ABSTRACT
Development of plant vascular tissues involves tissue identity specification, growth, pattern formation and cell-type differentiation. Although later developmental steps are understood in some detail, it is still largely unknown how the tissue is initially specified. We used the early Arabidopsis embryo as a simple model to study this process. Using a large collection of marker genes, we found that vascular identity was specified in the 16-cell embryo. After a transient precursor state, however, there was no persistent uniform tissue identity. Auxin is intimately connected to vascular tissue development. We found that, although an AUXIN RESPONSE FACTOR5/MONOPTEROS (ARF5/MP)-dependent auxin response was required, it was not sufficient for tissue specification. We therefore used a large-scale enhanced yeast one-hybrid assay to identify potential regulators of vascular identity. Network and functional analysis of candidate regulators suggest that vascular identity is under robust, complex control. We found that one candidate regulator, the G-class bZIP transcription factor GBF2, can modulate vascular gene expression by tuning MP output through direct interaction. Our work uncovers components of a gene regulatory network that controls the initial specification of vascular tissue identity.
INTRODUCTION
Vascular tissues play a central role in plant growth and development by providing plants with transport capabilities and structural support. The various steps of vascular tissue development have been studied in some detail, mainly in the Arabidopsis leaf (Donner et al., 2009; Gardiner et al., 2011; Krogan et al., 2012), shoot (Etchells et al., 2013; Hirakawa et al., 2010; Smetana et al., 2019; McConnell et al., 2001; Han et al., 2018) and root (Scheres et al., 1995; Miyashima et al., 2019; De Rybel et al., 2014). From this wealth of studies, a picture emerges in which dedicated regulatory modules function to create a properly sized and patterned transport bundle. Several steps can be recognized in this process: specification of vascular tissue identity, cell proliferation to generate a bundle of cells, patterning into xylem, phloem and cambium cell types, and finally differentiation into functional transport cells. The regulators and effectors of all but the first step have been dissected in some detail.
The rate of proliferation by periclinal cell divisions determines the width of a vascular bundle. Periclinal cell divisions in the vascular cells are controlled by several pathways: one directed by the xylem-expressed TARGET OF MONOPTEROS 5 (TMO5)/LONESOME HIGHWAY (LHW) dimer (De Rybel et al., 2013, 2014; Ohashi-Ito et al., 2014, 2013), another regulated by the phloem-expressed PHLOEM EARLY DOFs (PEARs) (Miyashima et al., 2019) and a third depending on the activity of WUSCHEL-LIKE HOMEOBOX 4 and 14 in the cambium (WOX4/14) (Etchells et al., 2013; Fisher and Turner, 2007; Hirakawa et al., 2010; Smit et al., 2020). In concert with proliferation, cells in the vascular bundle develop a pattern of distinct sub-identities. Xylem development is associated with high auxin signaling, and further specification of protoxylem or metaxylem identity depends on a combination of the cytokinin response and the activity of HD-ZIP III transcription factors (Baima et al., 2001; Bishopp et al., 2011; Carlsbecker et al., 2010; Mähönen et al., 2006; McConnell et al., 2001). Conversely, the specification of phloem identity is associated with high cytokinin activity and the presence of, among others, ALTERED PHLOEM DEVELOPMENT (APL) (Bonke et al., 2003). Located between the phloem and xylem, the meristem-like (pro)cambium has been shown to contribute to both the xylem and the phloem cell populations (Smetana et al., 2019).
Finally, several vascular cell types undergo irreversible terminal differentiation. The differentiation of xylem vessel elements can be triggered when a gene regulatory network under the control of VASCULAR-RELATED NAC-DOMAIN6 (VND6) and VND7 is initiated (Kubo et al., 2005; Yamaguchi et al., 2010; McCarthy et al., 2009). Although no differentiation-inducing factor has yet been found to trigger phloem-like differentiation (Blob et al., 2018), several factors necessary for phloem differentiation have been identified (Ruiz Sola et al., 2017; Rodriguez-Villalon et al., 2014; Wallner et al., 2017).
Most of the studied regulators of vascular development are expressed only, or preferentially, in vascular cells, which suggests the existence of a robust genetic identity. However, how this vascular tissue identity is initially specified has so far remained elusive. De novo vascular identity specification occurs repeatedly during the plant life cycle as new organs develop or when tissues are wounded (León et al., 2001; Melnyk et al., 2015; Yin et al., 2012). Specification of tissue identities involves the local accumulation of a signaling molecule (small molecule, peptide or protein) that will either promote or suppress the activation of a cell type-specific gene regulatory network. Such mechanisms have been described in non-hair versus hair cells in the root (Lee and Schiefelbein, 1999; Bernhardt et al., 2005), in meristemoids versus stomatal-lineage ground cells in the stomatal lineage (Zhang et al., 2016; Yang et al., 2015; Lau et al., 2014) and in xylem versus phloem cells in the vascular bundle (Smetana et al., 2019; Mähönen et al., 2000; Baima et al., 2001).
A signaling molecule that is strongly correlated with vascular development is auxin. Vascular tissue formation can be triggered by treatments with auxin (Sachs, 1981; Wetmore and Rier, 1963) and, as a result, auxin maxima are often associated with vascular development (Brackmann et al., 2018; Mattsson et al., 2003; Miyashima et al., 2019; Scarpella et al., 2006; Wabnik et al., 2013). Conversely, lack of the AUXIN RESPONSE FACTOR5/MONOPTEROS (ARF5/MP) transcription factor causes impaired vascular development in the Arabidopsis embryo, seedling, leaf and stem (Berleth and Jürgens, 1993; Hamann et al., 1999; Hardtke and Berleth, 1998; Mayer et al., 1991; Przemeck et al., 1996). Indeed, MP controls a variety of vascular-specific genes and pathways (De Rybel et al., 2013; Donner et al., 2009; Möller et al., 2017; Schlereth et al., 2010; Yoshida et al., 2019). However, all reported perturbations of auxin activity (synthesis, transport and response) that affect vascular development also affect a range of other processes (Bennett et al., 1996; Cheng et al., 2006; Marchant, 1999; van den Berg and ten Tusscher, 2017). It is therefore difficult to separate a role for auxin in the specification of vascular identity from its many other functions, which warrants the use of a simple developmental model system for studying initial vascular tissue specification in the absence of a certain process, such as differentiation. The early embryo is an attractive model given that it lacks a confounding wound response or extensive proliferation. Its simplicity and predictable division pattern allow the detection of early developmental defects (Scheres et al., 1994), and available transcriptome resources (Palovaara et al., 2017; Belmonte et al., 2013; Schon and Nodine, 2017; Slane et al., 2014) enable in-depth investigations of vascular identity specification.
Here, we first use a suite of established and novel transcriptional reporters to track the stepwise specification of vascular tissue identity in the embryo. We find that the identity initially specified is unique to the embryo, transitioning to a mature and robust identity in the root. We show that auxin response is necessary but not sufficient to specify vascular identity. Via large-scale enhanced yeast one-hybrid (eY1H) assays, we identify common regulators of vascular genes, and we find that one of these, the bZIP transcription factor GBF2, can interact with ARFs and modifies ARF5/MP activity in the regulation of vascular-specific genes.
