Interactions between the Powdery Mildew Effector BEC1054 and Barley Proteins Identify Candidate Host Targets.

There are over 500 candidate secreted effector proteins (CSEPs) or Blumeria effector candidates (BECs) specific to the barley powdery mildew pathogen Blumeria graminis f.sp. hordei. The CSEP/BEC proteins are expressed and predicted to be secreted by biotrophic feeding structures called haustoria. Eight BECs are required for the formation of functional haustoria. These include the RNase-like effector BEC1054 (synonym CSEP0064). In order to identify host proteins targeted by BEC1054, recombinant BEC1054 was expressed in E. coli, solubilized, and used in pull-down assays from barley protein extracts. Many putative interactors were identified by LC-MS/MS after subtraction of unspecific binders in negative controls. Therefore, a directed yeast-2-hybrid assay, developed to measure the effectiveness of the interactions in yeast, was used to validate putative interactors. We conclude that BEC1054 may target several host proteins, including a glutathione-S-transferase, a malate dehydrogenase, and a pathogen-related-5 protein isoform, indicating a possible role for BEC1054 in compromising well-known key players of defense and response to pathogens. In addition, BEC1054 interacts with an elongation factor 1 gamma. This study already suggests that BEC1054 plays a central role in barley powdery mildew virulence by acting at several levels.


■ INTRODUCTION
Microbial pathogens secrete effector proteins into host tissues and cells to facilitate infection. Some of these effectors play a crucial role by targeting key proteins involved in host immunity. This is well documented for bacterial pathogens of plants and animals. 1 Plant pathogenic fungi also produce arsenals of diverse effectors. 2 Large families of protein effectors have been described in biotrophic fungi such as Puccinia triticina wheat rust 3 and Blumeria graminis cereal powdery mildews. 4,5 These pathogens, like many biotrophs and mutualistic symbionts, develop specialized feeding structures called haustoria. It is now becoming clear that, in addition to taking up nutrients from the host, haustoria also play a central role in effector delivery to host cells. 6 Powdery mildew fungi are pathogens which infect a large number of plant species. For example, Blumeria graminis attacks economically important cereals such as wheat and barley. The genomes of two B. graminis "formae speciales" and several strains thereof have been sequenced. 5,7,8 This research identified about 6500 manually curated genes 9 assisted by existing transcriptome data 10 and large-scale proteogenomics. 11−14 Bio-informatic predictions drove the initial characterization of Candidate Secreted Effector Proteins (CSEPs) specific to powdery mildews. 4 These studies were accompanied by proteomic analysis that identified B. graminis proteins specifically associated with host cells colonized by haustoria which we termed Blumeria Effector Candidates (BECs). An initial functional screen found eight genes that, when downregulated by "Host-Induced Gene Silencing" (HIGS), led to reduced formation of haustoria, with the strongest reduction of 60 to 70% observed for BEC1054 and BEC1011, respectively. 15 Structure prediction revealed that BEC1011 and BEC1054, may possess an RNase fold. 4,12 We therefore refer to these as RNase-Like Proteins expressed in Haustoria (RALPH) effectors. It is notable that, in the B. graminis f.sp. hordei genome, RALPH effectors constitute the largest number (at least 120) of effectors arranged in several CSEP families. BEC1011 and BEC1054 (syn CSEP0264 and CSEP0064, respectively) belong to a small subgroup (Family 21) encoded by 5 genes. 4 Studying effectors and the host factors that interact with them has the potential to advance our understanding of the mechanism underpinning immunity. 16 An emerging concept is that many pathogen-specific effectors target a few universal resistance plant proteins involved in Pathogen Associated Molecular Patterns (PAMPs) Triggered Immunity (PTI) or Effector Triggered Immunity (ETI) (reviewed for bacterial effectors by . 17 One of the best studied examples is the RAR1-SGT1-HSP90 protein complex, where RAR1 is the first characterized gene in barley required for MLA10-based resistance against powdery mildew 18,19 and SGT1 associates with the Skp1-Cullin-F-box protein (SCF) ubiquitin ligase complex. 20 RAR1, which is conserved in plants and animals, is directly targeted by Pseudomonas syringae pv. tomato AvrB 21 or indirectly by effectors such as HopF2, AvrPto, AvrPtoB, AvrRPT2 or AvrRPM1 targeting the RAR1-associated protein RIN4 (for review see Deslandes and Rivas 17 ). Targeting of the RAR1-SGT1-HSP90-RIN4 complex by so many effectors suggests it plays a crucial role in resistance. Indeed, the complex is involved in surveillance and expression level control of a broad range of specific resistance (R) genes. 22,23 Based on the assumption that effectors target key components required for basal or gene-specific resistance, several studies have used protein-effector interactions to discover plant proteins that are involved in host resistance. Perhaps the most striking example is the identification of "interaction hubs" by large-scale protein interaction studies between Arabidopsis proteins and effectors from the bacterium Pseudomonas syringae, from the oomycete Hyaloperonospora arabidopsidis 24 or from the Arabidopsis powdery mildew fungus Golovinomyces orontii. 25 In these studies, over a hundred Arabidopsis proteins associate with many effectors including some already known to be important for the plant immune response, such as the TCP transcription factors and JAZ3.
In the study presented here, we investigated the barley interactome of the barley powdery mildew RALPH effector BEC1054. We used a recombinant BEC1054 protein for pulldown assays with extracts from whole leaves or the epidermis of healthy or powdery mildew-infected barley followed by the identification of host interactors by mass spectrometry (MS). To validate the interactions, we developed a targeted, quantitative, one-to-one yeast-2-hybrid (Y2H) approach. Our studies led us to identify barley proteins that interact with BEC1054. These findings contribute to the understanding of the mode of action of BEC1054 and reveal some key players of barley immunity.

