Nicotiana benthamiana XYLEM CYSTEINE PROTEASE genes facilitate tracheary element formation in interfamily grafting

Abstract Grafting is a plant propagation technique widely used in agriculture. A recent discovery of the capability of interfamily grafting in Nicotiana has expanded the potential combinations of grafting. In this study, we showed that xylem connection is essential for the achievement of interfamily grafting and investigated the molecular basis of xylem formation at the graft junction. Transcriptome and gene network analyses revealed gene modules for tracheary element (TE) formation during grafting that include genes associated with xylem cell differentiation and immune response. The reliability of the drawn network was validated by examining the role of the Nicotiana benthamiana XYLEM CYSTEINE PROTEASE (NbXCP) genes in TE formation during interfamily grafting. Promoter activities of NbXCP1 and NbXCP2 genes were found in differentiating TE cells in the stem and callus tissues at the graft junction. Analysis of a Nbxcp1;Nbxcp2 loss-of-function mutant indicated that NbXCPs control the timing of de novo TE formation at the graft junction. Moreover, grafts of the NbXCP1 overexpressor increased the scion growth rate as well as the fruit size. Thus, we identified gene modules for TE formation at the graft boundary and demonstrated potential ways to enhance Nicotiana interfamily grafting.


Introduction
Grafting is a technique used to unite two plant segments, following tissue adhesion and vascular reconnection. It has been widely practiced in horticulture, applied to propagate ornamental and fruit trees from ancient times. Moreover, grafting has been used for cultivation of vegetables in recent decades [1,2]. A graft establishment depends on the successful formation of graft union that includes wound response, cell-cell adhesion, callus proliferation, and vascular formation [3,4]. Although there is a range in the extent of graft viability and hardiness, successful grafts achieve subsequent scion growth and/or fruit setting eventually. Previous observations clearly showed that vascular connection between the scion and stock is a determinant of grafting success [5]. In general, graft compatibility is observed between phylogenetically closed relatives, limiting the range of plant combinations available for grafting. However, recent studies have shown potential possibilities to expand graft capability even among interfamily relations in eudicots and monocots [6,7]. Subsequent studies showed that a part of Solanaceae and a part of parasitic plants in Orobanchaceae and Convolvulaceae exhibit tissue adhesion ability with interfamily partner [8,9] and are called interfamily partner accepting grafts (iPAG) [9]. In Nicotiana iPAG, new xylem connections were observed at the graft junction, as xylem bridges were seen in parasitism between interfamily hostparasite combinations [10]. However, phloem connections are not achieved in many cases of parasitism and are rarely established in Nicotiana iPAG, whereas symplastic connections, probably through plasmodesmata, are established in both cases. The grafted and parasitic plants can survive and complete their life cycle until seed setting, implying essential roles of xylem bridge/connection in parasitism/grafting through these tissue adhesive conditions. Tracheary elements (TEs), main components of xylem, are dead hollow cells with secondary cell walls that form vessels or tracheids [11]. TEs are conducting tissues that transport water and dissolved minerals from the roots to shoots to maintain normal growth [12]. In addition, proteins, hormones, and metabolites transported in the xylem have been evidenced to achieve various roles in cabbage and maize plants [13,14]. The in vitro xylem vessel element formation system revealed that certain plantspecific NAC-domain (no apical meristem, NAM; Arabidopsis transcription activation factor, ATAF; up-shaped Cotyledon 2, CUC2) transcription factors were identified as VND (Vascularrelated NAC-Domain) genes [15]. It has been suggested that in Arabidopsis VND6 and VND7 are the primary determinants of TE differentiation and VND1-VND5 are the weaker determinants [15][16][17][18]. Moreover, VND7, a master switch in xylem vessel formation, has been demonstrated to regulate the expression of a broad range of genes in root and shoot 19,20 . TEs undergo programmed cell death (PCD), removing the entire protoplast and forming a hollow vessel for substance transportation. Metacaspase 9 (AtMC9), XCP1, and XCP2 degrade cellular remnants within the cell lumen after vacuolar rupture and assist in the final process of PCD [21][22][23].
In this study, we investigated significance of xylem formation at the graft junction by two means; a comparison between successful and unsuccessful interfamily grafting and a comparison between with or without an auxin transport inhibitor, 2,3,5triiodobenzonic acid (TIBA), that blocks xylem formation. Through the transcriptomic analysis, we have described the expression patterns of xylem-associated genes in successful and unsuccessful interfamily grafting. Using 36 transcriptome datasets and a super computer, a Bayesian network analysis was conducted to reveal the causal relationship between gene expression. Out of genes included in core modules for xylem formation, we focused on NbXCP genes to investigate how de novo TEs at the graft junction is important for the Nicotiana iPAG.

