Specific Transcriptional Regulation Controls Plant Organ‐Specific Infection by the Oomycete Pathogen Phytophthora sojae Authors: Long Lin, Yang Wang, Hui Qian, Jiawei Wu, Yachun Lin, Yeqiang Xia, Suomeng Dong, Wenwu Ye, Yuanchao Wang Published: 2024 ABSTRACT ABSTRACTThe organs of a plant species vary in cell structure, metabolism and defence responses. However, the mechanisms that enable a single pathogen to colonise different plant organs remain unclear. Here we compared the transcriptome of the oomycete pathogen Phytophthora sojae during infection of roots versus leaves of soybeans. We found differences in the transcript levels of hundreds of pathogenicity‐related genes, particularly genes encoding carbohydrate‐active enzymes, secreted (effector) proteins, oxidoreductase‐related proteins and transporters. To identify the key regulator for root‐specific infection, we knocked out root‐specific transcription factors (TFs) and found the mutants of PsBZPc29, which encodes a member of an oomycete‐specific class of basic leucine zipper (bZIP) TFs, displayed reduced virulence on soybean roots but not on leaves. More than 60% of the root‐specific genes showed reduced expression in the mutants during root infection. The results suggest that transcriptional regulation underlies the organ‐specific infection by P. sojae, and that a bZIP TF plays a key role in root‐specific transcriptional regulation. Introduction Disease resistance and microbial pathogenicity mechanisms are continually evolving in plants and their pathogens, respectively (Rausher 2001). To protect themselves from pathogen infection, plants have developed numerous physical and chemical barriers (Jones and Dangl 2006). For example, the plant cell wall is the main line of defence against invading pathogens. Furthermore, the aerial parts of plants are often covered with a hydrophobic waxy cuticle that prevents pathogen penetration (Yeats and Rose 2013), and stomatal closure blocks a potential route of pathogen invasion (Melotto et al. 2006). Plants also activate multiple defence responses after detecting an invading pathogen, such as oxidative bursts and secretion of pathogen‐degrading enzymes and toxic chemicals (Bindschedler et al. 2006; van den Burg et al. 2006; Schmelz et al. 2014). The organs of a plant are markedly different in structure, metabolism and defence responses. For example, waxy cuticles and stomata are present on leaves and stems, but not roots. In Arabidopsis thaliana , the transcriptomes of roots and aerial organs are distinct (He et al. 2016). Infection by pathogens such as the fungus Fusarium oxysporum can trigger further specific transcriptional reprogramming in different organs of A. thaliana (Lyons et al. 2015). For example, in A. thaliana , regulators of auxin and abscisic acid (ABA) signalling, mannose‐binding lectins, and peroxidases showed strong up‐regulation in roots, while defensins and genes associated with the response to auxin, cold and senescence were strongly induced in leaves (Lyons et al. 2015). During infection by Colletotrichum graminicola, maize roots exhibited a more rapid response and accumulated higher levels of defence‐related hormones than leaves, as well as higher basal levels of antimicrobial flavonoids, suggestive of an enhanced basal defensive state (Balmer et al. 2013). Therefore, in order to successfully infect different plant organs, pathogens must evolve pathogenesis strategies to overcome a diverse set of physical and chemical challenges. Phytophthora sojae and other oomycetes, including many plant and animal pathogens, form a distinct phylogenetic lineage of fungus‐like eukaryotic microorganisms that share an ancestor with nonpathogenic organisms such as brown algae and diatoms. Oomycetes have emerged relatively recently (250 million years ago) as pathogens (Matari and Blair 2014), yet in that time have evolved the ability to attack thousands of plant species. In particular, the genus Phytophthora has been estimated to have radiated into different species around 20 million years ago (Matari and Blair 2014). The genus encompasses over 160 species infecting roots, stems, leaves, fruits and flowers of thousands of plant species. This rapid and successful radiation implies that these pathogens have efficient mechanisms for generating genetic diversity as fodder for selection, while at the same time evolving the transcriptional regulatory circuits needed to effectively and flexibly deploy the changing repertoire of potential virulence genes. In addition to physiological and metabolic processes to protect pathogens from plant defence, pathogens, including Phytophthora species, secrete a large number of virulence factors (e.g., effectors) to weaken plant defence systems and promote host susceptibility (Jiang and Tyler 2012; Wang and Wang 2018; Wang, Tyler, and Wang 2019). The self‐protective processes and virulence factors both probably contribute to organ specificity. For example, a gene expression and silencing study on Phytophthora infestans nutrient transporters indicated that several genes in the nitrate reductase pathway were required for infection of tomato leaves rather than tubers; this may facilitate amino acid biosynthesis and protect the pathogen from natural or fertilisation‐induced nitrate and nitrite toxicity (Abrahamian et al. 2016). In the case of the fungus Ustilago maydis , the pathogen possesses effectors contributing to virulence in an organ‐specific manner (Skibbe et al. 2010; Schilling et al. 2014). However, no