■ EXPERIMENTAL PROCEDURES Plants and fungi used in this study
Barley (Hordeum vulgare cv. Golden Promise) was used for protein affinity pull-down, gene cloning and the propagation of Blumeria graminis f.sp. hordei (Bgh) isolate DH14. Plants were grown as described previously, and 7 day-old plants were inoculated with B. graminis f.sp. hordei, for maintenance and for experimental work. 15 The fungal tissues and barley epidermal strips were prepared as previously described. 10 Prior to extraction of nucleic acids and proteins, the material was ground to a fine powder in liquid nitrogen with some quartz sand using a mortar and pestle. All sampled biological material was stored frozen at −80°C.
Cloning of BEC1005 and BEC1054 into E. coli expression vectors RNA was extracted with guanidine isothiocyanate 10 from 3 or 4 barley leaves infected with B. graminis f.sp. hordei and collected 5 days post infection (dpi). The RNA was further purified using the RNAClean XP beads (Beckman Coulter, High Wycombe, UK). Poly(A) + RNA was then treated with DNase I and reverse-transcribed by oligo (dT) and random priming into cDNA, using the iScript synthesis kit (Bio-Rad, Hemel Hempstead, UK).
The cDNA was used as template to amplify BEC1005 and BEC1054 full length Coding DNA Sequences (CDS) (IDs CCU82697 and CCU83233 respectively; e! Ensembl Fungi database, http://fungi.ensembl.org/index.html), but excluding the signal peptide (as predicted by the SignalP 4.0 server; http://www.cbs.dtu.dk/services/SignalP/). PCRs were performed in 50-μL reactions with the AccuPrime Taq DNA Polymerase High Fidelity (Life Technologies, Life Technologies, Paisley, UK) in the presence of 5% DMSO. The PCR products were then cleaned using the QIAquick PCR Purification kit (QIAGEN, Manchester, UK). The purified PCR fragments were cloned into the pENTRY pCR8 vector, using the pCR8/GW/TOPO TA Cloning Kit (Life Technologies), followed by transformation into TOP10 E. coli cells.
BEC1005 and BEC1054 CDS were transferred from the pENTRY pCR8 vector into the Nova pET53-DEST Expression vector (Novagen, Merck Millipore, Watford, UK), using the LR Clonase II enzyme mix according to the manufacturer's instructions (Gateway, Life Technologies), but adjusting incubation of the recombination reaction to 4 h, followed by transformation into TOP10 E. coli. The pET53 vector (Novagen) is designed for expression of recombinant proteins with a short N-terminus 6 x His tag and a short C-terminus Strep-tag II. For the vector expressing BEC1054, stop codon was introduced upstream of the Strep-tag II. All inserts were checked by sequencing (GATC Biotech, London, UK).

Expression, solubilization and purification of recombinant BEC1005 and BEC1054 proteins
For recombinant BEC1005 and BEC1054 protein expression, the corresponding pET53 vectors were transformed into SoluBL21 E. coli cells for BEC1054 and BL21 (DE3)-pLysS E. coli cells for BEC1005. Selected colonies were grown overnight in 50 mL 2xYT medium supplemented with antibiotics (100 μg/mL ampicillin for BEC1054 and BEC1005 and 34 μg/mL chloramphenicol for BEC1005). In the morning, 1 L medium supplemented with antibiotics was inoculated with 22 mL of overnight cultures and cells were grown to OD 600 = 0.5−0.6, prior to induction in 1 mM IPTG and further growth for 3h at 220 rpm at 37°C. Bacteria were recovered by centrifugation at 3000 g. The bacterial pellets were flash frozen in liquid nitrogen and kept at −20°C until further use.
Bacterial pellets were resuspended in 30 mL of 20-mM Tris-HCl buffer pH 7.9 for BEC1054 and pH 8.5 for BEC1005 containing 150 mM NaCl, 0.1% Tween-20 and 1/2 a tablet of protease inhibitor cocktail (Roche, Burgess Hill, UK). Cells were lysed on ice with a sonication probe at 40% amplitude, in bursts of 2 s on, 2 s off. DNase I (3 μL at 5 mg/mL; Sigma-Aldrich, Gillingham, UK) was then added to the bacterial lysates, incubated on ice for 15 min. Insoluble particulates were then separated from the soluble fraction by centrifugation at 12,000 g, at 4°C for 30 min.

Article
Both BEC1054 and BEC1005 proteins were found in the insoluble fractions containing inclusion bodies. These were resuspended in 30 mL of their respective lysis buffer supplemented with 8 M urea. Inclusion bodies were resolubilised overnight on a rotating platform at 4°C prior to transfer into a dialysis tube (MWCO 12,000; Sigma). Proteins were then refolded by decreasing the urea concentration in solution in a stepwise fashion, using a modified ultrafiltration centrifugal dialysis protocol. 26 The dialysis bag was sealed and placed in a centrifuge bottle containing 800 mL of the lysis buffer, supplemented with successive decreasing concentrations of urea (6, 4, 2, 1, 0.5, 0.25, and 0 M). This was then centrifuged for 30 min in a SX4750 rotor (Beckman Coulter) at 3000 rpm at 4°C for each of the steps. Residual urea was then removed by dialysis overnight, against 3 L of the original lysis buffer (20 mM Tris HCl pH 7.9 and pH 8.5 for BEC1054 and BEC1005, respectively, with 150 mM NaCl and 0.1% Tween-20). Finally, particulates were removed by centrifugation at 12,000 g for 30 min at 4°C.
His-tagged BEC1005 and BEC1054 proteins were then purified from the refolded protein samples using affinity chromatography (Supporting Information Figures S-1 and S-2). The samples were loaded in 5-or 10-mL batches onto 2 × 1-mL HisTalon columns (Clontech) coupled to an AKTA Purifier FPLC system at 4°C (GE Healthcare, Little Chalfont, UK), pre-equilibrated in 20 mM Tris-HCl pH 7.9 and 8.5 (for BEC1054 and BEC1005, respectively), 150 mM NaCl, 0.1% Tween-20 and 20 mM imidazole. All purifications were performed at a constant flow rate of 0.5 mL/min. Then the affinity columns were washed in the loading buffer until the absorbance at 254 nm (A 254 ) baseline was stabilized. A 25 min gradient in loading buffer with increasing imidazole concentration from 20 to 600 mM was applied and 0.5-mL fractions were collected. Both BEC1054 and BEC1005 His-tagged BEC proteins eluted at 250−300 mM imidazole, as monitored by SDS PAGE (Supporting Information Figure S-2). BEC1054 was further purified by gel filtration on a Hi-Prep 16/60 Sephacryl S-300 column (GE Healthcare), pre-equilibrated in 50 mM Na-Acetate pH 4.6, 150 mM NaCl, using an AKTA Purifier (GE Healthcare) at 4°C.

