The Arabidopsis leaf provascular cell transcriptome is enriched in genes with roles in vein patterning

Authors


For Correspondence (e-mail timothy.nelson@yale.edu).

Summary

Several classes of genes have been associated, by mutant phenotypes or cell biology, with the formation of vein patterns during early leaf development, including genes for certain transcription factors, auxin transport and response factors, endomembrane traffic components and other signaling pathway components. The majority of these are expressed with spatial and temporal specificity that includes expression in the precursors of vascular cells – provascular (PV) and procambial (PC) cells – suggesting that other PV/PC-specific genes might have roles in vein patterning. We inventoried the PV/PC transcriptome of Arabidopsis leaves using a combination of laser microdissection and microarray expression profiling, and determined the phenotypes of knock-outs of previously uncharacterized PV/PC-specific genes. As examples, we observed vein pattern defects in knock-out lines of KEG and a CCCH zinc finger protein. This strategy of gene discovery, based on the identification of a gene set co-expressed in the same cells during the targeted developmental event, appears to be an efficient means of identifying genes functionally relevant to the event. In the case of vein patterning, this strategy would have identified many or most of the genes previously obtained by labor-intensive screening for pattern-defective mutants.

Introduction

The formation of veins in leaves and other foliar organs is progressive, beginning with the midvein (primary vein) and continuing with higher order veins in a hierarchical two-dimensional pattern that is distinct for each species. Ground cells in the path of an incipient vein are recruited into the provascular (PV) fate that leads to aligned vascular cells. This recruitment is first visibly evident by the procambial (PC) stage of cell differentiation, in which originally isodiametric cells become elongated and polarized, and divide longitudinally along the new vein axis. An earlier PV-committed state, sometimes called ‘pre-procambial’, has been detected experimentally and through the use of molecular markers, but cannot be distinguished from ground cells simply on the basis of cell morphology. The regulation of the progression from uncommitted ground cell through PV stages to mature aligned vascular cells is crucial, as it produces not only a functioning vascular system but also apparent positional landmarks that influence the differentiation of other leaf cell types. This in turn requires the correct regulation of intracellular vesicle traffic, polar auxin transport, and a variety of other molecular and cellular features governing cell polarity and communication (Fukuda, 2004; Caño-Delgado et al., 2010; Ohashi-Ito and Fukuda, 2010; Donner et al., 2010; Hirakawa et al., 2011).

The dispersed and transient nature of PV stages of differentiating vascular cells has made it challenging to identify genes and processes associated with individual stages. Several strategies have been used to identify genes influencing vascular patterning and differentiation, including mutant screens for defective vein patterns and transcript profiling of vascular stages from dissected vascular cambium, from in vitro differentiating tracheary elements, vascular overgrowths by auxin transport inhibitor or from cells sorted from zones of root vasculature (Demura et al., 2002; Birnbaum et al., 2003; Zhao et al., 2005; Brady et al., 2007; Wenzel et al., 2008). Based on the genes identified thus far, auxin, brassinosteroid (BR) and cytokinin signals may all be needed for the correct positioning of veins (Donner et al., 2009; Caño-Delgado et al., 2010; Bishopp et al., 2011). Other factors implicated are small RNAs, peptides, transcriptional regulators, post-transcriptional regulators and translational machinery (Hirakawa et al., 2010, 2011).

We report here the systematic comparison of the transcriptomes from PV/PC and mature vascular cells from developing leaf veins, along with those of leaf mesophyll and guard cells. Individual stages were identified by molecular markers and isolated by laser microdissection for subsequent transcript profiling. We used the Athb8:GUS reporter to isolate the PV/PC cells from the first pair of emerging leaf primordia. ATHB8 is a Class-III homeodomain leucine zipper (HD–ZIPIII) protein that has previously been shown to be an early and irreversible marker of PV and PC cells in the embryo and developing organs (Baima et al., 1995). Athb8 expression is directly controlled by the auxin response factor MP/ARF5 (Donner et al., 2009). We show that the PV/PC-specific transcriptome includes many genes exhibiting strong vein pattern defects when mutated by reverse genetics, including several not previously identified. We propose that the characterization and comparison of transcriptomes between/among stages in the differentiation of specific cell types will prove to be an efficient and generalizable means of identifying genes with critical developmental roles.

