Chorismate mutase and isochorismatase, two potential effectors of the migratory nematode Hirschmanniella oryzae, increase host susceptibility by manipulating secondary metabolite content of rice

Abstract Hirschmanniella oryzae is one of the most devastating nematodes on rice, leading to substantial yield losses. Effector proteins aid the nematode during the infection process by subduing plant defence responses. In this research we characterized two potential H. oryzae effector proteins, chorismate mutase (HoCM) and isochorismatase (HoICM), and investigated their enzymatic activity and their role in plant immunity. Both HoCM and HoICM proved to be enzymatically active in complementation tests in mutant Escherichia coli strains. Infection success by the migratory nematode H. oryzae was significantly higher in transgenic rice lines constitutively expressing HoCM or HoICM. Expression of HoCM, but not HoICM, increased rice susceptibility against the sedentary nematode Meloidogyne graminicola also. Transcriptome and metabolome analyses indicated reductions in secondary metabolites in the transgenic rice plants expressing the potential nematode effectors. The results presented here demonstrate that both HoCM and HoICM suppress the host immune system and that this may be accomplished by lowering secondary metabolite levels in the plant.

their entire life cycle. They keep on migrating through the root whilst feeding. Although this lifestyle might enable them to outrun local plant defences and effector proteins inhibiting plant defence might seem unnecessary, several potential effector proteins have been predicted from the transcriptome of plant-parasitic migratory nematodes Nicol et al., 2012;Bauters et al., 2014). This implies that migratory nematode species also invest energy in attenuating the defence responses of their host.
The plant hormone salicylic acid (SA) plays an important role as a signalling molecule during the defence reaction on pathogen infection. SA can be produced through two distinct pathways starting from chorismate, the final product of the shikimate pathway: the isochorismate synthase (ICS) pathway and the phenylalanine ammonia lyase (PAL) pathway (leading to SA and phenylpropanoids). Several chorismate-derived compounds play a role in plant physiology and the defence system, for example, SA, auxin, lignin, and flavonoids. Changes in concentrations of these compounds can affect the plant defence against pathogen infection (Jones et al., 2007;Nahar et al., 2012). Interestingly, transcripts encoding chorismate mutase and isochorismatase were detected in the transcriptome of H. oryzae (Bauters et al., 2014). Both enzymes can interfere in the host production of salicylic acid and/ or phenylpropanoids. Chorismate mutase catalyses the conversion of chorismate to prephenate. This enzyme is present in plants and bacteria but had not been reported in animals until its discovery in Meloidogyne javanica (Lambert et al., 1999). Since then it has been reported in several other plant-parasitic nematodes (Jones et al., 2003;Huang et al., 2005;Vanholme et al., 2009). It has been shown that chorismate mutase, secreted by plant-pathogenic organisms, is able to alter plant cell development by impairing the development of lateral roots and vascular tissue (Doyle and Lambert, 2003). More recent research provided experimental evidence that a secreted chorismate mutase originating from Ustilago maydis reduced SA levels in host plants (Djamei et al., 2011). A decrease in SA content was also observed in tobacco leaves transiently expressing a chorismate mutase from Meloidogyne incognita (Mi-CM3) upon infection by the oomycete Phytophthora capsici. In addition, constitutive expression of Mi-CM3 increased susceptibility of tobacco plants to M. incognita infection (Wang et al., 2018).
The gene encoding isochorismatase has not been characterized in nematodes before, but it was reported to be present in the genome of Meloidogyne hapla and the transcriptome of Rotylenchulus reniformis (Opperman et al., 2008;Wubben et al., 2010). Moreover, isochorismatase was detected in the secretome of plant-pathogenic fungi, but not in nonpathogenic species (Soanes et al., 2008) and it was reported to reduce SA content in plants on infection.
Isochorismatase converts isochorismate to 2,3-dihydroxy-2,3-dihydrobenzoate and pyruvate. Isochorismate is an intermediate in the biosynthesis of SA, hence isochorismatase is capable of reducing the pool of isochorismate available for SA synthesis (Liu et al., 2014).
In this research, the genetic structure and protein properties of chorismate mutase (HoCM) and isochorismatase (HoICM) isolated from H. oryzae are described. Their activity is demonstrated by com- All exon/intron transitions contain the consensus splice site GT/ AG, except for one HoICM noncanonical splice site GT/CG. The first 60 nucleotides of HoCM code for a secretion signal, targeting the encoded protein to the classical secretory pathway. Expression in the secretory glands was shown by in situ hybridization in previous research (Bauters et al., 2014). No secretion signal was detected in HoICM. Several attempts were made to perform an in situ hybridization for HoICM transcripts, but all of them were negative or showed ambiguous results. HoICM was PCR-amplified from genomic DNA extracted from a single nematode, three times, from an independent single hand-picked nematode. HoICM could be amplified in all three experiments, supporting that this is an endogenous H. oryzae gene rather than being derived from any contaminating material. The fact that this type of isochorismatase (ICM) is specific to plant-parasitic species (Bauters et al., 2014) and that it is secreted by filamentous pathogens through a nonclassical secreted pathway (Liu et al., 2014) points towards an effector function in nematodes as well.

