Regulation of preprocambial cell state acquisition by auxin signaling in Arabidopsis leaves

The vascular system of plants is composed of bundles of cell files that extend and intersect throughout all organs(Esau, 1965). Vascular bundles differentiate from procambial cells: narrow, cytoplasm-dense cells,characteristically arranged in continuous strands(Esau, 1943), which in leaves seem to emerge de novo from within the morphologically homogeneous population of apparently naïve ground cells(Foster, 1952; Pray, 1955).

Although the molecular details are not entirely clear, a role for the polarly transported plant signaling molecule auxin in the selection of leaf ground cells that will elongate to acquire procambial cell identity has increasingly gained experimental support(Mattsson et al., 2003; Mattsson et al., 1999; Sachs, 1981; Sachs, 1989; Scarpella et al., 2006; Sieburth, 1999; Wenzel et al., 2007). During leaf development, polygonal, isodiametric ground cells are shunted towards procambial fate through induction of broad domains of expression of the PIN1 auxin exporter (Scarpella et al.,2006; Wenzel et al.,2007). Decay of PIN1 expression and associated relapse to ground state occur in some of the cells initially expressing PIN1, and domains of PIN1 expression are eventually curtailed to individual files of cells that will stretch into procambial cells(Scarpella et al., 2006; Wenzel et al., 2007). Because formation of leaf vascular bundles, or veins, is reiteratively propagated in leaf development (reviewed by Nelson and Dengler, 1997), cells that have reverted to ground state may have other opportunities to assume procambial identity before adopting the alternative mesophyll fate (Scarpella et al., 2004; Scarpella et al.,2006; Wenzel et al.,2007).

While onset of PIN1 expression marks an unstable and reversible state in vein formation, files of PIN1-expressing ground cells that are stabilized towards procambial fate initiate expression of the homeodomain-leucine zipper(HD-Zip) III gene ATHB8 (Baima et al., 1995; Kang and Dengler,2004; Sawchuk et al.,2007; Scarpella et al.,2004). Available evidence suggests that ATHB8 expression identifies a crucial and typically irreversible stage in procambial cell fate acquisition: under both undisturbed and perturbed conditions, adoption of the ATHB8 `preprocambial' cell state accurately predicts sites of vascular differentiation (Alonso-Peral et al., 2006; Carland and Nelson,2004; Cnops et al.,2006; Kang and Dengler,2004; Koizumi et al.,2000; Petricka and Nelson,2007; Pineau et al.,2005; Sawchuk et al.,2007; Scarpella et al.,2004; Scarpella et al.,2006), and the ATHB8 preprocambial state is mutually exclusive with a `premesophyll' cell state that presages mesophyll fate assignment (Sawchuk et al.,2008). Whereas the preprocambial state is defined by the onset of ATHB8 expression, differential ATHB8 expression is only the object of a cell state transition and therefore does not necessarily provide information about the underlying patterning mechanism. Instead, knowledge of the set of transcription factors that determine initiation of ATHB8expression at the correct spatiotemporal coordinates in leaf development might provide insight into how the preprocambial cell state arises at defined positions and stages during leaf development.

Here, we show that ATHB8 is required to stabilize preprocambial cell specification against auxin transport perturbations, to constrict preprocambial cell state acquisition to narrow zones and to synchronize procambial cell identity assignment within and between veins. Further, we show that ATHB8 preprocambial expression is directly and positively controlled by the auxin-response transcription factor MONOPTEROS (MP) through an auxin-response element in the ATHB8 promoter. Finally, we show that ATHB8 functions in vein formation strictly depend on MPactivity. Our results suggest a molecular mechanism through which general auxin signal transduction is specifically translated into leaf vascular patterning inputs.

Sequences of primers and details of cloning are available on request. To generate the DR5Rev(9x)_pro:ECFP-Nuc construct, nine copies of the DR5Rev sequence (Ulmasov et al., 1997) upstream of the -46 cauliflower mosaic virus 35S promoter (Fang et al.,1989) were recombined into the pBGCN vector(Kubo et al., 2005). To generate the ATHB8_pro:ATHB8:mCherry construct, the mCherry coding sequence (Shaner et al., 2004)was cloned downstream of the fragment of the ATHB8 gene from -1997 to+4233. Functionality of the construct was tested by transformation into homozygous athb8-11 plants and by assessing normalization of sensitivity towards 5 μM 1-N-naphthylphthalamic acid (NPA; Sigma Aldrich,St Louis, MO, USA) in two independent, single-insertion transgenic lines. To generate the ATHB8 promoter deletion and mutation constructs,amplified fragments were recombined into the pFYTAG vector(Zhang et al., 2005). To generate the MP_pro:MP:ECFP construct, the ECFP coding sequence(Clontech Laboratories, Mountain View, CA, USA) was cloned at position +3815 of the fragment of the MP gene from -3311 to +4301. Functionality of the construct was tested by crossing two independent, single-insertion lines to heterozygous mp^U55 plants and by assessing rescue of the root phenotype in the F2 generation. To generate the UBQ10_pro:MP:GR construct, the sequence encoding amino acids 508-795 of the rat glucocorticoid receptor(Sablowski and Meyerowitz,1998) was cloned downstream of the fragment of the MPcDNA (Hardtke and Berleth,1998) from +1 to +2696 and controlled by the UBQ10promoter (Sawchuk et al.,2008).

