Evaluation of the cDNA prey library

Our cDNA prey library was commercially made (Norclone Biotech Laboratories, Ontario) from RNA extracted from uninfected 4-5 week old Arabidopsis rosettes, as well as plants infiltrated with a virulent pathogen (P. syringae pv. tomato DC3000, Pto DC3000), a non-virulent strain lacking the T3SS apparatus (Pto DC3000 ΔhrcC), and an avirulent strain translocating a T3SE recognized by an Arabidopsis R protein (Pto DC3000 with AvrRpm1, which is recognized by RPM1). Although it is common to amplify primary cDNA libraries after their initial construction, this step can potentially introduce representation biases that may influence the interactor screen. We amplified the primary library by semi-solid amplification as this method is believed to reduce overall amplification bias [22]. We first sequenced the primary and secondary cDNA prey libraries to assess representation and bias arising from the initial library construction and subsequent amplification. Amplified DNA was sequenced on an Illumina GA-IIx using standard protocols at the University of Toronto Centre for the Analysis of Genome Evolution and Function (CAGEF). Sequencing of the primary- and amplified libraries yielded 59.8 M and 5.8 M reads respectively, which read-mapped to ~11 K Arabidopsis genes (4,119 M and 213 M bases of data, Additional file 1). The cDNAs ranged from ~40 nt up to ~2,800 nt, with most not being full length since they were generated by random hexamer priming (Additional file 2A, B). A scatterplot of the hits/locus for the two libraries revealed very high congruence (R² = 0.96, Additional file 2C), indicating that very little bias was introduced during the amplification process.

To further evaluate the range of genes represented in the library, we analyzed the gene ontology (GO) annotations of the cDNAs recovered (Figure 1A). Many biological processes were represented including metabolism, response to stress or stimulus, development, transport and signal transduction, although unknown was the most common (42%). The subcellular localization of many of the cDNAs was also unknown; however, loci were identified in virtually every cellular compartment including 13% associated with membranes. Genes with known molecular functions included those involved in protein binding, hydrolase activity, nucleic acid metabolism, transcription factors, transporters, and kinases. We further examined publicly available microarray data available through the CAGEF Bio-Array Resource (BAR, http://bar.utoronto.ca, [23]) to determine whether loci represented in our library were differentially regulated in responses to biotic stress from bacteria (P. syringae), oomycetes (Botrytis cinerea, Phytophthora infestans, Golonivomyces orontii), or elicitors of innate immunity (harpins, lipopolysaccharides, and an oomycete elicitor NPP1). 38% of genes were upregulated more than 2-fold in response to biotic stress while 46% did not respond to biotic stress (Figure 1B). A further 16% did not have probe sets to detect transcriptional changes arising from biotic stress. None of our library loci were downregulated in response to biotic stress.

Quantitative Interaction Screen Sequencing (QIS-Seq)

Since most members of the HopZ family of T3SE proteins are membrane-associated by myristoylation [11], we adapted the split-ubiquitin yeast two-hybrid system that was developed for transmembrane bait proteins (Figure 2A and Additional file 3) [21]. Based on membrane-association studies of K-Ras, we constructed a bait vector with a C-terminal prenylation signal and polybasic sequence in order to stably associate the protein with membranes (Additional file 4) [24, 25].

We screened the Arabidopsis prey library by cloning catalytic mutants of all five of the HopZ alleles, as well as the Arabidopsis R protein ZAR1 that recognizes HopZ1a into our membrane-tethered bait vector (Figure 2A,Additional files 3, 4). We used the catalytic mutants of the HopZ T3SEs in order to prevent processing of putative substrates and potentially stabilize transient interactions that would occur with the active enzyme. This method has been termed inactive catalytic domain capture (ICDC) and has proven successful for stabilizing enzyme-substrate interactions [26, 27]. We also screened the prey library with HopF2 _Pto , which is a P. syringae T3SE that is unrelated, both with respect to function and sequence, to the HopZ family, as well as luciferase as a non-specific protein control since it should not specifically interact with any Arabidopsis proteins. We transformed our baits of interest by lithium acetate/PEG transformation and screened them on plates with dropout media lacking histidine, one of the interaction reporter genes. Approximately 2000 colonies of each bait were collected and then restreaked twice on dropout media lacking tryptophan in order to preferentially retain the prey plasmid and lose the bait plasmid thereby reducing the complexity of the plasmid pool for next-generation sequencing [28]. The colonies were harvested en masse, digested with lyticase, and then plasmids were purified using alkaline lysis. Finally, the prey-plasmid inserts were amplified with vector-specific primers using low-cycle PCR to reduce amplification bias and the PCR product was Illumina sequenced. Each bait provided 4.7 M to 33.6 M quality reads (176 M to 2,544 M bases of data, Additional File 1) which were read-mapped to the Arabidopsis reference using Novoalign (http://Novocraft.com), which performs base-quality aware, global alignments with affine gap penalties using full implementation of the Needleman-Wunsch algorithm. The number of reads per Arabidopsis coding sequence were converted to reads per million in order to normalize across samples.