Results

Comparison of cell- and stage-specific transcriptomes reveal the expression dynamics of vascular cell differentiation

We isolated four cell types by laser microdissection – mesophyll, guard cell, PV/PC and mature vascular – and obtained transcriptome data by whole-genome microarray analysis, as described in the Experimental procedures (Figure 1a–l). To distinguish PV/PC cells from mesophyll and ground cells, we visualized the PV/PC-specific expression of GUS under the control of the Athb8 promoter (Figure 1, m–o). Athb8 is a class–III HD–ZIP transcription factor that is expressed early in PV cells of emerging leaf primordia before these cells can be anatomically and histologically distinguished from the neighboring ground cells. Its expression continues into the PC cells of the developing leaf (Baima et al., 1995; Scarpella et al., 2004). We optimized the GUS staining, fixation and embedding conditions to preserve intact RNA yet to stain sufficiently to guide cell-specific laser microdissection. The RNA isolated from these cells was linearly amplified and hybridized to a 22K Arabidopisis oligo array (Agilent), as detailed in Experimental procedures.

Figure 1.

Laser microdissection of vascular bundle (a–c), guard cell (d–f) and mesophyll (g–i) from expanding third- and fourth-youngest rosette leaves of 1–month-old Arabidopsis plant. (j–l) Laser-capture microdissection (LCM) of provascular/procambial cells from first pair of emerging leaves (P1 and P2) of 7–day-old ATHB8:GUS seedlings. Shown in each set are images of a field before and after microdissection, and the cells captured from the same field onto the harvesting cap.

Each of the four isolated leaf cell types expressed a total of about 10 000 genes above background (Tables 1 and S1). A total of 1487 genes were expressed in PV/PC but not in mesophyll and guard cells. Of these, 1052 genes are expressed exclusively in the earlier PV/PC cells, and 435 genes continue expression in mature vascular cells. These two cell-specific gene sets included genes previously shown by other techniques to be expressed in the vasculature or to disrupt vein patterning when defective. Examples of these are presented as digital leaf cell in situ expression patterns in Figure S1. For several genes, an independent validation of specific expression pattern was possible by comparison with the corresponding Cold Spring Harbor Laboratory enhancer/gene trap lines (http://genetrap.cshl.org). Vascular cell expression was confirmed by both methods for the GATA-like transcription factor, 60S ribosomal protein and an AOP2 dioxygenase (Figure S2). We found that transcripts of several genes previously described as ubiquitously expressed housekeeping genes, such as nucleolin (Petricka and Nelson, 2007) and ribosomal protein STV1 (Nishimura et al., 2005), actually accumulate with distinct temporal and spatial patterns in PV or vascular leaf cell types, and that gene disruptions in these result in vein-pattern defects. This suggests that their products have specific roles in the function of these cells, rather than ubiquitous housekeeping roles.

Table 1. Genes expressed in leaf cell types
Cell typeTotal number of genes expressedCell specific/enriched
  1. a

    Includes genes that were downregulated compared with the reference control.

Provascular Cells10 1781052a
Vein9894641
Mesophyll10 625307
Guard cell9676121

Transcriptome dynamics accurately represent the expression patterns of individual genes in PV/PC and vascular cells

The ability to detect highly localized and transient expression is an advantage of cell-specific transcriptome analyses. To validate the accuracy of cell-specific transcriptome data, we observed the expression patterns of GUS reporter constructs made from representative genes from the PV/PC cell transcriptomes (Table S1). We confirmed that expression began in the PV/PC stage and extended into differentiating veins for AT1G15340 (MBD10 methylCpG binding domain containing protein), AT2G26290 (ARSK1; protein kinase), AT1G32940 (ATSBT3.5; similar to subtilase family protein), AT2G21200 (putative auxin-responsive protein) and AT1G76520 (annotated as an auxin efflux protein) (Figure 2a–d). The expression of AT2G26290 (ARSK1) has previously been reported to be specific to the roots (Hwang and Goodman, 1995) (and eFP browser http://bbc.botany.utoronto.ca/efp/cgi-bin/efpWeb.cgi). Since transcriptome data revealed this gene to be expressed in both early-stage and mature vascular cells of leaves, we tested the activity pattern in leaves of a reporter construct that included a longer promoter than previously published (3.035 kb from the 5′ upstream region of ARSK1 versus 2.3 kb from the downstream region). With the longer upstream region, we found the previously reported root expression pattern, plus dynamic expression in the leaf and emerging primordia in the cells along the path of developing veins (Figure 2b). As the gene was expressed in only a few cells at a given time point in a leaf, this highly restricted expression is unlikely to be detected via microarray analysis of the entire leaf or shoot.