RESULTS
Specification of vascular tissue identity is a multi-step process
As cell-type identity is often specified at the level of gene activity, it can be inferred using gene expression markers. Each gene expression marker acts as a proxy for an aspect of cell identity and cells that express multiple markers can confidently be identified as having the corresponding identity. We operationally define markers as positively or negatively marking a cell type, but should note that the true absence of marker expression is difficult to show or prove, and is perhaps not biologically realistic. Therefore, we address expression patterns, as well as relative expression strength across cells, to help define markers and cell types. Here, we used a large and diverse set of established cell type markers to ask when vascular tissue specification occurs during embryogenesis and to determine the ontogeny of the tissue. We selected SHORTROOT (SHR) and ATHB8 as early vascular markers in root and leaf development (Baima et al., 2001; Gardiner et al., 2011; Long et al., 2015). ZWILLE (ZLL) shows vascular-specific expression in the root and embryo (Radoeva et al., 2016; Tucker et al., 2008; Haseloff, 1999). WOODEN LEG (WOL), PEAR1 and DOF6 are all associated with cytokinin-responsive growth in the vascular bundle of the root (Mähönen et al., 2000; Miyashima et al., 2019). In contrast, expression of TMO5, TMO5-LIKE 1 (T5L1), TMO6, IQ-DOMAIN 15 (IQD15), SOSEKI 1 (SOK1) and WRKY17, was shown to depend on MP activity (Schlereth et al., 2010; Möller et al., 2017; Yoshida et al., 2019).
Lineage tracing has suggested that the first vascular cells are specified in the early globular stage embryo (Scheres et al., 1994), when the embryo first contains three distinct cell layers in the lower tier (Fig. 1A). Indeed, we found almost all vascular reporters to be expressed in the lower inner cells at this stage, except for TMO6 and T5L1 (Fig. 1B). However, at this stage, most reporters were expressed in both the central cell layer and the surrounding ground tissue cells: ATHB8, DOF6, PEAR1, WOL and ZLL were expressed at apparently equal levels in both tissue types. In contrast, IQD15 and SOK1 showed lower levels of expression in the ground tissue (Fig. 1B). The only markers that were restricted to the innermost cell layer were TMO5 and SHR, the expression of which could not be detected in the ground tissue (Fig. 1B). The broad expression of these vascular marker genes was only transient: within several cell division rounds, all vascular markers were restricted to the innermost cells, a pattern that was maintained in the postembryonic root (Fig. 1B). This suggests that, rather than being immediately restricted to a small number of innermost cells, the gene expression program associated with vascular identity is initially broad but becomes limited to inner cells over time. Although all 12 marker genes eventually become restricted to the vascular cells, it appears that not all are exclusively expressed in the vascular cells and there are multiple trajectories to their vascular-specific expression pattern.
Expression patterns of previously described vascular reporters. (A) Overview of stages of early embryogenesis. Cells previously discussed as vascular are marked in green. (B) Expression patterns of previously published vascular reporters in root and three stages of the early embryo. Images for WOL and ZLL are from the same root/embryo from a reporter line that carries both pWOL-sYFP and pZLL-mTurq. All reporters are transcriptional reporters except those for SHR and TMO6, which are translational fusions. Fluorescent protein signals are displayed in green, and cell wall staining in magenta. Roots are stained with PI and embryos with Renaissance 2200. Scale bars: 50 µm (root); 10 µm (embryo).
Expression patterns of previously described vascular reporters. (A) Overview of stages of early embryogenesis. Cells previously discussed as vascular are marked in green. (B) Expression patterns of previously published vascular reporters in root and three stages of the early embryo. Images for WOL and ZLL are from the same root/embryo from a reporter line that carries both pWOL-sYFP and pZLL-mTurq. All reporters are transcriptional reporters except those for SHR and TMO6, which are translational fusions. Fluorescent protein signals are displayed in green, and cell wall staining in magenta. Roots are stained with PI and embryos with Renaissance 2200. Scale bars: 50 µm (root); 10 µm (embryo).
The route to a vascular-specific expression pattern starts in the 16-cell stage embryo rather than at the early globular stage as previous reports have suggested. In previous work, inner lower tier cells at the 16-cell stage were shown to resemble their vascular daughter cells at the globular stage in terms of gene ontology term enrichment in transcriptomes (Palovaara et al., 2017). Indeed, we found that many vascular-specific marker genes start expression at the 16-cell stage, in which some showed expression in all cells (ATHB and ZLL) but most were exclusively expressed in the inner cells: DOF6, IQD15, PEAR1, SOK1 and WOL (Fig. 1B; Fig. S1). Thus, as the eight-cell embryo divides to generate outer and inner cell layers at the 16-cell stage, inner cells activate vascular markers. Their ground tissue daughters initially retain the expression of some vascular markers and switch these off later.
Many of the vascular marker genes were originally identified as targets of auxin signaling, often regulated by MP (Schlereth et al., 2010; Möller et al., 2017). As a result, there is a bias towards auxin-regulated genes among the well-studied vascular marker genes. We therefore searched for novel marker genes in an unbiased manner. Vascular-enriched genes were selected based on their expression in the early vascular cells, using a cell type-specific embryo transcriptome atlas (Palovaara et al., 2017) and additional publicly available vascular-specific transcriptome datasets (Fig. 2A) (Brady et al., 2007; Belmonte et al., 2013; Kondo et al., 2015; Melnyk et al., 2018). Transcriptional reporter lines were constructed to test the expression pattern of 36 potential marker genes and eventually five qualified as markers of vascular identity during embryogenesis (Fig. 2B). Expression of the remaining 31 genes could either not be detected during embryogenesis or was not limited to vascular cells in the embryo, while being specific to vascular tissue in the root (Fig. S2). Of the five selected reporters, GATA20 and AP2B3 expression started at the 16-cell stage, and at the early globular stage both were enriched in vascular cells (Fig. 2B). AP2B3 expression peaked in vascular cells but could also be detected in surrounding cell layers (Fig. 2B), whereas GATA20 expression showed vascular specificity (Fig. 2B). In the root tip, GATA20 has been shown to be expressed in the phloem (Lee et al., 2006) and we found that this expression was broader in the vascular cells close to the quiescent center (Fig. S1B). The other three selected reporters, MEE45, MIR171B and MSS3, were expressed at the dermatogen stage in all cells, but at lower levels in the vascular cells within the embryo, thus marking vascular cells with their expression minimum. MIR171B and MSS3, like WRKY17, were first expressed at the eight-cell stage and thus appeared to be repressed in the first vascular cells (Fig. S1). Because these genes have an inverted expression pattern compared with the expectation for a vascular gene, we will refer to these genes as ‘inverse’ markers of vascular identity in the embryo, indicating that their expression minimum, rather than their maximum, marks the vascular cells. This pattern is similar to that of WRKY17, a target of MP (Möller et al., 2017) (Fig. 1B). However, although WRKY17 was expressed broadly in the root meristem, MEE45, MIR171B and MSS3 showed tissue-specific expression in the root (Fig. 2B). Beyond resolving the ontogeny of vascular tissue identity in the embryo, our detailed analysis of vascular-specific markers also shows that there are significant differences in gene expression restriction to the vascular tissue between embryo and root. Hence, initial specification and maintenance of tissue identity seem to be associated with different gene expression patterns.
