A reporter-tagged T3E is often expressed in planta to check full-length expression of the fusion construct and, additionally, to provide information on its subcellular localization. For phytopathogens, the go-to transient expression system is tobacco infiltration with an Agrobacterium tumefaciens strain carrying a binary vector for the selection and expression of a tagged effector gene. Other in planta expression systems might also be used, such as expression in transgenic roots obtained through A. rhizogenes transformation or transgenic A. tumefaciens-transformed A. thaliana (Deslandes et al., 2003; Tasset et al., 2010; Ron et al., 2014; Sun et al., 2017). Lastly, infection with a bacterial mutant strain encoding an endogenously fluorescently tagged T3E (by means of allelic replacement or recombination) or transformed with an alike T3E expression construct can provide insight into host effector localization in a physiological setting. In the case of fluorescent or luminescent reporter expression constructs, confocal microscopy or bioluminescence microscopy, potentially supported by the expression of subcellular markers, has been utilized to localize T3Es in different subcellular compartments (Denne et al., 2021). Whereas no direct information on T3E functioning is obtained this way, it can provide contextual information for the interpretation of the results acquired from other methods applied. RipE1, for example, localizes to the cytoplasm and nucleus of N. benthamiana cells (Nakano and Mukaihara, 2019). A yeast 2-hybrid (Y2H) screen with a cDNA library revealed an interaction of RipE1 with JASMONATE-ZIM DOMAIN (JAZ) proteins, known to be predominantly localized in the cell nucleus, thus matching the subcellular RipE1 localization (Nakano and Mukaihara, 2019).

Whereas native delivery of the T3E reporter by the phytopathogen is desirable over its (over)expression in the host, native translocation is limited, for instance, by the size or mechanical stability (i.e., the ease of protein unfolding under force) of the reporter tag used. This restriction was shown in the context of the green fluorescent protein (GFP) that displayed a comparable thermodynamic stability, but a higher mechanical stability compared to two Salmonella effectors (SptP and SopE2), implying that mechanical instability enhances secretion through the T3SS (LeBlanc et al., 2021; Braet et al., 2022). Some efforts have been made to overcome the passage drawbacks through the T3SS. For instance, the T3E RipP2 from R. solanacearum had been cloned to the C-terminal β strand of GFP (GFP₁₁), instead of the commonly used full-length 11 β barrel strand-consisting GFP (Henry et al., 2017). The complementary part of GFP (GFP_1-10) was expressed under a constitutive promoter in transformed Arabidopsis plants. After native delivery of the RipP2-GFP₁₁ protein upon infection and complementation with GFP_1-10 expressed in a ripP2-deletion mutant, a GFP signal was observed in the nucleus of Arabidopsis cells (Henry et al., 2017), a methodology commonly referred to as split-GFP. Other methods to study T3E subcellular localization, whether or not by native delivery, have recently been extensively reviewed (O'Boyle et al., 2018; Braet et al., 2022).

For the RSSC, the subcellular host localization of many T3Es has been reported. Recently, subcellular localization of a subset of 19 T3Es was assessed in tomato leaves and hairy roots, underlining the diverse destinations of different T3Es, including cell periphery, nucleus, tonoplast, and peroxisomes (Denne et al., 2021) and highlighting the importance of in planta localization studies, rather than relying solely on in silico localization predictions. Interestingly, subcellular localization prediction tools were accurate when the T3E was predicted to localize in the nucleus or plasma membrane, but less accurate for observed organellar or cytoskeleton localizations.

Many T3Es have been shown to be involved in avoidance or suppression of the innate immune response. This function can be analyzed through several assays, such as an infection assay, in which, in the case of R. solanacearum, wilting symptoms are compared between a wild type and a (multiple) T3E deletion mutant. For example, for R. solanacearum strain OE_1-1, a multiple effector mutant lacking 42 T3Es was generated and the disease progression of infected tobacco plants measured in a virulence assay, revealing that some T3E families, such as the RipA T3E family members, had a higher impact on virulence than others, e.g., RipG or RipH T3E families (Lei et al., 2020). Instead of examining the effector virulence through effector deletion (loss-of-function), the involvement of single effectors can also be tested in a gain-of-function experiment in an effectorless polymutant of P. syringae DC3000 strain (Wei et al., 2018), as was done to study the effect of five core Ralstonia T3Es (RipAE, RipAQ, RipC1, RipU and RipW) on plant immune responses (Cong et al., 2022). Here, RipU was shown to reduce the ROS burst, and to upregulate MAPK cascades in tobacco leaves. In addition to disease progression measurement, bacterial enumeration and, thus, replication inside the host are also used as a proxy to estimate the influence of one or more T3Es on virulence (Deslandes et al., 2003; Sun et al., 2017; Yu et al., 2020).