Xylem reconnection is required for the scion growth
Nicotiana benthamiana grafted onto the inf lorescence stem of Arabidopsis (Nb/At) constituted a successful interfamily graft combination. However, Glycine max and Arabidopsis (Gm/At) form an incompatible interfamily graft [6]. Xylem cells differentiated from N. benthamiana scions and attached to the cells in Arabidopsis stem tissues in Nb/At graft unions. We observed xylem bridges connecting the scion and stock at around 14 days after grafting (DAG). Fewer xylem cells differentiated from G. max scions in Gm/At graft unions. The xylem bridge was not formed even at 14 DAG ( Figure 1A). In addition, we observed the development of secondary xylem in N. benthamiana and G. max stems from neighboring graft boundaries. Secondary xylem tissues were developed in N. benthamiana scion stems but were rarely observed in G. max scion stems (Supplemental Figure S1). Scion growth and fruit setting were observed in Nb/At grafts but scions did not grow in Gm/At grafts (Figure 1, B and C). In Nb/At interfamily grafting, tracheary elements (TEs), which have spiral patterns on cell walls ( Figure 1D), were differentiated in the callus of the N. benthamiana scion. The de novo TEs were differentiated in the callus at the graft boundary and eventually connected the xylem tissues originally located in the scion and stock plants and formed xylem bridges between the scion and stock. Next, TE formation in grafting was blocked to examine the effect of de novo TE formation on scion growth by applying 2,3,5-triiodobenzonic acid (TIBA), an auxin transport inhibitor. TIBA inhibited TEs formation in a Zinnia elegans cell culture system [24]. Similarly, TIBA inhibited de novo TE formation during N. benthamiana grafting (Figure 1, E and F). This inhibition was also ref lected in scion growth after 14 DAG (Figure 1, G and H). These results suggested that scion growth relies on de novo TE formation and xylem reconnection.

Xylem formation genes were expressed during graft establishment
We analyzed the transcription of genes related to xylem formation to investigate the molecular mechanism of xylem formation during interfamily grafting. In previous research on Arabidopsis, VND1-VND7 were considered master transcription factors that initiate xylem formation processes, such as cell autolysis and cell wall synthesis, through the regulation of downstream genes ( Figure 2A) [15,16,18,20]. Before analyzing gene expression patterns, we examined the timing of xylem formation initiation. We conducted Nb/At grafting to test the establishment of water transport by adding toluidine blue solution to Arabidopsis stock plants and observing dye transport to the scion part by making a section. At 14 DAG, all grafts exhibited xylem bridges (a representative image is shown in Figure 1A). The dye transport experiments were performed at 1, 2, 3, 7, and 10 DAG, in which 8-10 grafts were examined at each time point. Dye transport was first detected at 3 DAG, 20% (2 out of 10) of grafts. The detection frequency did not change at 7 DAG and increased to 62% (6 out of 8) of grafts at 10 DAG ( Figure 2B), suggesting that xylem formation started around 3 DAG and kept developing with time. We examined the gene expression patterns at 0, 1, 3, and 7 DAG in successful Nb/At and unsuccessful Gm/At interfamily grafts (Figure 2, C and D) based on this observation. All four VND7 homologous genes in N. benthamiana (NbVND7s) were upregulated more than the other VND homologs in the Nb/At grafts, indicating that these genes are important for de novo xylem formation at the graft boundary in Nicotiana ( Figure 2C). In the Gm/At grafts, one of the two VND7 homologous genes in G. max was upregulated but the other was not ( Figure 2D). To examine the expression patterns of downstream genes (marked in pink, blue and green in Figure 2C, D) of VND7s in these grafts, homologs of 19 genes known as VND7 downstream genes in Arabidopsis [15,[18][19][20] were identified in N. benthamiana and G. max by phylogenetic analysis. Most genes involved in xylem formation were upregulated in the Nb/At grafts but downregulated in the Gm/At grafts (Figure 2, C and D). Thus, the morphological features of TE formation at the graft junction are consistent with the expression patterns of xylem formationassociated genes.