We then assessed for overall enrichment of each candidate interactor by scaling the number of hits observed between our bait of interest and each candidate interactor relative to the frequency of those candidates in the primary library. The candidates' general 'stickiness' was also assessed by the number of times it was recovered using luciferase, our non-specific bait protein. Specifically, enrichment was calculated as (T3SE-luciferase)/(T3SE+luciferase+library) *100, where each term is scaled as the number of mapped-reads per million (Figure 2B). This enrichment measure scales from 0 to 100 with higher values corresponding to those candidate interactors that do not bind luciferase (are not sticky) and are rare in the library (Additional file 5).

Functional analysis of HopZ2-interactors

We measured P. syringae growth in Arabidopsis lines carrying T-DNA insertions for each HopZ2^C/A specific interactor to determine if the candidate HopZ2 interacting proteins played any role in P. syringae disease or resistance. We focused on interactors for which there were confirmed homozygous T-DNA insertion lines available and that were predicted to have the T-DNA insertion in an exon of the gene, and thus be more likely to interrupt the protein (5/33 loci, Table 1). We assayed for changes in immunity by infiltrating the T-DNA insertion lines with the virulent pathogen Pto DC3000 and evaluating bacterial growth after three days. Insertions in genes At4g35750, At5g20700, At4g00430 and At1g68440, showed no difference in Pto DC3000 growth compared to the wild-type Col-0; however, an interruption in gene At1g11310 (line mlo2-7), encoding MLO2 showed a ten-fold decrease in Pto DC3000 growth (Figure 3A). To further assess if this locus plays a role in resistance of Arabidopsis to Pto DC3000 we tested an additional T-DNA insertion line in At1g11310 (mlo2-6, Figure 3B). Bacterial growth was reduced by approximately 10-fold in mlo2-6 compared to Col-0 wildtype (Additional file 6A) indicating that the mlo2 mutation increases resistance to Pto DC3000.

The mlo2-11 (pmr2-1) mutant of Arabidopsis, which is a point mutant (D287N) in MLO2 Figure 3B), was previously identified in a genetic screen for mutants showing enhanced resistance to powdery mildew [29, 30]. However, unlike mlo2-6 and mlo2-7, but consistent with the previous report [29], mlo2-11 (pmr2-1) does not exhibit any significant change in P. syringae growth compared to Col-0 (Additional file 6B).

HopZ2 targets Arabidopsis MLO2

We further examined whether knocking out MLO2 would affect the virulence advantage normally conferred by HopZ2 when carried by the weak Arabidopsis pathogen P. syringae pv. cilantro 0788-9 (Pci 0788-9; [11]). We infiltrated Arabidopsis Col-0 and mlo2-7 leaves with Pci 0788-9 carrying either an empty vector (EV) or HopZ2. As previously observed [11], HopZ2 conferred a virulence advantage to Pci 0788-9 in Col-0, but the HopZ2-associated virulence advantage was compromised in the mlo2-7 line (Figure 3C), indicating MLO2 is required for HopZ2-associated virulence.

To determine whether HopZ2 and MLO2 interact in planta, we created fusions between both HopZ2^C/A and MLO2 to the N- or C-terminus of YFP (nYFP or cYFP) in a glucocorticoid-inducible conditional expression vector. We used a partial clone of MLO2 beginning at amino acid residue 281 of the full-length protein and containing the 4^th, 5^th, 6^th and 7^th transmembrane domains as well as the C-terminal cytosolic tail of the protein (MLO2^Δ1-280), corresponding to the fragment of the clone in our cDNA prey library (Figure 3B).