The involvement of a T3E in virulence can also be investigated by overexpression of the T3E within plant cells and by determination of the effect on mechanisms known to be implicated in immunity. Most often the development of ensuing immunity-related effects is evaluated, for instance, by performing cell death assays in tobacco leaves with an overexpression construct after agro-infiltration, HR induction assessment, ion leakage assays, electrolyte leakage monitoring from leaves as a measure for the cell death severity, or ROS assays (Table 3; Sang et al., 2018, 2020; Nakano and Mukaihara, 2019; Yu et al., 2020). Suppression of PTI by a T3E can also be examined in a PTI inhibition assay, based on the principle that previous PTI induction by a nonpathogenic bacterium can dampen the localized ETI necrotic response elicited by a pathogenic bacterial strain. However, when the nonpathogenic bacterium delivers a T3E that suppresses PTI, an ETI necrotic response occurs where the pathogenic strain was infiltrated (Badel et al., 2013; Le Roux et al., 2015). Additionally, expression of PTI/ETI marker genes, such as FLG22-INDUCED RECEPTOR-LIKE KINASE 1 (FLK1) or PATHOGENESIS-RELATED GENE 1 (PR1), is also often monitored (for instance, by reverse transcription-quantitative polymerase chain reaction (qRT-PCR)). PTI marker expression levels are regularly compared between leaves treated with the flagellin-22 (flg22) peptide (a bacterial PAMP, positive control) and leaves expressing the T3E or a mutant version (Le Roux et al., 2015).

The impact of T3Es on the host transcriptome is nowadays frequently investigated by means of RNA-sequencing (RNA-seq), although previously, microarrays had been used as well. In addition, the spatiotemporal expression of T3Es can also provide necessary information on the infection process itself and on the putative function of the T3E.

For instance, the spatiotemporal expression of T3Es examined by an RNA-seq approach in potato (de Pedro-Jové et al., 2021), retrieved RNA from potato plants infected with the cold-adapted R. solanacearum strain UY₀₃₁ at three different infection stages, namely at an early, middle, and late stage, from the apoplast, xylem of asymptotic plants, and xylem of wilted plants, respectively. After normalization for bacterial counts, the T3E expression generally became more prominent at the middle and late stages, challenging the view that T3Es would mainly be important at early infection stages, although sensitivity at the earliest stages of infection might still be (too) limited. Nevertheless, some T3Es were found to have exceptional expression patterns, such as RipD and RipAD, both expressed at all infection stages (de Pedro-Jové et al., 2021).

The host can be monitored alongside the pathogen’s transcriptome with dual RNA-seq. This approach was used to observe both the T3E of R. solanacearum strain Rs-SY1 and the pepper gene expression during infection (Du et al., 2021). Compared to noninfected pepper plants, many pepper genes were differentially expressed 1, 3, 5, and 7 days after inoculation. Among others, at all stages, the genome ontology (GO) terms were enriched in the upregulated differentially expressed genes “defense response to bacterium,” “ethylene biosynthetic process” and “isoprenoid biosynthetic process” and in the downregulated differentially expressed genes “photosynthesis, light harvesting” and “response to abiotic stimulus.” Only genes involved in “sesquiterpenoid and triterpenoid biosynthesis” were gradually upregulated with disease progression. After mining the genome of R. solanacearum Rs-SY1, 84 T3E coding sequences were predicted, of which only seven had a higher expression in early infected pepper hypocotyls than that in the control, i.e., nutrient agar-grown R. solanacearum, hinting at the potential involvement of these T3Es at the onset of the disease (Du et al., 2021). Additionally, because only 7 out of 84 predicted T3Es were differentially expressed, the question was raised whether the other T3Es might have a function at later infection stages.