Figure 2.

Promoter-GUS reporter lines for validation of the cell specificity of microarray data. (a) At1 g15340:GUS, MBD10 methyl-CpG-binding domain. (b) At2 g26290:GUS, putative protein kinase. (c) At1 g32940:GUS, subtilisin-like serine protease. (d) At2 g21200:GUS, auxin-responsive protein. (a), (c) and (d) are dark-field images, and (b) is a bright-field image; the pink/blue color shows the expression of the gene. Arrows indicate example regions of intense vascular staining. The diffuse staining pattern in immature regions (c) becomes more restricted as sites mature.

We did not observe vein defects in single insertion knock-out lines for any of the above genes, probably because each belongs to a gene family that might mask a single knock-out phenotype.

The provascular/procambial transcriptome is enriched in genes that affect vascular differentiation

Genes with previously characterized roles in vein patterning and/or expressed in the vasculature are highly represented in the PV/PC transcriptome, suggesting that the extreme spatial and temporal specificity of the samples might provide a filter for identifying other genes essential for vein formation and vascular differentiation. Previously characterized genes that are represented exclusively in the PV/PC cell transcriptome included ACL5, CVP1, BRL1, AGD1, AGD4 and VEP1. Those in which expression is detected only in the vascular cells from a mature and expanding leaf included APL, XYLOGEN, IRX1 and others. Those expressed beginning in PV/PC cells and continuing into the differentiating vascular cells included MP, IRX5, PARL1, PINHEAD/ZWILLE/AGO10, PINOID and others (Figure S3). This suggests that these leaf cell- and stage-specific data sets can be mined to recover previously unidentified developmental genes based on their highly restricted co-expression patterns.

Many of these known vein-patterning genes could have been identified by their vein-defective phenotypes in a reverse genetic strategy starting only with their PV/PC- or vascular-specific expression patterns. To test whether previously uncharacterized PV/PC-specific genes have roles in leaf vein patterning, we evaluated several corresponding insertional knock-out lines. As predicted, some of these exhibited vein defects. An example is KEG (KEEP ON GROWING), which is an E3 ligase/kinase that has a RING-HCa domain, ankyrin and HERC2-like repeats, and a functional kinase domain (Stone et al., 2006). keg mutants are hypersensitive to sugars and ABA, and accumulate high levels of ABI5. ABI5 physically interacts with KEG: the loss of ABI5 substantially, but incompletely, rescues the growth-arrest phenotype of keg mutant seedlings (Stone et al., 2006). Two insertions in the KEG gene (At5 g13530, SALK_049542 and SALK_018105) exhibited growth arrest (Figure 3a) and delayed or no vascular differentiation in the cotyledons, whereas vascular differentiation in the root remained unaffected (Figure 3b). The differentiation of vascular cells correlated with the greening of cotyledon tissue, which occurred either in one (7%) or both (15%) cotyledons of the mutant seedlings. The cotyledons that did not turn green exhibited apparently arrested PC cells that did not differentiate into vascular cells. To confirm the expression of KEG in PV/PC cells by an alternative method, we generated a KEG promoter:GUS reporter fusion construct and evaluated its expression pattern in transformed wild-type (WT/Col) plants. The KEG promoter was active throughout the emerging primordia (stages P1 and P2), and was gradually limited in activity to the PV/PC cells at stages P3 and P4, and to the veins of later leaf stages (Figure 3c,d). In roots, the KEG promoter was strongly active in the root tip and became restricted in activity to vascular tissue in the elongation and the maturation zones (data not shown).

Figure 3.

Vascular expression of KEG. (a) Wild-type Arabidopsis (left) and keg mutants (right). (b) Reduced vascular differentiation in a 7–day-old cotyledon of a keg mutant seedling. Scale bar: 10 μm. Inset shows the wild type at the same age. Scale bar: 100 μm. (c) GUS expression in the leaf of a KEG promoter:GUS reporter line. (d) Provascular expression of KEG in the emerging leaf primordia in 7–day-old seedlings with the KEG promoter:GUS reporter construct. The arrow indicates stained provascular cells in leaf primordium.

