SHR4z, a novel decoy effector from the haustorium of the parasitic weed Striga gesnerioides, suppresses host plant immunity

Summary

• Cowpea (Vigna unguiculata) cultivar B301 is resistant to races SG4 and SG3 of the root parasitic weed Striga gesnerioides, developing a hypersensitive response (HR) at the site of parasite attachment. By contrast, race SG4z overcomes B301 resistance and successfully parasitises the plant.

• Comparative transcriptomics and in silico analysis identified a small secreted effector protein dubbed Suppressor of Host Resistance 4z (SHR4z) in the SG4z haustorium that upon transfer to the host roots causes a loss of host immunity (i.e. decreased HR and increased parasite growth). SHR4z has significant homology to the short leucine‐rich repeat (LRR) domain of SOMATIC EMBRYOGENESIS RECEPTOR‐LIKE KINASE (SERK) family proteins and functions by binding to VuPOB1, a host BTB‐BACK domain‐containing ubiquitin E3 ligase homologue, leading to its rapid turnover.

• VuPOB1 is shown to be a positive regulator of HR since silencing of VuPOB1 expression in transgenic B301 roots lowers the frequency of HR and increases the levels of successful SG4 parasitism and overexpression decreases parasitism by SG4z.

• These findings provide new insights into how parasitic weeds overcome host defences and could potentially contribute to the development of novel strategies for controlling Striga and other parasitic weeds thereby enhancing crop productivity and food security globally.

Short abstract

See also the Commentary on this article by https://doi.org/10.1111/nph.16504.

Introduction

Parasitism has evolved multiple times within the angiosperm kingdom and today parasitic weeds are found in a wide variety of ecological habitats around the world (Barkman et al., 2007; Westwood et al., 2010). Parasitic weeds derive all (holoparasitic) or part (hemiparasitic) of their nutrition from their host plants by attaching to either the roots or aerial parts with a unique organ, the haustorium, which forms a physical and physiological conduit between host and parasite (Kuijt, 1969). While most parasitic weeds remain innocuous members of plant communities, some have established themselves as noxious and persistent pests in agricultural fields and pose a serious constraint to crop productivity. Among the latter are the witchweeds (Striga spp.), a genus of root hemiparasites that currently infest about two‐thirds of the farmland under cultivation in sub‐Saharan Africa (Parker, 2009). It has been estimated that witchweed infestations cause losses of yield in excess of 10 billion USD annually (Scholes & Press, 2008), directly affecting the lives and livelihoods of > 300 million small share‐holder, low‐input farmers in this region.

Striga species form two main groupings based on host preference (Mohamed et al., 2001). One group that includes S. hermonthica, S. asiatica, and S. aspera parasitises members of the Poaceae (grass family) including the agronomically important food and forage grains: maize (Zea mays), sorghum (Sorghum bicolor), rice (Oryza sativa), and pearl and finger millets (Pennisetum glaucum, Eleusine coracana). The second group includes S. gesnerioides, the most morphologically variable and widely distributed of the witchweeds, which preferentially attacks dicots, including wild and cultivated legumes, and members of the Convolvulaceae, Euphorbiaceae and Solanaceae (Mohamed et al., 2001). Among the primary hosts for S. gesnerioides is cowpea (Vigna unguiculata L. Walp), the most important food and forage legume in the African Sahel (Timko & Singh, 2008; Singh, 2014).

While most cowpea cultivars are susceptible to parasitism by S. gesnerioides, several resistant landraces and local accessions have been identified (Lane et al., 1993; Timko & Singh, 2008). Resistance in these cowpea genotypes takes on one of two different phenotypic manifestations: a hypersensitive response (HR) at the site of parasite attachment to the host root with a subsequent discoloration and death of the parasite within 3–4 d, or little to no host root response at the attachment site but subsequent arrested growth of the parasite seedling early in the tubercle swelling stage of development (Lane et al., 1993; Botanga & Timko, 2005). In the case of tubercle arrest, the parasite fails to make vascular connections with the host and shows no evidence of cotyledon expansion. Lane et al. (1993) were the first to document that some cowpea genotypes (cultivars and local accessions) showed a differential ability to be parasitised by S. gesnerioides isolates collected from different geographical locations, leading to the suggestion that distinct races of the parasite exist in West Africa (Lane et al., 1993, 1994). Subsequent studies of host differential resistance responses and molecular phylogenetic analysis of parasite isolates led Botanga & Timko (2005) to propose that at least seven distinct races of S. gesnerioides parasitic on cowpea are present West Africa. These were designated: SG1 (Burkina Faso), SG2 (Mali), SG3 (Nigeria and Niger), SG4 (Benin), SG4z (localised to the Zakpota region of Benin), SG5 (Cameroon), and SG6 (Sénégal). SG4 and SG4z are virtually indistinguishable based on their molecular genetic profiles (Botanga & Timko, 2006) and SG4z appears to be a recently evolved variant of SG4 that originated in farmer's fields after prolonged cultivation of B301, a cowpea cultivar widely grown because of its prior resistance to all known races of S. gesnerioides in West Africa (Supporting Information Fig. S1). The breakdown of resistance under host‐driven selective pressure is anticipated, as in all reported studies thus far, race‐specific resistance is conferred by single dominant genes distributed in two gene clusters in the cowpea genome (Singh & Emechebe, 1990; Timko & Singh, 2008). Using a molecular marker‐assisted positional cloning strategy, Li & Timko (2009) subsequently isolated the RSG3‐301 gene from cowpea that confers resistance to S. gesnerioides race SG3 and showed that it encodes a typical nucleotide‐binding domain and leucine‐rich repeat containing (NLR) protein with a N‐terminal coiled‐coil domain (CC), followed by a central nucleotide‐binding site (NBS) and a C‐terminal leucine‐rich repeat (LRR) domain. The characterisation of RSG3‐301 led to the suggestion that race‐specific Striga resistance in cowpea is an example of effector‐triggered immunity (ETI) in which intracellular NLR proteins (such as RSG3‐301) are activated either directly or indirectly upon recognition of pathogen/parasite effectors (Oren et al., 2016; Jubic et al., 2019).

The nature of the gene expression changes occurring during compatible and incompatible Striga–cowpea associations was examined by Huang et al. (2012) who documented dramatically different expression profiles in the roots of B301 plants being attacked by S. gesnerioides races SG3 and SG4z. These investigators specifically noted that many of the genes and pathways upregulated in the RSG3‐301‐mediated HR to SG3 attack were unaltered or repressed in the susceptible interactions to SG4z, suggesting that the parasite is targeting specific components of host defence mechanism (Huang et al., 2012).

In this study, comparative transcriptomics and in silico computational analysis was used to identify a small soluble effector protein that is highly expressed in the SG4z haustoria transferred to the host root, and capable of suppressing the host innate immunity by binding to a host BTB‐BACK domain‐containing ubiquitin E3 ligase homologue (VuPOB1). We show that silencing of VuPOB1 expression in transgenic B301 lowers the frequency of HR and increases the levels of successful parasitism by SG4, whereas overexpression of VuPOB1 resulted in decreased parasitism by SG4z suggesting that VuPOB1 functions as a positive regulator of the HR response. These studies provide novel insight into the role of secreted effectors as part of the strategy used by parasitic weeds to overcome host immunity to complete their life cycles.

Materials and Methods

Plant materials

All experiments involving viable S. gesnerioides seeds, developing parasites and host–parasite interactions, were performed in an APHIS‐approved quarantine facility in the Department of Biology at the University of Virginia (Facility Number 669; APHIS Plant Protection and Quarantine Permit no. P526P‐14‐03013). S. gesnerioides seeds were collected from plants growing on cowpea cultivars in the following locations: SG3 (Maiduguri, Nigeria – 2008), SG4 (Cana, Benin – 2009) and SG4z (Zakpota, Benin – 2006) (Fig. S1). Seeds of the multirace resistant cowpea cultivar B301 were provided by Lucky Omoigui (University of Agriculture Mukurdi, Nigeria) and California Blackeye no. 5 (BE) seeds were purchased commercially (W. Atlee Burpee & Co., Warminster, PA, USA).

