ULTRAPETALA trxG Genes Interact with KANADI Transcription Factor Genes to Regulate Arabidopsis Gynoecium Patterning

Organ formation relies upon precise patterns of gene expression that are under tight spatial and temporal regulation. Transcription patterns are speci ﬁ ed by several cellular processes during development, including chromatin remodeling, but little is known about how chromatin-remodeling factors contribute to plant organogenesis. We demonstrate that the trithorax group (trxG) gene ULTRAPETALA1 ( ULT1 ) and the GARP transcription factor gene KANADI1 ( KAN1 ) organize the Arabidopsis thaliana gynoecium along two distinct polarit y axes. We show that ULT1 activity is required for the kan1 adaxialized polarity defect, indicating that ULT1 and KAN1 act oppositely to regulate the adaxial-abaxial axis. Conversely, ULT1 and KAN1 together establish apical-basal polarity by promoting basal cell fate in the gynoecium, restricting the expression domain of the basic helix-loop-helix transcription factor gene SPATULA . Finally, we show that ult alleles display dose-dependent genetic interactions with kan alleles and that ULT and KAN proteins can associate physically. Our ﬁ ndings identify a dual role for plant trxG factors in organ patterning, with ULT1 and KAN1 acting antagonistically to pattern the adaxial-abaxial polarity axis but jointly to pattern the apical-basal axis. Our data indicate that the ULT proteins function to link chromatin-remodeling factors with DNA binding transcription factors to regulate target gene expression.


INTRODUCTION
In multicellular organisms, developmental processes that specify different tissues and organs result from precise patterns of gene expression that are tightly regulated in a spatial and temporal manner. Maintenance of these specific transcription patterns can be controlled at the level of chromatin structure, and this epigenetic regulation of gene expression is important for the correct deployment of developmental programs and the maintenance of cell fates (Köhler and Hennig, 2010). Polycomb group (PcG) and trithorax group (trxG) proteins sustain the stable transcription patterns of key developmental regulatory genes by modifying histones to maintain chromatin in the inactive and active transcriptional states, respectively (Schwartz and Pirrotta, 2007). PcG proteins act in large complexes to sustain transcriptional repression, and their action is restricted to defined sets of target genes dependent on tissue-and development-specific contexts. This is achieved by trxG proteins that antagonize PcG complex function and exclude target genes from PcG-mediated repression (Schwartz and Pirrotta, 2007). trxG complexes direct active epigenetic histone modifications such as H3K4 trimethylation that induce an open chromatin configuration, thereby facilitating transcription at specific target loci (Schuettengruber et al., 2011). Although several multiprotein trxG complexes have been identified from Drosophila extracts, including a trithorax acetylation complex (TAC1), a BRM complex, and an ASH complex, relatively little is known about how these trxG factors are recruited in a sequence-specific fashion to regulatory sites on the target DNA (Schuettengruber et al., 2007;Smith and Shilatifard, 2010).
In contrast with animals, few plant trxG chromatin-associated transcriptional activators have been characterized, and their biochemical activities and mechanisms of recruitment to DNA are poorly defined. The Arabidopsis thaliana genome encodes five ARABIDOPSIS TRITHORAX proteins (ATX1 to ATX5) and seven ARABIDOPSIS TRITHORAX-RELATED proteins (ATXR1 to ATXR7). Among these, ATX1 displays H3K4 methyltransferase activity (Alvarez-Venegas et al., 2003) and recruits TATA Binding Protein (TBP) and RNA Polymerase II (Pol II) to target promoters (Ding et al., 2011). ATX1 directly interacts with WDR5a, which is also part of a second, ASH2R H3K4 methyltransferase complex (Jiang et al., 2011). The H3K4 and H3K36 methyltransferase EFS/ASHH2 (Kim et al., 2005;Zhao et al., 2005) and the closely related SWI2/SNP2 ATPases SYD and BRM (Wu et al., 2012) also display trxG activity. The chromatin remodeler PICKLE is required to retain the transcriptional activity of PcG target genes in Arabidopsis and thus has been proposed to have a trxG-like function (Aichinger et al., 2009).
Genetic and biochemical experiments have identified the ULTRAPETALA1 (ULT1) transcriptional regulatory protein as an Arabidopsis trxG factor that counteracts PcG gene function during development (Carles and Fletcher, 2009). ULT1 and the paralogous ULT2 gene encode closely related proteins containing a putative DNA binding SAND domain as well as a B-box-like domain that has been proposed to mediate protein-protein and RNA-protein interactions (Carles et al., 2005). Evidence indicates that ULT1 promotes transcriptional activation by limiting the deposition of repressive histone marks by PcG complexes (Carles and Fletcher, 2009). ULT1 interacts physically with ATX1, and it has been proposed that these proteins direct an open chromatin conformation to recruit proteins involved in transcriptional initiation and elongation to target loci .
The ULT1 and ULT2 expression patterns overlap in embryonic, inflorescence, and floral meristems (Carles et al., 2005). In developing flowers, transcripts from both genes are detected in stamens and the adaxial side of the gynoecia. ULT1 plays a major role in negatively regulating inflorescence and floral meristem cell accumulation and in maintaining floral meristem determinacy (Fletcher, 2001;Carles and Fletcher, 2009). ULT2 has a minor but overlapping role with ULT1 in regulating shoot and floral stem cell accumulation and can regulate many of the same developmental genes (Monfared et al., 2013). In addition, ULT1 and ULT2 have redundant functions in specifying gynoecium polarity along the apical-basal axis (Monfared et al., 2013). The ULT1 and ULT2 proteins interact physically in planta, suggesting that they may function together in a multifactor trxG complex (Monfared et al., 2013).
In this study, we used the Arabidopsis gynoecium, the female reproductive organ, to explore the role of the ULT1 and ULT2 trxG proteins in regulating developmental patterning. The gynoecium is a complex organ essential for fertilization, embryo development, and mature seed dispersal, and its patterning is under strict genetic control, making it an excellent system to study how differentiation events occur during development. Like the evolutionarily related leaves, the gynoecium tissues are arranged along three axes of polarity: adaxial-abaxial, medial-lateral, and apicalbasal (Larsson et al., 2013). The adaxial-abaxial polarity axis is patterned early during gynoecium development, with the placenta, ovules, and septum considered adaxial tissues and the replum and outer valve epidermis considered abaxial tissues. The KANADI1 (KAN1) and KAN2 genes, which encode GARP family transcription factors, specify abaxial polarity in both the gynoecium and the leaves Kerstetter et al., 2001), along with the closely related auxin response factor genes ETTIN/AUXIN RE-SPONSE FACTOR3 (ETT/ARF3) and ARF4 (Pekker et al., 2005).
