Variation and inheritance of the Xanthomonas gene cluster required for activation of XA21-mediated immunity

The rice XA21-mediated immune response is activated upon recognition of the RaxX peptide produced by the bacterium Xanthomonas oryzae pv. oryzae (Xoo). The 60 residue RaxX precursor is posttranslationally modified to form a sulfated tyrosine peptide that shares sequence and functional similarity with the plant sulfated tyrosine (PSY) peptide hormones. The five kb raxX-raxSTAB gene cluster of Xoo encodes RaxX, the RaxST tyrosylprotein sulfotransferase, and the RaxA and RaxB components of a predicted type one secretion system. The identified the complete raxX-raxSTAB gene cluster is present only in Xanthomonas spp., in five distinct lineages in addition to X. oryzae. The phylogenetic distribution of the raxX-raxSTAB gene cluster is consistent with the occurrence of multiple lateral transfer events during Xanthomonas speciation. RaxX variants representing each of the five lineages, and three Xoo RaxX variants, fail to activate the XA21-mediated immune response yet retain peptide hormone activity. These RaxX variants contain a restricted set of missense mutations, consistent with the hypothesis that selection acts to maintain peptide hormone-like function. These observations are also consistent with the hypothesis that the XA21 receptor evolved specifically to recognize Xoo RaxX.


INTRODUCTION
Host receptors activate innate immunity pathways upon pathogen recognition (Ronald & Beutler, 15 2010). The gene encoding the rice XA21 receptor kinase (Song et al., 1995) confers resistance 16 against most strains of the gamma-proteobacterium Xanthomonas oryzae pv. oryzae (Xoo) 17 (Wang et al., 1996). This well-studied XA21-Xoo interaction provides a basis from which to 18 understand molecular and evolutionary mechanisms of host-microbe interactions. 19 Four Xoo genes that are required for activation of XA21-mediated immunity, are located in the 20 raxX-raxSTAB gene cluster (Fig. 1). The 60-residue RaxX predicted precursor protein variants examined in this study, together with two representative PSY sequences for comparison. 47

Boundaries flanking the raxX-raxSTAB gene cluster and adjacent genes suggest lateral transfer through general recombination
The raxX-raxSTAB gene cluster lies between two core (housekeeping) genes (Fig. 1). One, 121 gcvP, encodes the pyridoxal-phosphate subunit of glycine dehydrogenase. An approximately 122 170 nt riboswitch (gcvR in Fig. 1 Comparing the gcvP -[raxX-raxSTAB] -"mfsX" region from the reference genomes reveals 127 sharp boundaries flanking the position of the raxX-raxSTAB gene cluster. On the left flank, 128 substantial nucleotide identity spans the gcvP gene, the gcvR riboswitch, and a predicted gcvR 129 promoter -10 element (Mitchell et al., 2003) (Fig. S2). On the right flank, identity begins 130 shortly after the "mfsX" initiation codon. Accordingly, upstream sequence elements for initiating 131 "mfsX" gene transcription (Mitchell et al., 2003) and translation (Ma et al., 2002) are conserved 132 within, but not between, raxX-raxSTAB gene cluster-positive and -negative sequences (Fig. S2). 133 Between these boundaries in raxX-raxSTAB gene cluster-negative species, the compact (≤ 200 134 nt) gcvP-"mfsX" intergenic sequence is modestly conserved in most genomes (about 60-80% 135 overall identity; Fig. S2). Much of this identity comes from the "mfsX" potential transcription 136 and translation initiation sequences described above. The overall intergenic sequence is less 137 conserved in the early-branching species (X. albilineans, X. hyacinthi and X. sacchari), 138 displaying about 50-65% overall identity. 139 We hypothesize that raxX-raxSTAB gene cluster phylogenetic distribution results from general 140 recombination between conserved genes flanking each side (e.g., in or beyond the gcvP and 141 "mfsX" genes). Two observations are consistent with the hypothesis, First, we observed that the 142 sequences flanking the raxX-raxSTAB gene cluster are different from the gcvP-"mfsX" intergenic 143 sequence in raxX-raxSTAB gene cluster-negative strains (Fig. S2). This argues against models in 144 which the raxX-raxSTAB gene cluster has integrated into the gcvP-"mfsX" intergenic sequence 145 during lateral transfer events. 146 The second observation consistent with lateral transfer via general recombination is that gcvP 147 length polymorphisms (Fig 1 and Fig. S3) do not align with Xanthomonas phylogenetic 148 relationships (Fig. 3). Inheritance patterns such as this often result from general recombination 149 in the vicinity (Nelson et al., 1997). 150 Notably, this gcvP-"mfsX" intergenic region conserved is also conserved in the X. citri lineage 151 (Fig. S2). If the raxX-raxSTAB gene cluster was lost during formation of this lineage (see 152 above), then general recombination would replace the resident raxX-raxSTAB gene cluster with a 153 donor conserved gcvP-"mfsX" region. 154 raxST but not raxX homologs are present in genomes from diverse bacterial species Our GenBank database searches identified raxX homologs and the raxX-raxSTAB gene cluster 155 only in Xanthomonas spp. However, these searches did identify raxST homologs encoding 156 proteins with about 40% identity to, and approximately the same length as, the Xoo RaxST 157 protein. These sequences include the PAPS-binding motifs that define sulfotransferase activity 158 (da Silva et al., 2004, Negishi et al., 2001. Regardless of its current function, a raxST homolog 159 potentially could evolve to encode tyrosylprotein sulfotransferase activity. 160 None of these raxST homologs is associated with a raxX homolog, and most also are not 161 associated with raxA or raxB homologs. Presumably, the enzymes by these raxST homologs act 162 on substrates other than RaxX. These raxST homologs support the hypothesis that the raxSTAB 163 cluster arose from a new combination of pre-existing raxST, raxA, and raxB homologs. These raxST homologs are in diverse genetic contexts in a range of bacterial phyla including 167 Proteobacteria and Cyanobacteria (Fig. S4). Nevertheless, for most species represented by 168 multiple genome sequences, the raxST homolog was detected in a minority of individuals, so it is 169 not part of the core genome in these strains. Moreover, relationships between species in a raxST 170 gene phylogenetic tree bear no resemblance to those in the overall tree of bacterial species. For 171 example, in the raxST gene tree, sequences from Cyanobacteria are flanked on both sides by 172 sequences from Proteobacteria (Fig. S4). Together, these findings provide evidence for lateral 173 transfer of raxST homologs (Kuo & Ochman, 2009).