A transcriptomic approach can also investigate one or more T3Es under controlled expression instead of the broad T3E expression profiles during infection. The effect of T3E expression on transcriptomic profiles was studied in yeast cells ectopically expressing the T3E RipA5 (Popa et al., 2016). After measuring the genome-wide transcriptomic changes at several timepoints (2, 4, and 6 h) after induction of the RipA5 expression by means of DNA microarrays, 319 and 447 genes were discovered that were at least 2-fold induced and repressed, among which many nitrogen catabolite repression genes and ribosomal protein-encoding genes and genes involved in ribosome biogenesis, respectively. Intriguingly, this expression profile was fairly similar to that of rapamycin-treated yeast, a known inhibitor of the TOR COMPLEX 1 (TORC1) pathway. After some follow-up experiments, RipA5 could be shown to impact the TOR pathway in planta by acting as a TOR inhibitor in yeast and plant cells (Popa et al., 2016). A study on in planta function of the R. solanacearum T3E RipAB by RNA-seq (Zheng et al., 2019) revealed 417 differentially upregulated genes by comparison of three effector-expressing potato lines versus a control cultivar. A GO term enrichment analysis pointed to the involvement of RipAB in Ca²⁺ signaling (Zheng et al., 2019). In another experiment, RNA-seq carried out at different timepoints post inoculation (2 or 4 h) of N. tabacum cv. Xanthi leaves with wild-type R. solanacearum or with a ripAK deletion mutant identified regulated genes, including upregulation of photorespiration-related genes and downregulation of plant immunity-related genes, hinting at the involvement of RipAK in the modulation of early plant responses upon infection (Sun et al., 2017).

Transcription activator-like effectors (TALEs) also represent interesting targets for transcriptome-based analyses. TALEs represent a family of T3Es in the Xanthomonas genus, but also have one representative RSSC member, RipTAL1 (aka Brg11), that serve as transcription factors and activate the expression of certain virulence- or susceptibility-related genes by binding specific sites in target promoter regions. By differential RNA-seq of tomato plants infected for 24 h with X. euvesicatoria encoding either the R. solanacearum T3E RipTAL1 or RipTAL1 without its DNA-binding domain (DBD) under control of a constitutive promoter, the tomato ARGININE DECARBOXYLASE (ADC) genes SlADC1 and SlADC2 were identified as upregulated targets of wild-type RipTAL1 (Wu et al., 2019). Subsequently RipTAL1 was found to increase ADC activity and to boost plant polyamine levels, thereby potentially attenuating growth of R. solanacearum competitors.

For the inspection of differing protein expression profiles, a shotgun proteomics approach can be taken. Such a method was used to monitor proteome changes in early and later stages of P. syringae infection (Fan et al., 2019). Similarly, it can be applied to investigate the effect of a microbe on the posttranslational modifications of the host (Walley et al., 2018).

However, these methods do not provide information on the direct function of a T3E. As many T3Es exert their function by binding with host proteins or metabolites, knowledge on their interactions is thus of fundamental importance. In this section, we will focus on different methods that can be used to elucidate effector-host protein–protein interactions (EH-PPI), each with their advantages and disadvantages as recently extensively reviewed (Struk et al., 2019; De Meyer et al., 2020).

Although certainly valuable, functional protein microarrays have rarely been used to examine PPIs, mainly due to the complexity and the cost of the microarray design. In a protein microarray, individual proteins are immobilized on a microarray and a purified protein of interest is applied, resulting in a readable signal. In this manner, a proteome array probed with 15,000 proteins (56% proteome coverage) has been designed for Arabidopsis (Popescu et al., 2007; Manohar et al., 2014). Even though microarrays allow the detection of weak PPIs, the native context is missing, resulting in high false positive and negative rates because of the absence of chaperones (and thus also correct folding), posttranslational modifications, etc. Nevertheless, the interaction between a large set of human proteins and several type IV effectors from Legionella pneumophila (Yu et al., 2015, 2018) was successfully studied using protein microarrays, but, to our knowledge, proteome microarrays to investigate the interaction between T3Es and plant proteins are yet to be reported.