Identification of novel vascular reporters. (A) Overview of transcriptomics datasets used for the selection of new vascular reporters. The studied tissue/cell type in each experiment is marked in green. (B) Expression patterns of new vascular reporters for the embryo. Fluorescent protein signals are displayed in green, and cell wall staining in magenta. Roots are stained with PI, embryos with Renaissance 2200. Overlapping square symbols indicate images resulting from a stack. Scale bars: 50 µm (root); 10 µm (embryo).
Identification of novel vascular reporters. (A) Overview of transcriptomics datasets used for the selection of new vascular reporters. The studied tissue/cell type in each experiment is marked in green. (B) Expression patterns of new vascular reporters for the embryo. Fluorescent protein signals are displayed in green, and cell wall staining in magenta. Roots are stained with PI, embryos with Renaissance 2200. Overlapping square symbols indicate images resulting from a stack. Scale bars: 50 µm (root); 10 µm (embryo).
Auxin signaling through MP is necessary, but not sufficient, for initiation of vascular identity
Auxin signaling plays many key roles in plant development and one of the clearest is its contribution to vascular development. We sought to investigate the role of auxin signaling in the specification of vascular identity in the early embryo. To this end, we expressed a nondegradable version of the BDL (BODENLOS/IAA12) protein to block MP activity (Hamann et al., 2002; Weijers et al., 2006), while examining markers of vascular identity (Fig. 3A-B). As bdl expression in the entire embryo results in early developmental defects (Rademacher et al., 2012; Yoshida et al., 2014), we employed two-component gene activation and selectively expressed bdl in vascular cells using the Q0990-GAL4; UAS-erGFP driver line (Fig. 3A) (Haseloff, 1999). The GAL4 driver in the Q0990 line is inserted near the ZLL gene (Radoeva et al., 2016) and erGFP thus reports ZLL expression. Vascular markers were introduced into the Q0990 background and crossed with a line containing GAL4-dependent UAS-bdl (Weijers et al., 2006). The domain of bdl expression was marked by ER-localized GFP, whereas the vascular markers were reported by nuclear GFP fluorescence (Fig. 3A,B).
The role of the auxin response in embryonic vascular gene expression. (A) Diagram showing the experimental setup for Q0990≫bdl crosses. (B) Embryos resulting from crosses between a line containing Q0990::GAL4, UAS::erGFP and a vascular reporter, and either Col-0 or a line containing UAS::bdl. Numbers in the top left corner of each panel indicate the fraction of embryos observed with the pattern displayed. Green stars indicate altered ground tissue division orientation. (C) Diagram showing the experimental setup for pRPS5A≫UAS::MPΔPB1 crosses. (D) Embryos resulting from crosses between a line containing pRPS5A::GAL4 and a vascular reporter, and either Col-0 or a line containing UAS::MPΔPB1. Fluorescent protein signals are visualized using the Fire lookup table (see color key bar). Green stars indicate altered division planes in epidermal cells and in the hypophysis. Embryos are stained with Renaissance 2200 (white). Overlapping square symbols indicate images resulting from a stack. In A, vascular reporters are represented by nuclear signal, whereas the expression domain of the Q0990 reporter (UAS-erGFP) is marked by ER-localized GFP. Scale bars: 10 µm.
The role of the auxin response in embryonic vascular gene expression. (A) Diagram showing the experimental setup for Q0990≫bdl crosses. (B) Embryos resulting from crosses between a line containing Q0990::GAL4, UAS::erGFP and a vascular reporter, and either Col-0 or a line containing UAS::bdl. Numbers in the top left corner of each panel indicate the fraction of embryos observed with the pattern displayed. Green stars indicate altered ground tissue division orientation. (C) Diagram showing the experimental setup for pRPS5A≫UAS::MPΔPB1 crosses. (D) Embryos resulting from crosses between a line containing pRPS5A::GAL4 and a vascular reporter, and either Col-0 or a line containing UAS::MPΔPB1. Fluorescent protein signals are visualized using the Fire lookup table (see color key bar). Green stars indicate altered division planes in epidermal cells and in the hypophysis. Embryos are stained with Renaissance 2200 (white). Overlapping square symbols indicate images resulting from a stack. In A, vascular reporters are represented by nuclear signal, whereas the expression domain of the Q0990 reporter (UAS-erGFP) is marked by ER-localized GFP. Scale bars: 10 µm.
Embryos in which bdl was expressed in the vascular cells often showed altered ground tissue division orientation, as previously reported (Möller et al., 2017) (Fig. 3B), indicating that auxin signaling was successfully inhibited. However, erGFP expression was not affected in bdl-expressing embryos (Fig. 3B), indicating that maintenance of ZLL expression in presumptive vascular cells does not depend on the auxin response. Crosses of the Q0990-GAL4 line with Col-0 wild type resulted in no such changes in division orientation. bdl expression led to 96% (n=24) of the observed embryos lacking nuclear SOK1 expression, whereas almost all wild-type crossed embryos showed normal SOK1 expression. As SOK1 is regulated by MP (Möller et al., 2017; Yoshida et al., 2019), this further confirms the repression of MP activity. However, not all vascular characteristics were absent from the inner cells. In Q0990≫bdl embryos, GATA20 expression was absent in about half (54%, n=13) of the embryos, but remained present in the other half, indicating that, in many embryos, repression of vascular identity was incomplete, as was supported by normal ZLL expression. In addition, expression of the inverse marker MIR171B was mostly unchanged in Q0990≫bdl embryos (Fig. 3B). These findings indicate that, when auxin signaling is blocked in inner cells, vascular identity is compromised but not abolished. The remaining vascular program is insufficient for further proliferation and development (Berleth and Jürgens, 1993; Hamann et al., 1999; Weijers et al., 2006; De Rybel et al., 2013; Möller et al., 2017), thus we conclude that auxin signaling through MP is essential for functional vascular tissue specification.
After confirming that auxin signaling is required for specification of vascular identity, we next asked whether it is also sufficient. Although differences in auxin activity across cell layers in the early embryo, as measured by the R2D2 and DR5v2 reporters (Liao et al., 2015), are small, there is a clearly defined gradient with high levels in central and lower levels in peripheral cells (Möller et al., 2017) (Fig. S3). We asked if this small difference in auxin signaling between central and peripheral cells in the embryo is sufficient to restrict (vascular) identity to inner cells. We therefore expressed a version of MP that cannot be inhibited by auxin-dependent Aux/IAA proteins and that is hyperactive (MPΔPB1) (Krogan et al., 2012) from the ubiquitous RPS5A (RIBOSOMAL PROTEIN 5A) promoter (Weijers et al., 2001) using the GAL4-UAS system (Fig. 3C). Embryos with ubiquitous MPΔPB1 expression often showed altered division planes in epidermal cells and occasionally in the hypophysis (Fig. 3D), indicating the effectiveness of transgene expression. However, ectopic MPΔPB1 expression did not induce ectopic vascular marker expression. Expression of vascular genes (GATA20, SHR and SOK1) remained restricted to the vascular cells; likewise, inverse markers of identity (MIR171B, MSS3 and WRKY17) still showed an expression minimum in the vascular cells (Fig. 3D). These results show that the ectopic auxin response can induce changes in cell division orientation, but is insufficient for inducing vascular tissue specification in the early embryo. This suggests that unknown additional factors limit the domain of vascular identity.