RaxX protein sequence variants representing all six raxX-raxSTAB gene cluster-positive lineages
RaxX protein sequences from diverse Xanthomonas spp. assort into several sequence groups 175 differentiated by polymorphisms within the predicted mature peptide sequence (Fig. 2)  To assess the function of RaxX variants, we focused on frequently observed variants in species 182 represented by numerous genome sequences (Fig. S1). These include sequence groups A, B and 183 D from X. oryzae pv. oryzae and X. oryzae pv. oryzicola, as well as sequence groups E, G and H, 184 representing most genomes for the X. euvesicatoria and X. vasicola groups (Fig. 2). Finally, 185 sequence group K is most numerous among X. translucens genomes. The comparison reference 186 is the RaxX protein sequence from the Philippines Xoo strain PXO99 A (sequence group A). 187 Examples from lower frequency (mostly unique) sequence groups were analyzed by 188 complementation, as described below. 189 RaxX variants promote root growth but fail to activate the XA21-mediated immune response We generated and purified tyrosine-sulfated full-length (unprocessed) RaxX peptides for these 190 seven variants using an expanded genetic code approach (see methods) ( Fig. 2) Root lengths for seedlings grown without added peptide averaged 23.5 mm, whereas root lengths 199 for seedlings grown with 100 nM peptide were at least twice as long ( Fig. 5A and Fig. 5B). This 200 observation is consistent with the hypothesis that these peptides mimic PSY1 peptide hormone 201 activity. Note that three of these variants (groups D, E and G) were examined previously (Pruitt 202 et al., 2017) and are included here to facilitate direct comparisons as well as to monitor 203 consistency of results. 204 In the second assay, we tested each RaxX peptide for direct activation of XA21-mediated 205 immunity by assaying induction of the PR10b marker gene as a readout for immune activation 206 (Thomas et al., 2016, Pruitt et al., 2015. In contrast to results with the root growth assay, here 207 only the group A RaxX protein (from Xoo strain PXO99 A ) was able to induce XA21-mediated 208 PR10b marker gene expression (Fig. 5C). 209 In a separate test for activation of XA21-mediated immunity, we used a raxX deletion mutant 210 of Xoo strain PXO99 A as a host for genetic complementation. We tested each of the raxX alleles 211 shown in Fig. 2, which includes examples from lower frequency (mostly unique) sequence 212 groups. We introduced each raxX allele into the raxX test strain, and monitored disease 213 progression in leaves of whole plants. Only the group A raxX allele (from Xoo strain PXO99 A ) 214 was able to complement the Xoo PXO99 A raxX strain to activate XA21-mediated immunity 215 (Fig. 6). Expression of each raxX allele was confirmed by qPCR (Fig. S5). 216 Together, these results provide direct evidence that activation of XA21-mediated immunity is 217 restricted to RaxX proteins from sequence group A, found in most strains of Xoo. None of the 218 other X. oryzae RaxX variants tested (including RaxX from X. oryzae pv. oryzicola) was able to 219 activate XA21-mediated immunity. The observation that all RaxX proteins tested stimulated 220 Arabidopsis root growth suggests that the RaxX PSY peptide mimicry function is not restricted 221 to rice. 222 African Xoo strain AXO1947 RaxX and RaxST natural variants both lead to evasion of the