The apical-basal axis is specified later in gynoecium development as the primordium grows into a tube-shaped organ (Sundberg and Ferrandiz, 2009) and begins to differentiate along the longitudinal axis. The stigma and style are the most apical tissues, followed by the ovary and replum/septum, and the basal gynophore. Apicalbasal patterning is dependent on an auxin gradient that spans the gynoecium primordia (Nemhauser et al., 2000), where high auxin concentrations induce apical tissue formation, moderate levels specify the ovary, and low levels cause basal tissue formation (Nemhauser et al., 2000;Ståldal and Sundberg, 2009). ETT has been proposed to interpret the auxin gradient to establish the regional apical-basal boundaries in the gynoecium (Nemhauser et al., 2000).
ETT establishes the apical-basal axis by repressing the expression of the basic helix-loop-helix transcription factor gene SPATULA (SPT), which promotes carpel margin tissue, apical style, and stigmatic tissue formation (Heisler et al., 2001). A tight coordination between these factors seems to be crucial to pattern the longitudinal axis of the gynoecium.
Here, we demonstrate that the ULT and KAN genes interact to regulate Arabidopsis gynoecium development along several axes of polarity. We show that ULT1 acts antagonistically to KAN1 to pattern the adaxial-abaxial axis during gynoecium development but that the two genes act together to pattern the apical-basal axis. We find that mutations in the ULT1/2 and KAN1/2 genes cause dose-dependent genetic effects in the gynoecium and that the ULT and KAN proteins interact physically in yeast and in planta, suggesting that they may function in a complex to regulate gynoecium patterning. Our results reveal a role for the KAN genes in establishing the apical-basal axis of the gynoecium and demonstrate a novel link between chromatin regulatory proteins and transcription factors in the control of organ patterning.

RESULTS
Enhancer of the ult1-2 Flower Phenotype Is a kan1 Allele ULT1 and ULT2 play redundant roles in promoting basal gynoecium polarity in Arabidopsis (Monfared et al., 2013). To further investigate the function of the ULT1 trxG gene in gynoecium development, we performed an ethyl methanesulfonate chemical mutagenesis of ult1-2 plants and isolated an enhancer (line 59-1) that showed a novel gynoecium phenotype. Wild-type Arabidopsis (Landsberg erecta [Ler]) flowers have four sepals and petals and a gynoecium consisting of a stigma, style, two valves, a replum, and a gynophore from the apex to the base (Figures 1A and 2A). ult1-2 flowers typically contain five sepals and petals ( Figure 1B), and the gynoecia are patterned like the wild type but often consist of three or four valves (Fletcher, 2001). ult1-2 59-1 flowers contain gynoecia that produced large ectopic outgrowths along the replum from the apex to the base, with stigmatic tissue prominent along the top of the outgrowths ( Figure 1C). The double mutant plants were backcrossed to wild-type Ler plants, and single mutants were identified among the F2 progeny that displayed a gynoecium phenotype resembling kan1-12 null mutant plants with a ridge of tissue growing out from the base (Kerstetter et al., 2001). Double mutants of ult1-2 with the kan1-12 allele ( Figure 1D) displayed the same gynoecia phenotype as ult1-2 59-1 plants ( Figure 1C), suggesting that the ult1-2 59-1 phenotype was caused by a mutation in KAN1.
Sequencing the KAN1 gene from the ult1-2 59-1 plants, we identified a G-to-A transition in the first base pair of the first intron ( Figure 1E). To determine if this mutation altered the recognition of the first intron 59 splice site, we performed RT-PCR on RNA extracted from Ler and ult1-2 59-1 plants with primers spanning the KAN1 first intron ( Figure 1E). A 528-bp product was amplified from wild-type plants as well as two bands from ult1-2 59-1 plants. The upper band is ;100 bp larger than the expected wildtype product ( Figure 1F) and contains the 104-bp first intron of KAN1. The lower band contains a deletion of 38 bp upstream of 2 of 17 The Plant Cell the first intron splice site ( Figure 1F). Thus, this mutation in KAN1, which we denote kan1-13, causes the loss of the 59 splice site in the first intron, resulting in an inclusion of the first intron in the majority of the mRNA transcripts. The resulting mutant protein is truncated due to a premature stop codon upstream of the KAN1 DNA binding Myb domain. A minority of the RNA transcripts have a displacement of the first intron splice site 38 bp upstream of the wild-type splice site, also leading to a disruption of the KAN1 open reading frame upstream of the Myb domain. Because in both cases the mutant protein lacks the DNA binding domain and, therefore, is nonfunctional, and because the mutation causes the same phenotype as the kan1-12 mutation, we conclude that kan1-13 represents a null allele of the KAN1 gene. These results show that ULT1 has a synergistic role with the GARP domain transcription factor gene KAN1 in controlling gynoecium development, with the double mutants displaying severe ectopic tissue outgrowth and basal expansion of the stigmatic tissue.
ULT1 Maintains the Adaxial-Abaxial Polarity Axis Antagonistically to KAN1 KAN1 specifies abaxial cell identity in both carpels and leaves Kerstetter et al., 2001). kan1 gynoecia form ectopic projections and ectopic ovules on the outside of the organ, and both features have been interpreted as a partial transformation of the abaxial (external) tissues into adaxial (internal) tissues (Kerstetter et al., 2001). We used scanning electron microscopy to compare wild-type (Ler), ult1-2, kan1-12, and ult1-2 kan1-12 gynoecia. ult1-2 gynoecium morphology is indistinguishable from that of wild-type plants (Figures 2A, 2B, 2E, and 2F), whereas kan1-12 gynoecia form ectopic ovules at the external base of the organ as well as small ectopic outgrowths tipped with stigmatic tissue that arise from the replum ( Figures 2C  and 2G). ult1-2 kan1-12 gynoecia are surrounded by ectopic tissue with stigmatic tissue on top but completely lack ectopic ovules at the base, where only the gynophore is visible ( Figures  2D and 2H). This lack of ectopic ovules indicates that ult1 mutations suppress the kan1 adaxialization defect in the gynoecium. Wild-type rosette leaves consist of a nearly flat lamina that grows out along the adaxial-abaxial leaf axis ( Figure 2I). ult1-2 leaves are indistinguishable from wild-type leaves ( Figure 2J), whereas the first leaves of kan1 seedlings curl upward strongly ( Figure 2K) due to their partial adaxialization (Kerstetter et al., 2001). We investigated whether this leaf phenotype is also suppressed by the ult1-2 mutation and found that ult1-2 kan1-12 seedlings have flat leaves resembling the wild type ( Figure 2L). These data show that ult1-2 suppresses the kan1 adaxialized phenotype in both gynoecia and leaves. Our results reveal a role for ULT1 in regulating adaxial-abaxial polarity in the gynoecium and rosette leaves and indicate that ULT1 acts antagonistically with KAN1 to pattern this polarity axis.