Identification of transcriptional regulators of vascular gene expression
Given the co-expression of vascular marker genes in the embryo and in the post-embryonic vasculature, it is likely that there are common regulators. To identify transcription factors that bind multiple vascular gene promoters, we performed an eY1H assay (Gaudinier et al., 2011; Reece-Hoyes et al., 2011). Promoters from 14 of the aforementioned vascular reporter genes were screened against a customized collection of 2037 transcription factors and other DNA-binding proteins in an all-by-all setup (Table S1). Among the 32,592 interactions tested, 1111 were positive. Combining all interactions resulted in a network comprising 14 promoters and 382 transcription factors (Fig. S4; supplementary Dataset S1; Table 1). This network contained a large number of transcription factors that could bind to many of the vascular promoters. For in-depth analysis, we therefore selected the six vascular promoters that each had more than 25 interactors and showed the most prominent vascular specificity in the early embryo (Fig. 4A). These six consisted of: three vascular-specific marker genes (GATA20, SOK1 and ZLL); and three vascular inverse marker genes (MIR171B, MSS3 and WRKY17). The network that contains these promoters and their interactors comprises 221 transcription factors and 521 interactions (Fig. 4A). If there is a common vascular transcriptional program, it is likely that multiple vascular genes are regulated by a common set of transcription factors. We therefore parsed the interaction network to identify such common transcription factors as potential regulators of vascular identity specification.
Partial yeast one network and selection, and candidate regulators of vascular identity. (A) A yeast one-hybrid network showing all interactors of six out of 14 vascular promoters screened. Nodes representing transcription factors are colored according to their transcription factor family (see inset), and are grouped by outdegree. Nodes representing promoters are colored light (vascular specific) or dark (vascular inverse) gray. (B) Network overview of the 20 candidate regulators of vascular identity with all 16 vascular promoters screened. Colors are as in A. (C) Dendrogram resulting from hierarchical clustering of promoters by interactor set. Branch length indicates distance/similarity in the interactor set. Two promoters from an unrelated screen (TCA1 and TCA2) were included as an outgroup.
Partial yeast one network and selection, and candidate regulators of vascular identity. (A) A yeast one-hybrid network showing all interactors of six out of 14 vascular promoters screened. Nodes representing transcription factors are colored according to their transcription factor family (see inset), and are grouped by outdegree. Nodes representing promoters are colored light (vascular specific) or dark (vascular inverse) gray. (B) Network overview of the 20 candidate regulators of vascular identity with all 16 vascular promoters screened. Colors are as in A. (C) Dendrogram resulting from hierarchical clustering of promoters by interactor set. Branch length indicates distance/similarity in the interactor set. Two promoters from an unrelated screen (TCA1 and TCA2) were included as an outgroup.
A large number of transcription factors were identified to interact with the majority of these promoter sequences. The majority of these transcription factors bind to both sets of promoters (vascular specific and inverse), whereas only a few can bind to only one set. CUC2 (CUP-SHAPED COTYLEDON 2; Aida et al., 1997) and IDD12 INDETERMINATE-DOMAIN 12) can bind to three or more vascular-specific markers but no inverse markers. JAG (JAGGED; Ohno et al., 2004), DRN (DORNROSCHEN; Kirch et al., 2003) and ARR1 (RESPONSE REGULATOR 1; Sakai et al., 2001) bind to two or more vascular inverse markers but no vascular-specific markers (Fig. 4A). However, these transcription factors are a small minority: in general, the vascular-specific and -inverse promoters have highly similar sets of transcription factors binding to their promoters, despite having very different expression patterns. When we performed clustering on the promoters based on their interactor set, we found that vascular-specific and vascular-inverse markers did not have distinct sets of interactors (Fig. 4C). These findings suggest a large set of transcription factors that could act in a complex gene regulatory network (GRN) controlling vascular identity specification. It should be noted though that both sets of promoters show differential expression between vascular and nonvascular cells, and it is possible that the same transcription factors (or related proteins) could act as vascular-specific activators or repressors.
To identify candidate regulators of vascular identity specification, we selected 20 transcription factors from the network. This selection was performed in two steps. First, transcription factors were discarded that: (1) bound to few vascular promoters; (2) were expected to be false positives based on their promiscuous binding profiles in previous screens (Gaudinier et al., 2018; Taylor-Teeples et al., 2015); or (3) were most likely not expressed during embryogenesis based on transcriptomics data (Belmonte et al., 2013; Schlereth et al., 2010). This approach resulted in a list of 50 transcription factors. Next, each transcription factor was scored for: (1) expression during vascular development and early embryogenesis; (2) the number of vascular promoters bound; (3) diversity of expression patterns bound; and (4) vascular promoter binding in published DNA affinity purification sequencing (DAP-seq) data (O'Malley et al., 2016). For further analysis, we selected the top 20 transcription factors ranked by cumulative score (Table S3; Fig. 4B).
A candidate regulator of identity would be expected to be present at the time and place that specification takes place. To ascertain the presence of candidate regulators during embryogenesis, translational fusions of genomic fragments to YFP were created and observed for 17 of these 20 transcription factors. These revealed that ten candidate regulators were indeed present at the 16-cell stage in the pro-embryo (Fig. 5A). The remaining seven were either not detected during embryogenesis or not at the correct time or location (Fig. S5). The majority of the ten candidate regulators expressed in 16-cell embryos are present uniformly in the nucleus, except for members of the GeBP (GL1 ENHANCER BINDING PROTEIN) family, which accumulate in foci within the nucleus, similar to previous reports (Fig. 5A) (Curaba et al., 2003). No conspicuous differences between cell types could be found in the early embryo, in either protein quantity or localization. This indicates that, if these candidates contribute to specifying vascular identity, their cell-specific action is not the result of protein level or location. Instead. an unknown mechanism might contribute to cell-specific activity.
Protein localization and interactions of candidate regulators of vascular identity. (A) Translational reporter lines of ten candidate regulators in the root tip and early pro-embryo. Gray square indicates the inset highlighting STKL2 expression in foci within the nucleus. Fluorescent protein signals are displayed in green, and cell wall staining in magenta. Roots are stained with PI, and embryos with Renaissance 2200. (B) Overview of split-YFP (BiFC) results indicating that GBF2 can interact with the full-length protein and DBD of six different ARFs. Green indicates that interaction was observed, whereas gray indicates no interaction. (C) Selected images of split-YFP (BiFC) assays showing the interaction between the DBD of MP/ARF5 and GBF2. (D) Distribution of potential MP/bZIP68 composite elements within MP-binding regions taken from DAP-seq. x axis numbers reflect the number of nucleotides, F, full overlap; P, partial overlap; S, spacer. Top: MP/bZip68 everted composite element distribution. Bottom: MP/bZIP68 direct composite element distribution. Scale bars: 50 µm (root) or 10 µm (embryo) (A); 75 μm (tobacco leaf) (C).