Preparation of plant protein extracts for pull-down assays
Noninfected 7-day old barley seedlings of the cultivar "Golden Promise" were used directly or inoculated with B. graminis f.sp. hordei. Three different types of protein samples were prepared immediately before the pull-down assays and each experiment was repeated independently 3 times. For the infected epidermis sample (A), epidermal peels were harvested from infected primary leaves collected 48 h post inoculation (hpi). For the noninfected total leaf sample (B), noninfected primary leaves were collected. For the infected leaf sample (C), primary leaves were collected 5−7 dpi (Table 1).
For sample A, 0.5 g of powdered epidermal peels were resuspended in 1 mL of 50-mM Na-phosphate buffer pH 7.8, 300 mM NaCl, 0.01% Tween-20, 1 mM MgCl 2 , 1% polyvinyl polypyrrolidone (PVPP, w/v), in the presence of a protease inhibitor cocktail without EDTA (Roche) and 2 μL of 5 mg/ mL DNase I (Sigma-Aldrich) followed by two centrifugations at 12,000 g for 15 min at 4°C. The supernatant was collected and 500 μL were used in the pull-downs.
For sample C, 1.2 g of ground leaves were extracted in 10 mL of 50-mM Na-phosphate buffer pH 8.0, 150 mM NaCl, 1 mM β-mercaptoethanol, 2 mM MgCl 2 , 5% glycerol, 1% PVPP (w/v) supplemented with 1/500 protease cocktail inhibitor set IV (Calbiochem) and 2 μL of Benzonase Nuclease (stock concentration was >250 μL/ml; Sigma-Aldrich). All plant extracts (samples A, B and C) were incubated on ice for 15 min prior to the removal of insoluble particulates by successive 15 min centrifugations at 4°C, once at 4,000 g and twice at 20,000 g. Any particulates were further eliminated from the last supernatant by filtration through a 0.22-μm syringe filter. Protein concentrations (typically 1−2 mg/mL) were determined using a Bradford assay (Bio-Rad). Samples were kept on ice while setting up the pull-down assays to be performed on the same day.

Pull-down assays
All pull-down assays (summarized in Table 1

Journal of Proteome Research
Article were incubated on a rotating wheel with 1 mL cleared plant extract and 50 μg purified 6xHis-tagged BEC1054, BEC1005 (negative control), or in absence of recombinant tagged protein as supplementary negative control. Beads were incubated with rotation at 4°C for 1.5 h. After incubation, the magnetic beads were washed 4 times in 400 μL sample A lysis buffer (without PVPP). Interacting proteins were eluted in 150 μL lysis buffer supplemented with 300 mM imidazole, followed by another elution in 150 μL Laemmli sample buffer and boiling to check for any residual proteins binding to the NiNTA beads.
For sample B pull-downs, 1-mL of the purified His-tagged BEC1054 (1.5−1.8 mg) was first diluted with 3-mL of binding buffer prior to loading onto a 1-mL HisTrap-HP column coupled to an AKTA Purifier FPLC system, and equilibrated in the binding buffer at a flow rate of 0.5 mL/min. The binding buffer consisted of 20 mM Tris pH7.6, 150 mM NaCl supplemented with 50 mM imidazole. The column was washed with 80 mM imidazole until A 254 reached the initial baseline level. The freshly prepared plant extract of sample B was supplemented with imidazole pH 7.5 to a final concentration of 50 mM imidazole prior to loading it onto the column as 2 batches of 5 mL. The column was then washed with binding buffer containing 20 mM Tris pH 7.5, 500 mM NaCl and 80 mM imidazole until A 254 reached the original baseline. Interactors and BEC1054 were eluted in the same buffer but with 400 mM imidazole. A 254 was monitored and fractions with increased absorbance were further analyzed. The whole process was also carried out with no BEC1054 as negative control.
For sample C, 25 μg purified BEC1054 or BEC1005 were diluted in 700 μL of 50 mM Na-phosphate pH 7.8, 300 mM NaCl, 20 mM imidazole and 0.005% Tween 20, and bound to Ni-NTA-His DynaBeads (Life Technologies) equilibrated in the same buffer. A negative control with no BEC was also included. Following binding at room temperature for 10 min on a rotating wheel, beads were washed three times with 300 μL wash buffer. Beads were then equilibrated briefly in binding buffer (50 mM Na-phosphate pH 7.8, 150 mM NaCl, 0.005% Tween 20) prior to the addition of the protein extract and 20 min incubation. Following 3 washes in the wash buffer, binding proteins were directly eluted twice in 25 μL SDS-PAGE buffer, and denatured at 90°C prior to gel electrophoresis.