We infiltrated equivalent optical densities of Agrobacterium carrying HopZ2::nYFP, HopZ2^C/A::nYFP or HopZ1c::nYFP with MLO2^Δ1-280::cYFP, as well as the reciprocal combination. We used HopZ1c as a negative control because it did not interact with MLO2 in our yeast two-hybrid screening (Additional file 5, data not shown). Protein expression was induced by spraying the plants with dexamethasone post-infiltration. 72 and 96 hours after dexamethasone application we observed bright fluorescence in leaf sections co-infiltrated with HopZ2::nYFP or HopZ2^C/A::nYFP and MLO2^Δ1-280::cYFP, as well as the reciprocal combination (Figure 4; Additional file 7A). No fluorescence was observed with HopZ1c::nYFP and MLO2^Δ1-280::cYFP (or the reciprocal combination) at these time points. The interaction between MLO2^Δ1-280 and HopZ2 localized to the periphery of the cell suggestive of the plasma membrane as well as reticulate network reminiscent of the endoplasmic reticulum (ER; Figure 4). This localization pattern was also observed when an MLO2^Δ1-280::YFP fusion was transiently expressed in N. benthamiana (Additional file 7B).

Bacterial strains and routine culture conditions

Escherichia coli and Agrobacterium tumefaciens were grown in Luria-Bertani broth, and Pseudomonas syringae was grown in King's broth (KB). Antibiotics were used at the following concentrations: for E. coli, 50 μg/mL kanamycin, 100 μg/mL ampicillin; for A. tumefaciens: 100 μg/mL kanamycin, 100 μg/mL rifamcipin; and for P. syringae: 50 μg/mL kanamycin, 50 μg/mL rifamcipin and 50 μg/mL cycloheximide.

Plant growth conditions

Arabidopsis plants were grown with 9 h of light (~130 microeinsteins m^-2 s^-1) and 16 h of darkness at 22°C in Sunshine #1 soil supplemented with 20:20:20 fertilizer. Nicotiana benthamiana plants were grown with 9 h of light (~130 microeinsteins m^-2 s^-1) and 16 h of darkness at 22°C in Sunshine #1 soil supplemented with 20:20:20 fertilizer and osmocute.

Cloning

Pfu polymerase (Fermentas) was used for all cloning and all constructs were confirmed by sequencing. For the split-ubiquitin constructs, bait genes were amplified by PCR to contain an in-frame HA epitope, a polybasic region (K6 or K8) and a CAAX box, as well as appropriate unique restriction sites. The bait-HA-K6-CAAX genes were cloned into the pBT3-N vector (Dualsystems Biotech) using Sfi I. The bait-HA-K8-CAAX genes were cloned into the pTLB-1 vector (gift of Dr. Igor Stagljar, University of Toronto) using Nco I. The orientation of each gene in the vector was confirmed.

For the split-YFP constructs, the HopZ genes or the 3' end of MLO2 were amplified by PCR to contain an in-frame HA epitope and appropriate unique restriction sites. All of the genes for the split-YFP system were cloned using Xho I and Stu I into pBD-nYFP or pBD-cYFP. pBD-nYFP and pBD-cYFP were modified from pTA7002 [47] to add an HA tag and the N- or C-terminus of YFP between the Stu I and Spe I sites. The N-terminus of YFP encompasses residues 1-155 while the C-terminus of YFP includes residues 156 to the stop codon.

The constructs used for plant infectivity assays were previously described [11]. In brief, the HopZ allele is expressed under its native promoter and contains an in-frame HA tag.

cDNA library

Five week old Arabidopsis rosette leaves were infiltrated by hand with a needleless syringe with P. syringae pv. tomato DC3000 (Pto DC3000), Pto DC3000 carrying AvrRpm1 or the ΔhrcC mutant of Pto DC3000 at an optical density of 0.1 (~5 × 10⁷ CFU/mL) at 600 nm. Infiltrated leaves were harvested at 4 hpi (Pto DC3000, Pto DC3000 carrying AvrRpm1, or Pto DC3000 ΔhrcC) or 12 hpi (Pto DC3000, Pto DC3000 ΔhrcC). Uninfiltrated leaves were harvested at 4 pi and 12 hpi. RNA was extracted using Trizol (Invitrogen). mRNA was cloned into the pPR3-N vector (Dualsystems Biotech) using the Sfi I sites (Norclone Biotech Laboratories, Ontario) with the NubG at the N-terminus of the prey proteins. The library contained ~2.3 × 10⁹ clones, with an average size of 1.2 kB and was 90% recombinant. Amplification of the library was carried out by the semi-solid method [22]. 0.5 μL of the primary library in E. coli strain DH5α was inoculated into 2× LB broth with 0.3% Seaprep agarose (FMC, Rockland) and 100 μg/mL ampicillin. The inoculated cultures were then incubated in a water-ice bath for 1 hour. Subsequently, the inoculated cultures were incubated at 30°C for 44 hours without shaking. To sequence the primary and secondary libraries, low-cycle PCR amplification was carried out with pPR3-N vector-specific primers and a high-fidelity Taq/proofreading polymerase mix (Fermentas, Burlington). This pool of DNA was sheared and prepared for Illumina sequencing by standard methods.

