamiana plants were co-infiltrated with an A. tumefaciens strain carrying the P19 156 silencing suppressor [26] in combination with strains harbouring bait or prey fusion constructs 157 resuspended in infiltration buffer (   The most represented family of proteins in the screen is the prenylated rab acceptor (PRA1) family for which fragments of six members were identified and three members passed all selection criteria. PRA1 genes encode small transmembrane proteins that localize to the secretory pathway in Arabidopsis and are proposed to play a role in vesicular trafficking in plants [31]. Also, they have previously been described as targets of G. orontii effector candidates in Arabidopsis [9]. In lettuce the PRA1 protein family is composed of 17 members. The six Y2H identified PRA1 proteins were found in four clades of the phylogenetic tree (Fig 2A). To determine the specificity of the interactions between B. lactucae effectors and PRA1 proteins, a targeted Y2H assay was performed. All six identified prey constructs containing lettuce PRA1 gene fragments were cotransformed with three interacting and one non-interacting B. lactucae effector in yeast and plated on selective medium of increasing stringency. The negative control BLR09 did not interact with any of the PRA1 members. The interactions between effectors BLR27, BLR32 and BLN01 with individual PRA1 proteins as identified in the library screens were confirmed. Furthermore, BLR27 and BLR32 interacted with more PRA1 proteins than determined by the library screen, indicating that the Y2H screening was not exhaustive and not all possible interactions were captured. In contrast, BLN01 interacted robustly with LsPRA1.B1 but only weakly with other PRA1 proteins (Fig 2B). Two of the three PRA1 proteins that passed all selection criteria belong to clade F. Interestingly, the Arabidopsis PRA1 proteins that interacted with G. orontii and H. arabidopsidis effectors also belong to clade F [9]. Twelve plant targets were selected for validation of protein-protein interactions in planta, including the top ten ranked effector targets by total number of yeast colonies. In addition, LsBPM3 and LsERF093 were chosen because of their predicted functions in targeting proteins for degradation and transcriptional regulation respectively, as will be described in more detail later. Using domain prediction and orthologous relationships to Arabidopsis proteins, nine out of the twelve were given gene names.
To further explore the selected effector targets, a closer look was taken at the prey plasmidencoded gene fragments. DNA sequences obtained by Sanger sequencing were aligned to the corresponding lettuce coding sequences. Lettuce gene fragments that included the start codon of the lettuce coding sequence (Fig 3) were detected for eight preys. Yeast clones of the remaining four preys only contained fragments that started downstream of the predicted start codon. The use of a cDNA library that contains both full-length coding sequences and gene fragments likely increased the number of identified protein-protein interactions. Of the 15 effectors described here, four localized exclusively to the cytoplasm and two showed both cytoplasmic and nuclear localization (Fig 4 and Table 3). Immunoblots of YFP-BLR18 showed a large fraction of free YFP and a minor fraction corresponding to the intact fusion protein. Therefore, it is possible that the observed nucleocytoplasmic signal does not reflect YFP-BLR18 but mainly free YFP.
YFP-BLR28 and BLR38-YFP contain predicted nuclear localization signals (NLSs) (S2 Table) and localized solely to the nucleus (also see [12]). Six effectors contain a predicted C-terminal transmembrane domain (Table 1 and S3 Table) and were thus expected to localize to membranes. We were unable to detect BLN03 and BLN04 C-or N-terminal fusion proteins. YFP-BLR05 and YFP-BLR09 predominantly labeled the endoplasmic reticulum (Fig S1), whereas YFP-BLR08 was associated with ring-like structures of various sizes in line with our previously reported findings [32]. YFP-BLR12 appeared as punctate structures. Upon prolonged overexpression, YFP-BLR12 also labeled the endoplasmic reticulum suggesting that the punctate structures represent components of the secretory pathway.