XA21 immune receptor
The raxX alleles from Xoo strains IXO685 and AXO1947 failed to complement the raxX 223 mutant of Xoo strain PXO99 A for XA21 immune activation (Fig. 6). In addition to its variant 224 raxX allele (Fig. 2), we noted that Xoo strain AXO1947 (Huguet-Tapia et al., 2016) carries seven 225 missense polymorphisms in the raxST gene (Fig. S6) not present in other Xoo strains such as 226 IXO685. To determine if the variant raxST allele from strain AXO1947 encodes a functional 227 protein, we performed additional complementation tests. 228 We found that the raxX allele from strain PXO99 A conferred the XA21 immune activation 229 phenotype upon strain IXO685 but not upon strain AXO1947 (Fig. 7B). This result suggests that 230 the raxX variant allele is not the only factor that prevents strain AXO1947 from activating the 231 XA21 immune response. Consistent with this hypothesis, the raxST allele from strain PXO99 A 232 failed to confer the XA21 immune activation phenotype upon strain AXO1947 (Fig. 7D). In 233 contrast, addition of both the raxX and raxST alleles from strain PXO99 A was sufficient to confer 234 the XA21 immune activation phenotype upon strain AXO1947 (Fig. 7F). 235 Taken together, these results suggest that Xoo strain AXO1947 has mutant versions of both 236 genes, raxST and raxX. Analysis by qRT-PCR confirms that these genes were expressed in the 237 complemented strains (Fig. S7). 238

RaxST variants from Xoo strain AXO1947
To determine which of the RaxST missense polymorphisms is responsible for the apparent 239 reduction in enzyme activity, we used site-specific mutagenesis to introduce each individually 240 into the raxST gene from strain PXO99 A . Genes encoding two of these [His-50 to Asp (H50D) 241 and Arg-129 to Leu (R129L)] were unable to complement the raxST mutant of Xoo strain 242 PXO99 A for XA21 immune activation (Fig. 8), indicating that both His-50 and Arg-129 are 243 necessary for RaxST activity.  We generated a RaxST molecular model with the program iTasser (Yang & Zhang, 2015) using 252 the crystal structure of human tyrosylprotein sulfotransferase-2 (TPST2) as a template (PDB: 253 3AP1). The sequence alignment is shown in Fig. S8. TPST2 is a functional dimer (Teramoto et 254 al., 2013), which is replicated in the RaxST structural model (Fig. S9). The two essential 255 residues identified from Xoo strain AXO1947, His-50 and Arg-129, display surface exposed side 256 chains in close proximity to the corresponding position for the bound substrate peptide co-257 crystalized with TPST2. These residues are distal to the catalytic site. Therefore, we 258 hypothesize that these RaxST residues are involved in RaxX peptide binding. 259

DISCUSSION
We previously hypothesized that RaxX mimics the actions of PSY hormones, and that the XA21 This prediction is supported here by our finding that all the RaxX variants tested stimulate root 262 growth ( Fig. 5A and Fig. 5B) (Pruitt et al., 2017) but fail to activate the XA21-mediated immune 263 response ( Fig. 5C and Fig. 6). Thus, RaxX sequence determinants are more stringent for XA21-264 mediated immunity activation than for root growth stimulation. In this discussion, we consider 265 two questions: (1) What are potential selective pressures acting on RaxX that affect sequence 266 variation; and (2) How was the raxX-raxSTAB gene cluster inherited in Xanthomonas spp.? 267

Opposing selection pressures drive RaxX natural variation
Maintenance of the raxX-raxSTAB gene cluster (Fig. 3) suggests that RaxX provides fitness 268 benefits to diverse Xanthomonas spp., presumably during their interactions with hosts that 269 collectively encompass a range of monocot and dicot species. This hypothesis is supported by in Thus, it appears that some Xoo strains that evade activation of XA21-mediated immunity arise 288 from a restricted set of raxX missense substitution alleles encoding variants that retain PSY-like 289 function. This observation suggests that it may be possible to engineer novel XA21 variants that 290 recognize these variant RaxX proteins. If so, it may then be possible to engineer broad-spectrum 291 resistance against Xoo (and other raxX-raxSTAB gene cluster-positive Xanthomonas spp.) by 292 expressing multiple XA21 proteins that collectively recognize multiple RaxX variants. 293 We also have identified raxST and/or raxA gene loss of function alterations in Xoo field isolates 294 intact raxX-raxSTAB gene cluster originated in an ancestor to the lineage containing X. oryzae, 303 X. euvesicatoria, and related species, with subsequent gains or loss through lateral transfer (Fig.  304 2). Relatively few events appear to have been necessary to form the raxX-raxSTAB gene cluster. 305 The raxX gene might have evolved from the gene for the secreted peptide substrate of the 306 RaxAB ancestor. The complete cluster would result from incorporation of the ancestral raxST 307 gene, homologs of which are distributed broadly (Fig. S4). 308