Yeast two-hybrid screening

HopZ1a^C/A, HopZ1b^C/A, HopZ1c^C/A, HopZ2^C/A, and ZAR1^CC were expressed under the weak CYC1 promoter in the pBT3-N vector, while HopZ3^C/A was expressed under the strong TEF1 promoter in the pTLB-1 vector. AP-4 yeast [48] carrying the bait construct were transformed using the PEG/LiAc method. In brief, yeast carrying the bait construct were subcultured in 300 mL SD-Leu overnight to an optical density of 0.6 at 600 nm. Yeast were washed twice in sterile H₂O and resuspended in 1.5 mL. Transformations were performed with 1 μg of cDNA library, 200 μL of yeast cells and 600 μL of PEG/LiAc (50% PEG, 120 mM LiAc, 10 μL 10 mg/mL boiled salmon sperm DNA) by the heat shock method at 42°C for 45 min. Yeast were washed twice in sterile H₂O and plated on SD-LeuTrp and SD-LeuTrpHis + 3-amino-1,2,4-triazole (3-AT). Interacting colonies were identified by growth on SD-LeuTrpHis + 3-AT. The appropriate amount of 3-AT was determined for each bait by testing for growth when transformed with the positive control pFur4-NubI and a lack of growth with pFur4-NubG [48]. Screening was performed until ~2000 interacting colonies were identified. Colonies were restreaked twice on SD-Trp to preferentially lose the bait plasmid [10] and grown at 28°C. Prior to plasmid isolation, colonies were restreaked onto SD-Trp and grown at 28°C. Yeast were harvested en masse in SD-Trp and pelleted at 1000 g for 5 min. The pellet (~ 5 g) was washed in 0.1 M NaPO₄ pH 7.4 and 1.2 M sorbitol and resuspended in 7.1 mL of lyticase buffer [0.1 M NaPO₄ pH 7.4, 1.2 M sorbitol, 500 μL lyticase (Sigma), 50 μL of 10 mg/mL RNAseA]. 25 KU of lyticase was resuspended in 0.01 M NaPO₄ pH 7.4 and 50% glycerol and used immediately. Yeast were incubated in the lyticase buffer overnight at 37°C. The Qiagen alkaline lysis spin kit was used to extract the plasmid DNA. The lyticase buffer was used instead of buffer P1. Volumes of buffers P2 and N3 were scaled up to the typical Qiagen proportions based on the volume of the resuspended pellet (1 vol lyticase buffer: 1 vol P2: 1.4 vol N3). Lysis in the P2 buffer was done for 15 min at room temperature and 15 min at 65°C. Buffer N3 was chilled prior to use. After addition of buffer N3, the yeast were incubated on ice for 20 min. Yeast were pelleted at 14000 rpm for 30 min at 4°C. The supernatant was removed and cleared again by centrifugation at 14000 rpm for 15 min at 4°C. The supernatant was loaded onto multiple Qiagen spin columns to purify the plasmid DNA. The columns were washed with buffer PB and buffer PE. Plasmid DNA was eluted with 50 μL of buffer EB after a 1 min incubation. A second elution was performed with 35 μL of buffer EB after a 1 min incubation. To sequence the putative interactors, low-cycle PCR amplification was carried out with pPR3-N vector-specific primers and a high-fidelity Taq/proofreading polymerase mix (Fermentas, Burlington). This pool of DNA was sheared and prepared for Illumina sequencing by standard methods.