The ult1 Mutation Enhances the kan1 Ectopic Outgrowth Phenotype Although the ult1-2 mutation suppresses kan1-12 ectopic ovule formation, ult1-2 kan1-12 gynoecia have extensive ectopic outgrowths topped with stigmatic tissue that extend basally from the apex of the organ. To investigate the nature of these outgrowths, we analyzed gynoecium development in wild-type and single and double mutant plants. The wild-type gynoecium is derived from two congenitally fused carpels that arise centrally at stage 6 to form a hollow cylinder (Sessions et al., 1997). As development proceeds through stages 7 and 8, two opposing meristematic outgrowths form at medial positions, producing the placentae and the ovules. At stage 9 the gynoecium closes at its apical end and the valves, placentae, and style become morphologically distinct. Between stage 9 and stage 11, the stigmatic papillae differentiate at the top of the style to mediate pollen tube adherence and germination (Smyth et al., 1990;Sessions et al., 1997). Scanning electron microscopy of stage 9 gynoecia shows that the stigmatic papillae cells are positioned atop the developing wild-type ( Figure 3A) and kan1-12 ( Figure 3B) gynoecia. However, in ult1-2 kan1-12 gynoecia, the number of these cells is increased and they are positioned more basally ( Figure 3C). At later stages, compared with wild-type gynoecia ( Figures 3D and 3G), kan1-12 gynoecia develop basal ectopic outgrowths that can be topped with stigmatic tissue ( Figure 3E) and also have a lateral expansion of the apical style ( Figure 3H). Relative to kan1-12 gynoecia, ult1-2 kan1-12 gynoecia develop more, longer, and larger apical ectopic outgrowths that extend from the apex to the base as well as an increased amount of stigmatic tissue atop the outgrowths (Figures 3F and 3I). Therefore, the ult1-2 mutation enhances the ectopic apical and basal outgrowths of kan1-12 gynoecia.
We also examined the cellular morphology of the ectopic apical outgrowths in ult1-2 kan1-12 gynoecia. In wild-type gynoecia, the valves have highly elongated cells interspersed with stomata ( Figure 3J), whereas the style consists of rectangular but flattened cells with characteristic wax ridges, also interspersed with stomata ( Figure 3K). ult1-2 kan1-12 ectopic apical outgrowths show flattened and rectangular cells interspersed with stomata ( Figure 3L), similar to wild-type style tissue. Thus, the ectopic outgrowths visible in the medial and basal regions of ult1-2 kan1-12 gynoecia consist of both style and stigmatic tissue, which are normally found only at the apical end of the organ.
ULT1 and KAN1 Act Redundantly to Pattern the Apical-Basal Gynoecium Axis ult1-2 ult2-3 gynoecia display small outgrowths of tissue tipped with stigmatic tissue protruding from the basal region of the ovary (Monfared et al., 2013), a phenotype similar to that of kan1-12 gynoecia ( Figure 3E). The ult1-2 ult2-3 gynoecia exhibit a basalization of the internal vascular patterning, indicating that ULT1 and ULT2 redundantly promote basal gynoecium polarity (Monfared et al., 2013). In addition, our scanning electron microscopy observations indicate that the visible ectopic outgrowths in kan1-12 and ult1-2 kan1-12 gynoecia are a basal expansion of style and stigmatic tissue. These results suggest that the kan1-12 and ult1-2 kan1-12 gynoecia may likewise have an apical-basal patterning defect in addition to the well-characterized adaxial-abaxial patterning defect displayed by kan1 plants.
We tested this hypothesis by observing the vascular patterning in wild-type and mutant gynoecia at the anthesis stage. Aniline blue staining revealed the four main bundles of the wildtype vascular system ( Figure 4A). The bundles in the lateral plane terminate just below the style, whereas the medial vascular bundles enter the style and bifurcate into fans of xylem tissue ( Figure 4A) (Sessions and Zambryski, 1995). The apical bifurcation point of the medial vascular bundles is unaltered in ult1-2 ( Figure 4B) and kan1-12 ( Figure 4C) gynoecia. By contrast, the bifurcation point occurs in a very basal position in ult1-2 kan1-12 gynoecia, below the middle of the ovary ( Figure 4D). Thus, both ult1-2 ult2-3 (Monfared et al., 2013) and ult1-2 kan1-12 gynoecia have an alteration in apical-basal patterning, exhibiting basalization of the tissue types with respect to external cell identity and internal vascular patterning. However, ult1-2 kan1-12 gynoecia show much stronger patterning defects than ult1-2 ult2-3 gynoecia.
Next, we tested the hypothesis that the visible ectopic outgrowths in kan1-12 and ult1-2 kan1-12 gynoecia are an expansion of style tissue by analyzing the ProASA1:GUS molecular marker. This style-specific marker is a b-glucuronidase (GUS) reporter gene driven by the ANTHRANILATE SYNTHETASE1 (ASA1) promoter (Sessions and Zambryski, 1995). In wild-type and ult1-2 gynoecia, ProASA1:GUS activity is confined to the inner layers of the apical style following anthesis (Figures 4E and 4F), indicating that ULT1 does not independently regulate the expression pattern of this style marker. However, in kan1-12 gynoecia, ProASA1:GUS activity is detected in the small ectopic outgrowths ( Figure 4G), and in ult1-2 kan1-12 gynoecia, strong ectopic ProASA1:GUS activity occurs throughout the ectopic

of 17
The Plant Cell outgrowths from the apex to the base ( Figure 4H). These results confirm that the ectopic outgrowths consist of style as well as stigmatic tissue and indicate that KAN1 and ULT1 together restrict ASA1 promoter activity to the apical region of the gynoecium. Thus, the molecular and morphological data are consistent with ULT1 and KAN1 acting to promote basal gynoecium polarity by restricting style identity to the apical domain. Defects in apical-basal gynoecium patterning can be associated with reduction of the basal ovary regions (Sessions and Zambryski, 1995;Nishimura et al., 2005). Although we did not observe basal reduction in ult1-2 ult2-3 gynoecia (Monfared et al., 2013), we analyzed at the molecular level if basal ovary reduction occurred in kan1-12 and ult1-2 kan1-12 gynoecia. We used the GT142 gene trap line that marks the dehiscence zone along the valve margins and visually delimits the valve area (Sundaresan et al., 1995). Wild-type gynoecia show GT142-driven GUS activity in a narrow strip along the margin cells ringing the entire valves, terminating just above the gynophore at the base of the gynoecium ( Figure 4I). In both ult1-2 ( Figure 4J) and kan1-12 ( Figure 4K) gynoecia, the GT142 pattern is unaltered, indicating that, independently, neither ULT1 nor KAN1 affects ovary formation. However, ult1-2 kan1-12 gynoecia show a strong reduction in the bottom part of the ovary in one or more of the valves ( Figure 4L). Our data indicate that at the morphological and molecular levels, ULT1 and KAN1 act together to restrict the stigma and style tissue to the apical region of the gynoecium and to promote the formation of the ovary in the basal region. Both of these features are consistent with a role for the genes in regulating the apical-basal polarity axis of the Arabidopsis gynoecium.