Striga–host root interactions assays and statistical analysis

Seeds of the S. gesnerioides race SG4 and SG4z were surface sterilised with 10% bleach for 10 min and preconditioned at 30°C for 9 d. Seeds were pregerminated by incubation with cowpea root exudates for 2 d (Mellor et al., 2012) and then gently transferred to transgenic and nontransgenic cowpea roots using a paintbrush. At 10, 20, and 30 d post‐inoculation (dpi) the occurrence number of all parasite–host root interaction events (i.e. attachment, tubercle swelling, cotyledon expansion and HR) was scored and the percentage for each interaction type was determined. Ten or more individual plants were used per treatment and the occurrence percentage for each interaction event was subject to a paired t‐test for statistical significance using R v.3.2.2. (R Core Team, 2013).

Transcriptome profiling using RNA HiSeq analysis

For RNA‐seq library preparation, parasite seedlings were removed from the host roots directly into RNAlater (Ambion, Inc., Austin, TX, USA), quickly frozen in liquid nitrogen, and stored at –80C until used for RNA extraction. Experiments were conducted in triplicate for all parasite races at two host genotypes and three time‐point conditions. Host tissue was harvested as previously described (Huang et al., 2012). Library preparation for subsequent Illumina single‐end sequencing was carried out as previously described (Yang et al., 2015). Complete details on RNA‐seq library preparation from host and parasite, transcriptome assembly and gene annotation are provided in Methods S1.

Identification of differentially expressed race‐specific effector candidates

High‐quality Illumina reads from each stage‐specific library were mapped to the reference transcriptome via bowtie2 (Langmead & Salzberg, 2012). Transcript abundance was evaluated independently in each library with express v.1.5.1 (Roberts et al., 2011) using the default settings (except ‐‐r‐stranded) and Transcripts Per Million (TPM) were calculated to represent transcript abundance and used for expression visualisation. Raw read counts were normalised and processed for differentially gene expression analysis using deseq2 v1.6.3 (R v.3.2.2) (Love et al., 2014). The two factor full generalised linear model (GLM) was used to exam race differential expression at each developmental stage. Race‐specific expressed genes were defined as genes differentially expressed during the interactions with both resistant and susceptible hosts.

To identify potential candidate effectors, genes differentially expressed across different developmental stages were first subjected to intuitive classification based on stage pairwise comparison (Table S1). We expect effectors to be highly expressed during the initial interaction of parasites with hosts. We anticipate three potential gene expression groupings: (i) genes with continuous high expression during the two attachment stages; (ii) genes with expression peaks at the early attachment stage; and (iii) genes with expression peaks at the late attachment stage. Following expression analysis, all contigs with complete protein‐coding regions were examined for the presence of signal peptides and transmembrane domains using SignalP (Petersen et al., 2011) and Tmhmm (Krogh et al., 2001). Contigs predicted to contain signal peptides and no transmembrane domain were further examined for the presence of endoplasmic reticulum (ER) localisation sequences by scanning using the PROSITE database (Hofmann et al., 1999). Contigs containing signal peptides, lacking transmembrane domains, and lacking ER localisation domains (i.e. predicted to be transport to extracellular region) were included in the ‘secretome’ data set. Contigs in the secretome data set with expression patterns thought to match those of candidate effectors as described above were placed in the ‘effector candidates’ group Table S2). A large number of contigs did not have complete coding regions and in particular could not be analysed for signal peptides at their 5′ ends. Those contigs that met all of the other criteria (i.e. lacking transmembrane domains, ER retention signals, and haustorial specific expression peaks in post‐attachment stages of the host–parasite interaction) were also included in ‘effector candidate’ group. A detailed description of the methodology used in subsequent 3′ and 5′ RACE cloning of effector candidates from haustorial RNAs, and for determining effector gene structure in genomic DNA can be found in the Methods S1.

Construction of transgene expression plasmids and generation of ex vitro composite plants

The full‐length SHR4z protein was predicted based upon the first ATG at the 5′ end and longest open reading frame (ORF) to an in‐frame stop codon and this coding region was cloned and mobilised into Gateway pDONR™221 Vector (Invitrogen, Carlsbad, CA, USA) using the Gateway BP Clonase Enzyme mix (Invitrogen). A truncated version of the SHR4z full‐length protein in which the predicted signal peptide was removed (SHR4zΔSP) was also cloned into the Gateway pDONR™221 Vector. In SHR4zΔSP an ATG codon was inserted at 5′ end of the truncated sequence to allow for translation initiation. Both pDONR‐SHR4z and pDONR‐SHR4zΔSP were further recombined with pK7WG2D (Karimi et al., 2002) using the LR recombination reaction in the Gateway LR Clonase Enzyme mix (Invitrogen). pK7WG2D expressing small FLAG‐tag were used as control. All pK7WG2D plasmids (pK7WG2D‐SHR4z, pK7WG2D‐SHR4zΔSP and pK7WG2D‐FLAG) were transformed into Agrobacterium rhizogenes strain R1000 for heterogeneously expression in cowpea cultivar B301.

B301 seeds were sterilised with 10% (v/v) bleach for 10 min and pregerminated on sterile Rockwool. Ex vitro composite plants were then generated as previously described (Mellor et al., 2012) using A. rhizogenes strains carrying the pK7WG2D‐SHR4z, pK7WG2D‐SHR4zΔSP and pK7WG2D‐FLAG plasmids. Seedlings with regenerated roots were moved to rhizotrons 20 d after transformation and grown at 30°C for 14 d before Striga inoculation. At 10, 20, and 30 dpi parasite–host root interaction events were scored and the percentage for each interaction type was determined for nontransgenic (no visible GFP) and transgenic (GFP expressing) roots, respectively (see Methods S1).

Yeast‐two‐hybrid (Y2H) screening

Total RNA from Striga‐infected host roots was isolated and used to construct a normalised cDNA ‘prey’ library in a modified pDEST22 vector (Bio S&T Inc., St Laurent, QC, Canada). Y2H screening was performed using ProQuest system (Invitrogen). PDONR‐SHR4zΔSP was recombined with pDEST32 to generate DNA binding domain ‘bait’ fusion (pDEST32‐SHR4zΔSP). PDEST32‐SHR4zΔSP was transformed to yeast strain MaV203 and positive transformants selected by planting on Leu‐deficient yeast nitrogen base (YNB) plates (Leu‐YNB). A single verified transformant was selected for the preparation of competent cells and the competent cells were then transformed with the cowpea host root cDNA ‘prey’ library. The resulting transformants were plated on triple amino acid deficient YNB plates (Leu‐/Trp‐/Ura‐plates) to select interaction‐positive clones. The putative host interacting proteins candidates were further characterised and sequenced and one such candidate subject to further analysis (see Methods S1). A transcript (UP12_14781) encoding the full‐length cowpea VuPOB1 protein was retrieved from cowpea expressed sequence tag (EST) database (v.1.42 of ‘HarvEST:Cowpea’, University of California, Riverside, CA, USA). Coding frame of VuPOB1 was determined by selecting the longest coding sequence (CDS) with 100% match to Y2H in‐frame insertion and cloned from a B301 root cDNA library using gene‐specific primers (Table S3, BTB_ORF_F and BTB_ORFnoT_R).