The origins of the PIN1_pro:PIN1:EYFP, J1721:mGFP5er,Q0990:mGFP5er, ATHB8_pro:HTA6:EYFP, UBQ10_pro:EGFP:LTi6b,ATHB8_pro:ECFP-Nuc, athb8-11 and athb8-12 lines have been published (Xu et al.,2006; Sawchuk et al.,2007; Sawchuk et al.,2008; Prigge et al.,2005). The mp^U55 line contains a G-to-A transition at position +1237 that disrupts the splicing acceptor site of the sixth intron and is predicted to result in loss of sequences strictly required for DNA binding (Ulmasov et al.,1999). The proportion of mp^U55 seedlings with single cotyledons (55/116, 47%), fused cotyledons (27/116, 23%) and two cotyledons (34/116, 29%) meet established criteria that define strong mp alleles (Berleth and Jurgens,1993). Therefore, both molecular and morphological evidence indicate the extreme severity of the mp^U55 mutation. The SALK_021319 line was confirmed to contain a single T-DNA insertion at position+3422 of the MP (ARF5) gene and was therefore renamed arf5-2. Sequences of primers used for genotyping are available on request.

Seeds were sterilized and germinated, and seedlings and plants were grown,transformed and selected as described(Sawchuk et al., 2008; Sawchuk et al., 2007). For auxin transport inhibition, seeds were germinated on growth medium supplemented with NPA. For auxin or dexamethasone induction, seeds were germinated on growth medium, transferred at 3.5 days after germination (DAG)to liquid growth medium supplemented with 10 μM 2,4-dichlorophenoxyacetic acid (Sigma Aldrich) or 30 μM dexamethasone (Sigma Aldrich) and incubated with shaking at 50 rpm under normal growth conditions for 16 hours prior to imaging.

Dissected leaf primordia were mounted and imaged as described(Sawchuk et al., 2008; Sawchuk et al., 2007). Detailed information on imaging parameters is available upon request. To visualize xylem patterns, leaves were cleared as described(Scarpella et al., 2004) and viewed under dark-field illumination with an Olympus SZ61TR stereomicroscope(Olympus Corporation, Tokyo, Japan). Images were captured with an AxioCam HR camera (Carl Zeiss, Oberkochen, Germany).

Brightness and contrast were not altered for images of mock treatments and induced gene expression. For all other images, brightness and contrast were adjusted through linear stretching of the histogram in ImageJ (National Institutes of Health, http://rsb.info.nih.gov/ij). Signal levels were visualized by applying look-up tables (available on request). Signal colocalization was visualized as described(Sawchuk et al., 2008).

Chromatin immunoprecipitation was performed essentially as described(Jackson, 1978; Nagaki et al., 2003; Palma et al., 2007; Ponnusamy et al., 2008). Briefly, nuclei were isolated from ∼2000 4-DAG seedlings (equivalent to∼2.5 g) per genotype per biological replicate, nuclear proteins were crosslinked to DNA with formaldehyde, chromatin was digested with micrococcal nuclease, and DNA-crosslinked fluorescent proteins were isolated using theμMACS GFP Isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany). Enrichment of DNA of putative target regions in the ATHB8 promoter was determined as described (Schubert et al., 2006) using a sequence of the UBQ10 promoter to normalize results (Martin-Trillo et al.,2006). Details of primers are available upon request.

Null mutants of ATHB8 display no obvious alterations in the vein patterns of mature leaves (Baima et al.,2001; Prigge et al.,2005). We first asked whether ATHB8 could be assigned to any distinct function in vein formation or whether ATHB8 activity is completely dispensable for this process. In Arabidopsis, veins of subsequent orders become recognizable progressively later in the same area of the developing leaf primordium, and veins of the same order appear in a tip-to-base sequence during leaf development(Candela et al., 1999; Kang and Dengler, 2002; Kang and Dengler, 2004; Kinsman and Pyke, 1998; Mattsson et al., 2003; Mattsson et al., 1999; Scarpella et al., 2004; Scarpella et al., 2006; Sieburth, 1999; Steynen and Schultz, 2003; Telfer and Poethig, 1994; Wenzel et al., 2007). Fig. 1A-D schematically depict the temporal sequence of vascular development events in Arabidopsisleaf primordia and define the stages and terminology to which we refer throughout this study.

Selection of ground cells that will acquire a preprocambial state,visualized through the dynamics of PIN1 expression(Scarpella et al., 2006; Wenzel et al., 2007),proceeded similarly in leaf primordia of wild type and of the null athb8-11 mutant (see Fig. S1 in the supplementary material). However,we observed distinct anomalies in the assignment of the preprocambial cell state and procambial cell identity, marked by J1721:mGFP5er and Q0990:mGFP5er expression, respectively (Sawchuk et al.,2007), during athb8-11 leaf development (for the attributes used to assess the representative nature of all displayed features and derived reproducibility quotients, see Table S1 in the supplementary material).

In wild type, J1721:mGFP5er expression is initiated in files of individual ground cells of the leaf primordium that coexpress ATHB8 and that will successively elongate to acquire procambial cell identity(Sawchuk et al., 2007). Whereas at all stages of wild-type leaf development, J1721:mGFP5er expression was invariably constrained to narrow zones(Sawchuk et al., 2007)(Fig. 1E-H), newly emerged J1721:mGFP5er expression domains encompassed wide fields of cells in athb8 leaf primordia (Fig. 1I-L). At later stages of vein development in athb8leaves, J1721:mGFP5er expression was, nevertheless, confined to strands of one or very few cell files (Fig. 1J-L).

In wild-type leaf development, Q0990:mGFP5er expression first emerges in files of elongated, ATHB8-expressing procambial cells, and all cells initiate Q0990:mGFP5er expression simultaneously throughout the length of a developing vein (Sawchuk et al.,2007). During unperturbed development, Q0990:mGFP5er expression is activated in a coordinated fashion in loop-forming lateral and marginal veins,such that expression appears simultaneously along entire vein loops(Sawchuk et al., 2007)(Fig. 1O,P). In athb8leaf development, however, Q0990:mGFP5er expression was switched on separately in lateral and marginal veins, and ectopic foci of transient epidermal expression were detected at the leaf margin during the development of the midvein and all loops (Fig. 1Q-T).

