VND-INTERACTING2, a NAC Domain Transcription Factor, Negatively Regulates Xylem Vessel Formation in Arabidopsis

The Arabidopsis thaliana NAC domain transcription factor VASCULAR-RELATED NAC-DOMAIN7 (VND7) acts as a master regulator of xylem vessel differentiation. To understand the mechanism by which VND7 regulates xylem vessel differentiation, we used a yeast two-hybrid system to screen for proteins that interact with VND7 and identiﬁed cDNAs encoding two NAC domain proteins, VND-INTERACTING1 (VNI1) and VNI2. Binding assays demonstrated that VNI2 effectively interacts with VND7 and the VND family proteins, VND1-5, as well as with other NAC domain proteins at lower afﬁnity. VNI2 is expressed in both xylem and phloem cells in roots and inﬂorescence stems. The expression of VNI2 overlaps with that of VND7 in elongating vessel precursors in roots. VNI2 contains a predicted PEST motif and a C-terminally truncated VNI2 protein, which lacks part of the PEST motif, is more stable than full-length VNI2. Transient reporter assays showed that VNI2 is a transcriptional repressor and can repress the expression of vessel-speciﬁc genes regulated by VND7. Expression of C-terminally truncated VNI2 under the control of the VND7 promoter inhibited the normal development of xylem vessels in roots and aerial organs. These data suggest that VNI2 regulates xylem cell speciﬁcation as a transcriptional repressor that interacts with VND proteins and possibly also with other NAC domain proteins.


INTRODUCTION
The plant vascular system, transporting water and nutrients and supporting the plant body, consists of the xylem and phloem, which differentiate from the vascular cambium and procambium. The xylem is composed of water-conducting vessel elements, supportive fibers, and xylem parenchyma cells, whereas the phloem comprises sugar-conducting sieve elements, accompanying companion cells, and phloem parenchyma cells.
Several lines of genetic evidence also indicate that fate determination of the xylem and phloem and cell-type specification in each tissue are largely regulated by a number of transcription factors (reviewed in Carlsbecker and Helariutta, 2005;Sieburth and Deyholos, 2006;Demura and Fukuda, 2007;Turner et al., 2007). Analysis of loss-and gain-of-function phenotypes of the class III homeodomain-leucine zipper transcription factor (HD-Zip III) and GARP-type transcription factor KANADI (KAN) family members suggested that xylem/phloem determination is regulated by the antagonistic and complementary activities of the HD-Zip III and KAN gene products (Hawker and Bowman, 2004). A mutation in ALTERED PHLOEM DEVELOPMENT (APL), a MYB coiled-coil transcription factor, resulted in seedling lethality with defects in phloem-specific cell division and in sieve element differentiation in the roots, where ectopic differentiation of vessel elements occurred instead of the differentiation of sieve elements. However, APL overexpression prevents or delays vessel element formation, suggesting that APL is required for sieve element differentiation and to inhibit xylem differentiation (Bonke et al., 2003). Specification into distinctive xylem cells is regulated by several NAM/ATAF/CUC (NAC) domain transcription factors. VASCULAR-RELATED NAC-DOMAIN1 (VND1) to VND7 are preferentially expressed in differentiating xylem vessels (Kubo et al., 2005;Yamaguchi et al., 2008), and the overexpression of VND7 and VND6 can induce the ectopic differentiation of two different types of vessel elements: protoxylem vessels, typically with annular or spiral secondary wall thickenings, and metaxylem vessels, typically with reticulate or pitted secondary wall thickenings (Kubo et al., 2005). The functional suppression of VND7 and VND6 results in the inhibition of vessel element formation (Kubo et al., 2005;Yamaguchi et al., 2008). These results strongly suggest that VND6 and VND7 act as the key regulators of the differentiation of xylem vessels. NAC SECONDARY WALL THICKENING PROMOTING FACTOR1 (NST1) and NST3 (also called SECONDARY WALL-ASSOCIATED NAC DOMAIN PROTEIN1 [SND1] or ARABIDOPSIS NAC ) are expressed in the interfascicular fibers and xylem fibers and function redundantly in regulating secondary wall biosynthesis in fibers (Zhong et al., 2006(Zhong et al., , 2007bKo et al., 2007;Mitsuda et al., 2007). The gene encoding XYLEM NAC DOMAIN1 (XND1) is highly expressed in xylem (Zhao et al., 2005), and its overexpression causes extreme dwarfism associated with the complete suppression of vessel secondary wall biosynthesis and programmed cell death, suggesting that XND1 negatively regulates xylem vessel differentiation (Zhao et al., 2008). SND2, MYB46, MYB103, and KNOTTED1-LIKE HOMEODOMAIN PROTEIN7 transcription factors are direct targets of NST3/SND1/ANAC012 and possibly of NST1, VND6, and VND7 (Zhong et al., 2007a(Zhong et al., , 2008, suggesting the existence of a transcriptional network regulating xylem differentiation. A number of NAC domain transcription factors from several plant species play critical roles in diverse processes in addition to xylem cell specification, such as the establishment of the shoot apical meristem, lateral root formation, flowering time, senescence, abiotic stress responses, and defense responses (Delessert et al., 2005;Uauy et al., 2006;Yoo et al., 2007;Kim et al., 2008;Ogo et al., 2008;reviewed in Olsen et al., 2005). Recent analyses of the NAC genes have suggested that NAC function is regulated at the transcriptional, posttranscriptional, and posttranslational levels (reviewed in Olsen et al., 2005). Their posttranslational regulation via protein degradation may, as suggested by yeast two-hybrid screening, act through proteinprotein interactions (Xie et al., 2002;Greve et al., 2003). VND7 can form homodimers and heterodimers with other VND proteins, and the stability of VND7 is regulated by proteasomemediated degradation (Yamaguchi et al., 2008), suggesting that the transcriptional activity of VND7 is regulated by its interactions with other proteins.
In this study, to understand the mechanism by which xylem vessel differentiation is regulated, we used a yeast two-hybrid system to identify proteins that interact with VND7, which resulted in the isolation of a previously uncharacterized NAC domain transcription factor, VND-INTERACTING2 (VNI2). We show that VNI2 binds with high affinity to VND7 as well as to VND1 to VND5 proteins and with low affinity to some other NAC domain proteins and that VNI2 is expressed in immature xylem cells and vessel precursors. We also show that VNI2 is a transcriptional repressor whose stability is regulated by PEST motif-dependent proteolysis and that it suppresses the capacity of VND7 to activate transcription. Moreover, the misexpression of VNI2 under the control of the VND7 promoter causes strong repression of xylem vessel differentiation. Considered together, our findings indicate that VNI2 regulates xylem cell specification as a transcriptional repressor that interacts with VND proteins and possibly with other NAC domain proteins.