Protein identification by Gel-nLC-nESI MS/MS
Thirty microliters of the elution samples were separated on 2/3 of total gel length using 12% acrylamide/bis-acrylamide (37.5:1) tris-tricine SDS-PAGE minigels (Mini-Protean III system, BioRad) and stained with colloidal Coomassie blue (Supporting Information Figure S-3). Each lane was cut into 8 gel bands for in-gel digestion. Each gel band was cut into 2−4 mm 2 pieces and transferred to a 0.5 mL tube. Tryptic digestion was performed as described previously 27 with some modifications. In summary, the gel pieces were washed, each time for 10 min, twice in 100 μL of 25 mM ammonium bicarbonate (ABC) followed by 2 washes in 50 μL of ABC with 33% acetonitrile (ACN) and 3 washes in 40 μL of ABC with 50% ACN to ensure complete destaining of the gel pieces. The gel pieces were then dried in a vacuum dryer concentrator for 20 min. Gel pieces were rehydrated in 30 μL of 10 mM DTT in ABC and samples were incubated for 45 min at 50°C. After removing the excess of liquid, reduced cysteines were alkylated in the presence of 30 μL of 10 mg/mL iodoacetamide in ABC, and tubes were incubated at room temperature in darkness for 45 min. The remaining liquid was removed. Following three washes with 40 μL of 50% ACN in ABC for 10 min, samples were dried as described above. Thirty μL of trypsin at a concentration of 8 ng/ μL were added and the gel pieces were left to rehydrate for 15 min prior to overlaying 30 μL of ABC and overnight incubation at 37°C. Then peptides were extracted 3 times by incubating the digested gel pieces with 30 μL of 5% trifluoroacetic acid (TFA) in 50% ACN for 10 min and once with 2.5% TFA in 75% ACN. Eluates were collected, pooled, vacuum-dried and stored at −80°C until nESI MS/MS analysis for protein identification. Just prior to the analysis by nLC-nESI MS/MS, samples were resolubilised in 20 μL of 0.1% TFA, 2% ACN.
For the nanoLC run, an Ultimate 3000 RSLCnano system (Dionex/Thermo Fisher Scientific, Hemel Hempstead, UK) was used, loading 4 μL of each in-gel digest. Loaded samples were first washed and desalted for 5 min with 0.1% formic acid (FA)/ 2% ACN at a flow rate of 4 μL/min for 5 min on a Nano Trap Column (100 μm i.d. × 2 cm, packed with 5 μm diameter/100 Å porosity Acclaim PepMap100 C18 beads; Dionex/Thermo Fisher Scientific). The peptides were then separated on a 15 cm long UHPLC reverse phase analytical column (75 μm I.D., packed with Acclaim PepMap RSLC C18 2 μm/100 Å beads; Dionex/Thermo Fisher Scientific) by applying an ACN gradient in 0.1% FA at a flow rate of 250 nL/ min at 40°C, increasing the ACN content from 4% to 32% in 70 min followed by a sharper increase from 32% to 48% in 20 min. The column was connected online with a stainless steel emitter to spray the eluting peptides (Proxeon/Thermo Fisher Scientific) into an LTQ Orbitrap XL hybrid FT mass spectrometer (Thermo Fisher Scientific) as described previously. 12 The Orbitrap resolving power was set at 60,000 (at m/z 400) for MS scans spanning the m/z range from 400 to 2000 with an automatic gain control (AGC) of 1,000,000. Data dependent acquisition of the five most intense precursor ions, with z ≥ 2 and a minimum signal of 500, were sequentially isolated (with an isolation window of m/z of 2.0) and transferred to the LTQ linear trap with an AGC target of 20,000, fragmented by collision-induced dissociation (CID) and MS/MS scans were performed using a normalized collision energy activation of 35%, activation Q of 0.25 and the activation time was 30 ms. Ions were selected twice prior to a dynamic exclusion for 30 s. Data generated with the LTQ Orbitrap XL in *.RAW format were converted to Mascot generic format (*.mgf) files using the batch processor of ProteomeDiscoverer version 1.3 (Thermo Fisher Scientific) using settings suitable for Orbitrap MS analysis with full scan CID MS/MS in the ion trap in positive mode, including the following parameters: inclusion of 2+, 3+, 4+ and 5+ ions with a mass of 350−8000 Da, exclusion of nondeconvoluted ions as well as ions with a signal-to-noise ratio <3, and spectrum grouping function was not selected. For protein identification, batch searches were performed with an "in-house" server hosting the Mascot 2.4 search engine (initial searches were performed with version 2.3.02; MatrixScience, London, UK) using the batch processor Mascot Daemon 2.4 (MatrixScience). Data were first searched against the barley U35 and then the U36 EST assembly database (http://harvest.ucr.edu/, U36 contains 420888 sequences, 127065402 nucleotide residues), or the U36 + cDNA database described in Matsumoto et al. (2011) 28 containing 450594 sequences and 141089724 nucleotide residues). Later, when it became available, the predicted ORFeome derived from the annotated barley published genome 29 was downloaded from ftp://ftpmips.

Journal of Proteome Research
Article helmholtz-muenchen.de/plants/barley/public_data/. The database size was 10630728 amino acid residues and consisted of 26159 ORFs. Selected parameters for Mascot searches were tryptic peptides with a maximum of two missed-cleavages. Carbamidomethyl was selected as a fixed modification (Cys) and oxidation of Met and Pro were selected as variable modifications; mass tolerance was 15 ppm for MS and 0.8 Da for MS/MS; CID Ion trap was selected for the MS/MS protocol; the decoy function was selected for estimation of the false discovery rate (FDR). Peptide data were retrieved in CSV files as described in the Supporting Information section 1, selecting a significance threshold of P < 0.05, and a Percolator score cut off of 13, 30 conditions which gave a typical FDR below 1%, and rarely around 1%. A macro was designed in Excel (Microsoft) to select significant peptides and to remove duplicates, keeping peptides with the highest score in order to generate for each data sets a list of proteins with at least two significant peptides (as described in Supporting Information section 1. Proteins identified with the same set or with a subset of peptides were also included. Once the nonredundant list of proteins was generated for each interactome, these were compared to each other aided by the generation of pivot tables in Excel. Putative BEC1054 interactors considered were proteins that were only identified in the BEC1054 interactome ("BEC1054-only") but not in any of the BEC1005 or negative interactome samples within the same biological replicate or within the 3 biological replicates of the same type of experiment (Pull-downs A, B or C). Thus, BEC1054-only interacting protein lists were generated for the 3 types of samples: A (48h infected epidermis), B (noninfected leaves) and C (infected leaves 5−7 dpi). Proteins that were identified as such at least twice were selected for further analysis. Database accession numbers were used for protein selection, but since the harvest database may contain partial and redundant EST contigs, the UniRef90 description was also considered for comparison. From this list, candidates were selected for validation of interaction using a yeast-two-hybrid (Y2H) assay.