ULT1 and KAN1 Act through a Different Pathway from ETT and Regulate SPT
The plant hormone auxin plays an important role in establishing the apical-basal polarity axis in the gynoecium. The ARF-encoding ETT/ARF3 gene is proposed to interpret an apical-basal auxin concentration gradient in the gynoecium (Nemhauser et al., 2000), as ett null mutant gynoecia display basally expanded style and stigmatic tissue as well as strong valve reduction (Sessions and Zambryski, 1995;Sessions et al., 1997), similar to ult1-2 kan1-12 gynoecia. We investigated whether the ULT1 and KAN1 genes regulate gynoecium patterning through this auxin-responsive pathway by analyzing ETT mRNA levels in wild-type, ult1-2, kan1-12 and ult1-2 kan1-12 young flower clusters. Quantitative RT-PCR results show no reduction in ETT mRNA levels in ult1-2 or ult1-2 kan1-12 gynoecia and a slight but significant reduction in kan1-2 gynoecia that can be attributed to the reduced numbers of abaxial cells in plants of this genotype (Supplemental Figure 1). Introducing the ult1-2 mutation into the kan1-12 background suppresses the adaxialization phenotype, thereby restoring ETT transcript to wild-type levels. Thus, ULT1 and KAN1 together do not promote basal gynoecium polarity by regulating ETT mRNA accumulation.
We further investigated the potential interaction between ULT1 and KAN1 with ETT by genetic analysis. Compared with wild-type gynoecia, which have symmetric valves (Supplemental Figure 2A), the main characteristic produced by the weak ett-2 allele is a slight asymmetric valve reduction at the gynoecium base (Supplemental Figure 2B). Introducing the ett-2 mutation into either the ult1-2 (Supplemental Figure 2C) or the kan1-12 (Supplemental Figure 2D) background results in gynoecia with a severe reduction of basal valve tissue and an expansion of stigmatic tissue down the gynoecium along the replum (Supplemental Figures 2F and 2G). Introducing the ett-2 mutation into the ult1-2 kan1-12 background (Supplemental Figure  2E) produces gynoecia with poorly defined valves and more severe ectopic outgrowth of stigmatic tissue expanded radially toward the base of the gynoecium (Supplemental Figure 2H). Therefore, the ult1 and kan1 mutations enhance the weak ett-2 valve reduction phenotype, revealing that ULT1 and KAN1 promote the formation of basal valve tissue when ETT activity is compromised. In addition, the ett-2 mutation produces a basalized ectopic outgrowth phenotype in combination with either the ult1 or the kan1 mutation and enhances the outgrowth phenotype of ult1 kan1 plants. These data are consistent with ETT functioning in a separate genetic pathway from ULT1 and KAN1, with both pathways converging to promote basal cell fate along the apical-basal gynoecium axis.
ETT patterns the gynoecium by negatively regulating SPT expression (Heisler et al., 2001). Mutations in SPT produce gynoecia often unfused at the apical end, with reduced style, stigma, and septum growth and an absent transmitting tract, indicating a role for SPT in promoting the growth of the carpel margins and of the pollen tract tissues derived from them (Alvarez and Smyth, 1999;Heisler et al., 2001). In the gynoecium, spt mutations are largely epistatic to ett mutations and SPT is ectopically expressed in developing ett gynoecia, revealing SPT as a downstream target of ETT (Heisler et al., 2001). Because SPT plays a key role in mediating apical-basal patterning, we investigated whether SPT misregulation contributed to the ult1 kan1 gynoecium phenotype.
We examined the SPT expression pattern in developing wildtype, kan1-12, and ult1-2 kan1-12 gynoecia. SPT expression in wild-type Ler floral buds is observed from stage 4 in an inverted conical domain at the floral meristem apex, becoming most intense in the adaxial region at stage 6 ( Figure 5K) (Heisler et al., 2001). As the gynoecium primordium elongates at stage 7, SPT expression becomes restricted to the adaxial domain, with a wider distribution at the apex than at the base ( Figures 5A and 5L). This pattern of expression continues through stage 8 ( Figures 5B and 5M), stage 9 ( Figure 5C), and stage 10 ( Figure 5D), with SPT mRNA also being detected in the developing anthers during this period. In more mature flowers, SPT mRNA is limited to the stigmatic papillae, ovules, and transmitting tract ( Figures 5E and 5N). In developing kan1-12 buds, weak ectopic SPT expression is observed in the abaxial region of the gynoecium during stages 6 through 9 (Supplemental Figures 3A to 3D) and later in the thin ectopic outgrowths toward the base of the organ (Supplemental Figure 3E). In developing ult1-2 kan1-12 buds, ectopic SPT expression is detected throughout stage 6 gynoecia, although still most intensely in the adaxial domain ( Figure 5F). By stage 7, ectopic SPT expression in the outermost cells of the abaxial region is as strong as in the inner region ( Figures 5G and 5P), and by stage 8, it becomes resolved to the edge of the abaxial domain running from the apex to the base ( Figures 5H and 5Q). Beginning at stage 8, ectopic SPT expression is also clearly observed in the outgrowths developing along the abaxial surface of the valves (Figures 5H and 5Q). SPT expression continues in the ectopic outgrowths as the gynoecia mature, appearing primarily in the apical region corresponding to the developing stigmatic papillae ( Figures 5I, 5J, 5R, and 5S). These results show that SPT expression is negatively regulated by ULT1 and KAN1 in the abaxial and basal domains of the developing gynoecium.
