Many Gram-negative plant-pathogenic bacteria including most Xanthomonas species rely on a highly conserved Type III protein secretion system (T3SS) for pathogenicity on host plants. The T3SS delivers effector proteins (T3Es) directly into host cells. T3Es are thought to be exported through a hollow Type III pilus independent of a classical cleavage-dependent signal peptide used by most other bacterial secretion systems (Büttner and Bonas, 2010). T3Es are known to suppress host defenses and alter plant physiology to enhance infection (Büttner, 2016).

Genes encoding the T3SS are typically clustered in a ∼20 kb genomic island. In plant pathogens, the T3SS genes have been collectively called the hrp [hypersensitive response (HR) and pathogenicity] gene cluster since mutants in these genes were found to be impaired in their ability to cause the HR or programmed cell death on non-host plants and lost pathogenicity on host plants (Büttner and He, 2009). The first hrp gene clusters from Xanthomonas have been described in 1991 for Xanthomonas euvesicatoria (a.k.a. Xanthomonas campestris pv. vesicatoria) and for Xanthomonas campestris pv. campestris (Arlat et al., 1991; Bonas et al., 1991). hrp genes have been further subcategorized according to their broad conservation and subtle phenotypes of some of the corresponding mutants. Eleven hrp genes that are conserved among plant and animal pathogens and which are critical for causing disease have been renamed into hrc (HR conserved) genes (Bogdanove et al., 1996). And those genes that contribute to but are not essential for the plant–bacteria interaction are called hpa (hrp associated) genes. The core hrp cluster consists of six hrp operons, hrpA to hrpF, encompassing a total of eleven hrc, seven hrp, and three hpa genes with identical genetic organization in other Xanthomonas strains (Weber et al., 2007).

Two different types of hrp gene clusters have been described for plant pathogens, based on their gene repertoires, operon structures and mode of regulation (Büttner and Bonas, 2002a). Group I comprises Pseudomonas syringae and plant-pathogenic enterobacteria while group II includes bacteria in the genera Acidovorax and Xanthomonas and the Ralstonia solanacearum species complex. Group-II hrp genes and their co-regulated T3E genes are under control of two key regulatory genes, hrpG and hrpX (hrpB in R. solanacearum and Acidovorax) (Büttner and Bonas, 2010). HrpG shares characteristics with two-component response regulators of the OmpR family and induces the expression of hrpX. HrpX is an AraC-type transcriptional activator that binds to the plant-inducible promoter (PIP) box (TTCGB-N₁₅-TTCGB), which is present in front of most of the hrp operons and many T3E genes (Koebnik et al., 2006).

Expression of the hrp gene cluster results in the formation of the contiguous molecular syringe that spans the bacterial cell envelope (Büttner and He, 2009; Kay and Bonas, 2009). It is predicted to consist of a multi-ring structure that is embedded in both inner and outer membrane and linked to the ATPase complex in the cytosol (Lorenz et al., 2012). In contrast to animal pathogens with their needle complex, plant pathogens evolved a pilus-like structure (i.e., encoded by Xanthomonas hrpE gene) for protein delivery into the plant’s cells (Weber et al., 2005). Type III-secreted proteins were located near the Hrp pilus during their secretion, suggesting that Hrp pili serve as conduits for the translocation of T3Es (Romantschuk et al., 2001).

Entry into the host cells is a final step in protein delivery; for this purpose, animal pathogens have evolved a multi-protein pore-forming translocon complex consisting of YopB and YopD, or homologs thereof (Chatterjee et al., 2013). However, no homologs of YopB or YopD were found in plant pathogens. Based on genetics and biophysical experiments, HrpF was identified to fulfill the role of a translocon protein in Xanthomonas (Rossier et al., 2000; Büttner and Bonas, 2002b); similarly, its homolog PopF is required for protein delivery from R. solanacearum (Meyer et al., 2006).

To date, the largest and the most conserved clade of xanthomonads (clade-2) includes the well-studied model species X. euvesicatoria, X. campestris, and X. oryzae (Parkinson et al., 2007). All functional studies of the Xanthomonas hrp cluster have been performed in this group. By far most strains of this Xanthomonas clade contain a canonical hrp cluster with the same genetic organization as in the model strains 85–10 and 8004 (Arlat et al., 1991; Bonas et al., 1991). Only recently researchers demonstrated that some clade-2 strains isolated from barley or cannabis are pathogenic despite the absence of an Hrp T3SS (Ignatov et al., 2015; Jacobs et al., 2015).