Co‐immunoprecipitation and bimolecular fluorescence complementation (BiFC) analysis

Epitope tags myc‐ and HA‐ were added to C‐terminals of protein SHR4zΔSP and VuPOB1 by PCR. The purified PCR products were ligated into YFP vectors SPYCE and SPYNE to generate expression constructs myc‐ SHR4zΔSP‐YN and HA‐VuPOB1‐YC. Agroinfiltration was performed on three of the largest leaves of a 6‐wk‐old N. benthamiana plant as previously described (Wydro et al., 2006). Assays were conducted in the presence or absence of MG132 (100 μM) infiltrated into leaves 6 h before harvesting of tissue for co‐immunoprecipitation and immunoblotting (see Methods S1). BiFC and confocal microscopy was performed at 3 d post initial infiltration of leaf tissues with Agrobacterium harbouring the above constructs. Leaves were treated with MG132 (100 μM) for 6 h before harvesting and examined under Leica sp8 scanning confocal microscope. Split‐YFP was excited with 488 nm from laser with emissions collected between 497 and 544 nm.

Ectopic overexpression and RNAi silencing of VuPOB1 in B301 roots

An c. 250 bp fragment within the coding region of VuPOB1 was amplified and mobilised into the Gateway pDONR™221 Vector (Invitrogen) and then recombined into pK7GWIWG2D(II) (Karimi et al., 2002) to generate inverted repeat construct pK7GWIWG2D‐VuPOB1‐RNAi. The full‐length PCR amplicon of VuPOB1 was mobilised into a Gateway^® pDONR™221 Vector (Invitrogen) generating pDONR‐VuPOB1 (with C‐terminal stop codon) and subsequently recombined with pK7WG2D (Karimi et al., 2002) to generate overexpression construct pK7WG2D‐VuPOB1. The sequence verified vectors were then transformed into A. rhizogenes R1000 and used to generate ex vitro composite B301 plants. Transgenic and nontransgenic roots were analysed by qRT‐PCR to confirm silencing of VuPOB1 expression (see Methods S1) in B301 roots.

Cowpea composite plants expressing the pK7WG2D‐VuPOB1 overexpression construct or RNAi silencing construct pK7GWIWG2D‐VuPOB1‐RNAi were moved to rhizotrons and grown for 14 d at 30°C, inoculated with Striga seedlings, and the occurrence percentage for each interaction event score (see Striga–host root interactions assays and statistical analysis above and Methods S1).

Results

Differential response of cowpea cultivar B301 to different races of S .  gesneriodies

Seeds of S. gesnerioides germinate in response to the presence of chemical signals (strigolactones) in the rhizosphere released from the roots of potential host plants (Bouwmeester et al., 2003). Here, 2‐d‐old germinated seedlings of the various S. gesnerioides races parasitic on cowpea in West Africa (e.g. SG4, SG4z, SG3) are phenotypically indistinguishable before host contact (Figs 1a, S1). However, following attachment to the host root, differences in host and parasite phenotype become evident depending on the partners of the interaction (Lane et al., 1996; Botanga & Timko, 2005). California Blackeye no. 5 (BE) is susceptible to all races of S. gesnerioides identified thus far (Botanga & Timko, 2005) and by 3 dpi BE roots show significant numbers of attached Striga seedlings from all three races. No obvious discoloration was found on the BE roots at the site of parasite attachment at this early stage. By 10 dpi, all three races of the parasite (SG4, SG4z and SG3) have successfully penetrated the BE root cortex and established connections with the host vascular system. There is the formation of an enlarged tubercle and evidence of cotyledon expansion of the parasite, an indication of successful vascular connection between host and parasite (Fig. 1b).

Figure 1.

Differential response of cowpea cultivars B301 and California Blackeye no. 5 to parasitism by different races of Striga gesnerioides races from West Africa. (a) Representative photographs illustrating the phenotypic response of B301, a multirace resistant cultivar and California Blackeye no. 5 (BE) a susceptible cultivar, to parasitism by different races of S. gesnerioides. Shown are the appearance of 2‐d‐old germinated SG4, SG4z and SG3 seedlings before host contact (0 days post‐inoculation (dpi)) and the roots of cowpea cultivars B301 and BE at 3 and 10 dpi. Bars, 200 μm. (b) Measured frequency of each different phenotypic event category during the interaction of the three S. gesnerioides races, SG3, SG4 and SG4z with resistant (B301) cowpea roots at 10 dpi. The abbreviation of the phenotypic event categories are as follows: AT, attachment; HR, hypersensitive response; TS, tubercle swelling; and CE, cotyledon expansion. The interaction event ratio for each category was obtained by counting the number of each event category and dividing by the total number of phenotypic events occurring on each host plant. Statistical analysis was done using the ANOVA with 10 independent host plants inoculated with equivalent amounts of the different races. The pairwise comparison P‐value was determined by Tukey's honest significance (Tukey HSD) test following GLM. Bars represent ± SD of the biological replicates. An asterisk (*) indicates a P‐value < 0.05.

By contrast, a very different phenotypic response is observed in the interactions between the Striga races and B301, a cultivar from Botswana resistant to all Striga races except SG4z (Botanga & Timko, 2006). At 3 dpi, B301 roots show only very minor distinguishable phenotypic differences when challenged by Striga races SG4z, SG4 and SG3 with the only notable effect being the appearance of a slight discoloration or browning of the host root at the attachment interface of SG3 and SG4 (Fig. 1a). However, at 10 dpi, dramatically different phenotypic responses are observed among the races (analysis of variance (ANOVA) P‐value < 0.01 for AT, CE, HR and TS). Both SG4 and SG3 have elicited an easily visible HR with associated browning and necrosis of the B301 root at the site of parasite attachment. There is also arrested growth of the parasite and a browning of the parasite. The HR reaction of B301 to SG3 challenge is more pronounced than that observed with SG4 (Turkey's honest significant difference (HSD) adjusted P‐value = 0.02), with just slightly more arrested growth at the tubercle stage of parasite development than observed in SG4 (Fig. 1b). In stark contrast, SG4z seedlings fail to elicit a strong HR response on B301 roots. Similar to that observed on the susceptible BE, SG4z–B301 interactions have significantly more tubercle swelling events and more frequent observed seedlings with evidence of cotyledon expansion (a characteristic associated with the formation of successful vascular connections with the host) by comparison to SG3 and SG4–B301 interactions (Turkey's HSD adjusted P‐value < 0.001).

Transcriptomic profiling of Striga races during compatible and incompatible cowpea interactions

To uncover alterations in gene expression in SG4z that could potentially underlie its ability to overcome or bypass the innate immunity/defence pathways present in B301 activated by the closely related SG4, we generated transcriptome profiles of SG4 and SG4z, and a more distally related race, SG3, during compatible and incompatible interactions with BE and B301 roots. Three developmental stages were sampled: 2‐d‐old germinated seedlings before host contact; parasite seedlings at an early host attachment stage (3 dpi); and parasite seedlings at a late host attachment stage (10 dpi). In total, 45 different libraries were generated and sequenced using an Illumina RNA HiSeq platform yielding > 300 million 1 × 150 bp single‐end reads.

To functionalise these data, we built a hybrid transcriptome assembly (SGall) that combined reads from all races and developmental stages and annotated it to coding proteins using ESTscan (Iseli et al., 1999).The SGall assembly contains 145 407 protein‐coding contigs with a N50 contig length of 1218 bp and captures 89.55% to 98.88% of the conserved single copy genes (Table 1). The quality of the SGall assembly and gene capture frequency are very similar to the haustorial transcriptome assemblies reported from three other root parasitic plants including the closely related S. hermonthica (Yang et al., 2015). Therefore, we believe that SGall represents the large majority of the expressed genes in the S. gesnerioides haustorial transcriptome. When the SGall assembly was compared with comparable developmental stage‐specific transcriptome assemblies from S. hermonthica (i.e. Stages 1, 2 and 4) by reciprocal blastn (evalue < 1e‐10), c. 63.6% of S. hermonthica Unigenes found homologous counterparts in the SGall assembly.

Table 1.

Transcriptome assembly and gene capture statistics.