| Important catalytic residues are conserved in both HoCM and HoICM
The HoCM protein consists of three domains, including the signal peptide. The N-terminal domain of the mature protein has no similarities with other domains in public databases, but it has two motifs, rich in serine and histidine, respectively, hence the name serine/histidinerich domain (S/HRD). The C-terminal part is similar to the Pfam chorismate mutase (CM) type II domain (PF01817) (CMD). Alignment of HoCM and other CM sequences deduced from the genomes of cyst and root-knot nematodes showed conservation of catalytic residues as initially characterized in E. coli and Mycobacterium tuberculosis (Lee et al., 1995;Ökvist et al., 2006). The eight conserved catalytic residues are highlighted in Figure 1a. All residues are conserved in H. oryzae, but far less in cyst and root-knot nematodes.  (Parsons et al., 2003;Goral et al., 2012).

| Predicting the protein structure of HoCM and HoICM
CMs are classified into two main groups according to their structure: the AroH and AroQ classes. Proteins of the rare AroH class, represented by the CM of Bacillus subtilis, have both α-helices and β-sheets F I G U R E 1 Gene structure and partial protein sequence of HoCM and HoICM. (a) HoCM has a total length of 942 base pairs at the genomic level, containing two small introns. The protein consists of three different domains: an N-terminal signal peptide (SP), a serine/histidinerich domain (S/HRD), and a chorismate mutase domain (CMD) (PF01817). The alignment shows the conserved catalytic residues in several plant-parasitic nematode species (Mjav, Meloidogyne javanica; Mare, Meloidogyne arenaria; Minc, Meloidogyne incognita; GPLIN, Globodera pallida). Eight conserved putative catalytic residues characterized in chorismate mutase of Mycobacterium tuberculosis and/or Escherichia coli are marked with filled dots. (b) HoICM is 822 bp long and contains two small introns. It contains an isochorismatase domain (PF00857), but no predicted N-terminal signal peptide. The three putative catalytic residues are marked with filled dots. Sequence names correspond to the protein sequences encoded by different nematodes, downloaded from the Wormbase Parasite database (Ho, Hirschmanniella oryzae, ALI53582; Dd, Ditylenchus destructor, Dd_10059; Hg, Heterodera glycines, Hetgly.G000001356; Gr, Globodera rostochiensis, GROS_g01640. Contig1915.frz3.gene2). *Sequence adjusted (part of intron used as exon) due to probable mistake in gene prediction in their structure (Chook et al., 1993). On the other hand, proteins of the AroQ class are more abundant and are represented by the CM structure of E. coli, containing only α-helices (Lee et al., 1995).
Secondary and tertiary structure predictions showed that HoCM is composed of α-helices and loops, without β-sheets, indicating that HoCM is a member of the AroQ class. A 3D model was constructed using Phyre2 with the CM structures of M. tuberculosis (PDB code 2FP1) and Yersinia pestis (2GBB) as a template. The predicted tertiary structure of HoCM is shown in Figure 2a. The CM domain of HoCM has the same predicted topology as the global structure of an AroQ γ protein, which consists of six α-helices connected by loops.
The S/HRD region (N-terminal) of HoCM lacked homology and was modelled ab initio with only 72% of the residues modelled with over 90% confidence. for which 95% of the residues were modelled with more than 90% confidence with Phyre2. The model shows a molecule with a Rossman fold; a six-stranded parallel β-sheet flanked by three big α-helices at one side and two at the other. Next to these α-helices, there are four putative 3 10 -helices present. The presumed catalytic triad is shown in Figure 2b. The catalytic centre where isochorismate can bind has been described in Oleispira antarctica and is conserved in other plant-parasitic nematodes (PPNs) (Goral et al., 2012). PyMOL (v. 1.6) structural alignment assigns HoICM to the structural subgroup of the phenazine biosynthesis proteins (root mean square deviation of 0.437).