comprehensive study of organ‐specific virulence factors has been performed for oomycete pathogens. Appropriate regulation of organ‐specific virulence genes is presumably mediated through transcription factors (TFs) and/or epigenetic modifying factors (Soyer et al. 2014; van der Does et al. 2016), but knowledge of how phytopathogenic fungi and oomycetes regulate the expression of virulence genes has been lagging. In U. maydis , several leaf‐specific and tassel‐specific effectors have been identified as direct or indirect targets of the WOPR family TF Ros1, which is a master regulator of sporogenesis and late effector gene expression (Tollot et al. 2016); however, whether Ros1 is directly required for organ‐specific infection is unclear. Several TF families and novel promoter elements of Phytophthora have been implicated in stage‐specific expression of genes (Qiu et al. 2023; Situ et al. 2022; Seidl et al. 2012; Sheng et al. 2015). However, direct evidence for organ‐specific regulatory genes in oomycetes has evaded discovery so far. Phytophthora sojae, which is a model oomycete plant pathogen, causes root and stem rot in soybean, and is also capable of infecting young soybean leaves. In this study, we compared the P. sojae transcriptome during infection of soybean roots and leaves to identify genes associated with virulence. Among these genes, we identified PsBZPc29, which encodes a member of an oomycete‐specific class of basic leucine zipper (bZIP) TF. PsBZPc29 was required for root infection and full induction of more than 60% of the genes specifically induced during root infection. Results Dual Transcriptome Profiling of the Early Stage of the P. sojae–Soybean Interaction By monitoring the progress of P. sojae infection according to the biomass measured by quantitative real‐time PCR (qPCR) and microscopic observation using a green fluorescent protein (GFP)‐expressing strain, we found that P. sojae strain P6497 could infect both the roots and leaves of soybean cultivar Williams, and the infection dynamics during the early stage of infection were similar in roots and leaves (Figure S1a,b). P. sojae‐infected soybean roots and leaves were collected at 1.5, 3, 6, 9 and 12 h post‐infection (hpi). The samples were mixed to represent the early stages of root infection (R‐IN) and leaf infection (L‐IN). Non‐P. sojae‐infected soybean samples were also collected as soybean controls (R‐CK and L‐CK). Two independent biological replicates per treatment were analysed by RNA‐seq (Figure S2a,b). Among the infection libraries, the majority of the transcript reads could be mapped onto the soybean genome (75%–82%), while 3%–5% were mapped onto the P. sojae genome. The leaf‐infection libraries contained more soybean chloroplast reads and fewer mitochondrion reads than the root‐infection libraries (> 60% vs. 2% and 1%–2% vs. > 6% respectively). The similarity of the read‐mapping statistical data (Table S1) and strong correlation of gene expression levels (Figure S3a,b) between each pair of replicates indicated that the data are reliable. Infected Soybean Roots and Leaves Exhibited Different Transcriptional Responses The correlation between the soybean root and leaf transcriptomes was weaker than between the infected and noninfected organs (Figure S3a,b). Compared to the noninfection libraries, 2715 soybean genes showed a consistent expression pattern between root and leaf during P. sojae infection, while 3751 and 5259 soybean genes showed organ‐specific responses to infection in infected roots and leaves, respectively (Figure S3c). Genes that were significantly up‐regulated during P. sojae infection were significantly over‐represented within 23 functional groups of paralogues compared to their representation among all soybean genes. The majority of these paralogue groups had members that were induced only in roots or in leaves. Several groups were annotated with functions related to plant disease resistance, for example, nucleotide‐binding site‐leucine‐rich repeat (NBS‐LRR) family proteins, receptor‐like protein kinases (RLKs), WRKY family TFs, and plant U‐box or F‐box proteins (Figure S3d). These results revealed that P. sojae‐infected soybean roots and leaves show different defence responses at the transcript level, which raises the question of how P. sojae overcomes these different defence responses. Hundreds of P. sojae Pathogenicity‐Related Genes Were Specifically Induced During Infection of Soybean Roots or Leaves We identified 320 P. sojae genes with significantly higher transcript levels in libraries R‐IN1 and R‐IN2 than in L‐IN1 and L‐IN2, and 322 genes showed the reverse pattern (Table S2); these two sets of genes were termed root‐specific genes (RSGs) and leaf‐specific genes (LSGs), respectively, and collectively termed specific‐genes (SGs). In contrast, there were few genes showing different expression levels (log2fold change [FC] ≥ 1) within each pair of replicates (Figure S4a). We selected 31 genes with different transcript levels and R‐IN:L‐IN ratios for verification by reverse transcription‐qPCR (RT‐qPCR). Comparison of transcript levels (R = |0.83|) and R‐IN:L‐IN ratios (R = |0.62|) both showed a strong correlation between the RNA‐seq and RT‐qPCR results (Figure S2c,d). To confirm the organ‐specific expression of the RSGs and LSGs, we also compared the transcript patterns using pre‐existing (Ye et al. 2011) and newly generated RNA‐seq datasets for a series of P. sojae samples including mycelia, zoospores, germinated cysts, and