Genetic analysis confirms that the negative regulation of SPT by the ULT1 and KAN1 proteins is biologically relevant. spt-2 gynoecia display an unfused apical end with reduced stigmatic and style tissue ( Figure 6A). ult1-2 spt-2 gynoecia show an additive phenotype consisting of three carpels, characteristic of ult1-2 plants, and an unfused apical end, characteristic of spt-2 plants ( Figure 6B). Likewise, kan1-12 spt-2 gynoecia exhibit an additive phenotype consisting of ectopic abaxial ovules, which The Plant Cell are slightly elongated but generally characteristic of kan1-12 plants, and an unfused apical end ( Figure 6C). Remarkably, the strong ult1-2 kan1-12 apical ectopic outgrowth phenotype ( Figure 6D) is partially suppressed in an spt-2/+ heterozygous background ( Figure 6E), and eliminating SPT activity in ult1-2 kan1-12 spt-2 gynoecia causes complete suppression of apical ectopic outgrowth formation ( Figure 6F). These results demonstrate that the ult1-2 kan1-12 apical-basal patterning defect is rescued by the spt-2 mutation in a dosage-dependent manner. Genetic and molecular analyses are both consistent with SPT activity mediating the ult1 kan1 basalization phenotype. Therefore, we conclude that ULT1 and KAN1 together pattern the apical-basal axis of the gynoecium by negatively regulating SPT expression. Finally, the adaxialized ectopic ovule phenotype of kan1-12 plants is enhanced in the absence of ULT1 and SPT activity. Eliminating SPT function in the kan1-12 background leads to the formation of slightly elongated ovule-like structures at the kan1-12 spt-2 gynoecium base, below the valves ( Figure 6C). In ult1-2 kan1-12 spt-2/+ gynoecia, these most basal structures are larger and more elongated ( Figure 6E), indicating that eliminating ULT1 activity while reducing the dosage of SPT also enhances the kan1-12 adaxialization defect. ult1-2 kan1-12 spt-2 plants display the most severe phenotype, in which numerous thin, elongated outgrowths arise and bifurcate from the base of the gynoecium ( Figure 6F). These data reveal a role for KAN1, ULT1, and SPT in repressing tissue growth in the abaxial, basal region of the gynoecium.
ULT2 and KAN2 Maintain the Apical-Basal Axis Together with ULT1 and KAN1 ULT2 plays a redundant role with ULT1 in promoting basal gynoecium polarity (Monfared et al., 2013). In addition, KAN2 acts redundantly with KAN1 to maintain the adaxial-abaxial polarity axis ULT and KAN Regulate Fruit Polarity 7 of 17 Pekker et al., 2005). kan1 kan2/+ gynoecia have basally expanded style and stigmatic tissue (Pekker et al., 2005), a phenotype that resembles that of ult1 kan1 gynoecia, suggesting that KAN2 may also play a part in patterning the apical-basal axis of the gynoecium. To gain insight into the genetic interactions between ULT1, ULT2, KAN1, and KAN2, we generated plants with different combinations of mutations in the four genes. We classified the gynoecium phenotypes into six classes according to their severity, with class 0 the least severe and class 5 the most severe ( Figure 7A). Class 0 plants display wild-type-looking gynoecia with no apical-basal defect ( Figure 7A), consisting of the ult1-2 (Supplemental Figure 4A), ult2-3 (Supplemental Figure 4B), kan2-1 ( Figure 7B; Supplemental Figure 4C), and ult2-3 kan2-1 (Supplemental Figure 4D) genotypes. Using kan2-1 plants as a representative of this class, we observe that they have no ectopic activation of the ProASA1:GUS marker ( Figure 7C) compared with the wild type ( Figure 4E) and that the GT142 pattern ( Figure 7D) is also similar to that of the wild type ( Figure 4I).
Class 1 plants display gynoecia with a weak apical-basal defect, consisting of thin ectopic outgrowths of style and stigmatic tissue arising from the basal region of the ovary along the replum ( Figure   7A). A large number of genotypes fell into this class (Supplemental Figures 4E to 4M), of which ult1-2 ult2-3 is representative ( Figure  7B). These gynoecia show ectopic activation of ProASA1:GUS in the basal ectopic outgrowths ( Figure 7C) but an unaltered GT142 pattern ( Figure 7D), indicating normal valve morphology.
Class 2 plants display gynoecia with a moderate apical-basal axis defect, with expansion of the apical style as well as basal ectopic outgrowths consisting of style and stigmatic tissue that are larger than class 1 outgrowths ( Figure 7A). The phenotypes of plants in this class are shown in Supplemental Figures 4N to  4R. kan1-2 is representative of this class ( Figure 7B), and as described for kan1-12 ( Figure 4G), these gynoecia display ectopic ProASA1:GUS activity in the apical and basal regions of the ectopic outgrowths ( Figure 7C). The GT142 pattern is unaltered ( Figure 7D), indicating no valve reduction.
Class 3 plants display gynoecia with a strong apical-basal defect, with expansion of style and stigmatic tissue nearly to the base of the ovary as well as significant valve reduction ( Figure  7A). The phenotypes of these plants are shown in Supplemental Figures 4S to 4Y. kan1-2 kan2-1/+ ( Figure 7B) is representative of this class, which also includes ult1 kan1 plants with slightly milder phenotypes. These gynoecia display strong ectopic ProASA1:GUS activity throughout the large ectopic outgrowths ( Figure 7C), and in some cases this ectopic style and stigmatic tissue encircles the central portion of the organ. In addition, GT142 activity in the valve margin does not extend to the base of the gynoecium but instead terminates in the middle ( Figure  7D), as in ult1 kan1 gynoecia ( Figure 4L). Thus, the ovaries of class 3 gynoecia are reduced to nearly half the wild-type size.
Class 4 plants display gynoecia with an extreme apical-basal axis defect, with large expansions of apical style and stigmatic tissue that in most cases encircle the whole organ, as well as more extreme valve reduction than class 3 gynoecia. The genotypes in class 4 are ult1-2 ult2-3 kan1-2 (Supplemental Figure 4Z), ult1-2 kan1-2 kan2-1/+ ( Figure 7B; Supplemental Figure 4AA), and ult1-2 ult2-3 kan1-2 kan1-2/+ (Supplemental Figure 4BB). These gynoecia exhibit strong ectopic ProASA1:GUS activation in the medial instead of the apical region ( Figure 7C). Furthermore, GT142 activity is confined to a ring toward the apical end of the gynoecia ( Figure 7D). Thus, class 4 plants display a strong expansion of style and stigma throughout the entire apical portion of the gynoecia as well as ovaries that are reduced to small rings just beneath the apical style and stigmatic tissue.