| HoCM and HoICM complement E. coli mutants
The activity of two different HoCM constructs was assessed by a complementation assay in a CM-deficient E. coli strain (KA12/pKIMP-UAUC) grown on dropout medium without phenylalanine. One construct contained the full CM coding sequence without the predicted signal peptide, the second only consisted of the catalytic region. The positive control was the E. coli mutant expressing a B. subtilis CM. The results confirmed CM activity of the HoCM protein ( Figure 3a). The fact that mutants with both protein constructs grew equally well reveals that the S/HRD region is not necessary for CM activity.
The activity of HoICM was tested using an ICM (entB) deficient  Complementation assay of entB-mutated E. coli AN192. AN192 cells were complemented with an empty vector (pDEST17) (white) or HoICM (grey). Cells were grown in liquid medium with different concentrations of bipyridyl to create iron-limiting conditions. Bacterial growth was observed by measuring optical density (y axis) at three different time points (an average was taken from four bacterial cultures). Treatments were compared with the control (no bipyridyl) and statistically analysed by a Mann-Whitney test. Each experiment was performed twice with similar results. Different letters above the graph indicate the significant differences between the different treatments Almost 95% of all reads could be mapped to the rice genome (IRGSP-1.0, see Table S2). were found in plants expressing HoICM. Approximately half of these genes had a significant (bit-score > 50) hit when compared with blast against the SwissProt database. These annotated genes were classified into six groups according to their putative function (File S1). Expression of 20 genes was down-regulated in the transgenic plants expressing any of the three constructs, among which were two glutathione S-transferases, a mono-oxygenase, and geranylgeranyl pyrophosphate synthase. A nonprotein-coding transcript was up-regulated in all three constructs. Nine genes had the same trend in transgenic plants with either CM construct, among which were a down-regulation of genes involved in steroid biosynthesis and auxin transport. It is interesting to note that many genes involved in signalling, stress response, and secondary metabolite production were differentially expressed upon in planta expression of both nematode genes. An overview of the significantly differentially regulated genes can be found in File S1.
The reliability of the RNA-Seq results was validated by evaluating the expression of 12 randomly chosen DEGs by RT-qPCR. Eleven of the genes confirmed RNA-Seq results; only one gene showed a contrasting expression pattern: the gene coding for histone H2B.3 was significantly up-regulated according to the RNA-Seq data (log 2 FC = 1.6) in plants expressing HoCM_CAT, while it was down-regulated according to the RT-qPCR data. The expression of two genes, coding for glutathione S-transferase and a receptor-like protein kinase, was so strongly down-regulated that it was undetectable by RT-qPCR in the transgenic lines (see Figure S1). The clear correspondence with RT-qPCR results for these 12 genes (p = .006, binomial test) supports the reliability of the generated RNA-Seq data.