infected leaves or roots at 3, 6 and 12 hpi (Figure S2b). On the whole, the organ‐specific expression patterns of RSGs and LSGs were confirmed in these data (Figure 1a; Figure S4b). Compared to the noninfection stages, 72% (230) of the RSGs and 92% (296) of the LSGs showed significantly increased expression during infection of root and leaf, respectively (Figure 1b), suggesting that most of these genes function in the pathogenicity of P. sojae. (In this study we routinely used zoospores as the base‐line to check whether these genes were induced or suppressed, as zoospores are the most common natural starting point for infection). Analysis of root infection‐specific genes (RSGs) and leaf infection‐specific genes (LSGs) of Phytophthora sojae. (a) Average transcript levels, which are calculated as log2(FPKM), of the RSGs and LSGs during root and leaf infection. ‘MY’, ‘ZO’, ‘GC’, ‘3 h’, ‘6 h’ and ‘12 h’ indicate mycelia, zoospores, germinated cysts, 3, 6 and 12 h post‐infection (hpi), respectively. (b) Venn diagram shows the number of RSGs, LSGs and non‐SGs (nonspecific genes) that are transcriptionally elevated during infection compared to noninfection stages. (c) Paralogue groups with significantly over‐represented RSGs and/or LSGs (hypergeometric test: Bonferroni‐adjusted p < 0.001). Star, diamond, circle and triangle indicate secreted proteins, carbohydrate‐active enzymes, oxidoreductase‐related proteins and transporters, respectively. Group IDs in OrthoMCL database (orthomcl.org; the group PRG_000049 is specific in P. sojae) are shown in parentheses. (d) Proportions of RSGs and LSGs in six indicated categories related to pathogenesis. A gene ontology (GO) enrichment analysis indicated that the SGs shared enriched GO terms that include hydrolase activity, oxidoreductase activity, and carbohydrate metabolism (Figure S5). The 15 paralogue groups that contained significantly higher proportions of RSGs or LSGs (than all genes) are associated with four functional categories: secreted proteins (eight groups), carbohydrate‐active enzymes (CAZymes; four groups, three are also secreted proteins), oxidoreductase‐related proteins (three groups) and transporters (two groups) (Figure 1c). The proportions of RSGs and LSGs in the four categories were both significantly higher than in all P. sojae genes (Figure 1d). Phytophthora sojae exhibited differential regulation of different categories of genes for infection of different organs. For example, a greater number of genes encoding secreted proteins and transporters were over‐represented in RSGs than in LSGs (Figure 1c). The proportions of RSGs were also higher than LSGs in all secreted protein and transporter‐encoding genes (Figure 1d). In contrast, the categories of CAZymes and oxidoreductase‐related proteins showed the reverse pattern (Figure 1c,d). A Root‐Specific TF Is Involved in Specific Infection of Soybean Roots The proportions of RSGs (0.4%) and LSGs (0.3%) in the TF category were low (Figure 1d). Three RSGs were predicted to encode two bZIP TFs (Ps_132371 [PsBZPc29] and Ps_129699 [PsBZPc17]) and one Myb TF (Ps_138282); these three genes showed 16‐, 11‐ and 20‐fold, respectively, higher transcript levels during root infection than during leaf infection. Two LSGs corresponded to heat shock factor TFs (Ps_130849 and Ps_132607); these genes showed three‐ and four‐fold higher transcript levels during leaf infection than root infection, respectively (Figure 2a). The transcript patterns of these genes were confirmed by RT‐qPCR (Figure 2b). Analysis of Phytophthora sojae PsBZPc29 knockout mutants. Transcript levels of transcription factor (TF) genes during root infection (R‐IN, red) and leaf infection (L‐IN, blue) in wild‐type (WT) P. sojae, and during root infection (T‐R‐IN, light red) and leaf infection (T‐L‐IN, light blue) in PsBZPc29 knockout mutants, measured by RNA‐seq (a) and reverse transcription‐quantitative PCR (b). Data are represented as mean ± SD of two and at least three independent experiments, respectively. *, significant difference (t test, p < 0.05) compared to the R‐IN (red) or L‐IN (blue). Detached soybean leaves were inoculated with zoospores of the T2 and T10 mutants, the control strain (CK; transformant in which PsBZPc29 was not knocked out), and wild‐type (WT; P6497) P. sojae. The photographs were taken 2 days after inoculation, with UV light exposure (c). Roots of etiolated soybean seedlings were inoculated with P. sojae zoospores and incubated for 2 days. The photographs were taken 2 days after inoculation (e). P. sojae biomass in infected soybean roots (d) and leaves (f) as determined by genomic DNA qPCR. Data are represented as mean ± SD of at least three independent experiments. **Significant difference compared to the WT (t test, p < 0.01). (g) Comparison of gene transcript levels between the T2 and T10 mutants. 0.25% and 0.35% of P. sojae genes had significantly higher levels in T2 and T10, respectively. (h) Comparison of gene transcript levels between WT and mutants. The numbers of PsBZPc29‐regulated specific genes are indicated. (i) Changes in transcript levels of root infection‐specific genes (RSGs) (red) and leaf infection‐specific genes (LSGs) (blue) in the mutants. 