VNI2 Encodes a Novel NAC Domain Protein that Interacts with VND7
The Arabidopsis thaliana NAC domain transcription factor VND7 acts as a master regulator of xylem vessel differentiation (Kubo et al., 2005;Yamaguchi et al., 2008). To further understand the mechanism by which VND7 regulates xylem vessel differentiation, we screened an Arabidopsis root cDNA library for proteins that interact with VND7, using a yeast two-hybrid system ( Figure  1). Because full-length VND7 is a strong transcriptional activator in yeast (Yamaguchi et al., 2008), the N-terminal region of VND7 (amino acid residues 1 to 146), containing consensus subdomains I to IV of the NAC domain, was used as the bait for screening ( Figure 1B). After screening more than 1.5 3 10 6 independent transformants, we identified eight positive clones with the expression of reporter genes ( Figure 1A). Six and two cDNAs rescued from these clones encoded two distinct NAC domain proteins, ANAC082 and ANAC083, which were designated VNI1 and VNI2, respectively (Figure 1; see Supplemental Figure 1A online). All the rescued VNI1 cDNAs lacked a part of the NAC domain but contained the putative C-terminal transcriptional activation domain, which included a possible PEST proteolysis target motif (PEST score of +7.17; Rechsteiner and Rogers, 1996) and nine tandemly repeated sequences of 13 amino acids (see Supplemental Figure 1B online). Both VNI2 cDNAs included NAC subdomain V and the C-terminal transcriptional activation domain (amino acid residues 147 to 252), containing a possible PEST motif (amino acid residues 213 to 226, TTDLNLLPSSPSSD; PEST score of +4.87; Rechsteiner and Rogers, 1996; Figure 1B). Because our previous transcriptome data showed that VNI2 expression is significantly upregulated during in vitro vessel element transdifferentiation in Arabidopsis suspension cells ( Figure 1D; Kubo et al., 2005), whereas VNI1 expression was not related to vessel element differentiation (see Supplemental Figures 1C and 1D online), we decided to focus on the functional analysis of VNI2 in this study.