Preparation of yeast clones for Y2H assays
Selected coding sequences for putative BEC1054 interactors first identified by pull-down were retrieved from Uniprot (http://www.uniprot.org/). In order to generate for each data sets a list of proteins with at least two significant peptides (as CAGT-3′ and a specific reverse primer from the gene. Selected colonies were grown in liquid cultures containing 100 μM spectinomycin and plasmids prepared with the Plasmid Midi Kit (QIAGEN) for confirmation by sequencing (GATC Biotech). The CDS inserts were then transferred into the pDEST32 or the pDEST22 destination plasmids (to create pEXP32 and pEXP22 expression plasmids for Y2H) through LR recombination, using the Gateway LR Clonase II Enzyme mix (Life Technologies).
For Y2H, the ProQuest Two-Hybrid System with Gateway Technology was used (Life Technologies) following the manufacturer's instructions. Yeast strain MaV203 was made competent and transformed with plasmid pairs specified in Supporting Information Table S-3. For each yeast line, six separate transformed colonies were assayed.
Quantitative β-galactosidase assay to quantify Y2H protein interactions The chlorophenol red-β-D-galactopyranoside (CPRG) assay 31 was modified from the ProQuest Two-Hybrid kit manual. The following steps were performed: yeast lines were cultured on selective media plates, cells were recovered from individual colonies by aspiration, washed twice using 0.5 mL phosphatebuffered saline, and lysed through three freeze/thaw cycles before mixing 50 μL of the cleared supernatant to 100 μL of 2.23-mM CPRG reaction buffer in a 96-well microtiter plate. For each cell line, cell count was estimated by measuring OD 595 for 100 μL of intact cell suspension transferred in a 96-well plate. This was then used for normalization of the relative βgalactosidase activity calculation. An asymptotic exponential model for CPRG activity was fitted to the relative β-galactosidase activity. Its equation is y = a − e −cx (y, observed CPRG value at a time x; a, asymptotic value of y; y, maximum CPRG value; e, Euler's number; and c, inflection point of the curve). The curve was fitted using the nonlinear least-squares (nls) function in the R software (v3.0.2) and differentiated to calculate the maximum gradient at time zero, V i , which corresponds to the initial CPRG activity rate (Supporting Information Table S-6A). For each cell line, data sets were compared against their corresponding controls. For instance interaction data for the BEC1054+eEF1A(1) and the eEF1A(1)+BEC1054 transformed MaV203 yeast lines were compared to the BEC1054 only (MaV203 yeast lines transformed with plasmids pEXP32/BEC1054 and pDEST22) and eEF1A(1) only (pEXP32/eEF1a(1) with pDEST22), and are referred to hereafter as a data set. Bartlett tests were performed to determine whether the variance of the data for the yeast lines was homogeneous. 32 All but three of the data sets investigated showed nonhomogeneous variance (Supporting Information Table S6-B). For consistency, the data sets were therefore treated as though they all showed nonhomogeneous variance. Response variables for data sets containing negative values were transformed through the addition of a constant to all data points (+732), equal to the greatest negative value. A General Linear Model (GLM) was applied to each data set with "gamma" error distribution. 32 Once GLMs had been created, and it had been determined that the protein interaction tested has an effect on CPRG activity in the yeast line, Games-Howell posthoc tests 32 were performed to determine which pairs of means were significantly different (Supporting Information Table S6-C).
The interactions between BEC1054 and the selected barley candidates were also monitored with the ura3 and his3

Journal of Proteome Research
Article reporters, which encode auxotrophic markers for uracil and histidine, respectively. In the yeast line used, MaV203, basal expression of his3 allows sufficient histidine production in −His medium. HIS3 enzyme activity is inhibited by 3-amino-1,2,4triazole (3-AT). However, histidine levels are restored through activation of the his3 gene during a positive interaction, thus allowing colonies to grow in the presence of 3-AT. 33 The URA3 protein allows positive selection as an auxotrophic marker as well as negative selection by converting 5fluoroorotic acid (5FOA) to the toxic 5-fluorouracil product if there is interaction. 34 ■ RESULTS

Cloning, expression, solubilization and purification of BEC1054
The B. graminis f.sp. hordei RALPH effector BEC1054 (CSEP0064, Family 21) and the nonrelated effector BEC1005 glycosidase-like protein (used as negative control) 12,15 were cloned and expressed in E. coli with an Nterminal 6x His-tag to carry out in vitro protein affinity pulldowns with barley leaf extracts. Different induction conditions (temperature, bacterial strains, induction time) were attempted to express BEC1054 in a soluble form in E. coli. However, BEC1054 was expressed as a highly insoluble form. Therefore, a protocol from Hammond et al. (1980) 26 was adapted to produce a large quantity of renatured recombinant protein, prior to further purification (Supporting Information Figures S-1 and S-2).

In vitro affinity pull-down to identify barley proteins interacting with BEC1054
In vitro pull-downs allowed us to identify several barley proteins that bound to BEC1054. For this, purified recombinant N-Histagged BEC1054 (Supporting Information Figures S-1 and S-2) was used as bait to pull-down prey proteins from either infected, noninfected primary leaves or epidermis; these were the three experimental conditions (A,B,C) described in Table 1. Each experiment was performed in three independent replicates, including negative controls (no bait or BEC1005). SDS-PAGE analysis showed that many proteins were eluted in all of the 3 different pull-downs (Supporting Information Figure S-3). The bound barley proteins were then identified by nanoLC-ESI MS/MS from in-gel digests using the U36 Harvest EST database. For each biological replicate, "BEC1054 only pull-down proteins" were listed, removing the proteins identified in the corresponding negative controls. A list of putative interactors was then generated (as summarized in Table 2), selecting proteins identified in at least two replicates of the same experimental set or across at least 2 of the 3 experimental sets (A, B, C). From these, we tested two groups of putative interactors. The first set ( Table 2, light gray) includes proteins whose peptides were found only in the BEC1054 pull-down samples, i.e. an elongation factor 1-gamma 3 (EF1G3; Q5Z627), a glutathione-S-transferase (zGST IN2-1, Q8H8U5), and a Pathogenesis-Related protein 5 (PR5; O23997). The second set ( Table 2, dark gray) is formed of proteins found in some negative controls, but for which the UniRef90 descriptor was over-represented in the BEC1054 pull-downs compared to controls (Supporting Information Table S-4). From these, additional putative interactors selected for validation using Y2H included the malate dehydrogenase (Q6YWL3), ribosomal protein 40S S16 (Q0IQF7), an elongation factor EF1A (Q9LN13) and a nucleoside diphosphate protein kinase (NDPK, Q9LKM0). In particular, the malate dehydrogenase (MDH, Q6YWL3) was selected as a possible interactor because the U36 _10746 contig was specific to BEC1054 pull-downs and several other U36 contigs associated with the same MDH UniRef90 accession (Q6YWL3) were over-represented in more than 50% of all BEC1054 pull-downs, compared to less than 20% of all the negative control pull-downs (Table 2).
An additional list of "BEC1054 only" interacting proteins was generated at a later date, after reanalyzing the LC-MSMS data with the IBSC barley database because it contains better annotated nonredundant CDS 29 (Supporting Information Table S -5). This list included previously identified proteins such as a barley peroxidase 10 and different isoforms of PR5/ thaumatin-like protein and an elongation factor 1-gamma (eEF1G). We were unable to test several of these putative interactors because of uncertainties in the gene models of the respective genes and the resulting difficulties in amplifying unequivocally the cognate isoforms from barley. Further investigations to discriminate between isoforms of many of the proteins identified with the ISBC database is required prior to confirmation of their interaction with BEC1054 by Y2H.