As ABA and BR are known to act antagonistically (Mandava, 1988; Friedrichsen et al., 2002), and BR biosynthetic and signaling genes have been shown to have roles in vascular differentiation, we tested the effect of BR supplementation on the keg insertion mutant [SALK_010842]. We observed partial rescue on 0.1 and 1 μm 24–epibrassinolide (24–eBL). At 10 days with supplementation, keg plants produced a first pair of leaves with evident xylem differentiation (Figure S4). The keg mutant phenotype appears to be the result of an interplay of multiple hormones, as has been observed for numerous aspects of vascular differentiation (Fukuda, 2004). The constitutive overexpression of the KEG cDNA from a 35S promoter is lethal and dexamethasone (DEX) -induced expression of KEG triggers cell death (Wawrzynska et al., 2008). If KEG has a role in regulating programmed cell death for vascular differentiation, this would explain both the PV/PC pattern of KEG expression and the vascular defects in keg mutants.

The PV/PC transcriptome includes a specific subset of the auxin signal machinery

The PV/PC and mature vascular transcriptomes together include a co-expressed and dynamic set of genes related to auxin signaling. Prominent are a subset of ARF and IAA genes (Table S1), making it possible to infer cognate partners for the auxin signaling machinery dedicated to vascular patterning. Two ARFs with well-documented roles in vascular patterning – MP and ARF4 (Wenzel et al., 2007) – are prominent in the PV/PC transcriptome. We observe significant co-expression of ARF7, ARF5, IAA1 and IAA19 in PV/PC cells. This co-expression is likely to be functionally significant, because several independent studies have revealed specific interactions among ARF5/MP, ARF7, IAA1 and IAA19 proteins (Tatematsu et al., 2004; Muto et al., 2006; Vernoux et al., 2011). Also present are transcripts of several vascular patterning genes, including GNOM, SFC, PIN1, AGO1 and others, which contain auxin-response elements (AuxREs) in their promoter regions (Guilfoyle and Hagen, 2007).

The auxin receptor complex is represented in a specialized way in the PV/PC-specific transcriptome. In the complex, TIR1 associates with an IP6 molecule that appears to be required for the stability of the receptor (Calderon-Villalobos et al., 2010). Genes in the IP biosynthetic pathway such as CVP2 influence vein patterning (Carland and Nelson, 2004, 2009). We find that inositol polyphosphate 3-/6-/5- kinase (IPK2β), which participates in the synthesis of IP6, is PV/PC-specific, in agreement with previous reports of its vascular expression (Zhang et al., 2007; Xia et al., 2003). IPK2β activity is redundant with that of the close homologs IPK1 and IPK2α. A T–DNA insertion in the single gene IPK2β resulted in a WT phenotype, in agreement with the normal levels of IP6 measured in the mutant plants (Stevenson-Paulik et al., 2005). In the ipk1 ipk2β double mutant, significant declines are observed in IP6 and in the IP precursors IP4 and IP5 in seed and developing embryo, although no shoot/seedling phenotypes were reported. The same study showed that alterations in the molar ratio of IPK1 and IPK2β in vitro resulted in the appearance of several new IP4 and IP5 species. We observed aberrant cotyledon vascular patterning in ipk1 ipk2β double mutants (Figure 4), including a bulging of tissue and callus-type outgrowths on the tips (Figure 4a,b), and disconnected and unusually thickened veins (Figure 4c). Subsequent seedling growth is normal, suggesting that IPK2α activity fully complements the defects in ipk2β and ipk1 at the later stages of plant development. As overexpression of IPK2β has been shown previously to exhibit auxin-related phenotypes, it is plausible that IPK2β (α) and IPK1 together produce specific IP compounds for cell-specific developmentally regulated auxin signaling.

Figure 4.

Double mutant of ipk1 and ipk2β. (a, b) Callus-like growth and bulging cells (arrows) in the cotyledon of a 7–day-old seedling. (c) Cleared seedling showing discontinuous veins in the cotyledon (arrow) but normal patterning in the leaf. Scale bar: 500 μm.

Knock-out mutants of PV/PC-specific transcriptional regulators exhibit vein pattern defects

Forward-genetic screens for vein pattern defects have identified several transcription factors (TFs) with roles in vein patterning, including MP, HD–ZIP III TFs and LEUNIG, which regulate vein patterning, and VND6, VND7 and APL, which regulate the formation of xylem and phloem, respectively (Caño-Delgado et al., 2010). The PV/PC and mature vascular cell transcriptomes include 94 additional TFs that are not expressed in adjacent mesophyll cells (Table S2). These vascular-specific factors are distributed among many classes of factors making up larger gene families. Single gene knock-outs in many of these genes did not result in a phenotype, probably because of gene redundancy.