However, this requires confirmation using appropriate markers. Interestingly, BLG03 that is recognized in Dm2 lettuce lines [11], does not contain predicted transmembrane domains, but YFP-BLG03 was associated with the plasma membrane ( Fig S1) similarly to BLN06-YFP and BLR40-YFP [12]. It is possible that post-translational modifications (S4 Table) contribute to observed membrane localizations.  Subsequently, the localization of the tagged lettuce proteins was analyzed. Seven of the twelve tested proteins localized to the cytoplasm and/or nucleus (Fig 5 and (Fig S2). Upon co-expression of the effectors with lettuce proteins, complete or partial colocalization was observed for 10 out of 13 combinations (Fig 6 and Table 3). As expected, the cytoplasmic localized effectors YFP-BLR11, YFP-BLR27 and YFP-BLN01 colocalized upon co-expression with their cytoplasmic targets CFP-LsHSP90-11, CFP-Lsa015570.1 and CFP-LsFER3. Interestingly, in several cases coexpression induced a relocalization, specifically, to the nucleus. Relocalization of effector BLG02 occurred in the presence of the nuclear localized protein CFP-LsBPM3: YFP-BLG02 was mainly cytoplasmic and, in some cells, nuclear localized, but showed a consistent nuclear and cytoplasmic (nucleocytoplasmic) signal in cells that also expressed CFP-LsBPM3 ( Fig 6). The cytoplasmic effector YFP-BLR20 relocalized to the nucleus (Fig 6) upon co-expression of nuclear localized CFP-LsERF093 and the intensity of the YFP nuclear signal was dependent on CFP-LsERF093 signal intensity.
In other cases, effector expression led to target relocalization. Effector BLR38-YFP induced a relocalization of its target CFP-LsFLX-like2 from punctate cytoplasmic structures to the nucleus (Fig 6).
In summary, ten of the thirteen effector-lettuce protein pairs colocalized in N. benthamiana.
Furthermore, relocalization occurred upon expression of four effector-lettuce protein pairs providing evidence that the Y2H-identified interactions also occur in planta.

Features of the B. lactucae effectors
RxLR effectors are restricted to oomycetes belonging to the clade of the Peronosporales. In each species a number of these effectors contain structurally conserved WY domains that adopt an α-helical fold and are linked to W/Y/L sequence motifs [21,30]. The WY-domain is present in 44% of the Phytophthora RxLR effectors, but is less abundant in downy mildews: 18% of the Plasmopara halstedii [29] and 26 % of the H. arabidopsidis [21] RxLR effectors contain WY domains. Our analysis revealed that from the set of 46 B. lactucae effectors originating from isolate Bl:24 seven (15%) effectors have predicted WY domains. Orthologs of these seven effectors were also identified in isolate SF5 [33]. The differences in WY-domain abundance may reflect a bias in the WY-domain algorithm towards Phytophthora sequences; other downy mildew effectors may contain a, so far, unrecognized conserved fold.

Yeast two-hybrid interactions between B. lactucae effectors and lettuce proteins
The cDNA library-based Y2H screen resulted in lettuce targets for 21 (46% of screened) effectors. These effectors interacted on average with three lettuce proteins. The vast majority (83%) of identified lettuce proteins interacted only with a single B. lactucae effector. Previously published screenings with C. elegans and human fragment libraries resulted in interactors for 37% and 31% of bait proteins respectively corresponding to an average of 2.2 and 2 interactors per bait [34,35]. An average of 3.4 interactors per effector was obtained in screens with effectors from G. orontii, H. arabidopsidis and P.
syringae against a library of ~8000 immune-related full-length Arabidopsis proteins [8,9]. The number of obtained interactors in our screen is thus comparable to other published Y2H screens.
Based on the identification of Arabidopsis proteins that interacted with effectors from three pathogens [9], it has been proposed that effectors converge on conserved host proteins. This proposition fits with insights into the mechanisms by which independent bacterial type III effectors converge on immune-related proteins such as MAPK proteins [36], SERK3/BAK1 [37][38][39][40][41] and RIN4 [42].
We were interested to determine if those proteins identified as major hubs, i.e. interacting with effectors from multiple pathogens, would also emerge in our screens with B. lactucae effectors.