Network analysis revealed a core module for xylem differentiation during grafting
To investigate the genes involved in xylem formation during grafting, we performed gene network analysis. We first performed Weighted Gene Correlation Network Analysis (WGCNA) using transcriptome data from 3 replicates of intact N. benthamiana stems and Nb/At grafting sites at 2 hours after grafting, 1, 3, 5, and 7 DAG (Supplemental Figure S2). The iDEP 0.96 interface (http://bioinformatics.sdstate.edu/idep96/) was used to draw co-expression networks for 3000 most variable genes (Supplemental Figure S2A). A "soft threshold" of "9" and a "minimum module size" of "100" generated 6 well-defined modules consisting of 2818 genes (Supplemental Figure S2B-D). These 6 modules contained 11 genes related to canal formation as shown in Figure 2C (Supplemental Table S1). From the N. benthamiana VNDs were included in modules 2 and 6, we extracted 313 genes that were linked neighboring to the two respective VND7 within those modules (Supplemental Table S2). For those genes, gene ontology analysis was performed using Arabidopsis homolog annotation. We detected transcription factors in the molecular function category, cell wall-related genes in the cellular component category, and water deprivation and wound response genes in the biological process category (Supplemental Table S3). However, this analysis did not resolve the causal relationship of gene expression, so it is not possible to distinguish whether these co-expressed genes are regulated by VND7 or affect the expression of VND7. We next performed a Bayesian network analysis, which can reveal the causal relationship between gene expression, using the SiGN-BN NNSR program [25,26] with 36 N. benthamiana grafting transcriptome data [6]. All data from the whole-genome transcriptome were used as initial inputs to draw the gene network ( Figure 3A). In the network, 2038 N. benthamiana genes were associated with four NbVND7s within the sixth level (a direct connection is the first level, a connection bypassing one other gene is the second, and so on) (marked in yellow in Figure 3A) including LBD15, LBD30, ATMC9, Myb83 VND2, XCP1, XCP2 and various unidentified genes (Supplemental Table S4). To further analyze the gene regulatory network of xylem differentiation, we attempted to apply the Bayesian network using the SiGN-BN HC + Bootstrap program [25,26] that can describe upstream/downstream relation among genes. Since the upper limit of input in this program is about 1000 genes, we narrowed down the genes to focus on through gene ontology (GO) analysis (see detail in Supplemental Text 1 and Supplemental Table S5). The Bayesian network for four NbVND7s and 846 N. benthamiana genes categorized into four GO groups were drawn ( Figure 3B, Supplemental Table S6). By narrowing down the genes associated with de novo xylem formation during grafting, we highlighted a core structure from NbVND7s to genes for xylem formation among the entire network linking to NbVND7s. We compared the 846 N. benthamiana genes with the 189 N. benthamiana graftinduced genes at 1-3 DAG that were previously identified since xylem formation was initiated after 3 DAG ( Figure 2B) [6]. As nine genes overlapped in the comparison (Supplemental Table S7), we extracted the nodes and edges starting from NbVND7s and ending with the nine genes as a core network that potentially links to xylem formation ( Figure 3C, Supplemental Table S8). In the center of the core network, the overlapping gene  [15,16,18,20]. B, Measurement of the vascular tissue reconnection in the Nb/At interfamily grafts. The frequency of toluidine-blue detection in the scion was measured at 1, 3, 5, 7, and 10 DAG. Eight to ten grafts were examined for each time point. C, D, Expression profiles of xylem formation-related genes in N. benthamiana and Glycine max. The transcription levels of genes in Nb/At or Gm/At interfamily grafting were estimated by RNA-seq. Colors in the heat map represent the fold changes (log10) in the transcription level at 1, 3, and 7 DAG relative to the intact stem (0 DAG). Xylem formation-related genes in Arabidopsis were listed from previous study [18,20]. Corresponding genes in N. benthamiana and G. max were orthologous genes of Arabidopsis estimated by the tblastn program. At, Arabidopsis; Nb, N. benthamiana; Gm, G. max.
Niben101Scf01015g01002, a homolog of pathogenesis-related 4 [27] and named NbPR-4 is connected by several genes ( Figure 3D). One cluster consisting of eight overlapping and eight additional genes included genes expected to be directly involved in tracheary element differentiation, such as all four Arabidopsis xylem cysteine peptidase 1 and xylem cysteine peptidase 2 (XCP1 and XCP2) orthologs, named NbXCP1-4, a tracheary element differentiation-related 6 (TED6) ortholog, and a glycosyl hydrolase family 10 protein (ATXYN1) ortholog [22,28,29], and the other cluster connecting to NbPR-4 consists of an additional six genes, whose roles in xylem formation are unknown (listed in Figure 3E and see also Supplemental Text 2).