One that did produce a striking vascular phenotype was the PV/PC-specific CCCH class zinc finger protein At5 g58620/AtTZF9. Two independent insertion lines (SALK_010842 and SALK_012173) exhibited an incompletely penetrant phenotype of cotyledon fusion and vascular defects. We found 5% of the seedlings showed cotyledon defects, including fusions similar to cuc (cup-shaped cotyledon) and dornroschen (drn and drnl) mutants, in some cases resulting in a trumpet-shaped cotyledon (Figure 5a,b). The veins in some fused cotyledons exhibited axialization defects (non-alignment of vascular cells, Figure 5d), and resulted in extra veins that filled the entire organ (Figure 5c). A common feature of auxin signaling and cuc mutants is the incomplete penetrance of their cotyledon phenotypes, suggesting there is redundancy in the pathways that lead to bilateral symmetry and cotyledon establishment (Chandler et al., 2007). The activity of the corresponding AtTZF9/AT5G58620 promoter is strong in the young and developing regions of the plant, with high activity in trichomes and veins (Figure 5e). Some weak expression was also seen in the epidermal cells, and very strong expression was observed throughout the root. The gene product is localized in the nucleus of cells in the root elongation zone, and additionally in the cytoplasm in root-tip cells (Figure 5f–h).

Figure 5.

Vascular expression of CCCH zinc-finger protein. (a, b) Fused cotyledon and tubular (trumpet-shaped) cotyledon phenotype of Salk insertion lines SALK_010842 and SALK_012173, respectively. (c) Cleared seedling showing vein pattern defects in the fused cotyledon. (d) Axialization defects along the vein (ends fail to meet). (e) CCCH promoter:GUS reporter line shows vascular expression. (f) Subcellular localization of the CCCH zinc-finger cDNA fused to 35S:GFP. (g) 4′,6–Diamidino-2-phenylindole (DAPI) stained image. (h) Differential interference contrast (DIC) image. Scale bars: C, 500 μm; D, 250 μm; F, G, H, 100 μm.

Discussion

Cell-specific expression and co-expression as criteria for identifying developmental genes

We employed a highly selective laser microdissection workflow that enabled us to obtain transcriptome data sets from the transiently present PV and PC cells that represent developmental stages prior to vein formation in Arabidopsis leaves. By comparison with similarly obtained cell-specific transcriptomes from mesophyll, guard and vascular cells, we identified sets of genes exclusively expressed in PV and PC cells. As these sets included many genes known from prior studies to be involved in vein formation, we performed reverse genetic characterizations of previously unstudied genes from this set, and found several with vascular phenotypes within the modest subset we tested. This suggests that the filtering of genes based on their cell and developmental specificity of expression is an effective strategy for the discovery of candidate genes with roles specific to particular cells or developmental processes. In addition, this strategy provides a wealth of co-expression information that can be employed to associate cognate binding partners (e.g. Aux/IAA-ARF pairs) or multistep pathway members drawn from otherwise confounding multigene families, based on their expression within the same cell type within a narrow developmental time window.

Vascular development gene discovery via reverse genetics

In addition to the example PV/PC genes characterized in the Results, we identified others with candidate roles in vascular development, including putative spatial patterning genes, hormone signaling genes, pseudo-housekeeping genes, defense genes and TFs. As a validation of the microdissection/co-expression gene discovery strategy, we discuss a selection of the output we obtained.

Spatial patterning genes

Members of the HD–ZIP III gene family have roles in adaxial/abaxial leaf cell polarity and in vascular patterning and formation. Some are differentially regulated by microRNAs 165 and 166 (Zhong and Ye, 2007; Zhou et al., 2007). We found that ZWILLE expression in the leaf is limited to the PV/PC cells and the veins, whereas AGO1 is expressed throughout. ZWILLE/PINHEAD/AGO10 was implicated in miRNA-directed translational inhibition and repression of miR165/166 levels, thereby affecting leaf polarity (Liu et al., 2009). The AGO1-dependent microRNA regulation of CUC2, PHB and PHV is also involved in determining leaf polarity and boundaries (Kidner and Martienssen, 2005). ZWILLE/AGO10 negatively regulates the protein levels of AGO1 and thus affects RNA silencing (Mallory et al., 2009). ZWILLE expression is also required in the vascular primordium during embryogenesis (Tucker et al., 2008). As mutations in vascular patterning genes affecting axes of polarity are often associated with leaf patterning defects (Bowman and Floyd, 2008), a PV/PC-specific gene discovery strategy might benefit the understanding of both vein and leaf patterning.