Protein localization and interactions of candidate regulators of vascular identity. (A) Translational reporter lines of ten candidate regulators in the root tip and early pro-embryo. Gray square indicates the inset highlighting STKL2 expression in foci within the nucleus. Fluorescent protein signals are displayed in green, and cell wall staining in magenta. Roots are stained with PI, and embryos with Renaissance 2200. (B) Overview of split-YFP (BiFC) results indicating that GBF2 can interact with the full-length protein and DBD of six different ARFs. Green indicates that interaction was observed, whereas gray indicates no interaction. (C) Selected images of split-YFP (BiFC) assays showing the interaction between the DBD of MP/ARF5 and GBF2. (D) Distribution of potential MP/bZIP68 composite elements within MP-binding regions taken from DAP-seq. x axis numbers reflect the number of nucleotides, F, full overlap; P, partial overlap; S, spacer. Top: MP/bZip68 everted composite element distribution. Bottom: MP/bZIP68 direct composite element distribution. Scale bars: 50 µm (root) or 10 µm (embryo) (A); 75 μm (tobacco leaf) (C).
GBF1 and GBF2 can interact with MP
To determine whether any of the candidate regulators could induce or repress gene activity during vascular tissue specification, each was expressed in meristematic cells with the RPS5A promoter (Weijers et al., 2001), either as native cDNA or as a fusion with a dominant SRDX repressor motif. Misexpression of several resulted in either lethal or mild developmental phenotypes, but none of the identified candidate regulators suppressed or ectopically induced vascular tissue differentiation (Fig. S6). Therefore, it is unlikely that any of these candidates act in isolation to control vascular tissue initiation. A more likely explanation is that identity is controlled by a complex GRN, in which the unique interactions between individual regulators provide cell-type specificity.
In our efforts to understand the mechanisms of gene regulation by MP, we immunoprecipitated MP-containing protein complexes from root tips in an MP-GFP transgenic line whose functionality had been validated (Schlereth et al., 2010). In this experiment, we identified GBF2 (G-BOX BINDING FACTOR 2) as a potential interactor of MP-GFP (Table S2). As both MP and GBF2 are candidate regulators of vascular identity specification, we decided to follow up on this interaction. Immunoprecipitation of GBF2-YFP and its homolog GBF1-YFP, followed by mass spectrometry, did not recover MP, presumably due to the very low abundance of MP, but did confirm previous observations that G-class bZIP proteins can heterodimerize extensively (Fig. S7). To test MP-GBF interactions more directly, we performed split-YFP assays [bimolecular fluorescence complementation (BiFC)] (Hu et al., 2002; Walter et al., 2004) in Nicotiana benthamiana. Both GBF2 and GBF1 could interact with MP (Fig. 5B,C; Fig. S8). Interestingly, this interaction was not restricted to MP: ARFs of all three major classes [A, B and C: Okushima et al. (2005); Finet et al. (2013)] could interact with both GBF2 and GBF1 (Fig. 5B; Fig. S8), and the interaction domain was mapped to the ARF DNA-binding domain (Fig. 5B,C; Fig. S8).
GBF proteins have been reported to be involved in the responses to blue light and in leaf senescence (Singh et al., 2012; Smykowski et al., 2010; Mallappa et al., 2006; Giri et al., 2017). However, gbf1, gbf2 and gbf3 single and double mutants show no developmental phenotypes (Fig. S9). This is likely a result of genetic redundancy: the bZIP G-class contains five members and double mutants show increased expression of close homologs (Jakoby et al., 2002; Dröge-Laser et al., 2018) (Fig. S9). A triple mutant could not be recovered from plants that were homozygous for gbf1 and gbf3 or from plants that segregated gbf2, suggesting that a lack of all three proteins may result in lethality. Indeed, disruption of a GRN underlying vascular identity specification would likely result in early developmental arrest. Overexpression using RPS5A or 35S promoters caused pleiotropic developmental defects. pRPS5A≫GBF2-SRDX plants were often sterile, whereas 35S::GBF1/2 plants had round leaves and showed delayed flowering (Figs S5 and S9). However, no ectopic vascular tissue differentiation was observed. These findings indicate that GBF1/2 protein quantity alone does not specifically limit vascular tissue specification.
GBFs bind to Gboxes close to auxin response elements to modulate auxin responsive expression
GBF2 and GBF1 physically interact with the DNA-binding domain (DBD) of ARFs, and thus have the potential to co-regulate auxin-responsive genes. Interactions among transcription factors can lead to cooperative DNA binding if both transcription factors can bind to cognate DNA elements in close proximity. Indeed, Gbox motifs were found to be enriched in close proximity to auxin response elements (AuxREs) (Weiste and Dröge-Laser, 2014; Ulmasov et al., 1995) and they are overrepresented in auxin-responsive and ARF-binding regions (Berendzen et al., 2012; Cherenkov et al., 2018). To test the co-occurrence of ARFs and G-class bZIPs motifs, we applied the Motifs Co-Occurrence Tool (MCOT) (Levitsky et al., 2019) to ARF5 and ARF2 peaks taken from genome-wide DAP-seq profiles (O'Malley et al., 2016). We analyzed all possible combinations of AuxREs (ARF2/5 motifs) and Gboxes (GBF3 and bZIP16/68 motifs) with any overlap or spacer lengths below 30 nucleotides, and found that bZIP68 and ARF5 motifs overlap (P-value<5E-40) (Fig. 5D).
To investigate the function of these linked motifs, we selected three vascular promoters that contained clear AuxRE and Gbox motifs in close proximity: with an overlap (WRKY17), a short spacer (TMO5) and a long spacer (GATA20) (Fig. 6A). Transcriptional reporters with mutated promoters were generated to determine the contribution of the AuxRE-Gbox motif to vascular gene expression domain and level. Removing the complete AuxRE-Gbox motif from the promoters of GATA20, TMO5 and WRKY17 resulted in a strong and significant reduction of fluorescence in transgenic roots (Fig. 6C-E). For the WRKY17 reporter, this reduction was more significant in the vascular bundle compared with the rest of the root meristem, resulting in a changed vascular/nonvascular signal ratio (Fig. 6E-H). This suggests that, for this promoter, the overlapping AuxRE-Gbox motif controls expression levels specifically in vascular cells. Removing only the Gbox had a smaller effect (Fig. 6C-E). Mutated GATA20 and TMO5 promoters in which the adjoining Gbox had been removed did not show a significant decrease in fluorescence, but instead resulted in increased variation in expression level among transgenic lines (Fig. 6C,D). Thus, the Gboxes in these promoters appear to be contributing to the expression stability instead of to absolute vascular expression levels. As members of several transcription factor families can bind to Gboxes (Schindler et al., 1992; Lian et al., 2017; O'Malley et al., 2016), it is unclear whether GBF2 is the factor contributing to stability of vascular gene expression.