Transcriptome metrics	Value
Transcriptome size (Mbp)	115
Transcript/contig number	145 407
Trinity component number	114 231
Medium length (bp)	475
Contig N50 length (bp)	1218
Number of N50 contigs	27 168
PlantTribe gene capture number	899
PlantTribe gene capture percentage	93.74%
COSII gene capture number	197
COSII gene capture percentage	89.55%
UCOs gene capture number	353
UCOs gene capture percentage	98.88%

To determine which factor contributed most to expression profile differences among the parasite races, we mapped the reads from all 45 RNA‐seq libraries to the SGall hybrid transcriptome assembly and then performed expression correlation clustering across the 15 conditions (parasite race × developmental time × host genotype). This analysis showed that host genotype and parasite developmental stage contributed significantly to differences among the various parasite race expression profiles (Fig. S2). At the germinated seedling stage, all three parasite races clustered together indicating that while differences exist among the races, such differences are not pronounced before host interaction. As might be expected, given the recent evolutionary derivation of SG4z from SG4 (Lane et al., 1994; Botanga & Timko, 2006), SG4 and SG4z samples clustered closer to each other than either did to SG3. Similarly, samples from the same post‐attachment developmental stage clustered together, with the exception of those from the late attachment stages when compatible (susceptible) and incompatible (resistant) host–parasite pairing are being analysed and clear distinctions are evident.

Differences in the SG4 and SG4z transcriptome profiles identify effector candidates

Differences in gene expression have been suggested to underlie variations in virulence (Whitehead & Crawford, 2006) and therefore we hypothesised that among S. gesneriodies races differential expression would account for the differential resistance elicitation of SG4z compared with SG4 in its interaction with B301. Among the possible differences we anticipated that SG4z might have lost or significantly diminished levels of expression of one or more gene product functioning as an avirulence factor that is recognised by B301. This would preclude its detection by the host immunity system allowing SG4z to establish a compatible interaction with B301. Alternatively, SG4z may have altered the expression of an existing gene or gained new gene expression function (e.g. a new effector) that allows it to subvert the B301 defence pathways thereby yielding the host susceptible to parasitism.

To address these possibilities, we first examined the degree to which transcripts are differentially expressed between SG4z and SG4 at the various stages of interaction with BE and B301. Not surprisingly given their recent evolutionary divergence, relatively few contigs (135 in total) were found to be significantly differentially expressed (FDR < 0.01 log₂FoldChange > 1) between the SG4 and SG4z transcriptomes of germinated seedlings (0 dpi) and of these 135 contigs 50 were upregulated in SG4z and 85 were upregulated in SG4 (Table S1). In total, 259 contigs were differentially expressed between SG4 and SG4z at the early attachment stage (3 dpi) on B301, whereas a slightly larger number (300 contigs) is differentially expressed at 3 dpi on BE (Table S1; Fig. S3B, C). In addition, the magnitude of the expression differences between SG4z and SG4 were significantly increased at 3 dpi compared with 0 dpi (Fig. S3A; Wilcoxon rank sum test P‐value < 2.2e‐16). At the late attachment stage (10 dpi), 17 242 contigs (c. 11.9%) were differentially expressed between SG4z and SG4 parasitising B301, but only 191 contigs (0.13%) were differential expressed in the interaction with BE. This large difference in the number of differentially expressed genes at 10 dpi reflects the fact that one race (SG4z) is participating in a compatible interaction with B301, while the other (SG4) is in an incompatible interaction in which host penetration is arrested and Striga is dying (Table S1; Fig. S3B,C). In total, 205 and 150 contigs are shared between both interactions (B301 and BE) and are significantly differentially expressed between SG4z and SG4 at 3 dpi and 10 dpi, respectively (Fig. S3C). Combined, 267 contigs have race‐specific differential expression at either post‐attachment stages (Fig. S3C). Hierarchical clustering of these 267 contigs showed that 122 are upregulated in SG4z at either 3 dpi or 10 dpi, while 131 have a higher expression in SG4 (Fig. S3D).

As effectors and avirulence factors would be expected to be released from the haustorium upon interaction of the parasite with its potential host (serving to elicit host defence responses or functioning to suppress them) we hypothesised that there might be significant alterations in their expression relative to germinated parasite seedlings. We also expect that proteins moving from parasite to host would likely to have contained secretion signals and other structural features to facilitate this movement (Goritschnig et al., 2012; Hacquard et al., 2012; Rafiqi et al., 2013). Therefore, we developed a candidate effector identification pipeline to identify contigs differentially expressed in post‐attachment SG4z and SG4 haustoria that encoded proteins with the hallmarks of secreted protein effectors (Fig. S4). Thirty‐four of the 267 differentially expressed contigs met all or most of the criteria for effector candidates under our scheme (Figs 2a, S5A).

Figure 2.

Expression difference of effector candidates between SG4 and SG4z. (a) Heatmap of expression change of effector candidates over development. The expression level (the mean of log₂TPM) is represented by colour intensity from low (red) to high (blue) in heatmap. Upregulated effector candidates in SG4z (red bar) and upregulated effector candidates in SG4 (green bar) were grouped into different clusters. The blue bar includes the candidate effectors highly expressed in SG4z at the early attachment stage, while upregulated in SG4 at late attachment stages. Hierarchical clustering was performed based on pairwise correlation between contigs. (b) Shown is the relative abundance of the SGall_040908.1 transcript in germinated seedlings and haustoria of the three different races of Striga races (SG3, blue; SG4, green; and SG4z, red) at 3 and 10 d post‐inoculation (dpi) as determined by qRT‐PCR. Relative expression was normalised to endogenous control PP2A and calibrated relative to expression in SG3 at 0 dpi. Data shown are means ± SD of three technical replicates. The quantitative reverse transcriptase polymerase chain reaction (qRT‐PCR) expression trend confirmed RNA‐seq result.

To verify that the differential expression across developmental stages and interactions based on RNA‐seq analysis accurately changes in transcript abundance we carried out qRT‐PCR on a subset of the candidates (Fig. S5B). All of those tested confirmed their differential expression and one such validated candidate, contig SGall_040908.1, that reproducibly showed no detectable expression in SG4 and significantly higher (100‐fold or greater) expression in SG4z compared to SG3 at both 3 and 10 dpi (Fig. 2b). We selected this candidate for further detailed analysis.

Characterisation SHR4z, a small soluble protein effector that suppressed host immunity

Contig SGall_040908.1 contains only a partial coding sequence and therefore we performed 5′‐ and 3′‐RACE using total RNA isolated from SG4z haustoria growing on B301 roots to obtain the full length of 942 nt transcript encoding a 195 amino acid protein (Fig. S6A). The encoded full‐length protein has a 25 amino acid predicted extracellular (apoplastic) targeting signal peptide at the N‐terminus. There are four LRRs in the protein with one located adjacent to the signal sequence at the N‐terminus and three arranged in tandem near the C‐terminus of the protein (Fig. S6A). The predicted 3‐D structure of the protein indicated that three tandem LRRs forms a stack of beta‐bend ribbons that serve as a potential protein interaction domain (Fig. S6B). Phylogenetically, the short LRR domain of the protein has significant homology to members of the SOMATIC EMBRYOGENESIS RECEPTOR‐LIKE KINASE (SERK) gene family (Fig. S11). Based on its unique high level of expression in SG4z haustorium and its demonstrated role in mediating the loss of the HR in SG4z–B301 interactions (described below), we designated the gene as Suppressor of Host Resistance 4z (SHR4z).