Much less work has been performed on members of the clade-1, which comprises the five highly diverse species Xanthomonas albilineans, Xanthomonas sacchari, Xanthomonas theicola, Xanthomonas hyacinthi, and Xanthomonas translucens. For two of them, X. albilineans and X. sacchari, genomic analyses demonstrated that they do not contain an Hrp T3SS (Pieretti et al., 2009; Studholme et al., 2011). In contrast, the 48 available genome sequences from seven different pathovars of X. translucens (Wichmann et al., 2013; Gardiner et al., 2014; Pesce et al., 2015a,b; Hersemann et al., 2016, 2017; Jaenicke et al., 2016; Peng et al., 2016) and draft genome sequence of X. hyacinthi (Naushad et al., 2015) revealed that all of them contain an Hrp T3SS the genetic organization of which is at variance to those from the clade-2.

X. translucens forms a diverse group of bacteria, most of which are pathogenic to monocotyledonous plants, such as various grasses and small-grain cereals. Strains have been assigned to ten pathovar subgroups based on symptoms on different host plants. Mutants in the hrp structural genes hrcR and hrpE and in the major regulator gene hrpG of the forage grass pathogen X. translucens pv. graminis were impaired in their ability to cause symptoms when compared with the wild-type strain Xtg29 (Wichmann et al., 2013). Notably, unlike other clade-2 Xanthomonas spp., the X. translucens pv. graminis hrpG mutant, the only mutant that was complemented in this study, still caused clearly visible symptoms, and complementation was only partially achieved. Moreover, bacterial colonization of the hrp mutants in the plant tissue was only slightly impaired, and all mutants were still quantified in comparable numbers similar to wild-type bacteria; whereas clade-2 Xanthomonas hrp mutants are typically severely reduced in colonization (Wichmann et al., 2013). Hence, these observations were in sharp contrast to those with clade-2 xanthomonads, thus casting doubt on the importance of the Hrp T3SS for the pathogenicity of X. translucens.

In this study, we therefore compared the core hrp gene clusters and their flanking sequences of several X. translucens strains with those from other bacterial pathogens. We analyzed them for novel genes that might be important for pathogenicity. Knockout mutagenesis of a conserved hrc gene demonstrated the importance of the Hrp T3SS for pathogenicity of barley-pathogenic X. translucens strains. Two other conserved genes, initially called hgiA and hgiB (for hrpG-induced gene), which are encoded next to the core hrp gene cluster, were analyzed for their contribution to pathogenicity of two pathovars of X. translucens. One of them, hgiA, was found to be important for pathogenicity on barley and to contribute to the non-host HR on pepper plants and was therefore re-named hpaT (for hrp-associated gene T). Given the absence of hrpF homologs in strains of X. translucens, we provide evidence that hpaT might encode an undescribed translocon component of the X. translucens T3SS.

The first hrp gene clusters from Xanthomonas were discovered in the clade-2 xanthomonads X. euvesicatoria and X. campestris, which rapidly became model organisms to study the pathogenicity of Xanthomonas. Other species, such as the rice pathogen X. oryzae and the citrus pathogen X. citri pv. citri were studied in detail as well due to their high economic interest and concerns about associated diseases, namely bacterial leaf blight and leaf streak of rice and citrus canker, respectively.

Collectively, data has accumulated for these species over the last 25 years. However, much less is known for species of the clade-1. X. albilineans GPE PC73 was the first clade-1 strain whose genome was sequenced (Pieretti et al., 2009), followed by strains of X. sacchari (Studholme et al., 2011). It came as a surprise that these clade-1 strains did not have an hrp gene cluster because it was known from previous work that at least pathovars of X. translucens contain hrp genes and T3Es of the TAL family (Alizadeh et al., 1997; Bragard et al., 1997). Indeed, when the first X. translucens genome was sequenced, a complete, but noncanonical hrp gene cluster was found (Wichmann et al., 2013). In contrast to clade-2 strains, however, a knockout in a conserved hrp gene of the grass pathogen X. translucens pv. graminis did not lead to the typically observed drastic loss of virulence (Wichmann et al., 2013).