Illumina Sequencing

Illumina sequencing was performed either with 37 cycle single reads or 72 cycle paired-end reads (Additional file 1) following the manufacturer's protocol on an Illumina GAIIx and pipelined using the GA pipeline v1.4.

Bioinformatics

Illumina reads were mapped to Arabidopsis gene models downloaded from NCBI, using a short read mapping tool novoalign (novocraft.com). From the mapping data, the number of mapped reads and the total length of mapped regions for each gene were determined with in house scripts. Read numbers per gene were further normalized as reads per million (rpm) within each sample and compared among the samples. The enrichment of a specific interactor with a bait of interest was determined by considering the number of reads obtained with the bait and luciferase and normalizing against the abundance of reads for luciferase, the bait and the library (Additional file 5). The percentage of the mapped length was calculated using length of mapped regions and the theoretical length of the gene model. Gene Ontology terms for Arabidopsis genes were downloaded from the TAIR website, and assigned to the genes in the cDNA library. Up- or down-regulation of each gene in response to biotic stress was determined from microarray data available through the CAGEF Bio-Array Resource (BAR, http://bar.utoronto.ca, [23]. Biotic stress treatments in the BAR included inoculation with virulent, avirulent and non-host P. syringae, inoculation with oomycetes (Botrytis cinerea, Phytophthora infestans, Golonivomyces orontii) and inoculation with elicitors of innate immunity (harpins, lipopolysaccharides, and an oomycete elicitor NPP1).

Genotyping of T-DNA insertion lines

The following T-DNA insertion lines were used: SALK_060284C (At5g20700), SALK_082464C (At4g35750), SALK_079850C (At1g11310, mlo2-7), SALK_050191C (At1g11310, mlo2-6), SALK_024490C (At1g68440), SAIL_808_A10 (At4g00430). Genomic DNA was extracted from a leaf of 5-6 week old Arabidopsis plants. Primers were designed using the iSct feature in the SIGnAL database. Primer sequences are available upon request. PCR-based genotyping was employed to determine the homozygosity or heterozygosity of the individuals. PCR products were sequenced using Big Dye Terminator 3.1 on an ABI 3730 genetic analyzer.

Infectivity assays

For infiltration, P. syringae was resuspended to an optical density of 0.1 (~5 × 10⁷ CFU/mL) at 600 nm and diluted to a concentration of 1 × 10⁵ CFU/mL for growth curves. Plants were infiltrated by hand with a needleless syringe, as previously described [49]. Four disks (1 cm²) were harvested, ground in 10 mM MgCl₂, and plated on KB with rifampicin and cycloheximide on days 0 and 3 for colony counting.

Statistics

For growth assays, 8-10 plants were used for day 3 counts. Significance was determined using Fisher's Protected Least Significant Difference (PLSD) on the day 3 count data. Error bars indicate one standard deviation of the mean.

Agrobacterium transient expression assays and BiFC

Five-milliliter A. tumefaciens GV2260 cultures were grown overnight at 28°C in Luria-Bertani broth with kanamycin and rifampicin. The next day, the cultures were washed twice in induction medium (50 mM MES pH 5.6, 0.5% (w/v) glucose, 1.7 mM NaH₂PO₄, 20 mM NH₄Cl, 1.2 mM MgSO₄, 2 mM KCl, 17 μM FeSO₄, 70 μM CaCl₂, 200 μM acetosyringone) [50], and 3.75 mL was inoculated into 35 mL fresh induction medium to grow overnight. The following day, cultures were spun down, washed twice in 10 mM MES pH 5.6 with 200 μM acetosyringone and resuspended to an optical density of 0.4 at 600 nm. The culture containing the MLO2^Δ1-280-cYFP plasmid was mixed in equal volumes with a culture containing the HopZ1c-nYFP, HopZ2-nYFP or HopZ2^C/A-nYFP plasmid. The culture containing the MLO2^Δ1-280-nYFP plasmid was mixed in equal volumes with a culture containing the HopZ1c-cYFP, HopZ2-cYFP or HopZ2^C/A-cYFP plasmid. The underside of the leaves of 5- to 7-week-old N. benthamiana plants were infiltrated by hand with a needleless syringe. Plants were sprayed with 20 μM dexamethasone (Sigma) 1-2 hours after inoculation. Sections of leaves were imaged with a Leica SP5 microscope using Leica software at 24 hours (YFP fluorescence) or 72-96 hours (BiFC) post-dexamethasone induction.