To confirm the presence of the SHR4z gene in the parasite genome and define its structure, five pairs of primers were designed across the full‐length coding region and used in genomic PCR amplifications. The SHR4z gene was found to consist of six exons and five introns (Fig. S6C). As SHR4z transcripts were only detected in SG4z seedlings, we checked to see whether the SHR4z gene was present in other races of S. gesnerioides found in West Africa (Fig. S6C,D,E). Our data showed that while SHR4z is present in the genomes all of the S. gesnerioides races tested it is not expressed. Sequence analysis of the SHR4z genes amplified from SG4z and the other S. gesnerioides races did not uncover any variation in the protein‐coding regions of the genes and only a small number of sequence variations in the introns. Intriguingly, the ability to amplify SHR4z coding sequences from genomic DNA of all individuals within a given race varied, with SG4z, SG3 and SG5 having a significantly higher proportion of individuals (ANOVA P‐value < 0.001) (Fig. S6E) giving amplicons with the primer combinations tested. The reason for this is not known, but may simply reflect that minor variations exist in the gene coding regions that influence the ability of the primers to bind to their target efficiently.

Ectopic overexpression of SHR4z in B301 roots leads to suppression of parasite‐elicited HR

To test the hypothesis that SHR4z is a secreted effector involved in manipulating host root innate immunity, we ectopically overexpressed the full‐length SHR4z protein and a truncated version of SHR4z lacking the apoplastic targeting signal peptide (SHR4zΔSP) in B301 roots (Fig. S7A, B) using an ex vitro composite plant transformation system for cowpea (Mellor et al., 2012). We rationalised that the full‐length SHR4z is needed to place the protein on the surface of the haustorium where is can be transferred to the host. Thus, we might expect that the full length and truncated SHR4z proteins (minus the targeting signal) function differently depending on whether they were secreted to the cell apoplast or retained in the cytoplasm. As shown in Fig. 3(a), transgenic B301 roots expressing the C‐terminal mCherry fusion of full‐length SHR4z protein exhibit mCherry expression on the outer surface of the cowpea root cell consistent with the secretion signal targeting the protein to the apoplast. By contrast, transgenic roots expressing SHR4zΔSP, the truncated secretion signal‐less version of SHR4z, did not exhibit a similarly strong mCherry signal. However, in the presence of protease inhibitor MG132, the expression of SHR4zΔSP became visible and was mainly concentrated in cytoplasm, suggesting that the truncated protein is turned over rapidly in the cytoplasm.

Figure 3.

Ectopic overexpression of SHR4z suppresses cowpea host innate immunity. (a) Representative photographs showing the subcellular localisation of SHR4z‐ and SHR4z∆SP‐mCherry fusion proteins in transgenic cowpea B301 roots at 20 d post‐transformation as viewed by confocal microscopy. B301 plants were transformed with Agrobacterium rhizogenes containing the effector (pK7WG2D‐SHR4z‐mCherry and pK7WG2D‐SHR4zΔSP‐mCherry) and control (pK7WG2D‐mCherry) gene constructs. The roots of pK7WG2D‐SHR4zΔSP‐mCherry transgenic plants were incubated with 100 μM MG132 for 6 h before confocal imaging to prevent the rapid proteolytic degradation of truncated protein. Visualisation of mCherry fluorescence was performed with a 564–630 nm filter range and GFP fluorescence with a 497–544 nm filter range. Bars, 10 μm. (b) Representative images of HR and tubercle swelling (TS) on composite plants. B301 roots from composite plants (pK7WG2D‐SHR4zΔSP construct) were inoculated with pregerminated SG4 and SG4z Striga gesnerioides seedlings at 30 d after transformation. Interactions (CE, cotyledon expansion; HR, hypersensitive response; TS, tubercle swelling) were examined under light and fluorescence microscopy. The transgenic roots were visualised by GFP labelling. Bars, 200 μm. (c–e) Effects of ectopic overexpression of SHR4z∆SP on suppression of B301 innate immunity. Ex vitro composite B301 plants were generated by ectopically overexpressing the SHR4z∆SP transcript. Transgenic and nontransgenic roots were inoculated with 2‐d germinated SG4 (c) SG4z (d) and SG3 (e) seedlings, and the phenotypic responses of the roots were scored at 10, 20, and 30 d post‐inoculation (dpi). The interaction event ratio for each category was obtained by counting the number of each event category and dividing by the total number of phenotypic events occurring on transgenic (green) and nontransgenic (white) roots of each host plant. Statistical analysis was performed using the paired t‐test on > 10 independent host plant replicates (SG3‐10 dpi, n = 22; SG3‐20 dpi, n = 13; SG3‐30 dpi, n = 11; SG4‐10 dpi, n = 14; SG4‐20 dpi, n = 14; SG4‐30 dpi, n = 14; SG4z‐10 dpi, n = 12; SG4z‐20 dpi, n = 12 dpi; SG4z‐30 dpi, n = 11). Boxplots indicate the median (horizontal lines), 25^th and 75^th percentile range (boxes) and up to 1.5 × IQR (whiskers). An asterisk (*) indicates a P‐value < 0.05.

Transgenic B301 roots expressing the SHR4z and SHR4zΔSP proteins and nontransgenic roots were then challenged with Striga race SG4 known to elicit a strong HR in B301 roots (Fig. 3b,c). Transgenic B301 roots expressing SHR4zΔSP showed a significantly lower frequency of HR events than control nontransgenic roots with SG4 starting at 10 dpi (paired t‐test 10 dpi: P‐value = 0.03; 20 dpi: P‐value < 0.01; 30 dpi: P‐value = 0.01) and significantly more tubercle swelling events at 30 dpi (paired t‐test P‐value = 0.01) indicating that the overexpression of SHR4zΔSP in host root was suppressing the host resistance response. We did not observe any cotyledon expansion events on the transgenic B301 roots expressing SHR4zΔSP challenged with SG4. Cotyledon expansion is a well characterised phenotypic indicator of successful vascular connection between Striga and it host. The absence of cotyledon expansion on SHR4zΔSP expressing transgenic B301 roots suggests that despite its ability to suppress HR, SHR4z alone may not be sufficient to completely overcome host innate immunity to the extent that it fully allows successful parasite growth.

By contrast, B301 roots expressing the full‐length SHR4z did not show a corresponding decrease in HR frequency (Fig. S8). We infer that targeting the SHR4z to the apoplast of the host cell makes it unavailable to influence host defence pathways. Collectively, these findings support a role for SHR4z as a suppressor of the host resistance response in the host cells as we previously speculated.

To test whether SHR4z overexpression increases the susceptibility of B301 to SG4z parasitism, we compared the frequency of successful parasite growth on transgenic B301 roots expressing SHR4zΔSP to that of nontransgenic control roots (Fig. 3d). A significant reduction in frequency of HR events was found (paired t‐test 10 dpi: P‐value = 0.01; 20 dpi: P‐value < 0.01; 30 dpi: P‐value < 0.01), but in general, no significant differences were found in the frequency of cotyledon expansion events between transgenic and nontransgenic roots challenged with SG4z (Fig. 3d, paired t‐test 20 dpi: P‐value = 0.49; 30 dpi: P‐value = 0.45). Therefore, we infer from this finding that overexpressing SHR4zΔSP in B301 enhances host susceptibility to the already hypervirulent SG4z by suppressing HR but not by promoting parasite growth.

As resistance to Striga in B301 is conferred in a race‐specific manner involving a number of different race‐specific R genes, we also tested to see whether the resistance response of B301 to other races of S. gesnerioides is altered by SHR4zΔSP overexpression. To this end, we inoculated nontransgenic control and transgenic B301 roots expressing SHR4zΔSP with SG3 seedlings (Fig. 3e). Transgenic B301 roots overexpressing SHR4zΔSP had a significantly lower frequency of HR events than nontransgenic at 10 dpi (paired t‐test P‐value < 0.001), but no significant reduction in HR frequency at 20 and 30 dpi (paired t‐test 20 dpi: P‐value = 0.33; 30 dpi: P‐value = 0.059), suggesting that subtle differences in avirulence factors could also influence a parasite race's ability to differentially activate resistance mechanisms. The response difference between 10 dpi and later time points could also reflect the switch from basal resistance responses typified by pattern‐triggered immunity (PTI) to a more amplified and faster effector‐triggered responses (i.e. the nexus between PTI and ETI).