Y2H to validate a selection of interactors
Here, we tested putative barley interactors of BEC1054 identified by pull-down using a targeted yeast-2-hybrid (Y2H) approach. These were: a glutathione-S-transferase (GST), a malate dehydrogenase (MDH), a thaumatin-like PR5, an RNase-like PR10, translation elongation factors eEF1G, eEF1A(1), eEF1A(3), 40S ribosomal protein 16S and NDPK. Some of these were selected for Y2H validation because of their likely colocalization and their capacity to bind RNA (like the BEC1054 RALPH effector); thus, PR10 is an RNase-like protein; 35−37 40S 16S is an intrinsic ribosomal protein; eEF1A and eEF1G are associated with mRNA translation.
In the Y2H assay used here, we measured the interactions between BEC1054 and prey proteins through the reconstitution of an active transcription factor which in turn activated expression of the three genes (lacZ, ura3 and his3), each of which possesses an independent promoter. 33 A β-galactosidase assay was used as a quantitative indicator of the interaction Table 2. continued a UniRef90 identifier was used to search for and recover the full length CDS barley sequences available in Uniprot. Putative interactors further investigated are highlighted in gray. Light gray indicates that there was no identification observed in the negative controls, suggesting a possibly genuine interaction. Dark gray indicates that there were identifications of the protein in some negative controls. b "B54" columns: values indicate the number of times that a protein was identified in biological and technical replicates of an experimental set with the same U36 accession number in a B54 pull-down but not in the negative pull-downs (no bait or BEC1005) of the same corresponding biological replicate. c "Total" columns: values indicate the number of times that a protein was identified with the same U36 accession number in all the pull-downs (BEC1054, no bait or BEC1005) of the same corresponding biological replicate. If present in the negative control, the value of the "total" column is larger than for the "B54" column, indicating that the protein was seen in negative controls but of a different biological replicate. d The "Tot" columns indicate the total number of times that a protein was identified with the same U36 accession number in all the pull-downs (BEC1054, no bait or BEC1005).

Journal of Proteome Research
Article between prey and bait by determining the maximum conversion rate of the substrate chlorophenol red-ß-Dgalactopyranoside (CPRG) at the start of the reaction (Vi) (Figure 1). β-galactosidase activity quantification reflected the expression level of the cognate lacZ reporter gene. Transformed yeast lines were also analyzed by plating them on medium containing 5-fluoro-orotic acid (5FOA) and 3-amino triazole (3AT) (Figure 1). Growth on media lacking leucine, tryptophan and histidine (-His) is also shown for comparison. Inhibition of growth of 5FOA or growth on 3AT compared to controls indicates an interaction between bait and prey partners. Yeast lines containing pEXP32/BEC1054 and pDEST22 were used as a negative control. For each interaction tested, a yeast line expressing the recombinant barley putative interaction protein bound to the pDEST32 bait protein was used as an additional negative control.
Yeast clones transformed with the plasmid pEXP32/ BEC1054 grew poorly, and the transformation success rates were low for these yeast lines. This suggested that BEC1054 alone had a moderately toxic effect when expressed as a bait protein; this is not an uncommon observation. 38 The β-galactosidase activity (V i ) for BEC1054-zGST, BEC1054-MDH, and BEC1054-PR5 combinations in both bait−prey orientations was higher than the negative controls, as represented on a boxplot graph (Figure 1 and Supporting Information Tables S-6), indicating that BEC1054 interacted with barley zGST, MDH, and PR5 in yeast. Furthermore, the V i was significantly higher for BEC1054 with 40S 16, eEF1G, and eEF1A(1) in one bait−prey orientation but not in the other orientation ( Figure 1 and Supporting Information Tables S-6). There was no significant V i increase in yeast expressing BEC1054 with eEF1A(3), NDPK, or PR10 (Supporting Information Figure S-4 and Tables S-6). The increase in V i for PR10 (pEXP32/PR10 pDEST22) and NDPK (pEXP32/ NDPK pDEST22) alone compared to the controls indicated that PR10 and NDPK may act as transcriptional activators.
Yeast lines coexpressing GST, PR5, or MDH with BEC1054 grew on media containing 3AT and grew less on media containing 5FOA (Figure 1; Table 3). Yeast lines containing BEC1054-40S 16 (pEXP32/BEC1054 and pEXP22/40S 16) grew on media containing 3-AT, and grew weakly on media containing 5FOA, whereas the opposite bait−prey pair did not grow at all. These observations corroborated the galactosidase assay results, confirming that GST, PR5, and MDH interact (weakly) with BEC1054 in yeast and that 40S 16 interacted in one bait−prey orientation only. Yeast expressing BEC1054-eEF1G grew on media containing 3AT, whereas the opposite bait−prey pairing did not. Both yeast lines containing BEC1054