In this study, we demonstrated that a mutation of the conserved hrcT gene of the barley pathogens X. translucens pv. hordei and X. translucens pv. translucens resulted in a loss of pathogenicity. Mutants were strongly attenuated in symptom formation and could hardly colonize the leaf blade upon leaf clip inoculation, a typical phenotype that is shared with clade-2 strains. Surprisingly hrp mutants in hrcR and hrpE and in the regulatory gene hrpG in grass pathogen X. translucens pv. graminis had little effect on virulence. This seems largely inconsistent with the wide importance of the T3SS structural components and regulators in R. solanacearum, clade-2 Xanthomonas spp. and Pseudomonas syringae. Perhaps T3Es are more import for the colonization of small-grain cereals than the colonization of grasses. It will therefore be interesting to broadly compare the same hrp mutants (e.g., hrcT) in other pathovars of X. translucens, such as the small-grain pathovars undulosa and the grass pathovars arrhenateri, poae, phlei, and phleipratensis. It will also be worthwhile to analyze the repertoires of T3Es of all these X. translucens pathovars.

A comparison of five X. translucens hrp gene clusters revealed that they all have the same genetic organization, which to some extend resembles that of the hrp cluster from R. solanacearum. In both species, the two key regulatory genes, hrpG and hrpX/hrpB, are encoded on the left and right side of the hrp cluster. In general, both flanking regions are largely conserved among the five X. translucens strains (Figure 2). One of the few differences was the presence of a large open reading frame (1587 bp) beyond xopF in the X. translucens pv. translucens strain DSM 18974, which is homolog to xopM from X. euvesicatoria (Schulze et al., 2012) and has homologs in other xanthomonads and in some betaproteobacteria, such as Acidovorax, Collimonas, Ralstonia, and Rhizobacter. A homologous sequence was also found at the same position in the two other clade-1 xanthomonads, X. hyacinthi, and X. theicola, and frameshifted derivatives were observed in strains of X. translucens pv. graminis. Inactivation of genes at the boundaries of hrp clusters has been observed for clade-2 xanthomonads and is often linked to T3E genes. For instance, X. oryzae pv. oryzae encodes the LRR effector XopAE (syn. HpaF) beyond hrpF while the corresponding gene in X. euvesicatoria 85–10 suffers from a frameshift mutation (Sugio et al., 2005). Likewise, the candidate chaperone gene hpa3 in front of hpa4 (syn. xopF) from X. oryzae pv. oryzae contains a frameshift in the X. euvesicatoria homolog (Sugio et al., 2005). The broadly conserved T3E XopM can be considered as a core effector, a finding that is substantiated by the observation that the xopM gene is typically encoded in close vicinity to the T3SS gene cluster, not only in Xanthomonas (e.g., next to hpaH in X. euvesicatoria strain 85–10 and X. cannabis pv. phaseoli strain Nyagatare), but also in the betaproteobacteria Acidovorax and Collimonas.

In addition to the atypical location of the hrpG and hrpX genes, we detected two other genes in the hrp regions that are conserved among strains of X. translucens and in X. hyacinthi. Both genes, hpaT and hgiB, are preceded by a canonical plant-inducible promoter sequence (Koebnik et al., 2006). Quantitative RT-PCR experiments with strain UPB820^R revealed that both genes are indeed strongly activated by HrpG (and HrpX). Therefore, these genes have tentatively been named hgiA and hgiB, hgi for HrpG-induced, following the nomenclature of Noël et al. (2001). Owing to the phenotype of the hgiA mutant, however, hgiA was renamed hpaT.

Both hpaT and hgiB show considerable sequence diversity among the analyzed X. translucens strains (Figure 3), a finding that let us speculate that the genes may be under diversifying selection and that the gene products may be exposed to the plant surveillance system. Yet, the small number of available sequences does not allow testing this hypothesis. Using BLAST and PSI-BLAST searches in NCBI GenBank, homologs of hgiB have only been detected in X. translucens while more distantly related genes of hpaT were also found in several betaproteobacteria, such as Burkholderia, Collimonas, Massilia, Paraburkholderia, and Ralstonia, where they are typically encoded in the vicinity of a T3SS gene cluster.

Insertional knockout mutants in hgiB did not show any defect in virulence when tested in two different strains. Apparently, this gene does not obviously contribute to the type III secretion and effector translocation. Further work will address the question if hgiB encodes another T3E. In contrast, when hpaT was knocked-out in two different strains by insertion of a plasmid which contains a partial hpaT sequence by single homologous recombination, the mutant bacteria was non-pathogenic on the barley variety Morex.

Moreover, we analyzed two strains of X. translucens that caused a non-host HR in the pepper cultivar ECW-10R. ECW-10R carries the Bs1 resistance gene and produces an HR in response to the effector AvrBs1 from X. campestris pv. vesicatoria (Cook and Stall, 1963). However, neither UPB787^R nor UPB820^R has a homolog of avrBs1 (GenBank acc. no. JTEM00000000, and data not shown). It is therefore likely that another, yet unidentified T3E is recognized by pepper.