| Gene ontology enrichment analysis shows a decrease in defence compound biosynthesis terms
Parametric analysis of gene set enrichment (PAGE) was conducted using the log 2 fold changes of all genes as input. All levels of gene ontology (GO) in the category "Biological Process" were considered, with a significance level of 0.05 (FDR). GO terms with a calculated Looking at the child terms of this GO term, the repression was due to a reduction in the GO term "Phenylpropanoid metabolic process".
The GO term "Lignin catabolic process" is reduced in HoCM_FULL-and HoICM-expressing plants. "Diterpene phytoalexin metabolic process" is suppressed in plants expressing HoCM. Rice plants expressing either one of the HoCM constructs are also repressed in the general "Defense response", while all three lines are reduced in "Oxidation reduction".

| Metabolome analysis suggests that HoCM down-regulates phenylpropanoid biosynthesis
No differences in SA content in shoots or roots were detected in transgenic lines compared to the control ( Figure S2) using the protocol described by Haeck et al. (2018). To obtain more insight into the effects of either ICM or CM on phenolic metabolism in rice roots, a comparative liquid chromatography-mass spectrometry (LC-MS, see  (Table S3) and 17 (Table S4), respectively. In the HoCM set, 1 and 16 compounds displayed higher and lower levels in the transgenic lines as compared to the control lines. However, all of them were of low abundance, hindering the recording of good-quality MS/MS spectra for structural elucidation (Table S5) (Table S5 and Figures S3 and S4).
Because it proved to be difficult to characterize the unknown compounds that were statistically different in abundance between treatment and control, as an alternative approach only those metabolites that could be matched with a known compound were taken into consideration. To increase confidence in the proposed matches, only compounds with a fragmentation score higher than 0.8 were kept.
This pipeline resulted in a list of only 17 compounds, 16 of which were linked to the phenylpropanoid pathway. Although not significantly different, for HoCM-expressing lines 13 compounds had lower levels in the two independent lines per construct, compared to the control, suggesting down-regulation of the phenylpropanoid pathway. A two-way ANOVA using "relative abundance of each component" as variable and "control vs transgenic" and "independent line" as fixed and random factors, respectively, was conducted. A general lower abundance of compounds from the phenylpropanoid pathway in HoCM-expressing lines (p = .008) was observed without any effect of the "independent line" factor (p = .5). Because there was an interaction effect between the fixed and the random factor for HoICMexpressing plants, no clear conclusion could be drawn (see Table 1).