Journal of Proteome Research
Article and eEF1G showed decreased growth on 5FOA (Figure 1; Table 3). This data indicated weak interaction in one orientation, and some evidence for interaction in the opposite orientation, as was found in the galactosidase assay. The yeast lines expressing BEC1054 and the two eEF1a proteins, PR10 and NDPK, behaved like the negative controls on 3AT and 5FOA media, as they did not grow on media containing 3AT and grew well on media containing 5FOA (as summarized in Table 3). Therefore, this confirmed the results obtained with the galactosidase assay: there was no interaction between BEC1054 and either of the eEF1a, PR10, or NDPK proteins in yeast. None of the lines tested grew on selective media lacking uracil, suggesting only a weak interaction between BEC1054 and its preys (Supporting Information Table S-6 and data not shown).
In summary, the Y2H results showed that BEC1054 is capable of a direct physical interaction with several barley proteins. BEC1054 interacted with GST, MDH, and PR5, in both bait-prey orientations. BEC1054 also interacted with 40S 16 and eEF1G, but only in a single bait-prey orientation ( Figure 1; Table 3). This was mostly in agreement with the pull-down assays, as PR5, GST and eEF1G were only identified in BEC1054 pull-downs and in none of the negative controls (Table 2). MDH, was abundant (over-represented) in the Table 3. Summary of Evidence of Interaction Comparing Y2H Assays (as Detailed in Figure 1) with the Affinity Protein− Protein Interaction (PPI) by Pull-down/MS Identification Matching the U36_Harvest Database (as detailed in Table 2) a a The values in the "PPI B54" columns indicate the number of replicates within one of the A, B or C datasets (described in Table 1) for which a protein was identified (U36 Harvest database) only in the BEC1054 pull-downs but not in any of the negative controls of that same replicate.This was summed in the "PPI B54 ABC" column. The "PPI tot ABC" column indicates the number of times a protein has been observed, in "B54" or negative interactions. The "Y2H" columns described the outcome of the CPRG galactosidase, 5FOA and 3AT assays in both B54-x and x-B54 orientations, shown in Figure 1 and this is summarized in the "Y2H Interact. evidence" column. The general PPI and Y2H interaction outcomes are stated in the "PPI-Y2H Interact. evidence" column. Evidence of interactions are characterized as Y = Yes, W or P = weak or partial and N = no interaction. Assays showing partial or lack of evidence are gray shaded. PPI was performed for 3 biological replicates for each of the 3 data sets (A,B,C and summed in the ABC column, detailed in Table 1) and 2 technical replicates for A. Values in "PPI B54" columns indicate the number of times that a protein, as defined by a U36 identifier, was identified in biological and technical replicates of an experimental set with the same U36 accession number in a BEC1054 pull-down but not in the negative pull-downs (no bait or BEC1005) of the same corresponding biological replicate. Shaded gray cells indicate that the particular protein was observed in a negative control of other biological replicates. b Total of "B54" columns. C "PPI tot ABC" column: values indicate the number of times that a protein was identified with the same U36 accession number in all the pull-downs (BEC1054, no bait or BEC1005) of the same corresponding biological replicate. If present in the negative control, the value of the "total" column is larger than for the "B54" column, indicating that the protein was seen in negative controls but of a different biological replicate. If this was the case, it was then gray-shaded. d CPRG galactosidase assay shown in Figure 1 showed an increased V i value when there is interaction. Y = Gal activity is significantly higher than the negative controls and indicates interaction. N = Gal activity is identical to or not significantly different from negative control. e B54-x/x-B54 yeast lines were grown in medium containing 5FOA (Yeast growth inhibition indicates interaction). Y = less growth evidencing interaction. W = weak inhibition for weak interaction. N = no growth inhibition, i.e. no interaction. f B54-x/x-B54 yeast lines were grown in medium containing 10 mM 3AT (favors growth when there is interaction). Y = more growth evidencing interaction. W = weak growth suggesting weak interaction. N = no growth, i.e. no interaction. g Y2H summary conclusion for evidence of interaction between BEC1054 and protein x in B54x/x-B54 orientations. Y = evidence of interaction, P = partial evidence of interaction, N = no interaction. h Y2H assays in negative controls (PR10 or NDPK) gave a high value, suggesting that these porteins themselves act as transcription activators. i Shaded gray cells indicate that the particular protein was observed in negative controls of other biological replicates in the PPI assays, or that there were no significant differences with the negative controls for the various Y2H assays.

Journal of Proteome Research
Article BEC1054 pull-downs compared to the negative pull-downs, and was only identified in a small number of negative controls. The ribosomal protein 40S S16 interacted consistently, but weakly, in yeast in one bait-pray orientation only (Figure 1; Table 3) and was identified in numerous negative control pulldowns (Table 2); therefore, not enough experimental evidence was allowed to confirm interaction between BEC1054 and this ribosomal protein. There were no direct interactions between BEC1054 and eEF1A(1), eEF1A(3), NDPK, or PR10 in yeast, and these were identified in many negative control pull-downs; thus, data were not corroborated (Table 3).

Multiple barley proteins interact with BEC1054
In this study we found that the RALPH effector BEC1054 (CSEP0064), a virulence factor in barley powdery mildew, 15 interacts with several host proteins ( Figure 2). We used two orthogonal approaches to determine the protein−protein interactions with BEC1054: affinity precipitation/MS-based analysis for primary identification and targeted Y2H for validation. In our study, BEC1054 interacted with four proteins in both assays: a GST, a malate dehydrogenase (MDH), a thaumatin-like protein (PR5), and an elongation factor (eEF1G). In addition, a weak interaction was shown with the Y2H assays between BEC1054 and a ribosomal protein (40S 16S). However, this interaction was not entirely specific in the affinity purification assays. Other well-characterized bacterial, oomycete, and fungal effectors interact with many plant proteins. 24, 25,39 Large-scale studies demonstrated that effectors from three taxonomically unrelated pathogens interact with several host proteins, some of which overlap. The latter are considered to represent hubs of effector action. 24, 25 Our study leads to a new challenge: understanding the biological significance and mechanistic role of the affinities between BEC1054 and the host interactors. We discuss some of these possibilities below for those host proteins for which we have validated affinity by Y2H.

Glutathione-S-transferases (GST)
Glutathione-S-transferases (GSTs) conjugate glutathione with electrophilic molecules. 40 GSTs are involved in responses to stress in concert with glutathione peroxidase activity, providing protection by reducing cytotoxic hydroperoxides to alcohols. 40−44 However, the exact function of the majority of GSTs is still poorly understood. We found that BEC1054 interacts with a GST from the zeta class (zGST) in pull-down and Y2H assays. zGST possesses glutathione peroxidase activity in Arabidopsis, 45 maize, 46,47 soybean, 48 and wheat. 40,49 Other GSTs were identified in the affinity precipitation assays, including the elongation factor eEFG, which possesses a GST domain involved in protein interaction (through formation of disulfide bridges 50 ). In addition, GSTs act as antioxidant enzymes. Oxidative stresses, such as those produced during the hypersensitive response, increase GST expression or activity. 40,51−55 The interaction between the effector and the zGST could modulate the oxidative stress response within infected cells; the fungus may thus decrease the levels of cytotoxic hydroperoxides. If so, this could in turn prevent the death of the infected cell, allowing the biotrophic fungus to access nutrients from the host. Verifying this hypothesis will require an understanding of the exact function(s) of this particular zGST in barley.