NbXCP genes are expressed to form de novo TE during grafting
To validate the described gene regulatory network during grafting, as an example, we tested the role of NbXCP homologs that were supposed to degrade cellular contents through protease activity to accomplish mature TE formation [22] on xylem formation at the graft junction. NbXCP1 and NbXCP2 were most similar (99% similarity in coding sequence region), followed by NbXCP3 ( Figure 4A and Supplemental Figure S3A). NbXCP4 has a large C-terminal deletion and is likely to be a pseudogene that does not function as a protein (Supplemental Figure S3A). NbXCP1-3 were upregulated after interfamily grafting, but NbXCP3 expression was lower ( Figure 4B). NbXCP3 positively affected the expression of NbXCP1 and NbXCP2 ( Figure 3D), suggesting that the control of NbXCP3 expression may be different from those of NbXCP1 and NbXCP2. Therefore, we focused on NbXCP1 and NbXCP2. We conducted a heterogeneous yeast one-hybrid assay using the NbXCP1 promoter (as a representative) and Arabidopsis VNDs to verify the control of NbXCP gene expression by VND homolog transcription factors. Only VND7 could bind to the NbXCP1 promoter (Supplemental Figure S4, Figure 4C). The target sequence of VND7 [30] was identified in the NbXCP1 promoter ( Figure 4C). An equivalent sequence was present in the NbXCP2 promoter, although it contained a mismatch of one terminal nucleotide.
To investigate NbXCP1 and NbXCP2 expression patterns in plant tissues, we generated promoter GUS transgenic lines (Supplemental Figure S5). In pNbXCP1::GUS and pNbXCP2::GUS lines, GUS expression was found in the xylem tissues of seedlings (Supplemental Figure S6) and the xylem and internal phloem of the stems of 4-week-old plants ( Figure 4D, Supplemental Figure S7). GUS expression is restricted to TEs rather than to other xylematic cell types. We conducted a GUS expression analysis in Nb/At interfamily grafting. GUS expression was concentrated in the TEs newly generated in the callus at the graft junction ( Figure 4E and F, Supplemental Figure S7). To estimate protein function, we generated GFP fusion lines to examine the subcellular localization of NbXCP1 and NbXCP2 (Supplemental Figure S5). When NbXCP1-GFP and NbXCP2-GFP fusion proteins were expressed in the stem using their promoters, GFP f luorescence was detected in the cytoplasmic region of TE maturating cells ( Figure 4G, Supplemental Figure S7). Except for the xylem cells, we detected the GUS stains and GFP expression in the mature pollen (Supplemental Figure S8), the autolysis of which has been reported in Arabidopsis, Lycopersicum peruvianum, Olea europaea, Lolium perenne [31,32]. We performed particle bombardment to express GFP alone or NbXCP1 and GFP simultaneously in N. benthamiana leaf epidermal cells. For the GFP solely expressed, GFP f luorescence was detected in the cytoplasmic region and nucleus in all 116 cells observed. In contrast, for the NbXCP1 and GFP simultaneously expressed, ∼10% (12/116) of cells exhibited a planar GFP f luorescence pattern with strong GFP dots, indicating cytoplasmic degradation in the cells ( Figure 4H). Finally, we confirmed the function of NbXCP1 and NbXCP2 in TE maturation by analyzing a double knockout mutant of NbXCP1 and NbXCP2 generated using the CRISPR/Cas9 system (Supplemental Figure S9). Transmission electron microscopy revealed that the Nbxcp1;Nbxcp2 mutant showed a defect in cellular digestion in differentiating TEs of the stem compared with the wild-type (WT) (Figure 4I), as observed in the Arabidopsis xcp1;xcp2 mutant [22]. Overall, NbXCP1 and NbXCP2 play a conserved role in TE formation.

De novo TE formation is essential for graft establishment and post-scion growth
We investigated the function of NbXCP1 and NbXCP2 during grafting to determine whether xylem formation is important for the establishment of interfamily grafting. We conducted virus-induced gene silencing experiments as reported previously (Notaguchi et al., 2020). Cucumber mosaic virus (CMV) vectors containing a 295 bp sequence of NbXCP1 that is identical to that of NbXCP2 (CMV-NbXCP1/2) or a 292 bp sequence of GFP (CMV-GFP) were infected with N. benthamiana shoots and grafted onto the Arabidopsis stock. RT-PCR and qRT-PCR analyses demonstrated knockdown of NbXCP1 and NbXCP2 (Supplemental Figure S10, Figure 5A). Compared with the non-infected (NI) and CMV-GFPinfected controls, CMV-NbXCP1/2-infected N. benthamiana scions showed significantly lower survival rates at 14 DAG ( Figure 5B). We next examined interfamily grafting using the Nbxcp1;Nbxcp2 knockout mutant and NbXCP1 translationally enhanced (NbXCP1-OX) lines that maintained endogenous expression patterns by using own promoter (Supplemental Figure S9). Although there was a slight decrease of the fresh weight only in the intact Nbxcp1;Nbxcp2 mutant seedlings, not in the NbXCP1-OX lines, the growth retardation was no longer observed in the grown plants (Supplemental Figure S9). The survival rate decreased in the Nbxcp1;Nbxcp2/At grafts, while the NbXCP1-OX/At grafts tended to increase ( Figure 5C). The TE formation timing at the graft junction was measured for each graft combination. De novo TE formation began 3 DAG. The frequency of grafts that first formed TE at the graft junction was lower in the Nbxcp1;Nbxcp2 mutant and higher in the NbXCP1-OX line than in the WT at 3 and 4 DAG. All grafts formed TEs at 5 DAG or later in all graft combinations ( Figure 5D). Consistent with this observation, transport of isotope-labeled phosphorus-32 ( 32 P) to the N. benthamiana scions was lower in the Nbxcp1;Nbxcp2 mutant than in the WT and NbXCP1-OX plants at 7 DAG ( Figure 5E). This finding revealed that NbXCP1 and NbXCP2 affect de novo TE formation from the early phase of graft union formation and that the pace of de novo TE formation at graft junction affects the level of water transport from the rootstock to the scion and scion growth. We continued to grow the grafts and measured scion growth. The scion growth of the Nbxcp1;Nbxcp2 mutant was lower but the NbXCP1-OX scion grew more than the WT Nb scion ( Figure 5F). The scion growth was also ref lected in fruit production. The fresh weight of fruits was decreased in the scion of Nbxcp1;Nbxcp2 mutant compared to the wild-type N. benthamiana scion. In contrast, the fruit weight was increased in NbXCP1-OX scion ( Figure 5H). However, we did not find the fruit weight has different in intact plants of WT, Nbxcp1;Nbxcp2 mutant and NbXCP1-OX line. These findings suggest that NbXCP1 and NbXCP2 have a role in xylem formation at graft junction that eventually contribute to both scion survival and post-grafting growth.