VNI2 Interacts with VND proteins in Vitro
To confirm the interaction between VND7 and VNI2, we performed in vitro binding assays with poly-His-tagged VNI2 (His-VNI2) expressed in Escherichia coli and immobilized on Ni-conjugated agarose resin, and with several NAC domain proteins, including VND7, prepared in a rabbit reticulocyte lysate system in the presence of [ 35 S]Met. Proteins bound to the resin were separated by SDS-PAGE and detected by autoradiography; a mock-treated control lacking His-VNI2 showed no [ 35 S] Met-labeled proteins ( Figure 2A). Although all the NAC domain proteins bound to the His-VNI2 resin to some extent, VND7, VND1, VND2, and VND3 bound more efficiently than other NAC proteins, including VND6 and NAC1 (Xie et al., 2000), CUP-SHAPED COTYLEDON2 (CUC2; Aida et al., 1997), and VNI2 ( Figure 2A). Because VND4 and VND5 bound nonspecifically to the Ni-conjugated agarose resin, the binding capacity of VND4, VND5, and VND7 was confirmed with VNI2 fused to maltose binding protein (MBP-VNI2) immobilized on amylose resin (see Supplemental Figure 2 online). We further tested the binding specificity between VND7 and VNI2 using MBP fused to ANAC041, which has the highest sequence similarity to VNI2 ( Figure 1C; see Supplemental Data Set 1 online), as a reference, and found that VND7 selectively bound to VNI2 rather than to ANAC041 ( Figure 2B). These results demonstrate that VND7 and other VND proteins, VND1 to VND5, preferentially bind to VNI2 with moderate specificity. (A) Interaction between VNI2 and VND7 in yeast cells. Yeast cells carrying plasmids containing amino acid residues 1 to 146 of VND7 fused to GAL4-BD (BD-VND7 1-146 ) and amino acid residues 147 to 221 of VNI2 fused to GAL4-AD (#210), which was isolated by yeast two-hybrid screening, grew on selective medium lacking Trp, Leu, and His. Plasmids containing a multicloning site (MCS) fused to GAL4-BD or GAL4-AD were used as the negative controls, and pBD-wt and pAD-wt (Stratagene) were used as the positive controls. (B) Schematic structure of the VND7 and VNI2 proteins. The NAC domain, which contains subdomains I to V, is conserved at the N termini of the VND7 and VNI2 proteins. LP and WQ represent the LP and WQ boxes conserved in the IIb/OsNAC7 subgroup (Ko et al., 2007). VNI2 has a predicted PEST motif in the C-terminal region. Amino acid residues 1 to 146 of VND7, used as the bait, and residues 147 to 221 of VNI2, obtained by screening, are shown below each protein.
(C) Phylogenetic analysis of VNI1, VNI2, and other previously characterized Arabidopsis NAC domain proteins. (D) Expression patterns of VND7 and VNI2 genes during in vitro vessel element transdifferentiation in Arabidopsis suspension cells, as determined by microarray analysis (Kubo et al., 2005). In the aerial parts of the plant, VNI2 was also preferentially expressed in vascular cells, including the xylem parenchyma cells surrounding the mature vessel elements and several phloem cells in the inflorescence stems, and in guard cells in the leaves ( Figures 3E to 3G). VND7 was also expressed in vascular cells in Arabidopsis seedlings (Yamaguchi et al., 2008). To examine the cells in which the expression of VNI2 and VND7 overlapped, we generated transgenic plants carrying These data suggest that VNI2 expression precedes VND7 expression during xylem vessel differentiation. Expression of VNI2 and VND7 was also evident in the nonvessel cells of the root central cylinder ( Figure 3A; Yamaguchi et al., 2008) and in the xylem parenchyma cells as well as differentiating xylem cells in the inflorescence stem ( Figure 3G; see Supplemental Figure 5 online), suggesting a cooperative function of VND7 and VNI2 even in these nonvessel cells.

VNI2 Stability Is Regulated by the PEST Motif
Transgenic Arabidopsis plants expressing full-length VNI2 fused to a GUS reporter gene (VNI2 full -GUS) under the control of VNI2 pro were prepared to localize VNI2 expression more precisely. GUS staining was faint in the roots ( Figure 4A) and hardly detectable in the shoots ( Figure 4C), which differed from the GUS staining pattern in the VNI2 pro :GUS seedlings (Figure 3), suggesting the degradation of the VNI2-GUS fusion protein in the VNI2 pro :VNI2 full -GUS seedlings. Because VNI2 has a possible PEST motif in the C-terminal region (Figure 1), we investigated whether this motif plays a role in the degradation of VNI2-GUS. We generated transgenic Arabidopsis plants expressing truncated VNI2 (amino acid residues 1 to 221), which lacked a portion of the C-terminal region containing about half the PEST motif sequence ( Figure 1B) and was fused to the GUS reporter gene under the control of VNI2 pro (VNI2 pro :VNI2 1-221 -GUS). All the VNI2 pro :VNI2 1-221 -GUS seedlings analyzed showed stronger GUS staining in the roots ( Figure 4B) and visible GUS staining in the shoots, including in the shoot apical region ( Figure 4D), compared with that in the VNI2 pro :VNI2 full -GUS seedlings. Consistent with microscopy observations, GUS activity was higher in

of 15
The Plant Cell the VNI2 pro :VNI2 1-221 -GUS plants than in the VNI2 pro :VNI2 full -GUS plants ( Figure 4E). Furthermore, the VNI2 pro :VNI2 full -GUS seedlings treated with MG-132, a proteasome inhibitor, displayed higher GUS activity than seedlings treated with DMSO ( Figure 4F). These data suggest that the stability of VNI2 is regulated by proteasome-mediated proteolysis and that the C-terminal region of the VNI2 protein, containing a possible PEST motif, is involved in maintaining the stability of VNI2.