Enrichment of BR genes

The process of vascular patterning and differentiation requires the interplay of auxin, BR and cytokinins, each with well-characterized signaling pathways. We used the AraCyc pathway resource to test whether auxin, BR and CK pathway components are over-represented in PV/PC transcriptomes, compared with other leaf cell types. Of these, the BR pathway is notably emphasized (Figure S5). BRs regulate the transdifferentiation of xylem from mesophyll cells in Zinnia (Fukuda et al., 1994; Nagata et al., 2001; Fukuda, 2004; Ohashi-Ito and Fukuda, 2010), and regulate the size and number of vascular bundles in both rice and Arabidopsis (Szekeres et al., 1996; Choe et al., 1999; Caño-Delgado et al., 2004; Nakamura et al., 2006). BR and auxin together have documented roles in vein patterning and differentiation, and the regulatory regions of some BR-responsive genes are enriched with ARF binding sites (Goda et al., 2004; Nemhauser et al., 2004). Conversely, many auxin pathway genes are upregulated in response to BR (Goda et al., 2002, 2004). Auxin and BR signaling pathways appear to be interconnected at the transcriptional level via a negative feedback loop (Jung et al., 2010).

Ribosomal protein genes

Several ribosomal protein genes are expressed in the leaf exclusively in the PV/PC transcriptome. The expression pattern of At2 g44120 (60S ribosomal subunit, ET1101, gene trap line) mimics the pattern of DR5 (synthetic reporter of auxin levels) (Ulmasov et al., 1997). STV1 (short valve 1) and several other ribosomal genes are also PV/PC-specific, suggesting that they tailor ribosomes to vascular-specific processes (Nishimura et al., 2005). It is possible that the selective translation of auxin response factors by specific ribosomal proteins facilitates the microRNA-mediated regulation of these genes. The cell specificity of transcripts for these ribosomal targeting proteins may be one element of this regulation.

Defense genes

The PV/PC and vascular cell transcriptomes are enriched in certain defense response genes. The KEG gene, which we show influences vascular differentiation, was previously identified as a suppressor of EDR1 (enhanced disease resistance 1), which confers enhanced resistance to infection by powdery mildew (Wawrzynska et al., 2008). Other vein patterning and differentiation genes may have induced roles in resistance to pathogens that attack or spread through the vascular system.

Transcription factors

Few TFs have been associated with vascular patterning by mutant phenotypes (Ohashi-Ito and Fukuda, 2010), probably because plant TFs tend to belong to gene families with potential functional redundancy. The PV/PC and mature vascular transcriptomes include TFs not previously associated with vascular functions (Table  S2). Of these, about 50% have ARF-binding motifs (TGTCTC element) within 1000 bp upstream of their transcriptional start site.

The vascular-specific AtTZF9/AT5G58620 that we identified and characterized above is one such TF. Similar CCCH zinc-finger proteins have been associated with mRNA processing in Arabidopsis. AT5G58620 belongs to subfamily IX (11 members) of the 68-gene CCCH zinc-finger protein family in Arabidopsis, the members of which contain the TZF (tandem zinc finger) motif (Wang et al., 2008; Pomeranz et al., 2010). Others in this class are PEI1, AtSZF1, AtSZF2 and SOMNUS, with roles in embryogenesis (Li and Thomas, 1998), salt stress response (Sun et al., 2007) and light-dependent seed germination. TZF proteins can recruit and activate mRNA decay enzymes for silencing mRNAs with AU-rich elements (ARE) in the 3′ untranslated region (3′–UTR). AtTZF1 shuttles between the nucleus and cytoplasmic processing bodies (PBs), which are sites of storage of mRNA protein complexes and of micro-RNA-mediated mRNA decay, and co-localize with AGO1, DCP2 and XRN4 (RNA processing enzymes) (Pomeranz et al., 2010). DCP1, DCP2 and Varicose together form a decapping complex (Xu et al., 2006). It is therefore plausible that TZF9 along with VCS and DCP genes might be involved in mRNA decay specific to vascular patterning and differentiation.