HR elicitation by an avr gene depends on a functional T3SS, as shown by the phenotypes of the hrcT mutant. Full HR was also dependent on hpaT and hpaH. Altogether, the observed phenotypes with respect to hpaT and hpaH correspond to the typical phenotype of an hrp mutant, which is non-pathogenic on susceptible host plants and cannot cause a non-host HR. This strongly suggests that hpaT and hpaH contribute to the delivery of T3Es into plant cells.

Strains of X. translucens do not possess the translocon protein HrpF, which is ubiquitous in all clade-2 xanthomonads that have an Hrp system. Wichmann et al. (2013) speculated that cells of X. translucens pv. graminis may use breaches to gain direct access to the xylem cells in which they reside, and as these cells do not contain a cell membrane, the hrpF gene may be dispensable for effector translocation and, consequently, may have been lost during evolution. However, using GFP-labeled bacteria of X. translucens pv. translucens we could detect the bacteria in the vascular bundle upon leaf clipping, clearly demonstrating that the bacteria colonize the intercellular space of barley leaves (unpublished results). We therefore favor the second alternative that was put forward by Wichmann et al., namely that a distantly related hrpF gene (or better, an anolog of it) might be present, but cannot be recognized on the basis of sequence identity.

Comparison of T3SS gene clusters with a similar genetic organization and mode of regulation (via HrpX) from various bacteria revealed that all the strains that do not have a detectable hrpF gene do instead possess a homolog of hpaT. We took this mutual exclusivity as circumstantial evidence that hpaT might encode a functional analog of HrpF. No transmembrane alpha-helical segment was predicted for HpaT by any of the tested algorithms, a feature that is shared with other putative translocon proteins, HrpF from X. euvesicatoria and PopF1 and PopF2 from R. solanacearum (data not shown). Moreover, both protein families, the HrpF family and the HpaT family, share the presence of repeated sequence motifs as a common structural feature. HrpF from X. campestris pv. campestris has three imperfect repeats of ∼110 amino acids, whereas the prototype HrpF protein from X. euvesicatoria has only two repeats, and the corresponding homolog in Sinorhizobium fredii, NolX, has only one copy of the repeated sequences (Sugio et al., 2005). Another polymorphic region is found in the C-terminal region of HrpF proteins where 5 to 15 copies of an imperfect tetrapeptide repeat are present in different species of Xanthomonas (unpublished data). Similarly, HpaT proteins from X. translucens contain several repeated sequence motifs (Supplementary Figure S1). It is tempting to speculate whether or not these variations function in the adaptation to specific host plants and/or evolved to escape from detection by the plant immune system.

To demonstrate functional equivalence with the putative translocon protein HrpF (Büttner et al., 2002), we tested whether or not an hpaT mutant can be complemented by hrpF from X. euvesicatoria, and vice versa, whether an hrpF mutant can be complemented by hpaT from X. translucens strain CFBP 2541. For the assay we infiltrated pepper leaves of the cultivar ECW-10R (containing the resistance gene Bs1), which trigger a race-specific HR to X. euvesicatoria strains expressing the avrBs1 gene and a non-host HR to some strains of X. translucens (e.g., UPB787 and UPB820). While the hrpF mutant could hardly be complemented, as indicated by the occasional presence of little brown spots in the area of infiltration, we observed strong complementation of the hpaT mutant by hrpF. This result was rather surprising given the fact that the HrpF homolog NolX, which is 48% identical to HrpF, was not able to complement an hrpF mutant (Huguet and Bonas, 1997). Notably, HpaT and HrpF belong to distinct protein families, which share only 16% or less amino acid sequence identity in pairwise comparisons (Supplementary Figure S8 and Table S3). From these results, we conclude that HpaT and HrpF may be analogous translocon proteins, which are to some degree functionally interchangeable.

Surprisingly, the polar insertion mutant in hpaT could be partially complemented by hpaT, but also by the downstream gene hpaH. This finding is puzzling since HpaH was initially predicted to act in the periplasmic space and may be involved in remodeling of the peptidoglycan layer thanks to its predicted lytic transglycosylase activity, thus helping to build up the T3SS machinery (Noël et al., 2002; Zhang et al., 2008). To that effect, an N-terminal signal peptide for the general Sec-dependent secretion pathway is predicted for HpaH using Phobius, and enzymatic assays with Hpa2 from X. oryzae revealed lytic activity against the bacterial cell walls (Zhang et al., 2008).