VNI2 Is a Transcriptional Repressor
Most NAC domain proteins have been characterized as transcriptional activators (reviewed in Olsen et al., 2005). To examine  the transcriptional function of VNI2, we used transient reporter assays ( Figure 5). A reporter plasmid containing firefly luc linked to GAL4 binding sites, and effector plasmids containing VND7, VNI2 full , or VNI2 1-221 fused to GAL4-BD under the control of the cauliflower mosaic virus (CaMV) 35S promoter (35S pro ), were delivered into Arabidopsis leaves by particle bombardment ( Figure 5A). Although an increase of more than fourfold in Luc activity was induced by the expression of GAL4-BD-VND7 compared with that induced by the GAL4-BD control, as reported previously (Yamaguchi et al., 2008), GAL4-BD-VNI2 full did not induce this increase in Luc activity ( Figure 5B), suggesting that VNI2 is not a transcriptional activator. However, it remained possible that the unstable nature of VNI2 led to the low level of Luc activity. Therefore, we examined Luc activity after the expression of the possibly more stable derivative, GAL4-BD-VNI2 1-221 ; however, this did not enhance Luc activity ( Figure 5B). These data prompted us to examine whether VNI2 has transcriptional repression activity. We used luc under the control of 35S pro , with GAL4 binding sites to increase the basal Luc activity (Mitsuda et al., 2005), as the reporter plasmid ( Figure 5A), which was delivered into Arabidopsis leaves with the effector plasmids. As expected, GAL4-BD fused to the SRDX strong repression domain (Hiratsu et al., 2002) severely reduced the relative Luc activity compared with that of the GAL4-BD control (to <10% of the control), and GAL4-BD-VNI2 full and -VNI2 1-221 significantly reduced the Luc activity (to ;40 to 60% of the control; Figure 5C). Taking these results together, we conclude that VNI2 is a transcriptional repressor and that the potential repression-related sequence of VNI2 is located between amino acid residues 1 and 221, upstream of the possible PEST motif.

Overexpression of VNI2 Causes Defects in Vessel Formation
To examine whether VNI2 is involved in the development of xylem tissues, we generated transgenic Arabidopsis plants overexpressing VNI2 or VNI2-SRDX under the control of 35S pro . 35S pro :VNI2 seedlings of the T1 generation showed no clear differences from the control seedlings in appearance, whereas overexpression of VNI2-SRDX retarded the growth of the seedlings (10 of 42 lines) ( Figure 6A). However, microscopy observations revealed defects in root vessel formation in the 35S pro :VNI2 seedlings (13 of 31 lines) ( Figures 6B to 6D). The control roots had two protoxylem vessels with helical secondary cell walls and three to four metaxylem vessels with reticulated secondary cell walls located between the protoxylem vessels in the mature region of the primary roots ( Figure 6B). By contrast, the 35S pro :VNI2 seedlings exhibited discontinuous vessel formation of one or both protoxylem vessel columns ( Figures 6C  and 6D). The same defect in vessel formation was observed when VNI2-SRDX was overexpressed (17 of 38 lines) ( Figure 6E), indicating that the defects in protoxylem vessel formation in the 35S pro :VNI2 seedlings were the result of the transcriptional repression activity of VNI2. Similar inhibition of vessel formation was previously shown by the overexpression of VND7-SRDX (Kubo et al., 2005) or C-terminally truncated VND7 (Yamaguchi et al., 2008), suggesting a functional interaction between VNI2 and VND7. . The reporter gene activity was normalized to the activity of Renilla luciferase, the gene that was cobombarded as a reference. Error bars indicate SD (n = 3 or 6). Asterisks indicate statistically significant differences (P < 0.05) relative to GAL4-BD.

of 15 The Plant Cell
To further substantiate this functional interaction, we generated transgenic Arabidopsis plants expressing VNI2 under the control of VND7 pro , which extends the VNI2 expression to the differentiating vessel elements at later stages. Among 28 firstgeneration transgenic plants carrying VND7 pro :VNI2 full , nine grew more slowly than control plants carrying VNI2 pro :VNI2 full . These slow-growing plants had an apparent defect in vessels, exhibiting discontinuous vessel formation in root proto-and metaxylem (Figures 7A to 7C; see Supplemental Figure 6 online). Because C-terminally truncated VNI2 (VNI2 1-221 ) is more stable than fulllength VNI2, it is likely that the expression of truncated VNI2 effectively causes the vessel defect. Therefore, we also generated transgenic Arabidopsis plants expressing VNI2 1-221 under VNI2 pro or VND7 pro . Although 30-d-old seedlings of the VNI2 pro : VNI2 1-221 plants (seven of 17 lines) showed slightly reduced growth, they grew and bolted normally thereafter ( Figure 7A; see Supplemental Figure 6 online). The vessel defect was enhanced markedly in VND7 pro :VNI2 1-221 plants (17 of 38 lines), which often showed a severe dwarf phenotype with downward curled, darkgreen rosette leaves and survived without bolting ( Figure 7A; see Supplemental Figure 6 online). The dwarf seedlings also showed discontinuous vessel formation in their aerial parts ( Figure 7E) as well as inhibited xylem vessel formation in their roots. These results demonstrate that the misexpression of VNI2 (i.e., coexpression with VND7 and the stabilization of VNI2 protein) leads to striking abnormalities in vessel formation.