Discussion
Water uptake from the soil through water-conducting tissues is essential for most land plants. Therefore, in grafting, new xylem formation at the graft junction is necessary for the survival of grafted plants. This study addressed this principle by comparing successful and unsuccessful interfamily grafting, where xylem formation was achieved and completely absent, respectively, and comparing mock and exogenous TIBA treatments to delay the timing of xylem formation. In Arabidopsis hypocotyl grafting, it was shown that TIBA suppresses the cell proliferation of vascular tissue during graft union formation [33]. Moreover, TIBA and the other auxin transport inhibitors, also blocked the formation of TEs and continuous vascular stands at the interface of the haustorium and the host tissues [10,24]. The experiments in this study confirmed that xylem formation is essential for the  NbVND7 transcriptional factors regulate the downstream genes including xylem formation genes and plant immunity genes during the TE differentiation. Bayesian network analysis reveals the candidate genes in graft union formation. Among them, the function of NbXCP1 and NbXCP2 on de novo TE formation has been verified in this study.
Nicotiana iPAG. The timing of xylem connections after grafting determine the post-grafting growth of scion plants ( Figure 1) and it was accelerated by the overexpression of NbXCP1, resulting in the enhancement of scion growth ( Figure 5). Transcriptomic analysis demonstrated that previously known xylem-associated genes exhibited consistent expression patterns (Figure 2). Previous transcriptome analyses on graft junctions also identified gene expression controls of xylem-associated genes [9,34,35]. In this study, we estimated a gene network related to xylem formation, whose accuracy was evaluated by analyzing a significant gene, NbXCP (Figures 4 and 5), and described the features of the gene modules ( Figure 3). In previous studies, two studies on gene network analysis for plant grafting have been reported. A co-expression network was used to identify genes involved in graft formation [36]. A Bayesian network has been used to detect gene regulatory networks especially involved in upstream event of grafting and successfully identified SlWOX4 as a key transcription factor to trigger downstream gene expressions [35]. In this network analysis, DEG and GO analyses were first conducted to narrow down the target genes for network estimation in order to match the limit of the number of genes that can be used as the input of the network analysis method used. This limitation in gene numbers for input required cutting out a part of gene population. To overcome this number limitation, the SiGN-BN HC + Bootstrap program [25,26] was applied to estimate Bayesian networks on a genome-wide scale. An entire network was successfully obtained where expression datasets of all genes of N. benthamiana were used as an initial input. However, since we could not solve the causal relationship on this scale, we reduced the number of genes to about 1000 by capturing our interested portion out of entire gene network, genes associated xylem formation in this study. To eliminate artificial bias as much as possible, the P value and population size were used as criterions in this study, rather than the contents of GOs, resulting in identification of an unexpected module ( Figure 3).
In this study, we proposed a model ( Figure 5H) for TE formation pathways of Nicotiana interfamily grafting. There are two reliable relationship gene modules under the control of NbVND7s, the master transcriptional element genes; a module involved in xylem formation, including NbXCPs and a module involved in plant immunity, centering NbPR-4. In the module of xylem formation, we have revealed the orthologs of tracheary element differentiation-related6 (TED6) and endo-β-1,4-xylanase (XYN1). TED6 was identified as secondary cell wall (SCW)-related membrane proteins, participating the formation of tracheary element. An interaction was found between the TED6 and a member of the SCW-Cellulose Synthase (CesA) complex [28,37]. XYN1 was verified expression on the predominantly in xylem tissues and encodes xylanase. The xylem transport patterns were affected in AtXYN1 modified lines [29]. In the module of xylem formation, genes associated with the generation and scavenging of reactive oxygen species (ROS) were found (Figure 3, Supplemental Table S7). These genes are considered to be directly involved in the xylem formation process. In Arabidopsis, secondary cell wall formation consisting of cellulose, hemicellulose, and lignin has been observed during xylem development. It has been shown that lignification proceeds by polymerization of monolignols, in which peroxidase uses hydrogen peroxide as a substrate [38,39]. Thus, this module can be highly associated with xylem cell differentiation.
The module involving NbPR-4 contains the orthologs of cationic amino acid transporter 6 (CAT6) and transthyretin-like protein (TTL). In Arabidopsis, AtCAT6 was nematode-induced in roots during the infestation, mediating transportation of amino acids to prevent plant injury [40]. Research also reported that CAT6 and CAT7 involve in oxidative stress response in cassava plants [41]. The TTL protein was found as scaffold proteins for a potential substrate of Brassinosteroid-insensitive 1 (BRI1) in Arabidopsis, promoting brassinosteroid responses [42]. Brassinosteroids play an important role in inhibiting pathogen infection, in which mediating growth directly antagonizes innate immune signaling [43]. However, three of the four genes positively regulating NbPR-4 encode the uncharacterized protein kinase or the proteins of unknown function. Further investigation may provide us new insights on the components involved in the relation between xylem formation and immune system in plants.
These findings may imply that there is an inevitable relationship between plant immunity and xylem formation at the graft boundary because it is reasonable that plants react after graft wounding to avoid pathogen infection. In fact, the tyloses from xylem parenchyma cells have been shown to give resistance against the pathogens [44]. Recent studies on pathogenesis in Arabidopsis reported that XCP genes may assist plant immunerelated gene expression and enhance resistance against xylem vessel pathogen [45,46]. Although XCP proteins did not directly act on plant pathogens [45,47], XCP1 activated the systemic immunity by proteolyzing Pathogenesis Related Protein 1 [46]. Thus, xylem formation, an enlargement of the apoplastic region inside of the plant body, may enhance the plant immune system since it gains potential risks for the exposure to endophytes and pathogenic bacteria transmitted through the xylem tissue. However, further elucidation is required for this hypothesis. The spatiotemporal gene expression patters of potential immunerelated genes will be important to understand the mechanism.
Thus, the gene network described in this study may include a set of cellular aspects of xylem formation and plant immune mechanism during grafting. In the future, it would be interesting to investigate newly identified genes during grafting and/or xylem formation in the development of other organs. In addition, the similarity of the network of TE formation between interfamily and intrafamily grafting using accumulated transcriptome data from other species is interesting. Moreover, network analysis of other biological processes during grafting would help to investigate the molecular events of grafting and identify crucial gene modules. This strategy will enhance our understanding of how plants achieve tissue reunion at graft wound sites.