The list of Arabidopsis hubs was dominated by TCP transcription factors [9]. Specifically, TCP13, TCP14 and TCP15 interacted with effectors originating from all three pathogens, and TCP19 and TCP21 interacted with effectors from at least one pathogen [9]. TCPs operate as transcriptional activators or repressors in plant growth and development [43,44]. Lettuce TCP family members were also detected as interactors of B. lactucae effectors. However, the effector-TCP interactions weakly activated the HIS3 reporter, were frequently observed as single colonies in Y2H screens and TCPs showed autoactivation in the Y2H system. Weßling and colleagues also classified Arabidopsis CSN5A as a hub due to its interaction with 12 P. syringae effectors, 11 H. arabidopsidis effectors and 9 G. orontii effectors in Y2H screens [8,9]. In our library screens, lettuce CSN5 was identified with two B. lactucae effectors, but also showed weak activation of the HIS3 reporter in the presence of an empty bait vector. Interaction between CSN5 and the GAL4 DNA binding domain was previously reported using multiple GAL4-DBD based vectors and yeast strains [45][46][47]. Rejecting CSN5 as false positive because of stronger reporter gene activation in the presence of B. lactucae effectors than empty vector is only partially defendable. Effectors may have unforeseen additive effects on reporter gene activation but these effects could be interaction independent. Thus, though some of the previously reported hubs may represent host proteins that play a prominent role in disease susceptibility, others may be false positives.

Several B. lactucae effectors localize to membranes in planta
We examined the localization of a subset of 15 B. lactucae effectors and assessed their stability using immunoblot. Analysis of the YFP-BLR20 fusion protein revealed multiple bands corresponding to monomeric, dimeric and multimeric forms of this effector. It is possible that a multimer represents the biologically active form of BLR20 as multimerization of two P. infestans effectors, PexRD2 and PiSF3, was shown to be required for interaction with their targets, a host MAP kinase and U-Box-kinase protein respectively [21,48,49]. The multimerization of BLR20 is possibly driven by the presence of WYdomains that have been proposed to support oligomerization [21]. This was previously shown for PiSF3 by mutation of two residues facing each other across the α-helices of PiSF3 that disrupted the formation of oligomers [48].
B. lactucae effectors BLG03, BLN06 and BLR40, which are recognized in specific lettuce lines, localized to the plasma membrane [11,12]. The plant membrane network was also targeted by 26% of tested H. arabidopsidis effectors [50] and 12% of Plasmopara viticola effectors [51]. Coiled-coil domains [52] and post-translational modifications such as N-myristoylation, could contribute to anchoring of effectors that lack transmembrane domains, to the plasma membrane [50,53,54]. Six effectors -BLN03, BLN04, BLR05, BLR08, BLR09 and BLR12 -were predicted to contain a single Cterminal transmembrane domain based on TMHMM analysis. Effectors BLR05, BLR08, BLR09 and BLR12 were indeed found to be associated with the endomembrane system: YFP-BLR05, YFP-BLR09 and YFP-BLR12 strongly labelled the endoplasmic reticulum, whereas BLR08 was associated with ringlike structures of varying sizes. The B. lactucae effectors were cloned without their signal peptide, which is required for co-translational ER targeting of transmembrane proteins by the signal recognition particle [55]. Instead membrane integration of tail-anchored proteins occurs post-translationally and is dependent on a C-terminal transmembrane domain that functions as a targeting signal. Some proteins with a moderately hydrophobic TM domain can insert unassisted in the ER, others with a highly hydrophobic TM domain require assistance. In yeast, the TM domain of tail-anchored proteins is recognized by components of the GET system including Get3 (TRC40 in mammals), delivered to the ER-resident Get1/Get2 receptor complex and inserted into the membrane [56][57][58][59]. Clearly, the ER localization of YFP-BLR05, YFP-BLR09 and YFP-BLR12 proves that these effectors associate with the membrane post-translationally. This immediately prompts the question how membrane-associated effectors are delivered to host cells. A role for extracellular vesicles in the delivery of soluble effectors has been hypothesized [60] and vesicular transport would provide a suitable vector for membrane-associated effectors as well. The C-terminal transmembrane domain of YFP-BLN03 and YFP-BLN04 may not have been sufficient for posttranslational insertion providing a possible explanation for the lack of fluorescence of these fusion proteins.
Effector BLG03 targeted the plasma membrane despite lacking a putative transmembrane domain. Post-translational modifications resulting in protein lipidation such as prenylation, Nmyristoylation or palmitoylation can affect membrane-protein interactions [61]. Putative Nmyristoylation sites were implicated in the membrane targeting of multiple bacterial type III effectors [54,62]. Alternatively, phospholipid conjugation [61] or phospholipid binding may contribute to membrane association. Multiple studies have explored the role of the N-terminal RXLR motif and Cterminal residues in phospholipid binding [63]. Though the relevance of phospholipid binding for effector uptake remains unclear, perhaps a role in effector localization in planta could be considered.