A Number of Vessel-Specific Genes Are Regulated Downstream of VND7
It is plausible that VNI2 binds to VND7 and/or other VND family proteins and represses their transcriptional activation of downstream vessel-related genes. We previously demonstrated using promoter analysis that the expression of three immature vesselspecific genes, At Xyn3, CesA7/IRX3, and LAC4/IRX12, are upregulated downstream of VND7 (Kubo et al., 2005). To identify additional genes downstream of VND7, we performed a microarray experiment with the GeneChip ATH1 Arabidopsis genome array (Affymetrix, http://www.affymetrix.com/index.affx) in transgenic Arabidopsis roots overexpressing YFP and VND7-YFP under the control of 35S pro . Data analysis, including statistical Student's t test (P < 0.05) and false discovery rate (Q < 0.05), revealed 89 and 43 genes that were up-and downregulated at least fourfold, respectively, by VND7-YFP overexpression (see Supplemental Data Set 2 online). Forty-one of the upregulated genes, including VND7, were transiently expressed during in vitro vessel element transdifferentiation in Arabidopsis suspension cells (Table 1; Kubo et al., 2005). In addition to VND7 itself, two other transcription factors were included among the 41 genes ( Table  1). Expression of LOB domain protein30 (LBD30)/ASYMMETRIC LEAVES2 LIKE19 (ASL19)/JAGGED LATERAL ORGANS (JLO) is regulated by VND6 and VND7, and LBD30/ASL19/JLO seems to be involved in a positive feedback loop for VND expression (Soyano et al., 2008). MYB46 is a direct target of  (Table 1). Next, we checked for changes in the expression levels of previously described vascular-related genes (see Supplemental Data Set 3 online). Although no substantial changes were observed for genes associated with procambium formation or vascular pattern formation (such as the HD-Zip III and KAN families, the brassinosteroid receptors, or xylogens; Fukuda, 2004), the expression of a number of genes responsible for IRX mutations (IRX1, -3, -5, -6, -8, -9, -10, -12, and -13) and known vesselspecific genes (encoding a glycosyl hydrolase family 10 protein At Xyn3; aspartyl protease; XCP1 and XCP2; subtilisin-type serine protease XSP1; and At Prx66) were significantly upregulated (at least twofold; P < 0.05; Q < 0.05) by VND7 overexpression.
We confirmed by quantitative RT-PCR that expression levels of VND7, LBD30/ASL19/JLO, MYB46, and XCP1 were higher in the 35S pro :VND7-YFP plants than in the 35S pro :YFP plants, which was consistent with the GeneChip analysis data (see Supplemental Figure 8A online). We next investigated whether VND7 can activate the transcription of these vessel-specific genes in a transient Luc reporter assay system. LBD30/ASL19/JLO, MYB46, and XCP1 were selected for the assay, and the promoter regions of these genes were linked to luc as reporter plasmids ( Figure 8A), which were delivered into Arabidopsis leaves with an effector plasmid containing 35S pro linked to VND7 cDNA ( Figure  8A). We demonstrated that VND7 effectively transactivated luc transcription via these three promoters ( Figure 8B).

VNI2 Represses the Transcriptional Activation Activity of VND7
To investigate whether VNI2 is sufficient to repress the expression of vessel-specific genes that are transactivated by VND7, we generated effector plasmids containing 35S pro linked to  The loss of VNI2 function was investigated with a T-DNA insertion line for VNI2 (vni2-1, SALK_143793: T-DNA insertion immediately upstream of the coding region), and the level of VNI2 transcripts was significantly reduced (see Supplemental Figure 7 online). Although the vni2-1 mutant showed no visibly aberrant phenotype in terms of xylem vessel formation in roots, hypocotyls, and inflorescence stems, the expression levels of LBD30/ ASL19/JLO and XCP1 in vni2-1 plants were increased compared with those of the wild type (see Supplemental Figure 7 online). These data suggested that VNI2 negatively regulates the expression of genes involved in xylem vessel formation.

VNI2 Is a Transcriptional Repressor
Most of the NAC domain transcription factors previously characterized are described as transcriptional activators based on assays in yeast and plant cells. However, a CALMODULIN BINDING NAC PROTEIN (CBNAC) was shown to act as a transcriptional repressor . Transcriptional repressors can be classified as passive or active repressors (reviewed in Thiel et al., 2004;Kazan, 2006). Our experiments suggested that VNI2 has features of both of these types of transcriptional repressor.
Passive repressors repress transcription either by competing with transcriptional activators for DNA binding, by interacting with transcriptional activators to form inactive heterodimers, or by binding to transcriptional coactivators to render transcriptional complexes inactive. The overexpression of C-terminally truncated VND7 protein, which lacks most of the transcriptional activation domain, exerts a dominant-negative effect on vessel formation, probably by acting as an artificial passive repressor and forming inactive heterodimers (Yamaguchi et al., 2008). Because the misexpression of VNI2 phenocopies the overexpression of C-terminally truncated VND7 (Figures 6 and 7), and VNI2 binds to VND proteins in vitro and possibly forms heterodimers in vivo (Figures 1 and 2  or XCP1 with a minimal CaMV 35S promoter (min pro) upstream from the firefly luciferase reporter gene. The effector constructs contained VND7 or full-length or C-terminally truncated VNI2 downstream from the CaMV 35S promoter. (B) Relative luciferase activities after cobombardment of one of the reporter constructs with the effector constructs. The reporter gene activity was normalized to the activity of Renilla luciferase, which was cobombarded as a reference. Error bars indicate SD (n = 3). Asterisks indicate statistically significant differences (P < 0.05) from values for the VND7 effector construct only.

of 15
The Plant Cell (Kazan, 2006). Our transient reporter assays showed that fulllength and C-terminally truncated VNI2 proteins significantly reduced luc reporter expression ( Figure 5C), suggesting that VNI2 might function as an active repressor, independent of VND7 binding. It is noteworthy that we found a sequence similar to the EAR motif ( L / F DLN L / F [x]P), which is conserved in class II ERFs and TFIIIA-type zinc finger proteins (Ohta et al., 2001), within the possible PEST motif (TTDLNLLPSSPSSD; the EAR motif-like sequence is underlined) in the C-terminal region of VNI2 ( Figure  1B). The functional significance of the EAR motif-like sequence in VNI2 is an important issue to be clarified in future research.