In summary, our results suggest that ATHB8 is required to circumscribe preprocambial cell state assignment to narrow domains of ground cells and to integrate procambium identity acquisition within and between veins.

The appearance of expanded zones of J1721:mGFP5er expression and the asynchronous emergence of Q0990:mGFP5er expression domains observed during athb8 leaf development are reminiscent of marker behavior under conditions of mild auxin transport inhibition(Sawchuk et al., 2007). Therefore, we next asked whether athb8 leaves displayed altered sensitivity to the auxin transport inhibitor 1-N-naphthylphthalamic acid(NPA). Leaves of plants germinated and grown in the presence of auxin transport inhibitors are characterized by a number of distinct anomalies in vascular organization (Mattsson et al.,1999; Sieburth,1999), most evident as great numbers of broad lateral veins and fusion of marginal veins to give rise to a continuous wide zone of vascular differentiation that extends along the entire margin of the leaf(Fig. 2C). Because these responses are quantifiable and NPA concentration dependent(Mattsson et al., 1999; Sieburth, 1999) (see Fig. S2 in the supplementary material), they can be used to assess sensitivity to auxin transport inhibition. At low concentrations of NPA, athb8-11leaves showed greater numbers of lateral veins, and a higher fraction of athb8-11 leaves displayed the formation of a marginal vascular differentiation zone than wild-type leaves(Fig. 2A-D; see Fig. S2 in the supplementary material), suggesting that vein development in athb8-11is more susceptible to auxin transport inhibition. Similarly, leaves of the weaker athb8-12 allele (Prigge et al., 2005) displayed hypersensitivity to NPA (see Fig. S2 in the supplementary material).

Wild-type leaves developing under conditions of reduced auxin transport display an expansion of PIN1 expression domains proportional to the level of auxin transport inhibition (Scarpella et al., 2006; Wenzel et al.,2007). We therefore asked whether the exaggerated response of vein patterns to auxin transport inhibition in athb8 was associated with enhanced broadening of PIN1 expression fields under the same conditions. Concentrations of NPA that evoked a maximum differential response of vein patterns in athb8 versus wild type resulted in higher levels and wider domains of PIN1_pro:PIN1:EYFP expression in athb8-11than in wild-type leaves (Fig. 2E-H). Because PIN1 expression in leaves is auxin inducible(Scarpella et al., 2006; Wenzel et al., 2007), we asked whether the exaggerated response of PIN1 expression to auxin transport inhibition in athb8 could be attributable to abnormal auxin sensitivity. In leaves, the synthetic DR5 promoter(Ulmasov et al., 1997b) serves as a cell type-independent reporter of auxin response(Mattsson et al., 2003). Levels and patterns of DR5Rev(9x)_pro:ECFP-Nuc expression in athb8-11 leaves, either under control conditions or upon treatment with exogenous auxin, were comparable to those in wild type(Fig. 2I-L).

We conclude that ATHB8 is required for normal sensitivity of PIN1 expression and vascular patterns to auxin transport inhibition in the leaf,but that ATHB8 appears expendable for leaf auxin response.

In agreement with previous observations(Baima et al., 1995; Kang and Dengler, 2004; Sawchuk et al., 2007; Scarpella et al., 2004), we found that the 2.0 kb sequence upstream of the ATHB8 translational start site is sufficient to drive expression of a nuclear-localized yellow fluorescent protein (HTA6:EYFP) (Zhang et al., 2005) in isodiametric cells of the leaf primordium that have been recruited towards Q0990:mGFP5er-labeled procambium formation(Fig. 3A,B) and that have therefore been designated as preprocambial cells(Mattsson et al., 2003). In leaf development, the intensity of ATHB8_pro:HTA6:EYFP signals was sustained in elongated procambial cells and eventually declined during late stages of vascular differentiation (Fig. 3B,C). As previously reported(Kang and Dengler, 2002; Scarpella et al., 2004), we additionally observed ephemeral activity of the ATHB8 promoter at the leaf tip and hydathodes, where not all ATHB8-expressing cells will differentiate into vasculature (Fig. 3D; see Fig. S3 in the supplementary material).

Post-transcriptional regulation has been shown to spatially constrain fields of ATHB8 expression in the root(Lee et al., 2006). To test whether post-transcriptional control impinges on domains of ATHB8expression in the leaf, we first visualized expression of a functional (see Materials and methods) translational fusion of ATHB8 with the red fluorescent protein mCherry (Shaner et al., 2004) during leaf development. Expression of ATHB8_pro:ATHB8:mCherry was initiated in polygonal cells of the leaf primordium (Fig. 3G),maintained in elongated Q0990:mGFP5er-marked procambial cells(Fig. 3H), extinguished during terminal vascular differentiation (Fig. 3I), and was always absent at the leaf tip and hydathodes(Fig. 3J; see Fig. S3 in the supplementary material). Therefore, with the exception of the leaf tip and hydathode nonvascular areas, ATHB8:mCherry accumulation profiles are accurately recapitulated by ATHB8 promoter-driven expression dynamics. We next asked whether imaging patterns of ATHB8 promoter activity and tagged ATHB8 protein localization within the same sample could reveal subtle differences that would go unnoticed in comparative analyses performed on separate samples. Covisualization of ATHB8_pro:HTA6:EYFP and ATHB8_pro:ATHB8:mCherry signals showed coincident expression of the two fluorescent proteins(Fig. 3M-O), suggesting that ATHB8 promoter activity parallels ATHB8 protein expression in vascular cells and that post-transcriptional regulation does not revise vein-associated domains of ATHB8 expression in the leaf.