Class 5 plants display gynoecia with no distinguishable apicalbasal polarity axis ( Figure 7A). Genotypes in this class are kan1-2 kan2-1 ( Figure 7B), ult1-2 kan1-2 kan2-1, ult2-3 kan1-2 kan2-1, and ult1-2 ult2-3 kan1-2 kan2-1. These plants have highly compact inflorescences in which it is very difficult to distinguish individual gynoecia. Those that can be detected lack defined valves and consist of a surface of style and stigmatic tissue interspersed with ectopic adaxialized ovules ( Figure 7B). Removing ULT1 and/or ULT2 activity in the kan1 kan2 background does not further increase the severity of the phenotype. Thus, KAN1 and KAN2 are required to establish the apical-basal polarity axis of the gynoecium, and ULT1 and ULT2 make no additional contribution to this process outside of the KAN pathway.
[See online article for color version of this figure.]

of 17
The Plant Cell the various mutant combinations can be divided into classes that reflect an increasing phenotypic severity along the apicalbasal polarity axis. Analysis of the two molecular markers shows that, moving from class 1 to class 4 gynoecia, the ectopic style and stigmatic tissue outgrowths expand basally and become more severe until they cover the entire organ, while the ovary size is gradually reduced. Thus, these four genes interact genetically to restrict the apical cell types to the tip of the gynoecium and to promote the growth of the basal ovary.

ULT1, ULT2, KAN1, and KAN2 Act in a Dose-Dependent Manner
Our analysis of the ULT and KAN alleles also revealed strong dosage-dependent genetic interactions. Although none of the alleles generated a gynoecium phenotype when heterozygous, reducing the dosage of one of the genes could induce an ectopic outgrowth phenotype in the mutant background of another. For example, kan2 flowers have a wild-type class 0 gynoecium phenotype ( Figure 8A), but when ULT1 is heterozygous in this background, the kan2 ult1/+ gynoecia form very thin ectopic outgrowths along the replum ( Figure 8B), placing this phenotype in class 1. Similarly, when KAN1 is heterozygous in this background, the kan2 kan1/+ gynoecia display a class 1 ectopic outgrowth phenotype ( Figure 8C). Thus, reducing the dose of either ULT1 or KAN1 in the kan2 background is sufficient to generate a weak phenotype. ult1 flowers often form extra carpels but never make ectopic gynoecium outgrowths ( Figure 8D), whereas when KAN1 and KAN2 are heterozygous in this background, the ult1 kan1/+ kan2/+ gynoecia display a class 2 phenotype of moderate outgrowth formation ( Figure 8E). Therefore, reducing the dose of both KAN1 and KAN2 by 50% in the ult1 background is sufficient to generate a gynoecium phenotype.
Reducing the dosage of one of the ULT or KAN genes can enhance the ectopic outgrowth phenotypes produced by alleles of others. kan1 flowers have a moderate class 2 gynoecium (A) Diagrams of the class 0 to class 5 gynoecium phenotypes. Yellow represents the stigmatic tissue, blue the style, green the replum, orange the valves, brown the gynophore, pale green the pedicel, and red the ectopic ovules. Diagrams of class 0 to class 4 show front views, and the diagram of class 5 shows a top view of the gynoecia. (B) Representative gynoecium phenotypes from each class: kan2-1 (class 0), ult1-2 ult2-3 (class 1), kan1-2 (class 2), kan1-2 kan2-1/+ (class 3), ult1-2 kan1-2 kan2-1/+ (class 4), and kan1-2 kan2-1 (class 5). Bar = 150 mm. phenotype ( Figure 8F), but when both ULT1 and ULT2 are heterozygous in this background, the kan1 ult1/+ ult2/+ gynoecia shows a stronger class 3 phenotype of ectopic outgrowths tipped with stigmatic tissue from the top to the base ( Figure 8G). Therefore, reducing the dose of ULT1 and ULT2 enhances the kan1 phenotype. ult1 ult2 kan2 flowers have a weak class 1 gynoecium phenotype, exhibiting a few ectopic outgrowths at the base ( Figure 8H). When KAN1 is heterozygous in this background, the ult1 ult2 kan2 kan1/+ gynoecia display a class 2 phenotype with broad ectopic outgrowths at the base ( Figure  8I). Thus, reducing the dose of KAN1 enhances the ult1 ult2 kan2 gynoecium phenotype. Finally, kan1 kan2/+ flowers have a class 3 gynoecium phenotype ( Figure 8J) with a stronger apical-basal defect than the kan1 class 2 gynoecia phenotype ( Figure 8F), indicating that reducing the dose of KAN2 enhances the kan1 gynoecium phenotype. Together, these data demonstrate that plants lacking any one of the four genes are sensitive to the dosage of the other three. The dosage-dependent interactions between the ULT and KAN genes reveal that trxG genes and DNA binding transcription factor genes function in the same genetic pathway to regulate the apical-basal polarity axis during gynoecium development.

ULT2 Interacts Physically with ULT1, KAN1, and KAN2
The dosage-dependent genetic interactions between ULT1, ULT2, KAN1, and KAN2 suggested that their products could be part of a complex that regulates the apical-basal axis polarity in the gynoecium. For this to be the case, the expression domains of the genes must overlap during gynoecium formation. KAN1 and KAN2 specify abaxial organ cell identity, and correspondingly, KAN1 expression is detected in the abaxial domain of developing rosette and cauline leaves and in the abaxial region of stage 7 gynoecia (Kerstetter et al., 2001). This early KAN1 gynoecium expression pattern is opposite that of ULT1 and ULT2, which are expressed specifically in the adaxial domain (Carles et al., 2005). However, by stage 9, KAN1 mRNA is cleared from the abaxial cells and is detected exclusively in the adaxial domain of the gynoecium (Kerstetter et al., 2001), coincident with ULT1 and ULT2.
To independently verify this KAN1 expression pattern, we performed in situ hybridization on developing wild-type Ler gynoecia. We confirmed that KAN1 is expressed only in the abaxial region of stage 6 (Supplemental Figure 5A) and stage 7 (Supplemental Figure  5B) gynoecia. At stage 8, KAN1 mRNA is cleared from the abaxial region and becomes expressed exclusively in the adaxial domain (Supplemental Figure 5C), where it resolves to the ovule primordia during stage 9 (Supplemental Figure 5D). Thus, KAN1 has a highly dynamic expression pattern in the developing gynoecium, initiating in the abaxial domain and then switching to the adaxial domain, where it overlaps with ULT1 and ULT2 expression during the formation of the apical-basal polarity axis.