Conflicting data exist, however, with respect to the export pathway and contribution to pathogenicity in the several Xanthomonas strains. Mutant analyses in hpaH (X. euvesicatoria) or its homolog hpa2 in X. oryzae had indicated only minor (hpaH from strain 85-10, hpa2 from strain RS105) or no (hpa2 from strain PXO99^A) effect on the interaction with its host plants (Zhu et al., 2000; Noël et al., 2002; Li et al., 2011a). However, others have demonstrated a significant reduction in symptoms and bacterial counts on rice when hpa2 was inactivated (strain PXO99^A) (Zhang et al., 2008). Similarly confusing is the observation that a hpa2 mutant in the X. oryzae strain PXO99^A completely loses its ability to cause a non-host HR on tobacco (Nicotiana benthamiana) (Zhang et al., 2008), while a hpa2 mutant in the X. oryzae strain RS105 triggered an HR similar to the wild type (Li et al., 2011a). Since it was shown by reporter fusions that HpaH contributes to the secretion of a subset of effectors by X. euvesicatoria (Büttner et al., 2007), subtle differences in the experimental conditions and/or differences in the sets of affected T3Es in the various strains may explain these results, which are otherwise hardly to reconcile.

The HpaH protein has been localized to the periplasmic space and in the extracellular milieu (Hausner et al., 2017). Conficting data exist whether export of HpaH/Hpa2 depends on the T3SS or not (Li et al., 2011a; Hausner et al., 2017). Similarly, while HpaH was reported to bind to the peptidoglycan as well as to periplasmic components of the T3SS (Hausner et al., 2017). Li et al. (2011a) reported that Hpa2 has the ability to bind to the host cell membrane. These apparently conflicting data may suggest that HpaH and Hpa2, which are 89% identical to each other at the amino acid level, act in a different manner. Since HpaH from X. translucens is basically equally distant (35% identity) to both HpaH from X. euvesicatoria and Hpa2 from X. oryzae one cannot derive at conclusions on its mode of action based on sequence similarity.

Our results suggest that both proteins, HpaT and HpaH, contribute to the function of the T3SS. Since the hpaT mutant could be partially complemented by a plasmid-borne copy of hpaH, HpaT is not absolutely required for the translocation of effector proteins into plant cells but rather serves as a facilitator of transport. To avoid multicopy effects, it would be interesting to perform complementation assays with chromosomal insertions of hpaT, hpaH, or both genes. Similar observations as with hpaT have been made for the other candidate translocon gene, hrpF from X. oryzae. Mutations in hrpF led to drastic loss of virulence but did not completely eliminate pathogenicity (Sugio et al., 2005; Cho et al., 2008; Li et al., 2011a). This finding is also reminiscent of the phenotype of a P. syringae mutant in hrpK, which shares low similarity with hrpF from X. euvesicatoria (Petnicki-Ocwieja et al., 2005). Strikingly, while single mutants in hrpF and hpa2 still provoked symptoms on rice, only a double mutant in both genes was non-pathogenic on its host (Li et al., 2011a). Moreover, it was shown by yeast two-hybrid and protein pull-down experiments that Hpa2 interacts with HrpF, and bimolecular fluorescence complementation assays using split YFP demonstrated that both proteins can interact in the plant cell membrane (Li et al., 2011a). From these results, along with data from several avirulence reporter experiments, Li et al. (2011a) concluded that both proteins, Hpa2 and HrpF, act together to translocate T3Es into plant cells, even if each protein alone has sufficient activity to allow effector translocation at much reduced rate.

Our complementation data suggest a mechanism for X. translucens where both HpaH and HpaT contribute to effector translocation, yet the presence of only one of the two protein is still sufficient to allow some effector delivery to the plant cell. A similar redundancy of translocon components was also found in P. syringae where a quintuple mutant in the candidate translocon components hrpK1, hrpW1, hrpZ1, hopP1, and hopAK1 could be complemented by each of these genes alone except for hopP1, which suggested that a consortium of semi-redundant translocators cooperate in the translocon formation (Kvitko et al., 2007). A similar multi-component translocon complex could operate in Xanthomonas, that may involve glycine-rich proteins (e.g., XopA, syn. Hpa1) as a third component (Sugio et al., 2005). These results challenge the view of a simple, single-protein translocon pore; more work is needed to understand the molecular details at the “port of entry” (Büttner and Bonas, 2002b).

RK and CB conceived the study. RK and CP performed the bioinformatic analyses. CP, EB, and TV generated mutants and constructed plasmids. CP, EB, JJ, MP, and RK conducted plant assays. JJ, MP, and RK designed, performed and analyzed the qRT-PCR experiments. All authors analyzed the data, and wrote the manuscript. All authors have read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.