62% and 20%, proportion of PsBZPc29‐regulated RSGs and LSGs, respectively. To investigate the regulation of root‐specific infection, we used the CRISPR/Cas9 system to individually knock out the three root‐specific TFs of P. sojae strain P6497 (Fang and Tyler 2016). We succeeded in knocking out PsBZPc17; the PsBZPc17 mutants showed similar virulence to the wild type (WT) (Figure S6). We also generated two gene knockout mutants (T2 and T10) replacing PsBZPc29 with NPTII, a geneticin‐resistance gene (Figure S7). Compared with the WT and control strain (CK; a transformant that failed to acquire PsBZPc29 knockout), the mutants exhibited a similar phenotype, namely significantly reduced virulence on soybean roots but no significant difference in mycelial growth rate or infection of leaves. The disease lesion sizes on WT‐ and CK‐infected leaves were similar to those on mutant‐infected leaves, while the disease symptoms were obvious on WT‐ and CK‐infected roots but weaker on mutant‐infected roots (Figure 2c,e). There was a 50%–70% reduction of P. sojae biomass in mutant‐infected roots compared to WT‐ and CK‐infected roots, while there was no significant difference during leaf infection (Figure 2d,f). PsBZPc29 Is Required for Correct Regulation of Many Organ‐Specific Genes, Including Other TF Genes To identify P. sojae genes regulated directly or indirectly by PsBZPc29 (PsBZPc29‐regulated genes), we did RNA‐seq analysis for the knockout mutants during root infection (Figure S2a). Between T2 and T10, only 0.6% (66/10986) of all detected genes were significantly different in transcript level, indicating consistent transcriptomes in the two PsBZPc29 deletion strains (Figure 2g). Compared to WT, more down‐regulated genes were identified than up‐regulated genes (1007 > 800) in PsBZPc29 mutants. The down‐regulated genes showed a significantly higher value of mean log2FC than up‐regulated genes (2.1 > 1.7, t test, p < 0.01) (Table S3). Twenty‐four per cent of genes that exhibited reduced transcript levels were the SGs, and the number of RSGs was almost four‐fold than that of LSGs, while very few SGs showed increased levels (Figure 2h). Among all 320 RSGs, 60% were significantly decreased in the mutants, and these RSGs were more highly elevated during WT P. sojae infection of roots. In contrast, the LSGs showed milder decreases in transcript levels in the mutants, and few of these genes were highly elevated during leaf infection (Figure 2i). These results indicate that PsBZPc29 is required for full induction of many RSGs and may also contribute to regulation of some LSGs. The deletion of PsBZPc29 also altered the regulation of the two other root‐specific TFs (PsBZPc17 and Ps_138282) and the two leaf‐specific TFs (Ps_130849 and Ps_132607). Based on the RNA‐seq and RT‐qPCR results (Figure 2a,b), we found that levels of PsBZPc17 and Ps_138282 were reduced during root infection but were elevated during leaf infection in the mutant compared to WT. Thus, PsBZPc29 appeared to have a major role in directly or indirectly suppressing leaf‐specific expression of these two root‐specific TF genes. In contrast, the two leaf‐specific TF genes, Ps_130849 and Ps_132607, showed significantly reduced expression during root infection and especially during leaf infection. Thus, all the four TF genes showed reduced organ specificity of transcript levels in the PsBZPc29 deletion mutants. The constitutively expressed bZIP TFs, PsBZPc31 (Ps_132422) and PsBZPc41 (Ps_135827) also were reduced over 10‐fold in the mutants during root infection (Table S3). PsBZPc29 contains an R‐R class bZIP domain characteristic of an oomycete‐specific class of bZIP TFs (Ye et al. 2013). At least one orthologue of PsBZPc29 was identified in three other species from Phytophthora clade 7 (Figure S8a). The majority of PsBZPc29 homologues in P. sojae and other Phytophthora species also contained the oomycete‐specific class of bZIP motifs (Figure S8b). These proteins formed a distinct cluster in a phylogenetic tree distant from those of non‐Phytophthora oomycete species and distant from the conventional N‐R class of bZIP TFs in P. sojae (Figure S8a). To predict binding site(s) of PsBZPc29, we used MEME software to discover conserved motifs in the promoter regions of the PsBZPc29‐regulated root‐specific genes (Figure S8c). Motif‐1 was consistent with Phytophthora core promoter element INR‐FPR (Roy, Poidevin, et al. 2013), and Motif‐2 was similar with sporangia‐enriched motif 86 (GTTGAAG) in P. infestans (Roy, Kagda, and Judelson, 2013). However, classic bZIP‐binding sites (core motif is ACGT) were not identified (Dröge‐Laser et al. 2018), indicating that PsBZPc29 may regulate these genes indirectly or PsBZPc29 binding sites differ from conserved bZIP TFs. Some Effector‐Coding Genes Exhibit Organ‐Specific Transcription and Were Altered by PsBZPc29 Deletion We found that genes of various putative effector families exhibited marked transcriptional variation between root and leaf infections. Some genes in the RxLR, YxSL[RK], protease, and small cysteine‐rich protein (SCP) family showed higher transcript levels during root infection, while some elicitin family genes showed higher transcript levels during leaf infection. For example, there were 10 RSGs and one LSG among the 155 P. sojae RxLR genes with detectable transcripts, while the elicitin family included one RSG and six LSGs (Figure 3b). In contrast, the NLP and CRN family genes showed relatively little variation, comparable to that of a set of 294 Core Eukaryotic Genes Mapping Approach‐identified genes (CEGMA genes) (Figure 3a,b; Figure S9a; Table S4). Transcriptional regulation of Phytophthora sojae effector‐coding genes. (a) Fold changes in transcript levels between infections of root and leaf. ‘RxLR’, ‘Elictin’, ‘YxSL[RK]’, ‘NLP’, ‘CRN’ and ‘SCP’ are different effector families. CEGMA genes are Core Eukaryotic Genes Mapping Approach‐identified genes. (b) Scatter diagram of the differences in transcript levels of the indicated effector families. x‐axis, log2 fold‐change (FC) of transcript levels between infection and noninfection zoospore stages (R, L‐IN vs. ZO1). y‐axis, log2FC between infections of root and leaf (R‐IN vs. L‐IN). Numbers of PsBZPc29‐regulated root infection‐specific genes (RSGs) and leaf infection‐specific genes (LSGs) are indicated in parentheses. (c, d) Comparison of degrees of host induction between the wild type (WT) (R‐IN vs. ZO1) and PsBZPc29 knockout mutants (T‐R‐IN vs. T‐ZO). Numbers in parentheses represent numbers of PsBZPc29‐regulated RSGs, LSGs, and non‐SGs. R‐IN, L‐IN and ZO1 indicate root‐infection, leaf‐infection and zoospore stages of WT. T‐R‐IN and T‐ZO indicate root‐infection and zoospore stages of PsBZPc29 mutants. RSGs, LSGs and non‐SGs stand for root‐specific genes, leaf‐specific genes and nonspecific genes, respectively. In the PsBZPc29 knockout mutants, compared to WT, expression of many effector‐coding RSGs and LSGs, and some non‐SGs were altered. For example, there were 21 genes (6 RSGs and 15 non‐SGs) in the RxLR family that exhibited significantly different transcript levels (Figure 3c). This was also true for 11 genes in the YxSL[RK] family, three elicitin genes, eight NLP genes and six SCP genes (Figure 3d; Figure S9b; Table S4). All analysed effector gene families exhibited overall weaker elevation during the early stages of root infection (compared to the zoospore stage). In fact, CRN genes showed overall reduced expression compared to zoospores in the mutants (Figure 3d). Although 43 non‐SGs and a LSG in the CEGMA group exhibited different expression in the mutants, transcript levels of genes in this family varied little between non‐infection and infection stages in both WT and mutants (Figure 3c,d). Organ‐Specific Expression of CAZyme‐Coding Genes CAZymes, including glycoside hydrolases (GHs), glycosyltransferases (GTs), polysaccharide lyases (PLs), carbohydrate esterases (CEs) and carbohydrate‐binding modules (CBMs), are involved in the synthesis, metabolism and transport of carbohydrates; many CAZymes are cell wall‐degrading enzymes. Of the various CAZyme families, 17 and 16 showed higher transcript levels during root and leaf infection, respectively (Figure 4a), and 42 RSGs and 53 LSGs were predicted to encode CAZymes (Figure 4b; Table S5). RSGs were significantly more abundant in the GH12, GH28, PL1 and PL3 families; proteins in these families mainly function in the degradation of pectin, cellulose and hemicellulose (Blackman, Cullerne, and Hardham 2014). LSGs were significantly more abundant in the GH3, GH5, GH78, CE5, CE10 and CE13 families; the proteins in these families degrade not only the above‐mentioned substrates, but also cutin, β‐1,3‐glucans, and arabinogalactan (Blackman, Cullerne, and Hardham 2014; Zhang et al. 2016). LSG‐encoding CAZymes probably degrade more substrates, and a greater number of LSG‐encoding CAZymes could degrade multiple substrates (Figure 4c). Transcriptional regulation of Phytophthora sojae CAZyme‐coding genes. Names of families refer to all three panels. (a) Transcript levels of the indicated gene families during infection of root and leaf. RSGs are root‐specific genes, and LSGs are leaf‐specific genes. (b) Numbers of genes in each CAZyme family in different categories. (c) Putative substrates of the CAZymes. Y: the organ‐specific genes are significantly over‐represented in the family compared to all genes (hypergeometric test: Bonferroni‐adjusted p < 0.001). Black spots indicate the substrates of CAZymes. Pectate lyases are responsible for eliminative cleavage of pectate (Marin‐Rodriguez, Orchard, and Seymour 2002), which is a significant component of the cell wall of higher plants (Ridley, O'Neill, and Mohnen 2001; Vincken et al. 2003). Among the 40 pectate lyase genes in P. sojae, 80% were induced during infection, 35% were RSGs and 68% had predicted signal peptides (Figure 5a), suggesting they may be secreted extracellularly to target the plant cell wall, and pectin‐associated structures of soybean, especially in the roots, function as barriers to P. sojae infection (Quiroz‐Figueroa et al. 2023; Ropitaux et al. 2020). Transcriptional regulation of pectate lyase and cutinase genes. (a) Transcript ratios of pectate lyase genes. S indicates predicted secreted proteins. (b) Transcript ratios and levels and phylogenetic tree of cutinase genes in Phytophthora sojae. (c) Phylogenetic analysis of cutinase genes in the indicated Phytophthora species. In (a) and (b), asterisks on the right side of heat map represent genes that were significant difference in transcript levels between indicated samples (adjusted p value < 0.05 and absolute log2 fold‐change ≥ 1). The trees were constructed using the full‐length protein sequences of cutinases in P. sojae (Ps), P. infestans (PITG), P. capsici (Pc), P. parasitica (PPTG), and P. ramorum (Pr). Bootstrap values greater than 60 are displayed. R‐IN, L‐IN and ZO1 indicate root‐infection, leaf‐infection and zoospore stages of wild type (WT). T‐R‐IN indicates root infection‐stage of PsBZPc29 mutants. Cutin is an essential component of the plant cuticle, which is a protective film covering the epidermis of leaves, young shoots, and other aerial plant organs lacking a periderm. Cutinases are a class of cutin hydrolases (Fich, Segerson, and Rose 