Survey of the RaxX, RaxST and the raxX-STAB genomic region in publicly available databases
We used the 5kb long Xoo PXO99 A raxX-raxSTAB genomic region, including 600 bp upstream 328 of raxST and 70 bp downstream of raxB, as query to search the following NCBI databases with 329 blastn and megablast using e-value cut-off of 1e-3; nr/nt, htgs, 330 refseq_genomic_representative_genomes, refseq_genomic, and gss. To identify RaxX homologs 331 we used the protein sequence of RaxX from Xoo PXO99 A as query to search the same databases 332 using tblastn with a PAM30 scoring matrix to account for the short sequence length of RaxX. In 333 case of raxST from Xoo PXO99 A we used the genomic coding sequence to search the same 334 databases using the same cut-offs. In addition, we used the RaxST protein sequence to search the 335 following database using blastp with an e-value cut-off of 1e-3 and a BLOSUM62 scoring Due to the small size of RaxX, tblastn was required to identify homologs (evalue cutoff of 1e-3). 346 Fasta files for each blast hit were generated using a custom python script (available upon

Rice growth and inoculation
Oryza sativa ssp. japonica rice varieties were TP309 and XA21-TP309, which is a 106-17- Six weeks after planting, rice pots were transferred to a growth chamber with the following 378 day/night settings: 28°C/24°C, 80%/85% humidity, and 14/10-hour lighting. Plants were 379 inoculated 2 to 3 days after transfer using the scissors clipping method (Song et al., 1995). 380 Bacteria for inoculation were taken from PSA plates and resuspended in water at a density of 381 approximately 8 × 10 8 CFU/ml. Water-soaked lesions were measured 14 days after inoculation. 382

Complementation tests
The Xoo strain PXO99 A marker-free deletions raxX and raxST were described previously 383 The raxX-raxSTAB gene cluster is located between the flanking gcvRP and "mfsX" genes. Gene 617 cluster acquisition through lateral transfer is hypothesized to occur by general recombination in 618 the flanking gcvR and "mfsX" sequences as described in the text. Sequences at the left and right 619 boundaries are shown in Fig. S2. Sequences for length polymorphisms in the gcvP gene are 620 shown in Fig. S3. 621  The Xanthomonas spp. cladogram is based on published phylogenetic trees; see text for 634 references. Red lines depict lineages for strains that lack the raxX-raxSTAB gene cluster, 635 whereas blue lines depict those that carry the cluster. Numbers indicate gcvP length 636 polymorphism in each species (see Fig. S3). Hypothetical events are: A, formation of the raxX-637 raxSTAB gene cluster; B, lateral transfer to X. translucens, relatively early during speciation 638 (indicated by the long blue line); C, lateral transfer to X. maliensis, relatively late during 639 speciation (indicated by the short blue line); D, loss from X. citri. Strain numbers denote sources 640 of RaxX proteins chosen for functional tests, as described in the text. 641

Fig. 4. Phylogenetic tree for raxX-raxSTAB nucleotide sequences.
The best scoring maximum likelihood tree for the catenated raxA, raxB, raxX and raxST coding 642 sequences. Numbers shown on branches represent the proportion of branches supported by 643 10,000 bootstrap replicates (0-100). Bootstraps are not shown for branches with less than 50% 644 support, nor for branches too short to easily distinguish. Species names are colored according to 645 phylogenetic group. 646  This tree was constructed by analysis of whole genome sequences as described in Materials and 692 Methods. Blue indicates genomes that contain the raxX-raxSTAB gene cluster; red indicates 693 genomes that do not. Group numbers are arbitrary. 694 Sequences are from the reference strains described in Table 1. Sequences conserved within a 695 group but different from other groups are colored green ("early-branching" species), brown 696   The relevant portion of the GcvP amino acid sequence is shown for each of the reference strains. 712 Species in red lack the raxX-raxSTAB gene cluster, whereas those in blue carry the cluster. 713 Numbers denote different allelic types for reference to Fig. 3  Distribution of raxST homologs across bacterial genera, including the major groups of 720 proteobacteria as well as cyanobacteria. The tree shown was constructed by neighbor-joining 721 with 1000 bootstrap replicates; branches with < 50% bootstrap support are not drawn. The raxST 722 sequence from Xoo strain PXO99 A was used as query for tBLASTn. 723   File. S1. Xanthomonas strains analyzed for whole-genome phylogeny.