SHR4z binds a cowpea homologue of the ubiquitin E3 ligase POB1

To identify host targets of SHR4z, we carried out a yeast‐two‐hybrid (Y2H) screen using SHR4zΔSP as bait and a cDNA prey library constructed from RNA isolated from B301 roots parasitised by SG4 and SG4z. From a screen of c. 2 × 10⁶ colonies, we identified five clones that gave a reproducibly strong positive signal (Fig. 4a). The inserts from these clones were recovered, sequenced, and annotated and all five were found to contain the same insert encoding a BTB (Bric‐a‐Brac, Tramtrack, and Broad Complex)‐BACK (BTB and C‐terminal Kelch) domain‐containing protein (Gingerich et al., 2007) with a high level of similarity to the Arabidopsis AtPOB1 and Nicotiana benthamiana NbPOB1 proteins (76% and 74% amino acid sequence identity, respectively) (Fig. S9A). Both AtPOB1 and NbPOB1 have been previously identified as essential ubiquitin‐protein ligases (E3s) involved in plant immune responses and are thought to ubiquitinate PUB17, a downstream component in the pathway, for subsequent degradation (Orosa et al., 2017).

Figure 4.

Identification of VuPOB1, a BTB‐BACK E3 ligase, as the cowpea host target of SHR4z. (a) Shown are the results of yeast‐two‐hybrid assays indicating that SHR4z without signal peptide (SHR4z∆SP) interacted with VuPOB1. The various indicated proteins were expressed as AD− and BD− fusions in MaV203 yeast cells. Transformed MaV203 cells were grown on selective medium (Leu−, Trp−, Ura−) and X‐gal medium to test for interacting partners and on nonselective medium (Leu−, Trp−) to test for the transformation efficiency. BD‐FLAG and AD‐FLAG fusion proteins were used as negative controls for interaction in yeast. (b) Bimolecular fluorescence complementation (BiFC) assay demonstrating the in planta interaction SHR4z∆SP with VuPOB1 in cytoplasm. Confocal microscopy was used to visualise the interaction of SHR4z∆SP with VuPOB1 following transient co‐expression of plasmids containing YN‐VuPOB1 and YC‐SHR4zΔSP in Nicotiana benthamiana leaves. Bars, 25 μm. (c) Co‐immunoprecipitation analysis indicated that myc‐SHR4z∆SP interacted with HA‐VuPOB1 in planta. Myc‐SHR4z∆SP and HA‐VuPOB1 fusions were transiently expressed in N. benthamiana leaves following Agrobacterium‐mediated transformation. Transformed tissue were subsequently infiltrated with 100 μM MG132 or H₂O as control 6 h before harvesting. Total protein was extracted 3 d after initial infiltration and incubated with anti‐myc beads to immunoprecipitate myc‐SHR4z∆SP. Both protein extract (3% input) and immunoprecipitates (IP) were analysed by immunoblot (IB) using anti‐HA and anti‐myc antibody.

Using HMM domain searching against the recently published cowpea reference genome assembly (Muñoz‐Amatriaín et al., 2017), we identified six BTB‐BACK proteins in cowpea. Phylogenetic comparisons (Fig. S9B) indicate that the cowpea BTB‐BACK protein identified in our screen clusters with the other cowpea BTB‐BACK proteins and three of six fall in a clade along with the Arabidopsis AtPOB1/AtLRB2, AtLRB1 and tobacco NtPOB1 proteins. Based on its high level of sequence similarity and phylogenetic relatedness, we confidently designated the cowpea interacting protein as VuPOB1 (Vigun10g140300.1). The two additional cowpea BTB‐BACK domains proteins that also fall within this cluster we have designated as VuPOB2 (Vigun10g140200.1) and VuPOB3 (Vigun06g120500.1).

The predicted macromolecular structures of SHR4z and VuPOB1 indicate a strong likelihood for interaction between the two proteins (Fig. S9C). To confirm a physical interaction between VuPOB1 and SHR4zΔSP in vivo, BiFC assays were performed using a split‐YFP system in N. benthamiana leaves. Single and paired plasmids expressing the SHR4zΔSP‐YN and VuPOB1‐YC constructs were Agro‐infiltrated into N. benthamiana leaves in the presence or absence of MG132, a cell‐permeable proteasome inhibitor that reduces the degradation of ubiquitin‐conjugated proteins. At 3 d post infiltration confocal microscopy was performed. Reconstituted YFP fluorescence could only be seen in the cytoplasm of the transgenic cells expressing both constructs (VuPOB1 and SHR4zΔSP) in the presence of MG132 indicating that these two proteins interacted with the host cell cytoplasm (Fig. 4b) but that the interaction led to their rapid turnover.

To further confirm the nature of this interaction co‐immunoprecipitation (Co‐IP) studies were performed by co‐expressing epitope‐tagged proteins VuPOB1 and SHR4zΔSP proteins (HA‐POB1 and myc‐SHR4zΔSP, respectively) in N. benthamiana leaves. Consistent with our previous observations (see Fig. 3a) suggesting that SHR4zΔSP is under rapid proteasome‐mediated degradation in plant cells, expression of myc‐SHR4zΔSP was barely detectable by immunoblot analysis (Fig. 4c: input). However, co‐infiltration with the proteasome inhibitor MG132 allows the detection of the SHR4zΔSP protein alone and as a co‐immunoprecipitate with VuPOB1 (Fig. 4c: IP) further confirming the interaction of the two proteins in planta.

VuPOB1 is a positive regulator of cowpea HR upon Striga parasitism

To uncover how VuPOB1 may be functioning in cowpea innate immunity, we first measured the levels of VuPOB1 transcripts in the various compatible and incompatible cowpea–Striga interactions. qRT‐PCR analysis of VuPOB1 transcript levels in B301 roots parasitised by SG4z and SG4 showed a substantial but transient increase in VuPOB1 transcript levels at 3 dpi compared with 0 and 10 dpi. By contrast, no significant expression change was observed in BE roots subjected to similar parasitism (Fig. 5a). Interestingly, in their analysis of differential gene expression during compatible and incompatible Striga–cowpea interactions, Huang et al. (2012) observed that among the genes most significantly upregulated (5.57‐fold and 6.04‐fold induced at 6 dpi and 13 dpi, respectively) in B301 roots parasitised by S. gesnerioides race SG3 was VuPOB2, the paralogue of VuPOB1 (Huang et al., 2012) (Fig. S9B). Cumulatively, these findings suggested a role for VuPOB1 (and perhaps the other paralogues) the cowpea resistance response to Striga.

Figure 5.

VuPOB1 functions as a positive regulator of plant defence. (a) Shown is the relative transcript abundance (level of expression) of VuPOB1 in the roots of B301 and BE cowpea before parasite attachment and 3 d post‐inoculation (dpi) and 10 dpi with Striga races SG4 and SG4z. The relative expression was normalised with endogenous control 18S and calibrated in relative to expression in BE at 0 dpi. Statistical analysis was done using unpaired t‐test with three technical replicates. Error bar represents ± SD. (b) Effect of RNAi silencing of VuPOB1 on SG4‐triggered host resistance response. Ex vitro composite B301 plants were generated by expressing constructs capable of RNAi silencing of VuPOB1 expression. Transgenic (green) and nontransgenic (white) roots were inoculated with 2‐d germinated SG4 seedling, and at 10, 20 and 30 dpi the phenotypic responses of the roots were scored. The abbreviation of the phenotypic event categories are as follows: CE, cotyledon expansion; HR, hypersensitive response; and TS, tubercle swelling. The interaction event ratio for each category was obtained by counting the number of each event category and dividing by the total number of phenotypic events occurring on each host plant. Statistical analysis was carried using the paired t‐test with > 10 independent host plant replicates (10 dpi, n = 16; 20 dpi, n = 15; 30 dpi, n = 15). An asterisk (*) indicates a P‐value < 0.05. (c) Effects of ectopic overexpression of VuPOB1 on SG4z‐mediated suppression of B301 innate immunity. Ex vitro composite B301 plants were generated by ectopically overexpressing the full‐length VuPOB1 transcript. Transgenic (green) and nontransgenic (white) roots were inoculated with 2‐d germinated SG4z seedlings, and at 10, 20, and 30 dpi the phenotypic responses of the roots were scored. The abbreviations for the phenotypic event categories and scoring of phenotypic categories are as above. Statistical analysis was done using the paired t‐test with >10 independent host plant replicates (10 dpi, n = 13; 20 dpi, n = 10; 30 dpi, n = 10). Boxplots indicate the median (horizontal lines), 25^th and 75^th percentile range (boxes) and up to 1.5 × IQR (whiskers). An asterisk (*) indicates a P‐value < 0.05.