Plant materials and growth conditions
Seeds of N. benthamiana and Arabidopsis were sterilized with 5% (w/v) sodium hypochlorite (NaClO) solution for 5 min, washed six times with sterile water, and incubated at 4 • C in the dark for three days. Seeds of N. benthamiana were sown on half-strength Murashige and Skoog medium supplemented with 0.5% (w/v) sucrose and 1% (w/v) agar. The pH was adjusted to 5.8 with 1 M KOH. Seedlings were cultured on medium for seven days in growth chambers and transplanted into the soil in a growth room. Arabidopsis seeds (ecotype Col-0) were directly surface-sown on the soil. Seedlings of N. benthamiana and Arabidopsis were grown at 27 • C and 23 • C with 70% and 30% relative humidity, respectively, and 100 μmol m −2 s −1 continuous illumination. G. max seeds were directly sown in the soil in a 23 • C growth room. Plants in the growth room were watered thrice per week.

Grafting
Wedge grafting was performed on the stems of 3-week-old G. max, 4-week-old N. benthamiana, and 5-week-old Arabidopsis plants, as described previously [6]. N. benthamiana and G. max stems were cut and trimmed to a V-shape. The inf lorescence stems of Arabidopsis were cut and split from the stem middle at the graft site. The Vshaped stem was inserted into a middle-split stem and fixed using wrapping film (Bemis Parafilm M) or clip to form a graft union. The scions were covered with water-sprayed plastic bags and grown in an incubator at 27 • C under continuous light (30 μmol m −2 s −1 ) for 10 d. Afterward, plastic bags were removed from the scions, the grafts were transferred to a 23 • C growth room with 30% relative humidity, and 100 μmol m −2 s −1 continuous illumination. CMVinfected grafts were grown in a 23 • C incubator for 10 days. The other experimental conditions were the same as those mentioned above. The scion outgrowth length was measured from the scion top node to the bottom node of the scion outgrow part (Figure 1, B and C) and recorded every week from 7 DAG (days after grafting) to the next several weeks.

Chemical treatment to graft region
An auxin transport inhibitor 2,3,5-triiodobenzoic acid (TIBA) was dissolved in dimethyl sulfoxide (DMSO) to prepare 100 mM TIBA stock solution. The stock solution was diluted to 100 μM using distill water. Equivalent concentration of DMSO (0.1%) solution was used as a control. A 50 μL drop of either of the chemical solution was added to graft region every two days until 14 DAG.

Construction of plasmid vectors
Genomic DNA was extracted from plant tissues using the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany). Target DNA was amplified by PCR using TaKaRa Ex Taq DNA Polymerase (TaKaRa Bio, Tokyo, Japan). DNA segments were purified using MonoFas DNA Purification Kit I (ANIMOS, Saitama, Japan). Total RNA was isolated from plant tissues using the RNeasy Plant Mini Kit (Qiagen). cDNA was synthesized using SuperScript III First-Strand Synthesis SuperMix (Thermo Fisher Scientific, Waltham, USA). DNA fragments were inserted into plasmid vectors using NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs, Massachusetts, USA). The primers used in this study are listed in the Supplemental Table S9. For the GUS-fusion construction (Supplemental Figure S5), the putative promoter sequence amplified from the N. benthamiana genome was fused upstream of the GUS (β-glucuronidase) reporter gene in pENTR vector (pENTR / D-TOPO Cloning Kit, Thermo Fisher