More frequently, Y2H-sequencing (Y2H-seq) and mass spectrometry (MS)-based approaches, such as affinity purification (AP) or biotin identification (BioID) are used to identify EH-PPIs. In Y2H, a bait protein (such as a T3E) is fused to a DBD and the prey (such as plant protein) to an activation domain (AD) or vice versa. Upon direct interaction, the DBD and AD are brought together, resulting in the reconstitution of a transcription factor that drives the reporter gene expression, often a gene responsible for the production of an essential amino acid. Conventional Y2H is frequently used for PPI validation, but also as screen by means of cDNA libraries, as in the case of Y2H-seq. In Y2H-seq, the bait (for instance, a T3E-expressing) yeast strain is supertransformed with a cDNA prey library, allowing the discovery of new PPIs. The cDNA library can be obtained by cDNA synthesis from the RNA of, for example, Arabidopsis plants infected with the phytopathogen of interest to reflect the expression profile more closely upon infection. cDNA libraries should preferably represent the expression profile upon infection, but might also be made after treatment of the host plant with biotic stress-related phytohormones, such as jasmonate or salicylic acid. The ensuing yeast colonies after supertransformation and selection are then pooled, the extracted plasmid DNA sequenced, and the resulting sequencing reads mapped to the host genome (in this case Arabidopsis; Erffelinck et al., 2018). The results of a large-scale Y2H-seq screening experiment for T3Es from two distinct vascular phytopathogens (R. solanacearum and X. campestris) against a cDNA library of Arabidopsis (González-Fuente et al., 2020) have been made public via an online interactive database² and show that T3Es from distinct pathogens share some interactors, commonly referred to as plant effector hubs, whereas others are specific to the pathogen. Although Y2H-seq is a high-throughput and relatively cheap method, it is also linked to high rates of false positives and negatives due to the yeast background, protein folding artefacts, underrepresentation of full-length prey proteins encoded in cDNA libraries -especially large proteins-, and membrane protein targets.

For these reasons, in planta methods, such as (tandem) AP/IP or BioID coupled to MS-based methods are often preferred. The increasing sensitivity of mass spectrometers allows label-free quantification (LFQ), thereby facilitating experimental setups. Postmetabolic labeling approaches, such as tandem mass tag (TMT) labeling, are commonly used for multiplexed quantification. A frequently used IP-MS method is GFP trapping, in which GFP is fused to the protein of interest and antibodies against GFP precipitate the interaction complex. In this manner, interaction partners of several T3Es have been identified, such as the h-type thioredoxin targets (NbTRX-h9, NbTRX-h10, NbTRX-h11, and NbTRX-h14) of the R. solanacearum T3E RipAY in tobacco plants (Sang et al., 2018). By contrast, BioID relies on the use of a promiscuous biotin ligase that biotinylates primary amines of vicinal proteins after biotin application. When fused to the bait of interest, the proxeome (proteins in the bait proximity) can be identified by streptavidin-based purification. Over the years, different versions of the biotin ligase have been engineered: Escherichia coli BirA* (R118G; 35.3 kDa), Aquifex aeolicus BioID2 (27 kDa), and Bacillus subtilis BASU (28 kDa; Choi-Rhee et al., 2004; Kim et al., 2016; Ramanathan et al., 2018). However, these biotin-dependent proximity labeling tags show slow kinetics, require long biotin labeling times, and thrive at an optimal temperature of 37°C, making them rather unsuited for usage in plants. Nevertheless, despite poor labeling kinetics, the BirA* tag was successfully applied in rice (Oryza sativa) protoplasts to study the rice transcription factor OsFD2 (Lin et al., 2017) and later in transgenic Arabidopsis to investigate the proxeome of HopF2, a T3E from P. syringae (Khan et al., 2018). An AP-MS to complement the BioID results revealed an overlap of 58% (11/19) of the putative targets identified, although some previously reported interactors of HopF2 were missed with the BioID approach, probably because of the inherent BioID limitations, such as unavailability of free lysine residues in interacting preys, etc. (Khan et al., 2018). To overcome some of the challenges of the BirA* tag, such as a long labeling time, high biotin concentrations, and optimal temperature of 37°C, two new versions were created that enabled more efficient in planta proximity labeling. By a directed evolution-based approach, TurboID and its shorter version miniTurboID were established, allowing faster labeling and working temperatures around 30°C or less (Branon et al., 2018). Compared to the other proximity-labeling tags, TurboID was the most effective in different plant models, working at temperatures ranging from 22 to 28°C, and necessitating the addition of biotin for shorter periods of time (~2 h) for the efficient capture of plasma membrane interactomes (Arora et al., 2020). TurboID was, for example, used in tomato hairy root cultures to investigate the R. rhizogenes effector RolB, leading to the identification of TOPLESS and Novel Interactor of JAZ (NINJA) as direct interactors of RolB (Gryffroy et al., 2023). UltraID (19.7 kDa), a directed evolution variant of BioID2, has recently been developed, requiring even shorter labeling times, and showing less background biotinylation in the absence of exogenous biotin (Kubitz et al., 2022) in the case of mammalian cell culture, yeast, and bacteria. The use of ultraID in planta, however, has not been reported yet.