To test whether VuPOB1 is necessary for the resistance response of B301, we used RNAi to silence VuPOB1 expression in B301 roots. As shown in Fig. 5(b), VuPOB1‐silenced B301 roots challenged with SG4 showed a significantly lower frequency of HR events (paired t‐test 10 dpi: P‐value < 0.001; 20 dpi: P‐value < 0.001; 30 dpi: P‐value < 0.001) and a greater frequency of tubercle swellings events (paired t‐test 20 dpi: P‐value = 0.04; 30 dpi: P‐value = 0.01) than nontransformed control B301 roots, indicating that VuPOB1 is a necessary and significant component of innate immunity. VuPOB1‐silenced B301 roots challenged with SG4z showed significantly fewer HR events (paired t‐test 10 dpi: P‐value < 0.01; 20 dpi: P‐value < 0.001; 30 dpi: P‐value = 0.01), more tubercle swelling events (t‐test 20 dpi: P‐value < 0.01), but comparable frequency of cotyledon expansion events (Fig. S10). These data suggested that silencing of VuPOB1 may only marginally enhance susceptibility of B301 to SG4z by suppressing HR but not promoting parasite growth.

When we examined the effects of ectopically overexpressing VuPOB1 on Striga–host interactions, a quite strikingly different outcome was observed in the normally successful parasitism of B301 by SG4z parasitism. Rather than the normal compatible interaction, B301 overexpressing VuPOB1 roots displayed a higher frequency of HR events (paired t‐test 10 dpi: P‐value < 0.01; 20 dpi: P‐value < 0.001; 30 dpi: P‐value < 0.001) and lower frequency of tubercle swelling (paired t‐test 10 dpi: P‐value = 0.049; 20 dpi: P‐value < 0.01; P‐value = 0.03). More significantly, there was no clear evidence of cotyledon expansion events among attached SH4z parasites (Fig. 5c). These data clearly indicated that overexpression of VuPOB1 resulted in decreased parasitism by SG4z suggesting that VuPOB1 functions as a positive regulator of HR and host immunity in cowpea.

Discussion

In this study we used high‐throughput RNA‐seq‐based transcriptomic profiling and transgenic expression analysis to identify a novel secreted effector protein, SHR4z, from a root parasitic angiosperm that was capable of modifying host plant immunity. The presence of secreted haustorial effectors was speculated on in our previous studies (Li & Timko, 2009) in which we showed that resistance in cowpea to S. gesnerioides is conferred by specific resistance genes, such as RSG3‐301, hallmark of ETI. The identification of SHR4z, whose unique high‐level expression in SG4z allows the suppression of the HR triggered by SG4, also underscored the fact that the evolution of new components in the arms race between cowpea host and their parasite Striga is a continually ongoing process. In this case, host‐driven selection of variants capable of overcoming host immunity within the SG4 population was likely to be facilitated by the repeated cropping of the multirace resistant cultivar B301 in farmer's fields in the Zakpota region of Benin over the several decades since its introduction (Lane et al., 1994). The occurrence of a mutation that resulted in high‐level expression of SHR4z and its associated suppression of the host plant's ability to mount a HR upon parasite attack, would have allowed these individuals to complete their life cycle. Because of the autogamous reproduction presented in S. gesnerioides, we expected that this trait would be rapidly fixed and build up within the region it occurred, presumably in and around Zakpota. The sensitivity of high‐throughput comparative transcriptomics provided the possibility for identifying rare transcripts differentially expressed between SG4z and is evolutionary precursor SG4 and thus potential candidate effectors involved in overcoming the race‐specific resistances present in B301.

The use of secreted effectors to modulate host resistance is widespread in nature and has been described in interactions between plants and a wide variety of pathogenic microbes, fungi and nematodes. This work extends the use of secreted effectors for host immunity manipulation to parasitic weeds thereby opening a new understanding of the mechanism of the intriguing plant–plant interactions. Our findings indicated that the SHR4z effector is formed in the parasite haustorium before host attachment and deployed in the early post‐attachment stages of the host–Striga interaction. The 25 amino acid secretion signal targets the SHR4z to the apoplast of the haustorial cell where it is available for entry into the host cell to carry out its suppressive function. Our data show that when the full‐length SHR4z protein is expressed in host roots (following Agrobacterium‐mediated transient transformation), the apoplastic targeting signal at N‐terminal of SHR4z moves the protein to the host cell apoplast where it is not functional in suppression of immunity. Thus, presence in the host cell cytoplasm is required for function. How the effector moves from parasite apoplast to host cytoplasm and what molecules are involved remain to be determined.

In oomycetes and fungus, it has been long known that the typical virulence effectors are among pathogenic secretome and have clear N‐terminal signal peptide to allow translocating from pathogen cells to apoplastic region. They can function in apoplast or engage cell entry by overrepresented host‐targeting motifs following N‐terminal signal peptide. The known host‐targeting motifs, like RXLR, LFLAK and CHXC amino acid sequences, were first discovered by examining sequence alignment of known AVR proteins (Rehmany et al., 2005; Whisson et al., 2007) and further tested by molecularly manipulating motifs for translocating detection (Albrecht et al., 2005). However, apart from its apoplast targeting signal, no obvious host‐targeting motifs are observed in SHR4z.

Since parasitic weeds and their hosts share most cellular features in common, unlike pathogenic microbes and fungi that require specialised structures for the transfer of effectors into their plant prey, the movement of small proteins between Striga and its host likely occurs without the development of specialised structures or protein transfer routes via plasmodesmata connections established after parasite attachment. When, where and how the apoplast targeting signal is removed from the SHR4z protein and mechanism of effector trafficking between parasitic plant and host is clearly an area in need of additional future study.

Regardless, once delivered to the host cell, SHR4z is capable of suppressing HR by interfering with the normal transduction pathway activated during plant immune responses (Fig. 6). The exact mechanism remains to be determined. SHR4z shares significant sequence and structural similarity with the N‐terminal region of the SERK subfamily of LRR‐RLKs (Fig. S11) and is predicted to interact with the BTB/POZ interface of VuPOB1 via the three LRRs at the C‐terminal of the protein (Fig. S9) SERK proteins are known to function as co‐receptors, physically interacting with numerous cell‐surface receptors to regulate various developmental and immune responses in plants including BRASSINOSTEROID signalling (Li et al., 2002; Nam & Li, 2002), male sporogenesis (Albrecht et al., 2005; Colcombet et al., 2005) and plant defence (Chinchilla et al., 2007; He et al., 2007; Heese et al., 2007; Kemmerling et al., 2007). For example, AtSERK3/BAK1 interacts with FLS2 immediately after perception of bacterial flagellum to trigger plant innate immunity. It also works redundantly with AtSERK4/BKK1 to control cell death. Similarly, the rice homologue of SERK1 (OsSERK1) is known to be involved in the defence response and significantly induced at the RNA level by rice blast fungus (Hu et al., 2005). However, unlike members of the SERK family proteins, SHR4z lacks the kinase domain typically responsible for the phosphorylation activity in SERKs. Therefore, SHR4z may serve as a decoy that interferes or competes with the interaction of one or more members of the SERK family with VuPOB1 but, by the phosphorylation/activation capacity of a SERK, is not able to activate the downstream pathway leading a HR (Fig. 6). Molecular mimicry modulating plant host responses to microbial pathogens and nematodes is well documented (Ronald & Joe, 2018).