Because ATHB8 expression is predicted by the activity of its upstream non-coding sequences, to identify regulatory elements required for preprocambial expression, we generated a series of ATHB8 promoter variants. All the promoter fragments were fused to the nuclear-localized HTA6:EYFP in the context of the pFYTAG binary vector(Zhang et al., 2005). HTA6:EYFP driven by the cauliflower mosaic virus 35S -47 minimal promoter(Fang et al., 1989) was not able to generate detectable levels of YFP fluorescence in transgenic plants(see Fig. S4 in the supplementary material), suggesting that the T-DNA in the pFYTAG binary vector does not contain cryptic regulatory elements. Three criteria were sequentially adopted to test preprocambial expression of the promoter fragments: (1) stereotypical expression in leaves 4 days after germination (DAG), as inferred by comparison with expression directed by the 2.0 kb promoter fragment; (2) isodiametric shape of cells first expressing the promoter fragments, as determined by simultaneous visualization of ubiquitously expressed plasma membrane-localized GFP(UBQ10_pro:EGFP:LTi6b) (Sawchuk et al., 2008); (3) colocalization of the onset of expression with that of a nuclear-localized CFP driven by the 2.0 kb ATHB8 promoter fragment (ATHB8_pro:ECFP-Nuc)(Sawchuk et al., 2007). Finally, vascular expression at stages later than preprocambial, as suggested by failure to satisfy criterion 1, was independently tested by simultaneous visualization of the procambial marker Q0990:mGFP5er(Sawchuk et al., 2007).

To initially demarcate the regulatory sequences that are required for ATHB8 preprocambial expression, we generated a series of 0.5 kb 5′ deletions of the 2.0 kb ATHB8 promoter (see Fig. S5 in the supplementary material). All deletions were designed so as not to interrupt putative cis-acting elements identified by available bioinformatics resources. The sequence of the ATHB8 promoter between -964 and -1 was the shortest promoter fragment able to direct preprocambial expression(Fig. 4A-H), as the -501 to -1 fragment did not promote any leaf expression(Fig. 4I). This suggests that the 463 bp region of the ATHB8 promoter between -964 and -501 is necessary for ATHB8 preprocambial expression.

To test whether the ATHB8 5′ UTR in the -964 to -1 promoter fragment is essential to drive preprocambial expression, we deleted the sequence immediately downstream of the ATHB8 transcriptional start site, as predicted by sequence alignment with the furthest upstream EST available (AV830211), while conserving a putative initiator sequence(Smale and Kadonaga, 2003)centered around the predicted transcription start site (see Fig. S5 in the supplementary material). The resulting 188 bp region of the ATHB8promoter between -964 and -776 was still able to impart preprocambial expression (Fig. 4J-L),suggesting that the ATHB8 leader sequence is dispensable for preprocambial expression.

To further define the regulatory sequences required for ATHB8preprocambial expression, we generated a 5′ deletion of the -964 to -776 region at position -927 (see Fig. S5 in the supplementary material). The -927 to -1 promoter fragment was not able to drive vascular expression at preprocambial stages (Fig. 4M-P), suggesting that the 37 bp region of the ATHB8promoter from -964 to -927 is necessary for preprocambial expression.

Finally, to more precisely map the sequences essential for ATHB8preprocambial expression, we generated progressive 5′ deletions of the-964 to -927 region (see Fig. S5 in the supplementary material). Because the-940 to -1 promoter fragment was not able to direct vascular expression at preprocambial stages (Fig. 4U-X), the region from -953 to -1 constitutes the shortest fragment that still promoted preprocambial expression(Fig. 4R-T). This suggests that the 13 bp sequence between -953 and -940 of the ATHB8 promoter is indispensable for ATHB8 preprocambial expression.

Interrogation of available databases of regulatory elements predicted the presence of core binding sites for MYB and GT1 transcription factors in the-953 to -940 ATHB8 promoter fragment that is required for preprocambial expression (see Fig. S5 in the supplementary material). By manual inspection, we further identified a TGTCTG motif, which is a variant of the TGTCTC auxin-response element (ARE)(Li et al., 1994; Liu et al., 1994) (see Fig. S5 in the supplementary material). To test whether any of these putative regulatory elements are necessary for ATHB8 preprocambial expression,we generated variants of the -953 to -1 ATHB8 promoter fragment in which each of the three elements was individually mutated so as to abolish binding of the predicted transcription factor as previously determined experimentally (Gubler et al.,1999; Ouwerkerk et al.,1999; Ulmasov et al.,1997a; Ulmasov et al.,1997b) (see Fig. S5 in the supplementary material). As shown in Fig. 5, mutations in the MYB or GT1 presumed binding sites had no effect on the activity of the -953 to -1 ATHB8 promoter fragment (Fig. 5A,C,D). However, mutation in the hypothetical ARE resulted in loss of expression at preprocambial stages(Fig. 5E-H). Furthermore, the expression induced by the -953 to -1 ATHB8 promoter fragment containing the mutated TGTCTG element was indistinguishable from that induced by the -940 to -1 fragment, which deleted the entire region containing the MYB and GT1 core recognition sequences and the putative ARE(Fig. 5B,E). This suggests that the TGTCTG sequence in the -953 to -940 region of the ATHB8 promoter is required for ATHB8 preprocambial expression.