When these EH-PPI methods are used, a list of putative interactors is acquired. The read count or the peptide count/protein intensity (LFQ) and the expression ratios obtained from the Y2H-seq or MS-based methods can be indicative for the reliability (and proximity) of a certain hit. Although Y2H-seq, AP-MS, and BioID are complementary, the overlap of identified proteins is rather limited, mainly, as mentioned, because of the method inherent limitations and the fact that Y2H returns binary interactions, whereas AP-MS and BioID also can provide proximal interactions. Therefore, further validation of direct and indirect interactions remains necessary. For direct interactions, the highest scoring hits can be verified with conventional Y2H, co-immunoprecipitation (Co-IP), Förster Resonance Energy Transfer-Fluorescence-lifetime imaging microscopy (FRET-FLIM), bimolecular fluorescence complementation (BiFC), or split-luciferase, each with their advantages and disadvantages (Struk et al., 2019). Validation of indirect interactions requires alternative follow-up studies. Recently, a new validation technique for application in plants, designated knocksideways (KSP) is based on the heterodimerization of two domains, the human FK506-binding protein (FKBP) and the FKBP12-rapamycin-binding domain of mTOR (FRB) upon the addition of rapamycin (Winkler et al., 2021). FKBP is typically fused to the bait protein and FRB to an organellar marker of choice that differs from the subcellular localization of the bait protein. Upon rapamycin treatment, both domains reconstitute at the chosen organelle, unless the delocalization of the bait protein is hindered by interaction with a prey protein of interest. All three studied proteins are labeled with different fluorophores, so the (change in) localization can be monitored. An image analysis script was developed to quantify the interactions. Thus, KSP allows in planta localization of bait and prey fluorophores as well as validation of multiprotein interactions and quantification of the interaction.

Besides their interaction with host proteins and/or alteration of host RNA levels, T3Es can also modify the host metabolism. Metabolomic approaches are frequently useful to describe microbe-host interactions. For example, an untargeted gas chromatography/MS metabolomics experiment revealed that 22 metabolites were enriched in the xylem upon infection of tomato plants with R. solanacearum, of which eight could be used as sole carbon or nitrogen source (Lowe-Power et al., 2018). Although in microbe-host metabolomic experiments, it is often challenging to determine which organism produces the identified metabolite, R. solanacearum was shown to be responsible for the production of putrescine, one of the 22 enriched metabolites found, because the pathogen harbors the corresponding biosynthesis and export genes. Putrescine was required for proliferation inside the xylem and an increased disease progression (Lowe-Power et al., 2018). Several attempts have been made to distinguish more easily the host from the microbe metabolites (Allwood et al., 2010; Pang et al., 2018). By labeling the metabolites from P. syringae DC3000 with stable heavy isotopes, Arabidopsis guard cell metabolites could be separated from microbial metabolites (Pang et al., 2018). This approach also allowed the separation of metabolic profiles at different timepoints after infection. Nevertheless, the role of T3Es in shaping the host metabolite landscape is still poorly understood. The use of single (or multiple) T3E deletion mutants could be insightful to determine their individual (or combined) roles in the generation of metabolomic changes during infection, but very few untargeted metabolomics have been done for single phytobacterial T3Es thus far. One example, however, is coumaroyl tyramine, a metabolic compound produced in the phenylpropanoid pathway that accumulates by the T3E WtsE from the maize (Zea mays) pathogen Pantoea stewartii spp. stewartia (Asselin et al., 2015). This compound was identified through LC–MS/MS by comparing the metabolite profiles of maize seedling leaves infected either with the wild-type or with the wtsE-defective strain. Instead of a metabolomic approach, many more targeted approaches have been utilized to determine the effect of a T3E on specific host metabolites, as, for instance, accumulation of salicylic acid, jasmonic acid and glutathione by the R. solanacearum T3E RipE1 was previously reported (Sang et al., 2020). After the identification of accumulated/depleted metabolites caused by T3E expression, a biochemical activity assay could be set up to pinpoint whether this T3E is directly (enzymatically) responsible for the accumulation/depletion of the compound.