Figure 6.

Proposed model of activation and suppression of cowpea host HR by different races of Striga gesnerioides. Parasitism of cowpea host B301 roots by Striga races SG3 or SG4 elicits a hypersensitive response (HR) at the site of attachment. The intracellular NLR protein RSG3‐301 is required for the response and may work directly or indirectly upon recognition of parasite effectors with members of the cowpea SERK family of proteins to bring about activation of the positive regulator of HR, VuPOB1. VuPOB1 then initiates activation of previously identified components of the HR pathway (see Huang et al., 2012; Mellor, 2013) possibly involving interaction with PUB E3 ligases. By contrast, apoplastically localised SHR4z secreted from the SG4z haustorium enters B301 host root cells via an unknown transport mechanism, where it interacts with VuPOB1 resulting in their subsequent ubiquitination and degradation and loss of HR in the host root.

E3 ligases (of which VuPOB1 is a homologue) have been known for some time to be common interactors with RKs or RLKs. The Arabidopsis Plant U‐Box E3 ubiquitin ligases PUB12 and PUB13 have been shown to be BAK1/SERK3 phosphorylation targets that lead to the degradation of FLS2 in order to attenuate Pathogen Associated Molecular Pattern (PAMP)‐triggered immunity (PTI) in a flg22‐dependent manner (Lu et al., 2011). The RING‐type E3 ligase XB3 is required for the accumulation of the receptor‐like kinase (RLK) protein XA21 and promotes downstream R‐mediated HR against bacterial blast Xanthomonas oryzae pv. oryzae (Wang et al., 2006). Two Plant U‐Box (PUB) families of putative E3 ubiquitin ligases (Os08g01900 and Os01g66130) were found to interact with four phylogenetically distant defence‐related RLKs (Os04g38480, Os07g35580, Os07g35260 and Os08g03020) in rice (Ding et al., 2009). E3 ligases regulate as well as are regulated by RLKs at all steps of plant immune responses and constitute the central of plant defence signalling pathways (Duplan & Rivas, 2014). As a result they have become the targets of pathogenic effectors. The effector AVR3a from the oomycete Phytophthora infestans targets and stabilises the U‐box‐type Ub‐ligase CMPG1/PUB20, which controls P. infestans elicitin INF1 triggered cell death (Bos et al., 2010). Effector AvrPiz‐t from the rice blast fungus Magnaporthe oryzae interacts and inhibits the rice RING‐type Ub‐ligase APIP6, which is normally a positive regulator of PTI. Binding of AvrPiz‐t and APIP6 results in degradation of both proteins in N. benthamiana and suppresses flg22‐induced ROS production (Park et al., 2012).

In tobacco, the VuPOB1 homologue NtPOB1/NbPOB1 was recently found to target the U‐box E3 ligase PUB17 for degradation suppressing Cf‐mediated HR‐PCD (Yang et al., 2006; Orosa et al., 2017). In Arabidopsis, there are 49 members in PUB E3 ligase group (Azevedo et al., 2001; Mudgil et al., 2004; Mazzucotelli et al., 2006) and several of these members have been shown to be involved in plant immunity. PUB12 and PUB13, for example, attenuate PTI by ubiquitinating the flg22 receptor FLS2 (Lu et al., 2011), and PUB 22, 23 and 24 act in combination to negatively regulate plant immunity by suppressing oxidative burst, the MPK3 activity, and transcriptional activation of defence regulatory genes (Trujillo et al., 2008). As VuPOB1 appears to be a positive regulator of HR, additional studies are clearly needed to determine its interaction with one or more members of the PUB E3 ligase family and how this interaction functions in immunity.

It is clear that the HR response in B301 requires the expression of a number of components acting downstream of effector recognition by RSG3‐301 (Fig. 6). Previous studies from our group (Huang et al., 2012; Mellor, 2013) have shown that RNAi silencing of the transcriptional activator VuGRF leads to a loss of the SG3‐induced HR. Similarly, silencing the expression of genes involved in cell wall modification (e.g. peroxidases, galacturonases, etc.), general defence (e.g. narbonin, oxylipin biosynthesis, etc.) and ROS release also decreases SG3‐activated HR and associated abiosis leading to parasite death post‐attachment (Mellor, 2013).

Finally, it is worth noting that the discovery of a secreted effector involved in mediating the interaction of S. gesnerioides with its cowpea host raises questions about the generality of this strategy among other parasitic weeds. Examinations of the haustorial transcriptomes of other parasitic members of the Orobanchaceae (Yang et al., 2015) revealed the presence of homologues of SHR4z in the transcriptomes of S. hermonthica and Phelipanche aegyptiaca, the former being a close relative to S. gesnerioides and the latter being much more evolutionarily diverged. Like SHR4z these homologues are also highly expressed in haustorial stages coincident with host contact. Whether these homologues are functional and acting in a manner similar to SHR4z has been addressed by Clarke et al. (C. R. Clarke, S‐Y. Park, Z. Yang, X. Jia, E. Wafula, L. Honaas, C. Yang, C. W. dePamphilis, J. H. Westwood, pers. comm.) who tested P. aegyptiaca secreted proteins for the ability to alter innate plant immunity. These investigators uncovered a set of five candidate effector proteins that suppress known plant defence pathways, including two LRR SERK‐like proteins that are similar to that identified here. Despite years of research, wide‐scale effective control methods for Striga based on breeding and selection of resistant cultivars and use of modified agronomic practices remain elusive. Understanding the molecular mechanism of Striga virulence and the evolutionary changes in the parasite that contribute to breaking host resistance enhanced our abilities to better solve this riddle. Thus our findings here on the nature of secreted effectors could potentially contribute to the development of novel strategies for controlling Striga and other parasitic weeds and thereby enhancing plant productivity and food security world‐wide.

Author contributions

MPT and CS conceived the research and designed the experiments; CS and HL performed the research; LH and EKW contributed new reagents/analytic tools; CS, HL, LH, EKW, CWdD and MPT analysed data; and MPT and CS wrote the paper; all authors approved. MPT (independently and as a member of the Parasitic Plant Genome Project) secured funding.

Supporting information

Please note: Wiley‐Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Fig. S1 Map of West Africa showing the general distribution of the various races of S. gesnerioides.

Fig. S2 Overall similarity of global gene expression profile. Heatmap showing the relationship of gene expression changes in comparisons of two cowpea cultivars.

Fig. S3 Differential expression analysis between SG4 and SG4z.

Fig. S4 The workflow of predicting effector candidates in S. gesnerioides using transcriptome data.

Fig. S5 Differentially expressed candidate effectors and qRT‐PCR verification of selected effector candidates.

Fig. S6 Gene and protein structure of SHR4z.

Fig. S7 Verification of SHR4z and VuPOB1 overexpression and RNAi silencing VuPOB1 in transgenic roots and nontransgenic roots.

Fig. S8 B301 roots expressing the full‐length SHR4z and control did not show difference between transgenic and nontransgenic roots.

Fig. S9 Protein structure and phylogenetic relatedness of cowpea VuPOB1.

Fig. S10 Effect of RNAi silencing of VuPOB1 on SG4z‐triggered host responses.

Fig. S11 Phylogenetic relationship among SHR4z and SERK homologues in Arabidopsis and cowpea.

Methods S1 Supplemental materials and methods and additional references.

Table S1 Summary of differential expressed gene statistics from between race comparisons at differential stages of the host–parasite interaction.

Table S2 Summary of contigs identified at different stages of the discovery process in the identification of effector candidates.

Table S3 List of oligonucleotide primers used in this study.