The TGTCTG element that we identified as being indispensable for ATHB8 preprocambial expression is very similar to the TGTCTC element necessary for the auxin response (Li et al., 1994; Liu et al.,1994), and ATHB8 expression is auxin inducible(Baima et al., 1995). To test whether ATHB8 auxin responsiveness depends on the TGTCTG element and whether such a sequence therefore represents a functional ARE, we monitored patterns of HTA6:EYFP fluorescence conferred by ATHB8 promoter variants in 4-DAG leaves after their exposure to the synthetic auxin 2,4-dichlorophenoxyacetic acid (2,4-D), comparing them to HTA6:EYFP expression in mock-treated samples. Auxin inducibility of the 2.0 kb ATHB8promoter (Fig. 6A,B) was largely retained by the -953 to -1 promoter fragment(Fig. 6C,D), suggesting that the region of the ATHB8 promoter between -1997 and -953 does not contribute significantly to auxin-regulated ATHB8 expression. Conversely, auxin responsiveness was lost in the -940 to -1 ATHB8promoter fragment (Fig. 6E,F),suggesting that the region of the ATHB8 promoter between -953 and-940 is responsible for auxin inducibility. Within this region of the ATHB8 promoter, mutation in the TGTCTG element, but not in the putative MYB and GT1 binding sites, eliminated auxin responsiveness(Fig. 6G-L). Therefore, we conclude that the TGTCTG element in the ATHB8 promoter is a functional ARE and that this sequence is required for both ATHB8preprocambial expression and auxin inducibility.

Transcription factors of the auxin-response factor (ARF) family have been shown to bind AREs in vitro (Guilfoyle and Hagen, 2001). Twenty-two ARFs have been identified in Arabidopsis (Guilfoyle and Hagen,2007), and the TGTCTG element in the ATHB8 promoter could be the target of several, if not all, of these ARFs (see below and Discussion). We focused on MONOPTEROS (MP; also known as ARF5) because of the reduced vascularization of mp leaves(Przemeck et al., 1996) and the decreased ATHB8 transcript abundance in mp seedlings(Mattsson et al., 2003).

If MP is a regulator of ATHB8 preprocambial expression, one would expect it to be at least partially coexpressed with ATHB8. To test this, we first monitored expression of a functional (see Materials and methods) translational fusion of MP with ECFP in leaf development. Unlike ATHB8, MP_pro:MP:ECFP expression was initiated in wide domains (Fig. 7A-C), but during leaf development these broad fields of expression resolved into narrower domains before subsiding to undetectable levels(Fig. 7B,C). To test whether MP expression domains represent locations of ATHB8 expression, we visualized fluorescence in leaves simultaneously harboring ATHB8_pro:HTA6:EYFP and MP_pro:MP:ECFP, and invariantly observed overlap of ATHB8_pro:HTA6:EYFP signals with MP_pro:MP:ECFP expression (Fig. 7D).

If MP is a positive regulator of ATHB8 expression at preprocambial stages, mutations in MP should at least reduce levels of ATHB8 preprocambial expression. Expression of ATHB8_pro:HTA6:EYFP, and of HTA6:EYFP when driven by the -953 to -1 fragment of the ATHB8 promoter, was initiated in preprocambial cells of wild-type leaves (Fig. 4A-D,R-T; Fig. 7E,F). However, the early stages of expression of these constructs were abolished in the background of the strong (see Materials and methods) mp mutant allele U55 (Fig. 5B,E-H; Fig. 7I,J). Furthermore, their expression in mp leaves was remarkably similar to that conferred in wild type by loss or mutation of the TGTCTG element in the ATHB8 promoter (Fig. 7G,H). Finally, expression of neither construct could be induced by exogenous 2,4-D in the mp mutant background(Fig. 7K,L). This suggests that MP is required for both ATHB8 preprocambial expression and auxin inducibility, and that MP function at the ATHB8 promoter is mediated by the TGTCTG element.

If MP activity is a limiting factor for ATHB8 expression,ubiquitous MP expression should result in expansion of ATHB8expression domains. To test this, we visualized fields of ATHB8_pro:HTA6:EYFP activity in dexamethasone-exposed UBQ10_pro:MP:GR leaves and compared them with those in mock-treated leaves. Broadened ATHB8_pro:HTA6:EYFP expression domains were only detected in dexamethasone-treated samples(Fig. 7M,N), and this response was dependent on the presence of the TGTCTG element in the ATHB8promoter (Fig. 7O,P).

Finally, we asked whether MP directly regulates ATHB8 expression. To test this, we immunoprecipitated chromatin-crosslinked ECFP in MP_pro:MP:ECFP and ATHB8_pro:ECFP-Nuc seedlings; we then assayed levels of co-precipitated ATHB8 promoter regions in MP_pro:MP:ECFP and, to control for the binding of nuclear ECFP to the ATHB8 promoter, in ATHB8_pro:ECFP-Nuc samples. We detected a statistically significant (P<0.001) 2.5-fold enrichment in the ATHB8 promoter fragment that contains the TGTCTG element in MP_pro:MP:ECFP versus ATHB8_pro:ECFP-Nuc chromatin immunoprecipitates (Fig. 7Q),suggesting that MP resides in vivo at the ATHB8 promoter.

In conclusion, our results suggest that the ARF MP is an essential, direct and positive regulator of ATHB8 preprocambial expression and auxin responsiveness.

Recognizable effects of loss of ATHB8 function in leaf vascular development are restricted to transient or conditional defects, but any additional regulatory potential of ATHB8 might be masked by wild-type MP activity in athb8 mutants. To test this, we compared vascular defects in mature leaves of the weak mp mutant allele arf5-2 with those of the arf5-2 athb8-11 double mutant. The arf5-2 allele carries a single T-DNA insertion at the 3′ end of the MP coding region (see Materials and methods) and displayed∼40% penetrance of the rootless phenotype (85/871 seedlings segregating from arf5-2 heterozygous parents). arf5-2 homozygous seedlings could form an embryonic root and be grown on soil, but they were invariably sterile (44/44 arf5-2 homozygous individuals found among 350 genotyped wild-type-looking plants segregating from arf5-2heterozygous parents). Most of the mature first leaves of rooted arf5-2 seedlings (49/54) showed a vascular pattern complexity similar to that of wild-type or athb8-11 first leaves(Fig. 8A,B,I). However, vein loops in arf5-2 leaves were located further away from the leaf margin than in wild-type or athb8-11 leaves (`centralized vasculature')(Fig. 8A,B,I). At maturity,approximately half of the first leaves of rootless arf5-2 seedlings(32/69) were characterized by a normally complex, but centralized vein pattern, whereas the remaining half of the leaves displayed a simpler vascular organization (Fig. 8D,E,I). Finally, in ∼60% of the leaves of either rooted or rootless arf5-2 seedlings (32/54 and 41/69, respectively), the midvein bifurcated at the leaf tip (Fig. 8C,E,I). The overall complexity of vein pattern was only slightly reduced by additional loss of ATHB8 function in the arf5-2background (Fig. 8I). However,∼80% of the leaves of either rooted (57/71) or rootless (35/46) arf5-2 athb8-11 seedlings developed a terminally branched midvein(Fig. 8I), suggesting an enhancement of arf5-2 leaf vascular defects in the double mutant,irrespective of its root phenotype.