Malate dehydrogenase (MDH)
A malate dehydrogenase (MDH) isoform also interacted with BEC1054, as it was over-represented in pull-downs with BEC1054, when compared with the negative controls, and the Y2H data clearly showed a direct interaction between MDH and BEC1054. MDH is a key enzyme of the citric acid cycle, where it reversibly catalyzes the oxidation of malate to oxaloacetate through the reduction of Nicotinamide Adenine Dinucleotide (NAD+) to NADH. 56,57 In addition to the wellknown role of MDH in primary metabolism, accumulation of malate during the apoplastic oxidative burst is observed following elicitation with Colletotrichum lindemuthianum fungal extracts. 52,58 Moreover, malate levels increase in susceptible plants infected with Magnaporthe oryzae, the causal agent of rice blast, 59 and in Arabidopsis cell cultures treated with fungal elicitors. 60 Following rice blast infection, malate accumulation in a susceptible interaction is associated with decreased NADP/ malate dehydrogenase activity and reduced H 2 O 2 production at the sites of penetration. 59 NAD(P)H produced by MDHs may be used as a reductant by NADPH oxidases or cell wall peroxidases for ROS production. 59,61−64 Likewise, malate metabolism plays a role in plant basal defense in Arabidopsis. In mutants lacking the NADP-malic enzyme, ROS production and callose papilla formation are impaired. 65,66 It is therefore possible that MDH is targeted by effectors, such as BEC1054, to modulate its activity, resulting in altered levels of ROS as discussed above for zGST.

Pathogenesis related (PR) proteins
BEC1054 interacts with a barley thaumatin-like protein or PR5. Interestingly, another barley powdery mildew effector protein, CSEP0055, interacts with barley PR1 and PR17. 67 PR proteins are up-regulated in plants following pathogen invasion. 66,68−71 In particular, PR5 accumulates in barley leaves following B. graminis infection. 72 The families of the PR proteins are defined by shared biochemical or biological properties, and many are associated with limiting pathogenic activity, spread, or growth. 69, 73 Enzymatic activities have been assigned to a number or PR proteins, for example endoglucanases, 74 Figure 2. Proposed model for the Blumeria effector BEC1054 virulence factor interacting with barley proteins. BEC1054 is secreted from the haustorium into the plant epidermis, where, to compromise the host defense or to maintain biotrophy, it targets a host malate dehydrogenase (MDH), a pathogenesis-related protein 5 (PR5), a glutathione-S-transferase (GST), and an elongation factor-1-gamma (eEF1G). This suggests that BEC1054 is likely to interfere with plant defense mechanisms, redox homeostasis, and ribosome functionality, perhaps at the level of protein translation.

Journal of Proteome Research
Article chitinases, 75 and proteinase inhibitors; 76 the functions of others, such as PR5, are less clear. A range of roles have been attributed to PR5, including antifungal activity. 77 Some PR5 proteins were shown to induce hyperpolarization of the Fusarium graminearum plasma membrane and possess 1,3-β-Dglucan binding activity. 78 One intriguing possibility is that fungal effectors, such as BEC1054 and CSEP0055, could target host PR proteins and moderate their antifungal activity; however, this is yet to be tested. Additional PR proteins were pulled down with BEC1054 and will need to be validated by Y2H or other means to consolidate the idea that one effector might interact with several PR proteins. These included the PR5 isoform (Q5MBN2) and PR4e (also called Win3 or Barwin), which has an endoglucanase (cellulase domain) and a chitin-binding domain (ProDom) as well as homology to the tobacco Hevein (http://www.uniprot.org/uniprot/Q6PWL8).

Ribosomal proteins
The ribosomal protein, eEF1G, binds to BEC1054 in both in vitro affinity pull-downs and in yeast in one bait−prey orientation only. Moreover it cannot be excluded that BEC1054 is interacting directly or indirectly with many more ribosomal proteins, but this still needs to be confirmed. eEF1G is involved in the formation of the eukaryotic elongation factor 1 complex, and it is associated with ribosome inactivating proteins (RIPs). 79 BEC1054 interaction with eEF1G suggests that the powdery mildew effector targets the host ribosome; this could alter the ribosome and interfere with its activity or may protect it from host RIPs activity. It remains to be seen what the consequences of these interactions are on disease development.
We originally also selected the 40S 16 protein (also known as Ribosomal Protein 16, RPS16). However, further reanalysis of the pull-down data with more recent databases showed unspecific binding, and the Y2H interacted in one prey−bait orientation only. We therefore believe there is scant evidence that BEC1054 targets this protein directly.

■ CONCLUSIONS
Our study indicates that the RALPH effector BEC1054 may act at different levels to compromise plant defense mechanisms through interactions with four barley proteins: a barley PR protein (PR5), a ribosomal protein (eEF1G), and proteins involved in primary metabolism and ROS production and signaling (MDH and zGST). These findings support the idea that one effector can have multiple targets, as observed in genome-scale Y2H studies of plant−pathogens protein interactions. 24, 25 The challenge for future work is to determine the functional significance of these protein−protein interactions in the development of powdery mildew disease. It also remains to be seen whether any of the closely related RALPH effectors of the CSEP Family 21, such as BEC1011, share these affinities and thus explain mechanistically how these central effectors work in concert to promote powdery mildew pathogenicity in barley infections.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jproteome.5b00732.
Method section with workflow for protein IDs and peptide information extracted from Mascot results and script for the macro "Mascot_csv_to_XL"; accessions of Blumeria effectors and barley putative BEC1054 interactors investigated in this study (Table S1); description of primers used for cloning of the Blumeria effectors B1005 and B1054 and putative interactors (Table S2); combinations of plasmids used to investigate protein−protein interactions in yeast-two-hybrid (Table  S3); BEC1054 putative interactors grouped as UniRef90 IDs (Table S4); putative BEC1054 interactors identified with the IBSC database in pull-downs A, B, or C (Table  S5); data for relative β-galactosidase activity using the CPRG assay (Tables S6); expression, extraction, renaturation, and purification of BEC1054 (1A) and BEC1005 (1B) (Fig. S1); chromatographic purification of recombinant BEC1054 (A) and BEC1005 (B) (Fig.  S2); SDS PAGE of protein affinity pull-downs (Fig. S3); graph of all interactions tested by Y2H with the relative β-galactosidase CPRG assay (Fig. S4)