2016). Compared to other Phytophthora species (most of which have three to six cutinase genes), P. sojae has 16 genes encoding cutinases. Among the 15 transcribed genes, six showed similar expression levels during root and leaf infection, and nine were LSGs (Figure 5b). Phylogenetic analysis revealed that the LSGs were separated from the other genes, and many were species‐specific (Figure 5c). Therefore, in soybean leaves cutin probably functions as a barrier to P. sojae infection, and the cutinase gene family may have formed through gene duplication and transcriptional regulation under host selection pressure. In the PsBZPc29 knockout mutants, 52% of CAZyme‐coding RSGs and 17% of CAZyme‐coding LSGs showed reduced expression during root infection (T‐R‐IN vs. R‐IN). These RSGs belonged to 12 CAZyme families, and the LSGs belonged to eight families (Figure 4b; Table S5). For example, 10 of the 13 pectate lyase‐coding RSGs showed reduced expression (Figure 5a). None of the nine cutinase LSGs was affected in the mutants; however, all the six genes, which exhibited high transcript levels during both root and leaf infection, exhibited significantly decrease of transcript levels (Figure 5b). Discussion Analysis of genes that are differentially regulated during infection of a variety of organs or hosts is an effective method of identifying genes potentially involved in the adaptation of the pathogen to those organs or hosts. In this study, the infection transcriptome of P. sojae showed considerable specialisation for the host organs (roots or leaves) under attack. The differentially induced genes included many genes previously implicated in infection such as genes encoding putative effectors and other secreted proteins. The overall transcript levels of genes encoding RxLR effectors, putative YxSL[RK] effectors, and SCPs were higher during root infection. The previously identified avirulence effector gene Avr3a (Qutob, Chapman, and Gijzen 2013; Shrestha et al. 2016) was one of the RxLR effector genes with the greatest elevation of transcript levels in roots compared to leaves. The YxSL[RK] family was previously suggested to be an effector family (Levesque et al. 2010) and the very high root‐specific transcript levels of several of these proteins supports this hypothesis, though no functional data to support this role is yet available. Elicitin is considered to be involved in transporting sterols that have a higher concentration in leaves compared to roots (Noman et al. 2020; Wewer et al. 2011). The leaf‐specific elicitins may act as sterol carriers to gain more nutrients from soybean leaves. The plant cell wall is a complex structure composed of multiple components such as cellulose, hemicellulose, pectin and glycoproteins, which can resist the invasion of pathogens (Somerville et al. 2004). The structure and composition of cell walls in different plants organs vary greatly (da Costa et al. 2017). Genes encoding CAZymes also showed particularly marked differences: distinctly different families of CAZymes were induced in roots compared to leaves and vice versa. Presumably these differences reflect adaptations to the different biochemical and physical environments of the leaf and root organs. For instance, cutin is often present in the leaf to prevent pathogen penetration (Yeats and Rose 2013). An expanded set of nine cutinase genes specific to P. sojae was strongly induced in leaf infection but not during root infection. These leaf‐specific cutinases may be involved in digesting and overcoming the cuticle of soybean leaves. Our results have also revealed that the bZIP TF PsBZPc29 of P. sojae plays a major role in regulating (directly or indirectly) genes expressed during infection of soybean roots as well as contributing to regulation of some genes expressed during infection of leaves. The PsBZPc29 knockout mutants also showed altered regulation of the two other root‐specific TFs and the two leaf‐specific TFs, suggesting that it may be a master regulator of a TF network. For these four TF genes, the major role of PsBZPc29 was to impose organ specificity through its effects on leaf‐specific expression; in the deletion mutant the two root‐specific TFs lost leaf‐specific repression while the two leaf‐specific TFs showed reduced leaf‐specific activation. A number of other potential regulatory protein genes that did not show organ‐specific induction were also affected by the deletion of PsBZPc29. These included the constitutively expressed bZIP TFs, PsBZPc31 and PsBZPc41, which were reduced more than 10‐fold in the PsBZPc29 mutants during root infection. Thus, a complex set of regulatory interactions among TFs including PsBZPc29 may be responsible for managing the expression of the infection‐induced genes, including the RSGs and LSGs. The bZIP TF family has an expanded number of genes in Phytophthora, and their functions are diverse and redundant (Sheng et al. 2021; Blanco and Judelson 2005; Ye et al. 2013). Many bZIP TFs are oomycete‐specific with novel bZIP motif variants, which may bind some novel cis‐elements in Phytophthora (Ye et al. 2013; Gamboa‐Melendez, Huerta, and Judelson 2013). PsBZPc29 encodes an R‐R type bZIP protein, and the PsBZPc29 homologues also occur in other clade 7 Phytophthora species. These PsBZPc29 homologues may regulate organ‐specific infection in each Phytophthora species. By analysing promoters of PsBZPc29‐regulated RSGs, we identified several promoter motifs, including Phytophthora core promoter element INR‐FPR (Roy, Poidevin, et al. 2013), and sporangia‐enriched motif (GTTGAAG) (Roy, Kagda, and Judelson, 2013), while no motif similar to the classic bZIP‐binding motif was found. Because PsBZPc29 has an oomycete‐specific bZIP domain with a variant DNA‐binding motif, no information is available regarding its possible DNA‐binding elements. Thus, currently our data do not indicate whether PsBZPc29 directly or indirectly regulates the RSGs and LSGs, including the other TF genes. Further work is needed to elucidate the mechanisms by which the network of TFs including PsBZPc29 impacts the regulation of virulencerelated genes, and how those changes collectively impact root‐specific virulence. In summary, our results showed that P. sojae uses different weapons to overcome the defence of soybean roots or leaves. During root infection, P. sojae secretes more RxLRs, YxSL[RK]s, SCPs, proteases and pectate lyases to attack plant cell wall, and pectin‐associated structures of soybean root. During leaf infection, more elicitins and cutinases are released by P. sojae to attack the plant cuticle covering the epidermis of leaves. PsBZPc29 may be a master regulator for suitable expression of both the RSGs and LSGs. Experimental Procedures Sampling of P. sojae Development Stages Phytophthora sojae strain P6497 was used as the wild type in this study. All P. sojae strains were grown in 10% vegetable juice (V8) agar medium at 25°C in the dark. Mycelia (MY) were cultivated in 10% V8 liquid medium at 25°C in the dark for 48 h, blotted dry with absorbent paper and preserved in liquid nitrogen for RNA isolation. Zoospores (ZO) were induced by repeatedly rinsing mycelial mats with sterile distilled water, concentrated by centrifugation at 400g at 4°C to ≥ 150 zoospores/μL, and then preserved in liquid nitrogen for RNA isolation. Cysts were produced by vortexing the zoospore suspension at room temperature for 30 s, and then germinated cysts (GC) were obtained by cultivating cysts in 5% V8 liquid medium at 25°C for 1 h and collected by centrifugation at 400g at 4°C. Sampling of P. sojae‐Infected Soybean Phytophthora sojae zoospore suspensions were adjusted to a final concentration of 100 zoospores/μL for inoculation. Susceptible soybean (cultivar Williams) plants were grown in a greenhouse at 22°C–25°C with a cycle of 16 h light and 8 h darkness. Leaves from the second‐leaf stage and roots of etiolated seedlings were soaked in a zoospore suspension for 20 min. Leaves were treated with 0.5% vol/vol Tween 20 before inoculation. At least five leaves and seedlings were moved to each of five plates, which were covered with moist filter paper for maintenance of humidity. Samples were collected at 1.5, 3, 6, 9 and 12 hpi for RNA‐seq. At the indicated time points, one leaf was removed from each plate, and 1.5‐cm‐diameter discs were produced using a hole punch; 5 cm lengths of root tips were removed by cutting the seedlings; the remaining procedure was as described for leaves (Figure S2a). Two biological replicates were performed for each treatment. The samples were immediately frozen in liquid nitrogen and stored at −80°C until RNA extraction. RNA Extraction and Sequencing RNA was extracted using the EZNA total RNA kit I (Omega BioTek). After removing ribosomal RNA (rRNA) from total RNA with the Ribo‐Zero rRNA Removal Kit, rRNA‐free RNA was used to produce sequencing libraries using the Illumina TruSeq Sample Preparation Kit v. 2 (for non‐mixed root infection samples) or the NEB Next Ultra Directional RNA Library Prep Kit for Illumina (for other samples), following the manufacturer's recommendations. RNA‐seq was conducted using an Illumina HiSeq 2000 system with 90 base pair (bp) paired‐end reads (for non‐mixed root infection samples), or an Illumina HiSeq 4000 system with 150 bp paired‐end reads (for non‐mixed MY, ZO, GC samples, and mixed WT samples), or an Illumina HiSeq X Ten system with 150 bp paired‐end reads (for mixed mutants samples). The names and abbreviations for the RNA‐seq samples analysed in this study are described in Figure S2b. Read Mapping and Gene Expression Analysis The raw reads were filtered by removing adaptor sequences and low‐quality reads. Then the filtered reads were mapped to ribosomal RNA sequences downloaded from the high‐quality ribosomal RNA database SILVA (www.arb‐silva.de) to remove reads of the remaining rRNA. The resulting high‐quality reads were aligned to the complete genome assemblies of soybean (Wm82.a2.v1 from cultivar Williams 82; the nuclear genome was from Phytozome [phytozome.jgi.doe.gov], and the non‐nuclear chloroplast and mitochondrion genomes were from NCBI [www.ncbi.nlm.nih.gov]) and P. sojae (v1.1 for isolate P6497; jgi.doe.gov) with TopHat v. 2.1.1 (ccb.jhu.edu). A total of two mismatches and gaps per read were allowed, and only when both of a pair of reads were successfully mapped was the data included in analyses. The mapping results for the RNA‐seq reads are provided in Table S1. To eliminate differences among soybean organ components, reads mapped to the chloroplast or mitochondrion were excluded from the analysis. Transcript abundance was indicated as fragments per kilobase of exon model per million mapped reads (FPKM). To identify differentially expressed genes, read counts for each gene model were obtained using featureCounts software (bioinf.wehi.edu.au), then the log2FC value and adjusted p value were calculated using DESeq2 software (www.bioconductor.org), and genes with an adjusted p value < 0.05 and an absolute log2FC ≥ 1 were considered differentially expressed. To assess variability among samples, we performed hierarchical clustering (using MEV 4.7.4; www.tm4.org) and principal component analysis (PCA; using R package; www.r‐project.org) based on the gene expression values. We also performed differential expression analysis between duplicate samples for all P. sojae genes using edgeR software (www.bioconductor.org) to demonstrate the reproducibility of the method. Promoter motifs (in 1000 bp upstream) of the PsBZPc29‐regulated RSG coding regions were analysed using MEME software (meme‐suite.org/meme/tools/meme). Functional Annotation and Analysis of Genes GO enrichment was analysed using web gene ontology annotation plot (WEGO; wego.genomics.org.cn). Paralogous groups of protein sequences were clustered using OrthoMCL (orthomcl.org). Predicted functions of genes were annotated by querying the non‐redundant protein sequences (nr) database of NCBI (www.ncbi.nlm.nih.gov) and the PFAM database (Finn et al. 2016). Phylogenetic trees were constructed based on protein sequences using MEGA software (v. 5.02; megasoftware.net; neighbour‐joining method, bootstrap value of 1000). Secreted proteins were predicted using software packages SignalP v. 3.0, TargetP v. 1.1, and TMHMM v. 2.0 available at www.services.healthtech.dtu.dk; candidates should simultaneously have a SignalP HMM score > 0.9, subcellular localization as secreted, and no transmembrane domain after the predicted signal peptide cleavage sites. With the exception of the ABC transporters (Morris and Phuntumart 2009), we used BLASTP with an E‐value of 1e−5 as the cut‐off to identify the other transporters in P. sojae, based on those identified in P. infestans (Abrahamian et al. 2016); PFAM was used to predict and validate the transporter domain. The proteins with the following GO terms were included in the oxidoreductase‐related family: GO IDs 0016491, 0016624, 0005777, 0008131, 0006979, 0016616, 0016628, 0015036, 0008137 and 0004601. For effectors annotation, RxLRs, CRNs, elicitins and NLPs were provided in the previous publications (Jiang et al. 2006; Haas et al. 2009; Wang et al. 2011; Dong et al. 2012). The SCPs were defined as secreted proteins that were 50–150 amino acids in length containing > 5% cysteine residues (Raffaele et al. 2010). Secreted proteins corresponding to GO term 0006508 were identified as proteases. YxSL[RK] effectors were identified according to those published in Globisporangium ultimum (Levesque et al. 2010), using BLASTP with an E‐value of 1e−5 as the cut‐off; the candidates were aligned with the reference sequences to determine the presence of typical motifs. The CAZyme families of proteins were clustered using the dbCAN tool (Yin et al. 2012). Quantitative PCR RT‐qPCR was performed to confirm the RNA‐seq results of selected genes. RNA was quantified using a spectrophotometer (ND‐1000; NanoDrop). To remove contaminating genomic DNA from the RNA samples, total RNA (3 μg) was treated with two units of RNase‐free DNase I (TaKaRa) at 37°C for 30 min. First‐strand cDNA was synthesised using Moloney murine leukaemia virus reverse transcriptase (RNase‐free) and an oligo(dT)18 primer (Invitrogen). qPCR was performed in 20 μL reactions containing 20 ng cDNA, 0.2 mM primers for the target gene or reference gene (Ps_108986), 10 μL of SYBR Premix ExTaq (TaKaRa), and 6.8 μL of deionised water. The primers are listed in Table S6. PCR was performed on an ABI Prism 7500 Fast Real‐Time PCR System (Applied Biosystems Inc.) under the following conditions: 95°C for 30 s; followed by 40 cycles of 95°C for 5 s and 60°C for 34 s; and finally, 95°C for 15 s, 60°C for 1 min and 95°C for 15 s. Relative expression levels were calculated using the 2−[Ct(target gene)−Ct(reference gene)] method. To quantify the P. sojae biomass in infected soybean tissue, samples were removed at 0.5, 1.5, 3, 6, 9, 12, 24 and 36 hpi and genomic DNA extracted using the DNA SECURE Plant Kit (TianGen). P. sojae actin (Ps_108986) and soybean CYP2 was quantified by qPCR, and the biomass ratio was calculated using the 2−[Ct(actin)−Ct(Cyp2)] method. Confocal Microscopy Observation Soybean leaves and root epidermal layers infected with GFP‐expressing P. sojae (a P6497 strain containing a pTOR‐GFP overexpression plasmid provided by Professor Francine Govers, Wageningen University) were observed using an LSM 710 laser scanning microscope with a 20× or 40× objective lens (Zeiss). GFP fluorescence was observed using excitation at 488 nm. Phytophthora Transformation and Virulence Assay Homologous gene replacement was performed using the CRISPR/Cas9 system (Fang and Tyler 2016). The mutants were confirmed by PCR and sequencing (Figures S6 and S7). The detailed sequences of PCR products were listed in Table S7. RNA‐seq of the PsBZPc29 mutants showed no PsBZPc29 transcript. The virulence of P. sojae transformants was determined by inoculating soybean roots and leaves with zoospores (ZOs). For leaf inoculation, we used a liquid drop that contained 100 ZOs, and for root infection, the roots were soaked in 5 mL ZO suspension that contained 100 ZOs. The plants were incubated in the dark at 25°C before assessment of infection. Virulence was determined by determining the P. sojae biomass in infected soybean tissue by qPCR. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1. Figure S2. Figure S3. Figure S4. Figure S5. Figure S6. Figure S7. Figure S8. Figure S9. Table S1. Table S2. Table S3. Table S4. Table S5. Table S6. Table S7.