We tested the ability of the ULT1, ULT2, KAN1, and KAN2 proteins to associate physically by performing yeast two-hybrid (Y2H) and bimolecular fluorescence complementation (BiFC) assays (Walter et al., 2004). For the Y2H assays, the full-length coding regions of ULT1, ULT2, KAN1, and KAN2 were cloned into both bait (containing the GAL4 DNA binding domain) and prey (containing the GAL4 activation domain) vectors. Physical interaction between the proteins cloned into the bait and prey vectors bring the GAL4 DNA binding domain (BD) and activation domain (AD) together to initiate transcription of a HIS3 reporter gene that allows the yeast to grow on selective medium lacking histidine (James et al., 1996). Control Y2H assays behave as expected (Supplemental Figures 6A to 6D and 6F to 6I), except that, as reported previously (Kelley et al., 2012), KAN2 protein fused with the GAL4-BD causes HIS3 activation in the presence of empty prey vector (Supplemental Figure 6E). Therefore, Y2H assays were not performed using KAN2-BD. Yeast containing the following pairs of fusion proteins are able to grow in medium lacking histidine: ULT2-BD and ULT1-AD ( Figure 9A), ULT2-BD and KAN1-AD ( Figure 9B), ULT2-BD and KAN2-AD ( Figure 9C), ULT1-BD and KAN1-AD ( Figure 9D), ULT1-BD and KAN2-AD ( Figure 9E), and KAN1-BD and KAN2-AD ( Figure 9F). For each pair of protein fusions, yeast growth was also observed on selective medium when the activation and binding domains were switched (Supplemental Figures 6J to 6L).

of 17
The Plant Cell Negative control cotransformations yield no fluorescent signals (Supplemental Figures 7D, 7G, 7J, and 7M). All cells were also cotransformed with a ProRecA:RFP (for red fluorescent protein) construct as an internal marker to identify transformed cells. We observed reconstituted YFP fluorescent signal in cells cotransformed with ULT1-YC and ULT2-YN fusion proteins in the nucleus as well as in the cytoplasm ( Figure 9G), confirming that ULT1 and ULT2 can interact in plant cells (Monfared et al., 2013). Cells cotransformed with the KAN1-YN and ULT2-YC fusion proteins display YFP signal specifically in the nucleus ( Figure 9J), as do cells cotransformed with the ULT2-YN and KAN1-YC fusion proteins in the opposite orientation (Supplemental Figure 7P). Similarly, cells cotransformed with the KAN2-YN and ULT2-YC fusion proteins reconstitute the YFP signal in the nucleus ( Figure   9M), as do cells cotransformed with the ULT2-YN and KAN2-YC combination (Supplemental Figure 7S). We did not detect physical interactions between ULT1 and either KAN protein using this assay (Supplemental Figures 7V and 7Y) or between KAN1 and KAN2 (Supplemental Figure 7BB). Together, these two independent assays demonstrate that the ULT1, ULT2, KAN1, and KAN2 proteins can interact physically with one another, consistent with our genetic findings that the each of these genes is sensitive to the dosage of the others.

DISCUSSION
Differential gene expression is at the core of development, and transcription regulation is a central component of differential gene expression. The correct deployment of developmental programs in higher plants and animals requires complex, multilayered gene regulatory systems (Chen and Rajewsky, 2007). Transcription processes are tightly regulated, often by complexes of factors, to ensure that genes are activated or repressed at precise moments during development. In this study, we show  how cooperation between chromatin-remodeling and transcription factors is crucial for the correct regulation of genes affecting leaf and gynoecium patterning during Arabidopsis development. Leaves and gynoecia develop along three different axes of polarity, and we show that the trxG genes ULT1 and ULT2 as well as the transcription factors KAN1 and KAN2 differentially regulate the establishment of two of these axes. We present evidence that ULT1 and KAN1 function antagonistically to pattern the gynoecium and leaves along the adaxial-abaxial axis but act together to pattern the gynoecium along the apical-basal axis.

ULT1 and KAN1 Oppositely Regulate Two Axes of Gynoecium Polarity
Our study unveils a role for the ULT1 trxG factor in maintaining adaxial identity in leaves and in early gynoecium development that is revealed in the absence of KAN1. A striking phenotype of ult1 kan1 plants is the suppression of the kan1 adaxialized phenotype in rosette leaves and gynoecia ( Figure 2). Therefore, ULT1 and KAN1 act antagonistically to establish the adaxialabaxial organ polarity axis. This is consistent with expression data showing that KAN1 is expressed on the abaxial side of very young leaves and the abaxial side of stage 7 gynoecium primordia (Supplemental Figure 5) (Kerstetter et al., 2001), whereas ULT1 is specifically expressed in the adaxial domain of the gynoecium primordia (Carles et al., 2005). Because adaxial-abaxial polarity in Arabidopsis is specified by the opposing activities of genes that separately specify either adaxial or abaxial identity (Bowman et al., 2002), we propose that ULT1 may be required for the misexpression of adaxial identity-promoting genes such as class III HD-ZIP genes (McConnell et al., 2001) in kan1 leaves and gynoecia. Failure to sustain ectopic expression of such genes in the absence of ULT1 would lead to suppression of the kan1 adaxialized phenotype in ult1 kan1 plants. However, ult1 mutations are not sufficient to suppress the more severe adaxialized phenotype of kan1 kan2 plants (Supplemental Figure  4). This is perhaps because in kan1 kan2 tissues, high levels of ectopic adaxial identity gene expression are maintained by other factors besides ULT1. Nonetheless, our data demonstrate that ULT1 and KAN1 function antagonistically to pattern the gynoecium along the adaxial-abaxial axis, uncovering a novel role for the trxG factor ULT1 in regulating this organ polarity axis.