| D ISCUSS I ON
Although the presence of a gene encoding CM has been reported in transcriptomes of migratory PPNs before (Bauters et al., 2014; F I G U R E 5 Gene ontology (GO) term enrichment analysis, showing the significantly enriched GOterms in the category "Biological process". GO terms for which at least one of the three constructs resulted in plant transcripts with significant difference with an absolute Z-score (y axis) value of 4 are shown in the graph. a and aʹ indicate an ancestor term (a) with its child terms (aʹ). All GO terms represented in this graph are significantly less abundant in the transgenic lines (Z-score < 0) compared to control plants (empty vector line). Nonsignificant data are not shown in the graph Haegeman et al., 2011), this is the first time it has been functionally characterized in a migratory nematode. Chorismate is the final product of the shikimate pathway, a pathway that has only been reported in plants and microorganisms, but not in animals (Herrmann, 1995). The first report of CM from a nematode was in M. javanica (Lambert et al., 1999), where it was thought to be involved in the establishment of a feeding site within the host. Indole-3-acetic acid (auxin, IAA) levels were reported to be lowered due to expression of this CM in plant roots, which had an influence on root development. This observation is in contradiction with the findings that IAA is important for feeding site initiation, while the nematode-secreted CM was hypothesized to be a vital element in this process (Goverse et al., 2000).  (Bauters et al., 2014), led us to believe it might be involved in plant parasitism. This hypothesis is strengthened by the fact that ICM is also present in the secretome and in the culture supernatants of phytopathogenic fungi, helping to increase susceptibility of the host (Soanes et al., 2008;El-Bebany et al., 2010;Ismail et al., 2014). ICM is probably secreted by a nonclassical secretory pathway in fungi (Liu et al., 2014), but some fungal ICMs are predicted to have a secretion signal (Ismail et al., 2014). Further research is necessary to provide experimental evidence of ICM secretion by nematodes.
Although it is generally assumed that SA mediates resistance against biotrophic pathogens (Glazebrook, 2005) and H. oryzae kills the cells it is feeding from, in contrast to sedentary nematodes (De Vleesschauwer et al., 2013), SA was shown to be important for an adequate defence response against H. oryzae infection in rice . On the other hand, the SA biosynthesis pathway appears changed after H. oryzae infection with up-regulation of OsPAL1 and down-regulation of OsICS in shoots . SA biosynthesis in plants can occur through two different pathways, both starting from chorismate as a precursor (Vlot et al., 2009;reviewed in Lefevere et al., 2020). A secreted CM from U.
maydis was able to reduce SA accumulation upon pathogen infection (Djamei et al., 2011). ICM, secreted by two fungal pathogens (Verticillium dahliae and Phytophthora sojae), reduces the SA con-  (Gao et al., 2008;Uehara et al., 2010;Feng et al., 2011;Khanam et al., 2018). In contrast to our observations, Djamei and colleagues (2011) reported an increase in phenylpropanoid biosynthesis products in plants infected with U. maydis, which indicates that chorismate is directed into the phenylpropanoid pathway. CM originating from U. maydis was also able to reduce the SA content of maize upon infection (Djamei et al., 2011). A recent study showed that transient expression of CM originating from the PPN M. incognita was able to reduce the SA content of tobacco leaves after Phytophthora infection (Wang et al., 2018). Our results showed that unchallenged rice plants expressing HoCM or HoICM did not change their SA content in roots or shoots. This observation was strengthened by the fact that no SA-responsive genes (e.g., OsWRKY45 or OsNPR1) were differentially regulated according to the transcriptome analysis.
Diterpenoid phytoalexins (DPs) have several functions in plants. Their antimicrobial properties have been well documented. DPs accumulate in infection spots where necrotic tissue starts to develop (Huffaker et al., 2011). Transgenic lines deficient in OsCPS4, a phytoalexin biosynthetic gene, showed increased susceptibility to rice blast fungus (Toyomasu et al., 2014). Also in nematodes it was shown that induction of DP biosynthesis increases resistance against nematodes (Verbeek et al., 2019). Some terpenoids might be involved in nematode resistance by accumulation in lesions near the infection site (Veech, 1982), so a reduced terpenoid content could influence susceptibility to nematodes. The mechanism by which expression of HoCM or HoICM influences the diterpenoid phytoalexin biosynthesis pathway remains unclear. So far, it was only suggested that the phenylpropanoid and terpenoid pathway compete for carbon and that carbon flow into these two pathways is tightly regulated (Xie et al., 2008).
The data generated in this paper suggest that the immune system of the host is altered by lowering secondary metabolite content upon secretion of CM and ICM by H. oryzae. Further research is needed to pinpoint the exact metabolites responsible for the drop in immunity and the mechanism by which HoCM and HoICM are able to do this.

| Nematode DNA extraction and gene amplification
Genomic DNA was extracted from a batch of mixed stages of H. oryzae or a single hand-picked nematode as previously described by Bolla et al. (1988), with some minor modifications. Briefly, nematodes were sonicated three times for 10 s to break cell membranes.
Samples were incubated for 1 hr at 65°C after which DNA was isolated using phenol and chloroform. Isolated DNA and a cDNA library (Bauters et al., 2014) were used to amplify HoCM and HoICM by PCR with gene-specific primers (see Table S6). PCR fragments were cloned into pGEM-T (Promega) according to the standard thymine/ adenine protocol and sequenced by the Sanger method (LGC) to verify the insert.