Gboxes can be bound by GBF2 and are needed for stable vascular expression. (A) Schematic overview of the promoter sequences of GATA20, TMO5 and WRKY17. X1 and X2 indicate regions containing TF-binding sites. Blue and red lines indicate control regions and Gbox regions used for ChIP-qPCR, respectively. (B) ChIP-qPCR performed on Arabidopsis cell cultures expressing either GBF1-YFP or GBF2-YFP. Relative enrichment of the BOX regions compared with CONTROL regions. Data are mean±s.e.m. (C-E) Boxplots displaying fluorescence intensity of transcriptional reporter lines. Each plot compares the mean fluorescence in the measured cells for T1 roots containing full-length or truncated promoters of GATA20 (C), TMO5 (D) and WKRY17 (E). Each point is the mean fluorescence in the early vascular cells measured from one independent T1 root. For the WRKY17 promoter, two areas were measured: the vascular bundle (white) and adjacent nonvascular cells (blue). For box plots, the box marks the first and third quartiles, split by the median; whiskers extend to a maximum of 1.5× the interquartile range beyond the box. *P<0.05, **P<0.001; two-sided, unpaired Student's t-test. One-way analysis of variance (ANOVA) followed by Tukey's honest significant difference (HSD) test was performed to compare differences in fluorescence intensity. Samples were classified into up to three categories per experiment (a/ab/b). Results from the Tukey's HSD test are listed in Table S4. (F) The ratio of WRKY17-driven GFP signal in the vascular cells compared with signal in the nonvascular cells. (G,H) Expression patterns in representative WRKY17 T1 roots. Boxes indicate the regions in which fluorescent signal was measured. Scale bars: 50 μm.
Gboxes can be bound by GBF2 and are needed for stable vascular expression. (A) Schematic overview of the promoter sequences of GATA20, TMO5 and WRKY17. X1 and X2 indicate regions containing TF-binding sites. Blue and red lines indicate control regions and Gbox regions used for ChIP-qPCR, respectively. (B) ChIP-qPCR performed on Arabidopsis cell cultures expressing either GBF1-YFP or GBF2-YFP. Relative enrichment of the BOX regions compared with CONTROL regions. Data are mean±s.e.m. (C-E) Boxplots displaying fluorescence intensity of transcriptional reporter lines. Each plot compares the mean fluorescence in the measured cells for T1 roots containing full-length or truncated promoters of GATA20 (C), TMO5 (D) and WKRY17 (E). Each point is the mean fluorescence in the early vascular cells measured from one independent T1 root. For the WRKY17 promoter, two areas were measured: the vascular bundle (white) and adjacent nonvascular cells (blue). For box plots, the box marks the first and third quartiles, split by the median; whiskers extend to a maximum of 1.5× the interquartile range beyond the box. *P<0.05, **P<0.001; two-sided, unpaired Student's t-test. One-way analysis of variance (ANOVA) followed by Tukey's honest significant difference (HSD) test was performed to compare differences in fluorescence intensity. Samples were classified into up to three categories per experiment (a/ab/b). Results from the Tukey's HSD test are listed in Table S4. (F) The ratio of WRKY17-driven GFP signal in the vascular cells compared with signal in the nonvascular cells. (G,H) Expression patterns in representative WRKY17 T1 roots. Boxes indicate the regions in which fluorescent signal was measured. Scale bars: 50 μm.
Next, we tested whether GBF2 alone could bind to the Gboxes present in vascular promoters. ChIP-qPCR on suspension cell cultures overexpressing GBF2-YFP confirmed that GBF2 could bind to the Gbox motif in the WRKY17 promoter, but could not confirm the same for the GATA20 and TMO5 promoters (Fig. 6B). Instead, it is possible that GBF2 and MP both need to be present to interact with the promoters of these two genes.
If GBF2 can bind to vascular promoters and co-regulate vascular gene expression, its overexpression should affect the regulation of vascular genes. To test this directly, we generated protoplasts from vascular reporter lines pVASC::n3GFP (VASC being either GATA20, TMO5 or WRKY17), and transfected these with a combination of 35S::GBF2-mTurquoise2, 35S::MPΔPB1-mScarlet-I and corresponding empty vectors to determine their effects on target promoter activity (Fig. 7D-F). MPΔPB1 was used to overcome any auxin-dependent inhibition. Overexpression of only GBF2-mTurquoise2 had no effect on the promoter activity of GATA20, TMO5 and WRKY17. In contrast, expression of MPΔPB1-mScarlet-I resulted in increased TMO5 promoter activity and decreased WRKY17 promoter activity (Fig. 7B,C). This effect disappeared when GBF2 was co-expressed with MPΔPB1, suggesting that GBF2 acts by restricting MP activity, probably via competitive binding with the overlapping AuxRE-Gbox motif. This was not the case for the GATA20 promoter, the activity of which was decreased by both MPΔPB1-mScarlet-I and by MPΔPB1-mScarlet-I combined with GBF2-mTurquoise2 (Fig. 7A). These findings suggest that the interaction between GBF2 and MP depends on promoter context, yet reflects a functional interaction that contributes to the regulation of vascular genes.
GBF2 modulates MPΔ induced vascular gene expression. (A-C) Boxplots displaying intensity of green fluorescence in nuclei of protoplasts transformed with two misexpression constructs that were either empty or contained GBF2 or MPΔPB1. Protoplasts were generated from leaves that expressed pGATA20::n3GFP (A), pTMO5::n3GFP (B) or pWRKY17::n3GFP (C). One-way analysis of variance followed by Tukey's honest significant difference (HSD) test was performed to compare changes in fluorescence intensity. Samples were classified into up to three categories per experiment (a/ab/b). Results from the Tukey's HSD test are listed in Table S4. Each assay was performed twice with similar results. (D-F) Fluorescence signals detected in protoplast assays. Pink, chloroplasts; green, GFP; cyan, mTurquoise2; red, mScarlet-I. Scale bars: 75 μm.
GBF2 modulates MPΔ induced vascular gene expression. (A-C) Boxplots displaying intensity of green fluorescence in nuclei of protoplasts transformed with two misexpression constructs that were either empty or contained GBF2 or MPΔPB1. Protoplasts were generated from leaves that expressed pGATA20::n3GFP (A), pTMO5::n3GFP (B) or pWRKY17::n3GFP (C). One-way analysis of variance followed by Tukey's honest significant difference (HSD) test was performed to compare changes in fluorescence intensity. Samples were classified into up to three categories per experiment (a/ab/b). Results from the Tukey's HSD test are listed in Table S4. Each assay was performed twice with similar results. (D-F) Fluorescence signals detected in protoplast assays. Pink, chloroplasts; green, GFP; cyan, mTurquoise2; red, mScarlet-I. Scale bars: 75 μm.
DISCUSSION
Vascular tissues play a central role in plant development. Anatomical, physiological and genetic studies contribute to our understanding of vascular tissue development and key aspects of its regulation. Although significant insights have been gained into the regulation of cell identity specification, pattern formation, growth and differentiation within the vascular tissue, a key unresolved issue is how this tissue is initially specified from nonvascular precursor cells. From lineage tracing in Arabidopsis, it is clear that the vascular lineage has its origins in the early embryo (Scheres et al., 1995); however, the origin of embryonic vascular tissue has thus far not been characterized molecularly. Here, we have used a panel of vascular marker genes to map the ontogeny of vascular tissue identity from the embryo to postembryonic tissues. First, we traced its initiation to the 16-cell stage embryo. At this stage, the outer cells acquire protoderm identity (Abe et al., 2003), and we found that the inner cells express multiple vascular marker genes, identifying these cells as the first with vascular attributes. None of the ground tissue markers we have analyzed was present in the inner lower tier cells of the 16-cell stage embryo. These findings are in line with results of a recent transcriptome study that found that the transcriptome of these (inner) cells is similar to that of the later vascular cells (Palovaara et al., 2017). This suggests that, instead of vascular and ground tissue identities emerging simultaneously, the first ground tissue cells are the daughters of the first vascular cells.