The cotyledon fusion phenotype of the AtTZF9 knock-out mutant is similar to that of dornroshen mutants (DRN and DRNL), which interact with the vascular-associated ARF MP directly through canonical ARF-binding motifs (Cole et al., 2009). An ARF-binding motif is also present in the AtTZF9/AT5G58620 promoter about 2.3 Kb upstream from the transcriptional start site. The BR signaling component BIM1 interacts with DRN, DRNL and PHV, perhaps as part of a multimeric TF complex affecting cotyledon development (Chandler et al., 2009). It is tempting to speculate that AtTZF9 participates in a similar or overlapping auxin-regulated complex dedicated to cotyledon development.

Several, and perhaps all, Arabidopsis CCCH zinc-finger domain proteins possess ribonuclease activity, similar to those in bacteria that act in programmed cell death, stress responses and in processes that require the reprogramming of gene expression (Addepalli and Hunt, 2008). In Arabidopsis, the roles of CCCH genes may be targeted to specific processes such as vein patterning.

Prospects

The use of cell-specific and temporal co-expression patterns to identify related sets of developmental genes should be successful for many localized developmental processes in plants, particularly when augmented with more sensitive methods of transcriptome analysis. We anticipate that the additional sensitivity of RNA-seq and similar transcriptome inventory methods will permit a deeper gene discovery strategy, based on cell and developmental specificity, including the characterization of small non-coding RNAs, the comparison of development of different vein orders, and the evaluation of embryonic vascularization. In the case of vascular development, based on the numbers we recovered by microarray analysis (Approximately 80% genome coverage with the Agilent 22k oligoarray), we estimate that RNA-seq transcriptome methods might add another 2000–3000 protein-coding genes to the list of expressed genes for each of the leaf cell types, of which 200–300 are likely to be specific to the PV/PC transcriptome.

Experimental procedures

Plant growth conditions

Seeds of Arabidopsis thaliana Columbia ecotype were surface-sterilized with 30% bleach and 0.1% Triton X–100, and were germinated on a medium containing 1% agar, full-strength MS media (Sigma M0404) and 10 g l−1 sucrose under continuous light. Selected seedlings were transferred to soil (2:1 mix of Fafard Super Fine Germination Mix:vermiculite; Conrad Fafard, http://www.fafard.com) and grown under photoperiodic cycles of 16 h of light and 8 h of dark at 21°C.

Marking of PV/PC cells for laser capture microdissection

Seven-day-old seedlings carrying the ATHB8 promoter:GUS transgene were transferred to a glass vial with 10 ml of GUS-staining solution (100 mm sodium phosphate buffer, pH 7.0, 10 mM EDTA, 3 mm potassium ferricyanide, 1% Nonidet-P40 or 0.2% Triton X–100 and 3 mm Xgluc; Gold Biotechnology, https://www.goldbio.com), vacuum infiltrated and incubated for 2 h at 37°C in the dark. Stained seedlings were washed twice with 0.1 M phosphate buffer, transferred to ice-cold 100% acetone solution and vacuum-infiltrated on ice. Acetone was changed immediately after vacuum infiltration, and again after an hour of incubation. Subsequent incubation was 1 h in 100% acetone, 2 h in 1:1 acetone:Histoclear and three changes of 1 h each in 100% Histoclear. Paraffin wax (Paraplast Xtra) pellets were added to the Histoclear solution to saturation, followed by several changes of wax solution at 65°C over a period of 2 days and embedding in microtome moulds. We designed this rapid staining, fixation and embedding protocol to reveal ATHB8-marked PV/PC cells, and to minimize RNA degradation during staining. The intactness of total RNA was evaluated at each step of staining, fixation and embedding by Bioanalyzer (Agilent, http://www.home.agilent.com) assay. Samples were subjected to laser microdissection, as for non-stained samples below.

Laser capture microdissection

For guard, mesophyll and differentiated vascular cells, third- and fourth-youngest rosette leaves were harvested from 1–month-old Arabidopsis plants and immediately fixed in ice-cold 100% acetone or 3:1 ethanol:acetic acid. The tissue was then embedded in paraffin wax after gradual changes of acetone into Histoclear to wax, and cut by microtome into 8–10–μm sections, as described previously (Kerk et al., 2003; Jiao et al., 2009). Guard cell, mesophyll and vascular bundle cells were captured using the Pix Cell IIE laser capture microdissection system (Arcturus,http://www.appliedbiosystems.com), as were the stained PV/PC cells from sections prepared as described above.