| Activity assays
Two HoCM fragments (HoCMFull and HoCMCat) were amplified by PCR using the full cDNA sequence as template (primer sequences can be found in Table S6). The first fragment comprised the full sequence of CM without sequence encoding the signal peptide, the second fragment only contained the sequence encoding the catalytic domain (base pair 258 to 747). The complementation assay was performed as described by Vanholme et al. (2009).

cells. AN192 cells are unable to grow in iron-limiting conditions.
HoICM was amplified by a two-step PCR starting from cDNA as template to add attB sites to the fragment (primer sequences can be found in Table S6). The fragment was cloned into the destina-

| Plant transformation
cDNA sequences of HoICM and HoCM were amplified by PCR on a cDNA library (Bauters et al., 2014). Primer sequences are provided in  (Benjamini and Hochberg, 1995) false discovery rates (FDR) were estimated to detect DEGs (FDR < 0.05).
GO enrichment analysis was performed using agriGO (Du et al., 2010). Gene identifiers with their corresponding Log 2 FC were used as input in a Parametric Analysis of Gene Set Enrichment (PAGE) (Kim and Volsky, 2005). The Hochberg multitest adjustment method was performed (p < .05).

| Phenolic profiling
Roots (approximately 5 mg of dry weight) were homogenized in liquid nitrogen and extracted with 1 ml of methanol. The methanol extract was then evaporated and the pellet dissolved in 200 μl of water/cyclohexane (1:1, vol/vol). Then, 10 μl of the aqueous phase was analysed via reverse-phase ultrahigh performance liquid chromatography using an Acquity UPLC BEH C18 column. In addition to full MS analysis, a pooled sample was subjected to datadependent MS/MS analysis (exclusion duration = 10 s). Integration and alignment of the m/z features were performed via Progenesis QI software v. 2.1 (Waters Corporation). The normalization was set on "external standards" and was based on the dry weight of the samples.
For statistical analysis, the data were split into two sets. A first set comprised the two independent HoICM expression lines (HoICM1 and HoICM4) and the corresponding control line (HoICM set). In the second set, one HoCM_FULL expression line, two independent HoCM_CAT expression lines (HoCM_CAT2 and HoCM_CAT3), and the corresponding control line were included (HoCM set). Both sets were subjected to a one-way ANOVA using the lm() function in R v. in-house mass spectral database (200 ppm fragment tolerance). As none of the MS/MS spectra of the differential m/z features could be identified via database matching, MS/MS spectral elucidation was attempted (see File S2) using gas phase fragmentation rules and in silico MS/MS elucidation software, that is, CSI:FingerID (Böcker and Dührkop, 2016;Dührkop et al., 2015) and CFM-ID (Allen et al., 2015). A more detailed description of the phenolic profiling protocol can be found in File S2.

| Nematode infection assay
The M. graminicola culture, originally isolated in the Philippines, was maintained on O. sativa "Nipponbare" in potting soil at 27°C (16/8 hr light regime). H. oryzae was obtained from infected rice roots sampled from fields in Myanmar. Nematodes were extracted from rice roots using a modified Baerman funnel technique. Rice seeds were first germinated on wet tissue paper at 27°C in the dark for 3 days.
Germinated seedlings were transferred to PVC tubes containing sand and absorbent polymer substrate (Reversat et al., 1999

DATA AVA I L A B I L I T Y S TAT E M E N T
Sequence data are available in GenBank at www.ncbi.nlm.nih.gov with accession numbers HoCM: KP297892, and HoICM: KP297893.
The data that support the findings of this study are available from the corresponding author upon reasonable request.