We found that the transcriptional dynamics and progression are vastly different among genes that later mark the vascular domain. Some genes, such as DOF6, GATA20, SOK1 and TMO5, quite strictly mark the vascular cells, whereas others, such as ATHB8 and ZLL, can be found in adjacent cell types during early embryogenesis. The strict vascular markers provide better tools for studying the emergence of vascular identity, at least in the context of embryogenesis. The dynamic expression patterns of vascular marker genes suggest that vascular tissue identity is not a uniform trait that exists across developmental stages, urging us to reconsider how we view the development of cell identity over time. Features that distinguish embryonic versus postembryonic vascular cells are the co-expression of xylem- and phloem-specific marker genes, and the lack of expression from vascular inverse markers. This suggests that initial vascular tissue identity in the embryo is a temporary state that does not persist. Indeed, in the postembryonic vascular cells, different vascular cell types are highly divergent, a trait that is emphasized by recent advances in single cell (sc)RNAseq. In scRNAseq experiments performed on roots, xylem and phloem cells form distinct clusters; however, vascular tissues as a whole do not form a cluster that is separated from the two other ‘major’ tissue identities: ground tissue and epidermis (Ryu et al., 2019; Shulse et al., 2019; Denyer et al., 2019). It is questionable whether cells in the postembryonic vascular bundle have, or need, a common identity. Instead, the brief existence of a ‘general’ primordial multipotent vascular identity may only be needed when new vascular bundles are initiated: to ensure proper placement of the vascular bundle as a whole and to establish the cambial cells. Future scRNAseq experiments on the embryo and, for example, on wounded stems or graft junctions (Melnyk et al., 2018) could help address this issue.
There is a strong connection between de novo vascular tissue formation and auxin activity. Lack of auxin signaling results in impaired vascular tissue development, whereas the application of exogenous auxin can induce the formation of new vascular bundles (Sachs, 1969; Krogan et al., 2012; Donner et al., 2009; Jacobs, 1952). Auxin and its key transcriptional effector MP are therefore often regarded as master regulators of vascular tissue development (Brackmann et al., 2018). Here, we asked whether this prominent role also pertains to the earliest steps in vascular tissue specification in the embryo. We found that auxin signaling in the central cells of the embryo is indeed required for the complete establishment of the vascular transcription program. This conclusion is supported by an earlier transcriptome analysis, in which many vascular genes were downregulated in embryos where the auxin response inhibitor bdl was expressed in the inner cells marked by the Q0990 driver (Möller et al., 2017; Radoeva et al., 2016). Interestingly though, several vascular marker genes, including the ZLL gene, persisted even upon auxin response inhibition. This suggests that for part of the vascular program, the auxin response is not required after the initial specification event. Conversely, we found that auxin signaling through MP is not sufficient to induce vascular tissue identity specification outside of the normal vascular domain. It should, however, be noted that a dominant-active version of MP lacking its C-terminal PB1 domains was used. It is possible that interactions with the PB1 domain, other than Aux/IAA inhibition – such as, for example, homo-oligomerization (Nanao et al., 2014) – are required for the activity of MP in vascular tissue development. Nonetheless, this result suggests that, in addition to the auxin response, there must be additional, yet undiscovered, factors that determine which cells acquire vascular tissue identity. Given the small size of the embryo, a system built on a set of regulators would also provide a more robust mechanism for the regulation of identity than a single master regulator of identity.
Using an eY1H assay, we identified ten transcription factors that can bind to a large number of vascular promoter sequences and are also present at the moment of vascular identity specification. Although the exact roles and relations of these transcription factors remain unclear, they could be part of the GRN controlling vascular tissue specification and provide a new entry point into studying its regulation. A complex GRN would provide a robust system for the initial specification of identity. Although it is clear that auxin signaling through MP is necessary, other candidate regulators of identity have not yet been untangled. As single perturbations of other components had little effect, we decided to explore the link between MP and a new candidate regulator: GBF2 and its close homolog GBF1.
GBF1 and GBF2 not only bind multiple vascular gene promoters but they also interact with the DBDs of ARF proteins of all three major classes [A, B and C; Okushima et al. (2005); Finet et al. (2013)]. Other previously identified ARF-interacting proteins have been shown to interact with the PB1 domain (Ripoll et al., 2015; Shin et al., 2007; Varaud et al., 2011) or middle region (Wu et al., 2015), which would likely modulate transcription activity. Instead, interaction of GBFs with the DBD could modify DNA-binding properties, either by exclusion or by cooperativity. GBF-binding motifs, Gboxes, were often found in close proximity to AuxREs (Berendzen et al., 2012; Cherenkov et al., 2018; Ulmasov et al., 1995), suggesting that GBFs and ARFs could regulate gene expression together. We show that the Gboxes in several vascular promoters affect the stability of expression levels in the vascular bundles and that GBF2 is able to prevent the effect of MPΔPB1 on several target genes. Thus, GBF1/2 is able to modulate and/or stabilize auxin-dependent regulation of vascular gene expression.
GBF1/2 is a strong candidate for being part of the GRN that controls the initiation of vascular identity specification. A gfb1 gbf2 gbf3 triple mutant could not be recovered, and its potential lethality highlights a common theme in investigating the regulation of basic cell identities. Further genetic analysis, e.g. using conditional mutant alleles, could help to define the role of these transcription factors in vascular tissue initiation. The broad expression of GBFs could suggest that, similar to MP (Möller et al., 2017), their predicted cell-specific activity is influenced by local signals, such as redox potential and phosphorylation (Shaikhali et al., 2012; Klimczak et al., 1992; Smykowski et al., 2016). In a larger GRN controlling vascular identity, GBF1/2 could contribute to limiting vascular identity to the innermost cells of the early embryo. Future research into these candidate regulators is needed to confirm such a role for GBF1/2.
In conclusion, our work identifies embryonic vascular tissue identity as a primordial state from which vascular cell types evolve in diverse patterns. We find that auxin response is necessary but not sufficient for specifying this initial vascular identity. We identify a range of potential regulators of vascular identity specification and suggest a complex GRN being in control of vascular identity regulation. One potential regulator, GBF1/2, could interact with MP to modulate vascular gene expression. We expect that further analysis of these and other candidate regulators will help identify the elusive mechanism that specifies vascular tissue identity.
MATERIALS AND METHODS
Plant material and growth conditions
All Arabidopsis plants used in this study were of the Col-0 ecotype, except for the cell cultures, which were Landsberg erecta. Reporter lines for DOF6 and TMO6 have previously been published (Miyashima et al., 2019). Transcriptional reporters for targets of MP, IQD15, SOK1, T5L1, TMO5 and WRKY17, have previously been published (Möller et al., 2017; Schlereth et al., 2010). The reporters for ATHB8 and SHR have previously been published (Donner et al., 2009; Long et al., 2015). New reporters were generated for WOL and ZLL using the primers documented in Table S5; these reproduce previously described expression patterns (Mähönen et al., 2000; Radoeva et al., 2016). All newly generated transcriptional and translational reporter constructs (see below) were transformed into the Arabidopsis Col-0 wild-type accession. Misexpression lines were generated by introducing UAS-gene constructs into a background containing the pRPS5A-GAL4 driver or by introducing 35S-driven constructs into the Col-0 background. T-DNA insertion lines gbf1 (SALK_027691), gbf2-1 (SALK_206654), gbf2-2 (SALK_205706) and gbf3 (SALK_067963) were obtained from the Nottingham Arabidopsis Stock Centre (Nottingham, UK) and the Arabidopsis Biological Resource Centre (OH, USA). Plants were genotyped using the primers listed in Table S5. Arabidopsis seeds were surface sterilized, plated on half-strength Murashige and Skoog medium with the appropriate antibiotic (50 mg/l kanamycin or 15 mg/l phosphinothricin) and underwent 2 days of stratification at 4°C before being placed in the growth chamber. Plants were grown at 22°C under standard long-day conditions [16 h light 110 µE m−2 s−1 (Philips Master TL-D HF 50W/840) and 8 h dark].