Degradation of VNI2
Plants use the degradation of positive and negative regulators to optimize the signaling pathways for growth and development (reviewed in Huq, 2006). The activity of several NAC domain proteins, including NAC1 and ANAC019, is known to be regulated by ubiquitin-dependent protein degradation (reviewed in Olsen et al., 2005). We have shown that the stability of VNI2 is increased by treatment with the proteasome inhibitor MG-132 (Figure 4), and VNI2 contains a predicted PEST motif ( Figure  1B), which is known to promote rapid protein degradation (Rechsteiner and Rogers, 1996). The C-terminally truncated VNI2 protein, which lacks part of the PEST motif, is more stable than full-length VNI2 ( Figure 4) and repressed vessel element differentiation as effectively as did full-length VNI2 (Figures 7 and  8). We also recently demonstrated that the stability of VND7 can be regulated by protein degradation, probably mediated by proteasomes (Yamaguchi et al., 2008). Thus, the sophisticated control of vessel differentiation may require the timely degradation of these negative and positive regulators.

Function of VNI2 in Vascular Development
We have shown that overexpression of VNI2 significantly inhibits normal vessel element differentiation (Figures 6 and 7). These defects are also observed in plants expressing dominant-negative forms of VND7, although loss of function of VND7 does not alter xylem development (Kubo et al., 2005;Yamaguchi et al., 2008). VND7 can form homodimers and heterodimers with other VND proteins (Yamaguchi et al., 2008), suggesting that the dominantnegative forms of VND7 affect not only VND7 itself but also other VNDs. Because VNI2 effectively binds to VND1 to VND5 proteins as well as to VND7 (Figure 2), it can regulate multiple VND proteins during xylem vessel differentiation. Initiation of VNI2 expression precedes that of VND7 expression in roots (Figure 3; see Supplemental Figure 4A online). We previously reported that VND1 to VND3 are expressed preferentially in procambial cells adjacent to root meristem (Kubo et al., 2005). Therefore, in early stages of xylem vessel differentiation before the onset of VND7 expression, VNI2 may bind to VND1 to VND3 and repress vesselspecific genes. Moreover, an in vitro binding assay suggests that VNI2 could bind not only to VND proteins but also to other NAC domain proteins (Figure 2). Therefore, it is possible that VNI2 may also regulate transcriptional activities of some other NAC domain proteins during xylem vessel differentiation.
Loss of function of VNI2 increases the expression of genes involved in xylem vessel formation, although the mutant does not show defects in xylem development (see Supplemental Figure 7 online). ANAC041, which has the highest sequence similarity to VNI2 ( Figure 1C), does not bind as effectively as VNI2 to VND7 ( Figure 2B). The T-DNA insertion mutation in ANAC041 does not affect the expression of vessel-specific genes regulated by VND7, in contrast with the vni2 mutant (see Supplemental Figure  7 online). These data suggest that ANAC041 does not function redundantly with VNI2 in xylem vessel differentiation. Recently, the xylem-specific NAC domain protein XND1 was shown to negatively regulate xylem vessel differentiation by suppressing the expression of vessel-related genes (Zhao et al., 2008). Although xnd1 mutants show slightly shorter xylem vessels than the wild type, patterned secondary wall deposition and autolysis in xylem vessels of the mutants occurred normally (Zhao et al., 2008), suggesting the existence of several distinct mechanisms to repress the expression of vessel-related genes, which could in turn account for why single mutations of negative regulators show only a weak phenotype.
VNI2 expression was observed in the xylem parenchyma cells surrounding mature vessel elements in inflorescence stems ( Figure 3G), in which VND7 is also expressed strongly (see Supplemental Figure 5 online), suggesting a role for VNI2 in repressing VND7 activity in these xylem parenchyma cells. The interruption of vascular bundles induces vascular regeneration (reviewed in Aloni, 1995). When the vascular bundles in the internodes of zinnia were severed, xylem parenchyma cells readily transdifferentiated into vessel elements (Nishitani et al., 2002). It is conceivable that the rapid disappearance of VNI2's repression activity against VND7 in these xylem parenchyma cells after vascular disruption ensures the quick transdifferentiation of the vessel elements.
VNI2, but not VND, is expressed in the phloem cells in roots and stems (Figure 3; see Supplemental Figure 3 online). In roots, the expression of VNI2 partially overlaps that of APL, which promotes and inhibits phloem and xylem differentiation, respectively (Bonke et al., 2003), suggesting a cooperative function for APL and VNI2 in inhibiting vessel element differentiation in phloem. Furthermore, VNI2 is expressed in locations as widely separated as the shoot apical region and the guard cells ( Figure  3). VNI2 may thus have pleiotropic functions in these regions or cells that are independent of its role in xylem vessel formation.