Driven by predictions and gathered data, experiments can be set up to pinpoint the exact function of the T3E of interest in planta. In the case of predicted protein domains with a known enzymatic function, targeted biochemical assays can be carried out. For instance, a ChaC domain that displays a GGCT activity in yeast and mammalian cells, was also found in the RipAY protein sequence and prompted the analysis of its GGCT activity and its ability to degrade glutathione (Fujiwara et al., 2016), revealing that RipAY had a robust GGCT activity in the presence of yeast or plant thioredoxins. Later on, RipAY was shown to interact with h-type thioredoxins in planta (Sang et al., 2018). Similarly, the discovery of the RipP2 acylation (i.e., (auto)acetylation) activity was steered by its homology to the YopJ-like family of effectors with a wide range of activities, such as (de-)sumoylation (AvrRxv and AvrXv4 from X. campestris pv. vesicatoria), de-ubiquitination and acetylation (YopJ from Yersinia spp.; Mukherjee et al., 2006, 2007; Roden et al., 2007), leading to the determination of the exact acetylation targets of RipP2 (Tasset et al., 2010; Le Roux et al., 2015). A third example is the R. solanacearum T3E RipTPS that presents homology with proteins harboring trehalose-6-phosphate (T6P) synthase (TPS) activity and really catalyzes T6P synthesis in yeast cells (Poueymiro et al., 2014). Besides relying on homology of protein domains to design follow-up experiments, known or newly identified EH-PPIs can also inform the selection of biochemical activity assays. For example, a catalase activity assay revealed that RipAK interacts with host catalases and pyruvate decarboxylases, thereby inhibiting host catalase activity and suppressing the plant’s HR (Sun et al., 2017; Wang et al., 2021).

The examples mentioned above indicated that information on the protein structure can aid successful functional characterization of T3Es. Determination of the structure of a protein (complex) is often a challenging task (Fujiwara et al., 2020). Nevertheless, the protein structures of some T3Es have been elucidated. So, the structure of Pto DC3000 T3E AvrPto has been solved in solution by nucleic magnetic resonance (NMR) spectroscopy (Lee et al., 2004) and the crystal structure of AvrB was determined by single wavelength anomalous dispersion (SAD; Wulf et al., 2004). For the RSSC, only the crystal structures of RipP2 (apo or in complex) have been described (Xia et al., 2021). Complexed protein structures are also interesting for the characterization of the protein function. In one case, the crystal structure was determined for RipP2, a known acetyltransferase, in complex with its host cofactor inositol hexaphosphate (IP₆), acetyl-coenzyme A (AcCoA), and the host resistance protein harboring a WRKY motif (RRS1-R_WRKY; Zhang et al., 2017). The WRKYGQK motif of the RRS1-R_WRKY substrate drives RipP2 to a lysine residue in the acetylation active site pocket. Additionally, IP₆ enhanced AcCoA and WRKY binding with RipP2, providing novel insights into the regulation of the RipP2 activity. To our knowledge, other RSSC T3E protein structures are still to be reported.

A potential solution for the challenging protein structure determination has recently been developed. The program, designated AlphaFold, relies on artificial intelligence to predict protein structures and has a high prediction accuracy (Senior et al., 2020). The AlphaFold database is currently vastly expanding, including additional proteomes and catalogued protein structures for even more accurate predictions. At the publication time, the AlphaFold source code is accessible online (Collabfold) and can be used to predict the protein structure of any protein of interest, as was done for the Ralstonia T3E RipE1 that harbors a cysteine protease domain (Tsakiri et al., 2022). More recently, the algorithm AlphaFold-Multimer has been created that could be used for the modelling of (effector-host protein) interactions (Evans et al., 2021).