RNA isolation, amplification and microarray hybridization

Total RNA was isolated from the individual cell types using a Picopure RNA Isolation Kit (Arcturus/Molecular devices) with DNAse treatment, and the mRNA was linearly amplified through two rounds using the RiboAmp OA kit (Arcturus/Molecular devices). The amplified RNA (aRNA) was labeled using amino allyl UTP (Ambion, now Invitrogen, http://www.invitrogen.com) in the second and last round of in vitro transcription (MessageAmp, kit 1750; Ambion), and hybridized to a 22K oligo microarray (Agilent 22 K Arabidopsis microarray, 21 500 features, 60 nucleotide probes and one probe per gene), representing 20 689 genes (Chen et al., 2007). Three biological replicates per cell type were used. The probe for each hybridization consisted of an equal volume of amino-allyl aRNA that was coupled with Cy3 or Cy5 mono-reactive NHS esters for 30 min in the dark according to the Ambion dye-coupling protocol, quenched with 4 M hydroxylamine HCl and separated from uncoupled dye molecules on RNeasy columns (Qiagen, http://www.qiagen.com). Probe elutions, microarray hybridizations (duplicate with dye swaps), scanning of arrays and statistical processing of data were performed as described previously (Tang et al., 2006).

Identification of T–DNA insertional lines and microscopy

T–DNA insertion lines were obtained from the ABRC stock center (http://abrc.osu.edu; Alonso et al., 2003). Whenever available, more than one insertion line was used to evaluate the phenotype corresponding to each gene. Cotyledons and the first pair of true leaves were harvested, fixed and cleared in cold 3:1 ethanol:acetic acid overnight, followed by transfer into 70, 90, 95 and 100% ethanol. The leaves were then cleared in Histoclear for 1–2 min and mounted in 2:1 Permount:xylenes. The tissue was viewed and photographed in a Zeiss Axiophot microscope under bright and dark fields (Zeiss, http://corporate.zeiss.com).

Promoter GUS cloning

The promoter region of the At5 g13530 gene (from –1 to –1046 bp, from the start codon of At5 g13530 upstream to the stop codon of the previous gene At1 g13520 in the opposite orientation) was PCR amplified from genomic DNA isolated from Arabidopsis ecotype Columbia (Col–0), and subcloned into pCR 2.1 TOPO vector (Invitrogen). It was then cloned into the binary vector pBI101.1 (Clontech, http://www.clontech.com) using SAL1 and XBA1 restriction sites. The plasmid was then transferred into Agrobacterium GV3101 for transformation into Col–0 using the floral-dip method (Clough and Bent, 1998). Transformants were selected on medium containing 50 mg mL−1 kanamycin; at least seven independent lines were analyzed.

Similarly, upstream regions of 790 bp, 2.874 kb, 910 bp and 3.037 kb were subcloned from At1 g15340, At1 g32940, At2 g21200 and At2 g26290, respectively, into the pCR 2.1 TOPO vector (Invitrogen), and were then cloned into the binary vector pBI101 using the restriction sites and primers indicated in Table S3. The plasmid was then transferred into Agrobacterium GV3101 for transformation into plants using the floral-dip method.

PCR primers were designed with additional attB sites at the ends and used to amplify 2.905 kb upstream from the initiation codon of the At5 g58620 gene from genomic DNA (Table S4). Amplified PCR product was recombined with the pDONR/Zeo vector (Invitrogen) using the BP clonase enzyme to get the entry vector that was recombined with the destination vector pGWB3 using the LR clonase–II enzyme. The recombined destination vector was then transferred into Agrobacterium GV3101 for transformation into Col–0 using the floral-dip method. Transformants were selected on agar media containing 50 mg mL−1 kanamycin.

GFP vector construction

At5 g58620 cDNA (stock U13953) was PCR amplified and cloned into SmaI/BamHI sites of the pEGAD binary vector (Cutler et al., 2000) in frame to generate the N–terminal GFP fusion p35S:GFP:At5 g58620 cDNA that was transferred into Agrobacterium GV3101 for transformation into Col–0. Transformants were selected on medium containing 50 mg mL−1 kanamycin.

Acknowledgments

This work was supported by an NSF award IOS-0718881 to TN. The authors thank the ABRC and J Stevenson-Paulik and JD York for providing IPK mutant seeds.

Ancillary