NAC Transcriptional Network Regulating Vessel Differentiation
In animals, the basic helix-loop-helix leucine zipper domain (bHLHZip) transcription factor family proteins Max, c-Myc, and Mad constitute a transcription network involved in cell proliferation, differentiation, and apoptosis by forming homodimers and heterodimers in response to upstream stimuli: c-Myc/Max and Mad/Max heterodimers act as transcriptional activators and repressors, respectively (reviewed in Lü scher, 2001;Amoutzias et al., 2008). The loss of Arabidopsis VND7 function produces no detectable defects in morphology (Kubo et al., 2005), and VND7 can form either homodimers or heterodimers with other VND members (Yamaguchi et al., 2008), suggesting that VND proteins function collaboratively and/or redundantly in xylem vessel differentiation. In this study, we also found that VNI2, a previously uncharacterized NAC domain transcriptional repressor, physically interacts with VND proteins. Therefore, VND proteins may well regulate xylem vessel differentiation by forming active or repressive complexes, similar to the c-Myc/Max/Mad bHLHZip transcriptional network. Moreover, as described above, XND1 was also shown to inhibit xylem vessel differentiation (Zhao et al., 2008). It is possible that VNDs, VNI2, and XND1 constitute a NAC transcriptional network that fine-tunes vessel formation-related gene expression.
Here, we identified a NAC domain protein, VNI2, which represses VND function at the early stage of vessel differentiation in roots and in the xylem parenchyma cells surrounding the mature vessels in stems (Figure 9). The differential expression pattern of VNDs and VNI2 seems to be regulated by different developmental signals (Figure 9). Because the CLE peptide TDIF is a strong suppressor of vessel element differentiation (Ito et al., 2006), it will be interesting to establish whether VNI2 functions downstream of TDIF. Further analysis of VNI2, including its misexpression under the control of promoters for cell-specific genes such as APL, its expression during vascular regeneration, the identification of its direct targets and DNA binding sequences, and the isolation of proteins with which it interacts, will undoubtedly shed more light on the biological roles of VNI2 in vascular development.
We also proposed a novel regulatory mechanism: some NAC domain proteins interact with other NAC domain proteins to impede their functions. Such NAC-NAC interactions may play important roles in various events during plant development.

Vector Construction
To generate the Gateway destination vectors for the expression of MBP fusion proteins in Escherichia coli and for the expression of proteins under the control of their own promoters in plants, pMAL-c2X (New England Biolabs) and pSMAB704 (Kubo et al., 2005) were digested with EcoRI and HindIII/SacI, respectively, and blunted using the BKL kit (Takara Bio). These linearized fragments were ligated to EcoRV-digested Gateway Reading Frame Cassette (GWRFC) B (Invitrogen). The resultant plasmids were designated pMAL-GWRFC and pBG, respectively. The promoter fragments and cDNAs of Arabidopsis thaliana genes were amplified from genomic DNA and cDNA pools of the Columbia-0 (Col-0) ecotype by PCR with gene-specific primer sets. These fragments were subcloned into the pENTR/D-TOPO vector (Invitrogen) and then integrated into Gateway destination vectors for transformation of plants (Kubo et al., 2005), for two-hybrid assays (Yamaguchi et al., 2008), and for expression in E. coli (pDEST14 and pDEST17; Invitrogen) using LR Clonase (Invitrogen). The GATEWAY destination vectors containing the nucleotide sequence of the MCS fragments (Yamaguchi et al., 2008) were used as controls. For the dual luciferase transient transfection assay, p35S-GAL4-BD, the effector plasmid, and the reporter and reference plasmids containing firefly luc and Renilla reniformis luc, respectively, were prepared as described previously . For the effector plasmid, p35SG, the GUS gene in pBI221 (Clontech) was removed and the tobacco mosaic virus sequence (Gallie et al., 1987) was inserted downstream from the CaMV 35S promoter (Fujimoto et al., 2000). VND7 and VNI2 cDNA fragments were inserted into the SmaI/SalI sites of the p35S-GAL4-DB and p35SG effector plasmids. The promoter fragments of LBD30/ASL19/ JLO, MYB46, and XCP1 were inserted into the HindIII-blunted/XbaI sites of the GAL4UAS:TATA:LUC reporter plasmid.

Yeast Two-Hybrid Screening
Total RNA was extracted from 4-d-old Arabidopsis (Col-0) primary roots, and poly(A) + RNA was isolated with the GenElute mRNA Miniprep Kit (Sigma-Aldrich) according to the manufacturer's instructions. A cDNA library for two-hybrid screening was constructed with the HybriZAP-2.1 XR library construction kit (Stratagene). A budding yeast (Saccharomyces cerevisiae) strain AH109 (Clontech) was transformed sequentially with pGAL4-BD-VND7 1-143 , which was constructed from the pGAL4BD-GWRFC vector by inserting the cDNA encoding the first 143 amino acids of VND7 by LR reaction (Yamaguchi et al., 2008) and then with the Arabidopsis root cDNA library. The plasmids from the positive yeast cells were isolated with the Yeastmaker yeast plasmid isolation kit (Clontech) and amplified in E. coli.