If ATHB8 preprocambial expression is contingent on any ARF activity additional to MP, the consequences of strongly reduced or complete loss of MP function on leaf vascular development should be further aggravated by additional deprivation of ATHB8 activity. To test this,we compared patterns of vascularization in leaves of mp^U55with those of the mp^U55 athb8-11 double mutant. First leaves of the invariably rootless mp^U55 seedlings displayed a dramatically simplified vascular organization, typically characterized by a bifurcated midvein and few additional vein fragments scattered across the lamina, and this phenotypic spectrum was not appreciably altered by supplemental loss of ATHB8 function(Fig. 8I).

Like athb8, mp seedlings display enhanced sensitivity to auxin transport inhibition (Schuetz et al.,2008). Because loss of ATHB8 function augments the effects of diminished MP activity on vein patterning, we asked whether the elevated response of athb8 to auxin transport inhibitors could be further exacerbated in a background of reduced MP function. To test this, we assessed the sensitivity to NPA of single and double mutant combinations of arf5-2 and athb8-11. Reduction in auxin transport frequently results in leaf fusion(Okada et al., 1991; Schuetz et al., 2008; Sieburth, 1999; Wang et al., 2005), a response that we first observed in wild type at 10 μM NPA (2/95) and in athb8-11 at 5 μM NPA (2/33). Approximately 10% (3/29) of arf5-2 seedlings displayed leaf fusion at 1 μM NPA, consistent with strong NPA hypersensitivity of mp mutants. At the same concentration of NPA, nearly 40% (10/27) of arf5-2 athb8-11 double mutants displayed leaf fusion (Fig. 9H). Because loss of ATHB8 function by itself did not result in leaf fusion at this concentration of NPA(Fig. 9F), we conclude that leaf separation defects elicited by reduced auxin transport in arf5-2are strongly enhanced by additional athb8 mutation.

In summary, our results suggest that non-conditional and conditional contributions of ATHB8 to leaf vascular patterning are covered by MP activity. Furthermore, our data suggest that any unique role of ATHB8 in vein patterning becomes largely inconsequential upon severe loss of MP function and, therefore, that MP is the primary regulator of ATHB8 non-redundant activities in leaf vascular patterning (see Discussion).

The molecular details of the mechanisms controlling the recruitment of ground cells in the leaf towards procambium formation are largely unknown. Substantial evidence has, however, been accumulating that implicates polarly transported auxin signals in leaf vascular patterning (reviewed by Berleth et al., 2000; Sachs, 1981). Near-ubiquitous expression of the auxin exporter PIN1 narrows to files of procambial cells during leaf development (Scarpella et al.,2006; Wenzel et al.,2007), but how cells that will acquire procambial identity are selected among the population of PIN1-expressing cells is not known. Nevertheless, these anatomically inconspicuous `preprocambial' cells can be identified by expression of the HD-Zip III gene ATHB8.

In this study, we have explored biological functions of ATHB8 in leaf vascular development and searched for regulatory elements and trans-acting factors required for ATHB8 preprocambial expression. We show that ATHB8 is necessary to stabilize preprocambial cell specification against perturbations in auxin transport, to confine preprocambial cell state acquisition to narrow regions and to coordinate procambial cell identity assignment within and between veins. Further, we find that ATHB8 expression in preprocambial cells depends on the presence of an ARE in its promoter. Finally, we show that the ATHB8preprocambial regulatory element is a direct target of the transcriptional regulator MP.

Loss of ATHB8 function leads to expanded expression of the preprocambial marker J1721 and asynchronous expression of the procambial marker Q0990. Expression of these reporters is strictly associated with zones of vascular differentiation in a variety of genetic backgrounds and under a number of experimental conditions (e.g. Dello Ioio et al., 2007; Leroy et al., 2007; Levesque et al., 2006; Sawchuk et al., 2007; Song et al., 2008; Weijers et al., 2005; Weijers et al., 2006),implicating ATHB8 in constraining preprocambial state acquisition to narrow fields of cells and in coordinating procambial cell identity assignment within and between veins. Although premature differentiation of procambial strands in athb8 leaves could simply represent a read-out of preprocambial defects, it is consistent with a proposed role for ATHB8 in maintaining the meristematic potential of vascular cells(Baima et al., 2001; Kang and Dengler, 2002).

The behavior of J1721 and Q0990 in athb8 strongly resembles that in wild type under conditions of reduced auxin transport(Sawchuk et al., 2007),suggesting a role for ATHB8 in promoting auxin flow during early stages of vein formation. According to this interpretation, loss of ATHB8 function would be expected to confer an enhanced response to auxin transport inhibitors. In the absence of alterations in auxin sensitivity, PIN1-labeled preprocambial cell specification and vein patterning are more sensitive to chemical obstruction of auxin flow in athb8than in wild type, suggesting the presence of auxin transport defects in the mutant. Alternatively, or additionally, the enhanced response of athb8 leaves could suggest a function for ATHB8 in stabilizing auxin flow against perturbations, consistent with the observed insensitivity of procambial strands to auxin transport inhibition(Mattsson et al., 1999).