Scientific, Waltham, USA) and transferred into the binary vector pGWB1 based on LR recombination reaction. For the GFP-fusion construction (Supplemental Figure S5), the promoter sequences and the coding sequences (CDS) were fused upstream of GFP (green f luorescence protein) reporter gene in pENTR vector and transferred into the binary vector pGWB1. For the NbXCP1 translationally enhanced (NbXCP1-OX) construction (Supplemental Figure S5), pNbXCP1::Ω-NbXCP1, an artificially synthesized Ω (omega) fragment was inserted between the promoter sequence and CDS. To the CRISPR/Cas9 vector, fragments of NbXCP1/2 were as single guide RNA (sgRNA) added to pKI1.1R [48], according to the abovedescribed methods.

Production of transgenic plants
The relevant plasmid vectors (pGWB1 and pKI1.1R) were transferred into N. benthamiana by the Agrobacterium-mediated transformation method for generating transformants. Transgenic plants of GUS-fusion lines, GFP-fusion lines and NbXCP1 translationally enhanced (NbXCP1-OX) lines were screened by antibiotic resistance in seedlings. CRISPR/Cas9-induced mutants were confirmed by sequencing. Homozygous lines of NbXCP1-OX lines and mutants were harvested in T2 or T3 generation, which were used in this study.

Chemical staining and microscopy
Phloroglucinol-HCl stain solution were composed of one volume of 1% (w/v) phloroglucinol dissolved in 70% ethanol and five volumes of 5 M HCl. The toluidine blue stain solution was 0.01% (w/v) dissolved in 0.1 M NaOAC at pH 4. The safranin-O stain solution was 2.5% (w/v) dissolved in 99% ethanol. Fresh tissue sections were directly cut by hand with a blade. Resin-embedded tissue sections were prepared with Technovit 7100 Kits (Kulzer, Wehrheim, Germany) and cut with a rotary microtome (RX-860, Yamato Kohki, Saitama, Japan). Tissue sections were soaked in stain solution for 5 min and observed using a microscope (BX53, Olympus, Tokyo, Japan) equipped with a digital camera (DP73, Olympus) for high-magnification images.
The GUS staining solution consisted of 10 mM PBS at pH 7, 0.1% Triton X-100, 10 mM EDTA, 0.5 mg/mL X-Gluc, 0.5 mM potassium ferricyanide, and 0.5 mM potassium ferrocyanide. Plant tissues were placed in a 90% acetone solution for 5 min, rinsed in 1 mM PBS, transferred to GUS stain solution, and incubated at 37 • C for 6-12 hours. Plant tissues were soaked in 70% ethanol to stop GUS staining and microscopically observed or prepared in resinembedded sections. GUS-stained seedlings were observed using a zoom stereomicroscope (SZX10, Olympus) equipped with a digital camera (DP22, Olympus) for photography.
For cellular GFP observation, tissue sections were treated with ClearSee (FUJIFILM Wako Chemicals, Miyazaki, Japan) solution using a fixative solution. A confocal laser scanning microscope (FV3000, Olympus; LSM5 Pascal, Zeiss, Jena, Germany) was set at 488 nm excitation wavelength. For tracheary element observation, the tissue sections were soaked in advance with a mixed solution composed of 20 μg/mL BF-170 (FUJIFILM Wako Chemicals) solution and 2 μg/mL propidium iodide (P1304MP, Thermo Fisher Scientific) solution. The excitation wavelength of the confocal laser scanning microscope (FV3000, Olympus; LSM5 Pascal, Zeiss) was set to 488 nm for BF-170 and 543 nm for propidium iodide. For transmission electron microscopy (TEM) observations, the plant samples were prepared in pieces and sent to Tokai Electron Microscopy, Inc. for photographs.