Plant growth methods
Wild-type Arabidopsis Landsberg erecta and transgenic PSB-D cell suspension cultures were maintained in Musharige and Skoog minimal organic medium in the dark at 25°C with gentle shaking at 130 rpm. Cells were subcultured every 7 days in a 1:10 dilution with fresh medium. Transformations were conducted without callus selection, as described by Van Leene et al. (2007).
BiFC was performed by infiltrating Nicotiana benthamiana leaves with Agrobacterium tumefaciens strains carrying the appropriate plasmids (see below). Two days after infiltration, leaf sections were cut and imaged by confocal microscopy.
Protoplasts were harvested with a tape sandwich (Wu et al., 2009) and transfection was performed as described by Russinova et al. (2004). Protoplasts were prepared from plants containing a stable vascular transcriptional reporter. Green fluorescence levels were measured in protoplasts with both red (mScarlet-I) and blue (mTurquoise) fluorescence 2 days after transfection using confocal microscopy.
Vector construction for plant transformation
All constructs for plant transformation were cloned using seamless ligation cloning extract (SliCE) cloning into previously published ligation-independent cloning vectors (Wendrich et al., 2015; Zhang et al., 2014). Promoters for transcriptional reporters were introduced into the pPLV04_v2 backbone (De Rybel et al., 2011). Translational fusion constructs were generated by amplifying up to 3 kb of the promoter and the gene up to but not including the stop codon and introducing this sequence into pPLV16_v2. UAS-gene-SRDX overexpression constructs were cloned by introducing the amplified cDNA sequence without stop codon into a modified pPLV32_v2 backbone containing a SRDX peptide using SLiCE cloning (Wendrich et al., 2015; Zhang et al., 2014). 35S overexpression constructs were generated by introducing the cDNA sequence into a modified pPLV26 containing a C-terminal YFP. All constructs were introduced using the simplified floral dip method as described by De Rybel et al. (2011). BiFC constructs were generated by introducing amplified cDNA sequences into modified pPLV26 vectors containing NtYFP or CtYFP, either before or after the insertion site. Binary vectors for misexpression in protoplasts were generated by introducing the cDNA of GBF2 or MPΔPB1 (first 766 amino acids) into pMON99 containing C-terminal mTurquoise or mScarlet-I. All primers used for cloning are listed in Table S5.
Microscopy
Confocal imaging was performed on a Leica SP5 II system equipped with hybrid detectors. Confocal microscopy was performed as described previously (Llavata-Peris et al., 2013). Cell walls were visualized by staining with propidium iodide (PI, for roots) or SCRI Renaissance Stain 2200 (Renaissance Chemicals R2200, for embryos).
Yeast one-hybrid
eY1H assays were performed as described previously (Gaudinier et al., 2017). The promoter used for the yeast reporter constructs (pMW2 and pMW3) was the same as the promoter used for reporting localization in Arabidopsis. The prey collection used was the complete Arabidopsis transcription factor collection available at the Brady lab (University of California, CA, USA) in July 2016 (Table S1). Network analysis was performed in Cytoscape (Shannon et al., 2003).
Affinity purification mass spectrometry sample preparation
For affinity purification, either 4 g root material or 50 ml of 3-day-old transgenic PSB-D cell suspension cultures were used, and protein extraction, pull down and sample preparation was performed as described previously (Wendrich et al., 2017). Peptides were applied to online nano LCMS/MS (Thermo Scientific) using a 60 min acetonitrile gradient from 5 to 50%. Spectra were recorded on a LTQ-XL mass spectrometer (Thermo Scientific) and analysed as described previously (Wendrich et al., 2017). Maxquant output proteingroups.txt was filtered in Perseus v. 1.6.2.3. Volcano plots were generated in R and further visualized in Adobe Illustrator.
Motif analysis
Analysis of potential binding sites presence was performed with position weight matrices taken from the Plant Transcription Factor Database (Jin et al., 2017) for GBF3 (MP00318), bZip16 (MP00291) and bZip68 (MP00173). Colocalization of binding sites with ARF-binding sites was analyzed with the MCOT package (Levitsky et al., 2019) using data on ARF2 (GSM1925138 and GSM1925826) and ARF5 (GSM1925827) binding regions from DAP-seq analysis (O'Malley et al., 2016).
ChIP-qPCR
ChIP-qPCR was performed on Arabidopsis cell cultures using a protocol adapted from Gendrel et al. (2005). Filtered cell culture material (3-4 g) was used as input material. After crosslinking and DNA fragmentation, the sample was split and GFP-Trap beads (Chromotek) were used to pull down GBF-YFP complexes, whereas Myc-Trap beads (Chromotek) were used for the negative control sample. qRT-PCR was performed using primers listed in Table S5. Ct values were then used to calculate fold enrichment and relative fold enrichment compared with the control regions.
Acknowledgements
The authors are grateful to Branimir Velinov, Koyan Bruggeling, Surabissree Merialu Diwakar, Sebastien Paque, Frederique Polder, Allison Gaudinier, Anne-Maarit Bagman, Joel Rodriguez-Medina, Pawel Roszak, Michelle Tang and Peter Etchells for help with experiments; to Weijers lab members for helpful discussion; and to Camila Lopez-Anido for comments on the manuscript.
Footnotes
Author contributions
Conceptualization: M.E.S., D.W.; Methodology: V.L.; Validation: M.S.; Formal analysis: M.E.S., V.L., D.W.; Investigation: M.E.S., C.L-P., M.R., H.v., D.N., V.L.; Resources: I.S., P.R., D.S., G.J., S.M.B.; Writing - original draft: M.E.S., D.W.; Writing - review & editing: M.E.S., C.L.-P., M.R., H.v., D.N., V.L., D.S., G.J., V.M., S.M.B., D.W.; Visualization: M.E.S.; Supervision: G.J., V.M., S.M.B., D.W.; Project administration: D.W.; Funding acquisition: M.E.S., V.M., D.W.
Funding
This work was supported by grants from the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO; 831.14.003 to M.E.S. and VICI 865.14.001 to D.W.); the European Molecular Biology Organization (EMBO ASTF-368-2016 to M.E.S.); the Russian Foundation for Basic Research (19-44-543006 to D.N.); a Wageningen Graduate Schools Sandwich PhD Scholarship (to D.N.); and Budget Project from the Russian Government (0324-2019-0040-C-01 to V.M., V.L. and D.N.). S.M.B. is a Howard Hughes Medical Institute faculty scholar.
Peer review history
The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.186130.reviewer-comments.pdf
References
Competing interests
The authors declare no competing or financial interests.