Phylogenetic Analysis
The deduced protein sequences of 30 NAC domain proteins were aligned with the MAFFT version 6.531b program (http://align.bmr.kyushu-u.ac.jp/ mafft/online/server/) using the L-INS-i strategy. The alignment was manually optimized with the MacClade 4.08 program (Sinauer Associates). The optimized alignment around the NAC domain (110 amino acids in length; see Supplemental Data Set 1 online) was used to produce a phylogenetic tree with the MAFFT version 6.531b program (neighborjoining method; JTT substitution model; ignoring heterogeneity among sites; 1000 bootstrap resampling). The tree was drawn with the FigTree version 1.1.2 program (http://tree.bio.ed.ac.uk/software/figtree/).

In Vitro Pull-Down Assay
VNI2 or ANAC041 cDNAs were integrated into the pDEST17 vector (Invitrogen) and pMAL-GWRFC vector for poly-His and MBP tags, respectively. The His-VNI2 fusion protein was expressed in E. coli BL21-AI (Invitrogen) grown at 308C in the presence of 0.2% (w/v) L-arabinose and then purified with Ni-agarose (Qiagen). The MBP-VNI2 and MBP-ANAC041 fusion proteins were expressed in E. coli BL21trxB (DE3; Cosmo Bio) in the presence of 0.1 mM isopropyl b-D-thiogalactoside and purified with amylase resin (New England Biolabs). VND and NAC cDNAs integrated into the pDEST14 vector (Invitrogen) were used for in vitro transcription and translation with the TNT Coupled Reticulocyte System (Promega) containing [ 35 S]Met according to the manufacturer's recommended method. The poly-His or MBP fusion proteins were immobilized 12 of 15 The Plant Cell with Ni-agarose or amylase resin, respectively, and then the affinity resins were incubated with the translated products for 90 min at 48C. The proteins retained on the resins were separated by SDS-PAGE.

Dual Luciferase Transient Transfection Assay
The effector, reporter, and reference plasmids were delivered into the rosette leaves of 4-week-old Arabidopsis plants by particle bombardment (GE Healthcare), and Luc activity was assayed with the Dual-Luciferase Reporter Assay System (Promega) using the Mithras LB940 (Berthold).

Plant Transformation
The resultant plasmids were electroporated into Agrobacterium tumefaciens strain GV3101/pMP90, which was used to transform Arabidopsis ecotype Col-0. The T1 generation of transgenic seedlings was selected on germination medium (Kubo et al., 2005)

Fluorometric Analysis of GUS Activity
VNI2 pro :VNI2-GUS transgenic plants (T3 generations) were grown for 7 or 8 d on the GM plate. The VNI2 pro :VNI2 full -GUS plants were soaked with the sterilized water with or without 100 mM MG-132 (Wako Pure Chemical) for 6 h. Total protein was extracted in the extraction buffer (Yoshizumi et al., 1999). A 50-mL aliquot of extract was mixed with an equal volume of 1 mM 4-methylumbelliferyl-b-D-glucuronide in the extraction buffer and incubated for 1 h at 378C. A 10-mL aliquot of each reaction was mixed with 1 mL of 0.2 M sodium carbonate, and the fluorescence was quantified by a Safire spectrophotometer (Tecan).

Expression Analysis
Total RNA from wild-type plants, transgenic plants, or T-DNA insertion lines was extracted using the Nucleospin RNA Plant System (Macherey-Nagel). cDNA was synthesized from the RNA with reverse transcriptase (SuperScript III; Invitrogen) and oligo(dT) primer and amplified by PCR with ExTaq DNA polymerase (Takara Bio) and primers specific for the target genes. Quantitative RT-PCR was performed in an ABI Prism 7000 sequence detection system (Applied Biosystems) using a QuantiTect SYBR Green PCR Kit (Qiagen). Primer sets used in the expression analysis are described in Supplemental Table 1 online. Microarray analysis was performed using GeneChip ATH1 Arabidopsis genome arrays (Affymetrix) of three independent biological replicates from the roots of 5-d-old 35S pro :YFP and 35S pro :VND7-YFP seedlings (T3 generations), as previously described (Kubo et al., 2005). GeneSpring GX 7.3 (Agilent Technologies) was used for the data analysis. P values for differences between 35S pro :YFP and 35S pro :VND7-YFP plants were calculated by Welch's t test, based on a two-tailed distribution (n = 3).
We calculated a Q value to estimate the false discovery rate from the P value described above using QVALUE software (Storey and Tibshirani, 2003) with the default setting. We considered genes with a Q value of <0.05 to be genes expressed at different levels in 35S pro :YFP and 35S pro : VND7-YFP plants.

In Situ Hybridization Analysis
Root explants derived from 6-d-old seedlings of Arabidopsis ecotype Col-0 were fixed in PBS containing 4% (w/v) paraformaldehyde. Paraffin sections were made from the fixed samples and hybridized with digoxigenin-labeled probes, followed by detection of signals using alkaline phosphatase-conjugated antidigoxigenin antibody, as described previously (Ohtani et al., 2008). Probes were prepared from cloned DNA sequences encoding the C-terminal regions of VNI2 (for primer information, see Supplemental Table 1 online).

Supplemental Data
The following materials are available in the online version of this article.