Gene expression analysis
RNA-seq data collected in a previous study [6] were used in this study. The RNA-Seq data are available from the DNA Data Bank of Japan (DDBJ; http://www.ddbj.nig.ac.jp/) under accession number DRA009936. qRT-PCR reactions were performed using the KAPA SYBR FAST qPCR Master Mix (2X) Kit (KAPA Biosystems, Wilmington, USA) on the QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific) with standard protocol. The primers used for PCR amplification are listed in the Supplemental Table S9.

Network analysis
Whole genome Bayesian network analysis in the normalized datasets of RNA-seq experiment in 24 Nb/At interfamily grafting, nine Nb/Nb self-grafting, and three intact N. benthamiana stems were performed using a Bayesian network estimation program, SiGN-BN (http://sign.hgc.jp/signbn/index.html) implemented on the supercomputer system at the Human Genome Center of the University of Tokyo (https://supcom.hgc.jp/english) [25,26]. The estimated gene network was analyzed using the gene network analysis software Cytoscape. GO enrichment analysis was performed with DAVID (https://david.ncifcrf.gov) using Arabidopsis gene IDs.

Yeast one-hybrid assay
A 1238 bp upstream sequence from the first nucleotide of the start codon was selected for the NbXCP1 putative promoter. This sequence was ligated to the pHISi vector. The CDS sequences of Arabidopsis VND1-VND7 were also ligated to the pDEST-GADT7 vector. The two plasmid vectors were co-transformed into yeast cells. Transformants were grown on SD (synthetic defined) medium lacking leucine or histidine as reported previously [49].

Virus-induced gene-silencing (VIGS) experiments
The plasmid vectors CY1, CMV-A1, and CY3 contained tripartite components of cucumber mosaic virus (CMV) genomic RNA (RNA1, RNA2, and RNA3, respectively). CMV-A1-NbXCP1/2 and CMV-A1-GFP (Supplemental Figure S10A) are genetically modified vectors for generating mutated RNA2. Plasmid vectors were linearized to generate templates for viral RNA in vitro transcription (T7 RNA Polymerase System, Takara Bio, Shiga, Japan). RNA1 to RNA3 were equally mixed to infect 3-week-old N. benthamiana plants using the mechanical inoculation method [50]. One week later, newly grown leaves with a mosaic phenotype were used to detect viral infection through RT-PCR. The primers used for PCR amplification are listed in the Supplemental Table S9. Virus-infected leaves were ground homogeneously and used for sub-inoculation on 3-weekold plants following the above experiment steps. These infected plants were then used for interfamily grafting after one week of growth.

Measurement of toluidine blue transport
Interfamily grafts of Nb/At were used to detect toluidine blue appearing on xylem at 1, 2, 3, 7 and 10 DAG. The bottom of Arabidopsis inf lorescence stems was cut and soaked into 0.5% (w/v) toluidine blue solution for 12 hours. Cross sections of scion stems were directly cut by hand with a blade and observed using a microscope (BX53, Olympus, Tokyo, Japan) equipped with a digital camera (DP73, Olympus) for high-magnification images.

Measurement of radioisotope phosphorus-32 ( 32 P) transport
Interfamily grafts of Nb/At were used to detect 32 P absorption 7 DAG. The bottom of Arabidopsis inf lorescence stems was injected with 2 mL of 0.1 μM Pi solution (H 3 32 PO 4 , 10 kBq/mL), followed by plant absorption for 6 h. Plants were transferred to an imaging plate for 2-3 hours of exposure. The radioactivity of 32 P was detected. Photo-stimulated luminescence (PSL) divided by area (PSL/area) indicates the numerical value of the 32 P luminescent signal. The final scion radioactivity value was calculated using the following formula: scion (PSL/area) / rootstock (PSL/area).