Genetic or pharmacological interference with auxin flow generates broad areas of vascular differentiation(Mattsson et al., 1999; Sieburth, 1999), while it is difficult to explain, based on current knowledge, how exuberant preprocambial state acquisition and incongruent procambial identity assignment would per se result in hypersensitivity to auxin transport inhibition. Nevertheless,because of feedback between auxin flow and vascular development(Sachs, 1981), it is not currently possible to assign a fixed position to either of these processes in a linear cause-effect relationship. As such, the exaggerated response of athb8 to auxin transport interference might underlie the altered marker behavior, or be a consequence of it, or the two might point to unrelated functions of ATHB8 in vein formation. Although the assignment of a role for ATHB8 in leaf vascular development at the molecular level will have to await the identification of its targets, the enhanced sensitivity of PIN1 expression and vascular patterns to obstruction of auxin flow in athb8 leaves suggests that ATHB8 is required to stabilize the selection of ground cells that will acquire a preprocambial state against perturbations in auxin transport.

Irregular vein formation in athb8 leaves is corrected and eventually resolves into a normal leaf vascular pattern(Baima et al., 2001; Prigge et al., 2005). Amelioration of early vascular defects during organ development is not unprecedented (e.g. Scarpella et al.,2003), and responses of the vascular system to local auxin application or auxin transport inhibition are more severe when evaluated at early stages of vein development(Scarpella et al., 2006; Wenzel et al., 2007). How are defects at early stages of vein formation normalized during athb8leaf development? One possibility is that ATHB8 has an ephemeral role that is confined to early stages of vascular strand formation and that it has an inconsequential function at later stages. Alternatively, or in addition,functional compensation among members of the HD-Zip III family could rectify defects due to loss of ATHB8 activity.

If transience is an intrinsic property of the biological role of ATHB8, effects of loss of ATHB8 function should not be expected to have long-lasting consequences in conditions of reduced activity of one of its regulators. The enhancement of vein pattern defects in arf5-2 athb8-11 double mutants as compared with those in the weak mp allele arf5-2 suggests, however, that ATHB8 can have permanent effects on vein patterning. In leaves of arf5-2, the midvein frequently bifurcates at the leaf apex, a response that is commonly evoked by defective auxin transport(Mattsson et al., 1999; Sieburth, 1999). The fraction of leaves displaying this phenotype increases in the strong mp^U55 allele, suggesting that the defect directly depends on MP function. In double-mutant combinations of athb8-11with arf5-2, no new phenotype class is observed; rather, the fraction of leaves displaying midvein bifurcation is increased to closely match that in the mp^U55 allele. Strong mp/arf5 alleles display an exaggerated response to auxin transport interference that results in obstruction of leaf formation (Schuetz et al., 2008), whereas enhanced sensitivity to defective auxin flow in the weak mp allele arf5-2 most conspicuously manifests in leaf fusion at very low concentrations of auxin transport inhibitors. Under these conditions, leaves of athb8 mutants show normal sensitivity to auxin flow inhibition, but additional loss of ATHB8 function greatly increases the occurrence of leaf separation defects in auxin transport-inhibited arf5-2 seedlings. Therefore, ATHB8 has functions in vein formation that extend beyond the evanescent contribution revealed by marker analysis and the moderate input exposed by auxin flow obstruction, but the regulatory potential of ATHB8 is concealed in the athb8 mutant background by the presence of functional MP activity.

If transience is not necessarily an inherent property of ATHB8function in leaf vascular development, then defects at early stages of vein formation in athb8 could be amended at successive stages of development through the overlapping activities of other members of the HD-Zip III family, as shown for other aspects of plant development(Prigge et al., 2005). Because correction of athb8 leaf vascular defects occurs to a lesser extent in the arf5-2 background, one function of MP in vein development could be the regulation of the entire HD-Zip III family. Post-transcriptional downregulation of all members of the HD-Zip III family through overexpression of the microRNA 165 (miR165) results in cotyledon vascular defects that are remarkably similar to those displayed by strong mp alleles (Zhou et al.,2007), and expression of other members of the HD-Zip III family in addition to ATHB8 is reduced in the mp background(Mattsson et al., 2003). If functional redundancy underlies amelioration of athb8 vascular defects, the inability of the athb8 mutation to shift the vein pattern complexity of arf5-2 towards the severe distribution typical of mp^U55 does not exclude a broader role for ATHB8 in the regulation of leaf vascular patterning. However, further studies will be required to unravel the overlapping and redundant roles of HD-Zip III genes in vein formation.

The ATHB8 promoter is activated in files of ground cells that are stabilized towards the procambial fate(Kang and Dengler, 2004; Scarpella et al., 2004). Additionally, the ATHB8 promoter induces transient expression at the tip of the leaf and at the hydathodes, where not all ATHB8-expressing cells will differentiate into vascular cells(Kang and Dengler, 2004; Scarpella et al., 2004). Expression of a functional ATHB8 translational fusion recapitulates all aspects of ATHB8 promoter activity with the exception of the nonvascular expression in the leaf tip and hydathode cells, suggesting the presence of post-transcriptional mechanisms that downregulate ATHB8expression at those locations. ATHB8 transcripts are predicted to be targets of miR165-mediated degradation(Rhoades et al., 2002), and miR165 is more abundantly expressed at the leaf tip(Li et al., 2005), suggesting that miR165-dependent post-transcriptional regulation of ATHB8 expression might occur at locations of non-overlap between ATHB8 promoter activity and expression of the ATHB8translational fusion.