Co-expression of effectors and interactors induces protein relocalization to the nucleus
Studies on H. arabidopsidis and P. viticola RXLR effector localization revealed a preference for nuclear (including nucleocytoplasmic) localization (66% and 83% respectively) [50,51]. In our research, only two effectors -BLR28 and BLR38 -were strictly nuclear localized in N. benthamiana. Strikingly, B.
lactucae RXLR effector BLR38 that is recognized in L. serriola LS102, induced a relocalization of its target LsFLX-like2 to the nucleus. LsFLX-like2 shows homology to the Arabidopsis FLOWERING LOCUS C EXPRESSOR (FLX) protein family of which two members are involved in flowering time control in Arabidopsis [64]. The punctate structures of CFP-LsFLX-like2 resemble the structures formed by cytoplasmic FLX homodimers in Arabidopsis [65]. Heterodimers of FLX with FRIGIDA or FRIGIDA ESSENTIAL 1 were restricted to the nucleus [65], suggesting that the observed relocalization of CFP-LsFLX-like2 upon co-expression with BLR38-YFP may reflect a different interaction state. BLR28 also induced relocalization of its target, DnaJ protein LsDjA2, to the nucleus. DnaJ proteins function as cochaperones to 70 kDa heat shock proteins to aid in protein folding, relocalization and degradation [66].
Furthermore, two predominantly cytoplasmic effectors, BLG02 and BLR20, displayed nucleocytoplasmic localization when co-expressed with their targets, LsBPM3 and LsERF093. LsERF093 contains an AP2/ERF domain that is typically found in members of the AP2/ERF transcription factor family. These proteins are known to be involved in responses to biotic and abiotic stress in Arabidopsis, tomato and rice [67,68]. LsBPM3 contains a BTB/POZ (broad complex, tram track, bric-a-brac ⁄POX virus and zinc finger) domain that is also found in 80 Arabidopsis proteins and is known to mediate protein-protein interactions. Specifically, the BTB/POZ domain acts as a substrate-specific adaptor for Cullin-RING E3 ubiquitin Ligases (CRLs) that ubiquitinate proteins targeted for degradation by the proteasome. Recruitment of substrates may occur via a Meprin and TRAF homology (MATH) domain [69,70]. The Arabidopsis MATH-BTB/POZ protein BPM3 localizes to the nucleus and promotes degradation of transcription factor ATHB6 that negatively regulates abscisic acid signaling [71].
The relocalization events that we observed, have two important implications. First, they suggest that physical interactions between RXLR effectors and targets occur in planta, which may affect the activity of the target and consequently, disease development. This principle was recently demonstrated for P. sojae effector Avh52 that recruits a cytoplasmic host transacetylase, GmTAP1, into the nucleus. Nuclear-localized GmTAP1 is able to transacetylase histones, which enhances susceptibility of N. benthamiana to P. capsici [73]. Also, Arabidopsis protein RD19 that was discovered as interactor of Ralstonia solanacearum type III effector Pop2 in a Y2H screen, relocalized to the nucleus in the presence of the effector and physically associated with Pop2. The effector target, RD19, was required for the resistance protein RRS1-R mediated immune responses [74] Thus, relocalization events are often associated with physical interactions between effector and target, and effector-target interactions can affect disease susceptibility. Functional analysis of the lettuce targets is required to provide further insights into the mechanisms underlying B. lactucae susceptibility.
The second implication is that screening by Y2H is a useful method to identify host targets.
However, the biological relevance of the observed interactions needs to be demonstrated in follow-up experiments and there is no universal rule (yet) for selecting the most promising candidates. We investigated LsFLX-like2/ BLR38 and LsDjA2/ BLR28 because they belonged to the top ten ranked interactions in the Y2H screen. In contrast, BLG02/ LsBPM3 and BLR20 / LsERF093 were selected based on literature demonstrating a role for the target candidate protein families in immune responses. It should furthermore be stressed that neither of these four targets function as hub in our Y2H screen: all of them interacted with a sole effector. Conversely, we have previously demonstrated that small networks identified in Y2H screens can be valuable i.e. LsNAC069 interacted with four -BLN04, BLR05, BLR08 and BLR09 -effectors in Y2H assays and was demonstrated to affect plant responses to both biotic and abiotic stress [32]. To conclude, our Y2H screens have uncovered multiple relevant targets whose precise function we are just beginning to unravel.