Expression of ATHB8 in preprocambial cells is strictly dependent on the presence of a TGTCTG element in its promoter. This element is a variant of the TGTCTC ARE (Li et al.,1994; Liu et al.,1994) and is required for auxin-induced ATHB8 expression. That both preprocambial expression of ATHB8 under unperturbed conditions and responsiveness of ATHB8 to auxin signals are contingent on the activity of a single regulatory element is uncommon. In fact, the presence of a functional ARE in the promoter is not usually necessary for tissue- or stage-specific gene expression, only for a ubiquitous response to auxin signals (Li et al.,1994; Liu et al.,1994). Nevertheless, a synthetic promoter composed of repeats of the TGTCTC ARE coupled to a minimal viral promoter (DR5)(Ulmasov et al., 1997b) is sufficient to drive expression in developing veins(Mattsson et al., 2003). Furthermore, fields of DR5 promoter activity in leaf primordia seem to overlap with ATHB8 preprocambial expression domains, although DR5 promoter-driven expression displays greater heterogeneity in onset, decay and level along individual veins (Mattsson et al., 2003; Scarpella et al.,2004). These observations suggest that in most auxin-responsive promoters, tissue-specific regulatory elements constrain the activity of AREs solely to auxin inducibility, whereas ATHB8 preprocambial expression might be the unrestrained read-out of auxin signal transduction.

A confounding multitude of genes in Arabidopsis (∼5000)contain a TGTCTG element in the 500 bp region immediately upstream of their coding sequence. This list includes genes expressed at early stages of vein development [e.g. CYCA2;1(Burssens et al., 2000), SCL3 (Ckurshumova et al.,2009), CESA2 (Beeckman et al., 2002)] and those with a proposed role in vascular development [e.g. VAN3 (SFC)(Deyholos et al., 2000; Koizumi et al., 2000), VAD1 (Lorrain et al.,2004), VND4 (Kubo et al., 2005)] or auxin response [e.g. ARF2(Li et al., 2004), AXR3 (IAA17) (Rouse et al., 1998), HAT2(Sawa et al., 2002)]. However,not all AREs present in promoters can bind ARFs in vitro(Inukai et al., 2005), and additional regulatory elements may constrain the regulatory potential of AREs to sole auxin responsiveness (Li et al.,1994; Liu et al.,1994). Therefore, we consider it unlikely that all these genes are expressed in preprocambial cells. If the presence of a TGTCTG element in the promoter is unlikely to be sufficient to predict expression in preprocambial cells, what other requirements are necessary for preprocambial expression?Nucleotides flanking the TGTCTC ARE seem to act as modifiers of ARE activity(Ulmasov et al., 1997a; Ulmasov et al., 1995; Ulmasov et al., 1997b). Vast systematic efforts will be necessary to test in vivo what, if any, boundary conditions are required for TGTC-containing elements to promote preprocambial expression.

Transcription factors of the ARF family have been shown to bind AREs in vitro (reviewed by Guilfoyle and Hagen,2001). Deletion or mutation of the TGTCTG ARE in the ATHB8 promoter eliminates preprocambial expression but does not induce expression in nonvascular cells. This observation suggests that the TGTCTG element is not the target of a repressor that normally extinguishes ATHB8 expression outside of the vasculature, but rather that an activator binds the TGTCTG element and induces ATHB8 expression in vascular cells. In Arabidopsis, the ARF family is encoded by 22 genes, of which five [MP (ARF5), ARF6, ARF7(NPH4), ARF8 and ARF19] function as transcriptional activators in transfected protoplasts, whereas the remaining 17 behave as repressors in a similar experimental context(Guilfoyle and Hagen, 2007)(and references therein). ATHB8 preprocambial expression is under the direct control of MP, which is consistent with reduced ATHB8transcript levels in an mp background(Mattsson et al., 2003). None of these observations, however, excludes the involvement of other ARFs in the control of ATHB8 preprocambial expression. At least three of the four remaining activating ARFs are expressed in domains that may overlap with those of ATHB8 (Hardtke et al.,2004; Li et al.,2006; Okushima et al.,2005; Tian et al.,2004; Wilmoth et al.,2005), and the class of activating ARFs is characterized by a high level of functional redundancy among its members(Hardtke et al., 2004; Nagpal et al., 2005; Okushima et al., 2005; Wilmoth et al., 2005). Further, because the tests for activation and repression of transcription by members of the ARF family rely upon transient expression assays in leaf mesophyll or suspension cell culture protoplasts, it remains possible that an ARF classified as a repressor could function as an activator, and vice versa,in certain cell types or environments (e.g. Okushima et al., 2005). Conditional manipulation of gene activity will be required to expose the overlapping and non-redundant roles of ARFs in the regulation of ATHB8 preprocambial expression, as any further reduction of ARF activity than that residual in strong mp backgrounds is likely to directly impinge on leaf primordium formation(Hardtke et al., 2004). However, deletion or mutation of the TGTCTG element in the ATHB8promoter confers expression in wild type that is indistinguishable from that of the full-length ATHB8 promoter in the strong mp^U55 allele. Moreover, leaf vascular defects in mp^U55 cannot be further enhanced by additional loss of ATHB8 function. Therefore, the contribution of ARFs other than MP to the control of ATHB8 preprocambial expression is probably subtle.

A functional MP translational fusion is at first detected in nearly all subepidermal cells of the young leaf primordium, and its expression is only subsequently confined to sites of vein formation. This is strikingly different from the activity of the ATHB8 promoter, which is initiated in single cell files before expression of the MP fusion protein has been restricted to the narrow sites of vein formation. Further, ubiquitous MP expression results in expanded domains of ATHB8expression, which do not, however, extend to include all cells in the leaf. Although it will be interesting in the future to understand how broad patterns of MP expression are translated into narrow sites of ATHB8activation, the identification of regulators of early vein development and of the transcription factors controlling their expression already assists in defining the contribution of auxin signal transduction to leaf vascular patterning at the molecular level.