Our data also show that KAN1 promotes basal identity, as kan1 gynoecia display a slight downward expansion of style tissue on the apical part of the organ as well as a small ectopic outgrowth near the base (Figures 3 and 4). Therefore, this work uncovers a role for KAN1, which had previously only been implicated in regulating adaxial-abaxial polarity, in patterning the gynoecium along the apical-basal axis. Interestingly, ult1 mutations strongly enhance the apical-basal polarity defect in kan1 plants. This enhancement is evident by the strong enlargement of the style ectopic outgrowths and expansion of the stigmatic tissue in ult1 kan1 gynoecia (Figure 3), the basalization of the medial vascular bundle bifurcation points, the reduction of the basal ovary (Figure 4), and the basalization of the style marker ProASA1:GUS (Figure 4). All of these features are associated with defects in apical-basal patterning (Sessions and Zambryski, 1995;Nishimura et al., 2005). These data indicate that ult1 kan1 gynoecia exhibit a basalization of the apical tissue types, demonstrating that ULT1 and KAN1 act together to restrict the style and stigma to the apical domain and to promote the formation of the basal ovary. Thus, these two genes play important and overlapping roles in patterning the gynoecium along the apicalbasal axis.
ULT1 and KAN1 act together to pattern the apical-basal axis by negatively regulating SPT transcription. This finding stems from our demonstrations that SPT is ectopically expressed in the abaxial and basal domains of ult1 kan1 gynoecia ( Figure 5) and that mutations in SPT completely suppress the ectopic outgrowths observed in ult1 kan1 gynoecia ( Figure 6). A similar scenario has been described for the relationship between ETT and SPT. spt mutations suppress the ett gynoecium phenotype, and ETT negatively regulates SPT expression (Heisler et al., 2001). Because the SPT promoter contains several putative auxin response element sites, it has been suggested that ETT may bind SPT at those sites and block its transcription in the gynoecium (Heisler et al., 2001). However, a scenario where the ULT1/KAN1 proteins also directly bind the SPT promoter is less likely. In contrast with the exclusivity of the ETT and SPT expression patterns (Sessions et al., 1997;Heisler et al., 2001), the ULT1 and KAN1 expression patterns overlap with that of SPT during stages 7 to 9 of gynoecium development (Heisler et al., 2001;Kerstetter et al., 2001;Carles et al., 2005). In addition, the genetic role of ULT1 as a trxG factor that functions as a transcription activator (Carles and Fletcher, 2009) makes it unlikely that ULT1 is a direct negative regulator of SPT transcription. An alternative scenario is that ULT1/KAN1 activates the transcription of a non-cell-autonomous intermediary factor that, in turn, represses SPT, an indirect regulatory pathway. Further experiments would be needed to discover the identity of this intermediary factor(s).
Our experiments show that reducing SPT activity has differential effects in different regions of the gynoecium. In ult1 kan1 spt gynoecia, ectopic outgrowth formation along the replum is suppressed, yet a greater extent of adaxialized tissue outgrowth occurs at the very base, below the valves ( Figure 6). Because SPT limits cell proliferation in rosette leaves (Ichihashi et al., 2010), we propose that the proliferation and outgrowth of the adaxialized ovule tissue at the base of ult1 kan1 gynoecia is likewise limited by the activity of SPT, such that when SPT function is eliminated, the tissues grow extensively. Interestingly, similar to what we observed for SPT and ULT1/KAN1, differential interactions among SPT and the CUP-SHAPED COTYLEDON (CUC) genes also regulate domain-specific aspects of gynoecium development (Nahar et al., 2012). SPT directs carpel fusion in the apical region by negatively regulating CUC1 and CUC2 expression, whereas all three genes promote the formation of margin-derived structures in the basal region. Therefore, region-specific regulatory modules appear to be repeatedly utilized during gynoecium formation.
The ult1 kan1 gynoecium phenotypes resemble those of ett gynoecia, and both ULT1/KAN1 and ETT regulate the SPT expression domain. Yet, despite these similarities, our data suggest that ULT1 and KAN1 may act through a separate pathway from ETT (Supplemental Figures 1 and 2). This is concordant with the mutually exclusive expression patterns of ETT and KAN1 at stages 7 to 9 of gynoecium formation and the mutually exclusive 12 of 17 The Plant Cell expression patterns of ETT and ULT1 throughout gynoecium development. It is also consistent with reports that KAN1 activity is not required for the transcription of ETT and ARF4 and that neither ETT nor ARF4 rescues the kan phenotypes (Pekker et al., 2005). ETT interacts physically with KAN1 in the Y2H system (Kelley et al., 2012), but no data yet confirm that this interaction occurs in planta or has biological relevance. Thus, the available evidence converges on the conclusion that ETT and ULT1/KAN1 act in separate pathways to pattern the apical-basal axis in Arabidopsis gynoecia.

trxG Proteins and DNA Binding Transcription Factors Associate Physically in Planta
Multiple examples exist in the literature of alleles that show quantitative interdependence, such as clv1 and clv3 (Clark et al., 1995), which presages the demonstration of a physical interaction between the proteins (Ogawa et al., 2008). In animals, genetic analysis showed that two members of the Polycomb family, M33 and bmi1, cooperate in a dosage-dependent manner, and immunoprecipitation experiments established that these PcG proteins are part of the same large multiprotein complex (Bel et al., 1998). Similarly, our genetic analysis reveals synergistic effects between the ult1, ult2, kan1, and kan2 alleles (Figure 7; Supplemental Figure 4) and a strong dosage sensitivity among alleles of all four genes with respect to the basalized gynoecium phenotype (Figure 8; Supplemental Figure 4). These dosage-dependent interactions are consistent with their products constituting a biochemical complex. Correspondingly, we find that the ULT1/2 and KAN1/2 proteins interact physically with one another in all pairwise combinations in yeast cells and that ULT2 associates with both KAN1 and KAN2 in planta ( Figure 9). We were unable to confirm using BiFC assays a physical interaction between ULT1 and either KAN protein or between KAN1 and KAN2, potentially because the interactions are too weak to detect by this method or because additional proteins are needed in onion epidermal cells to stabilize the interactions. Nonetheless, these findings indicate that ULT1, ULT2, KAN1, and KAN2 may act together as components of a complex in which chromatin-modifying factors and transcription factors associate to regulate target gene expression and establish the apical-basal polarity axis in the gynoecium. A similar association may occur between the SWI2/SNF2 family chromatin-remodeling protein GYMNOS and the YABBY family transcription factor CRABS CLAW, which interact genetically to pattern the adaxialabaxial axis of the gynoecium (Eshed et al., 1999). trxG proteins function in large complexes to modify chromatin during eukaryotic growth and development (Schuettengruber et al., 2011). It has been proposed that the highly specific regulation of gene expression by the ULT1 trxG factor (Carles and Fletcher, 2009) requires various interacting proteins that contribute capacities such as histone mark recognition, sequencespecific DNA binding, and transcription activation . ULT1 associates in planta with the trxG factor ATX1 (Carles and Fletcher, 2009), an H3K4 methyltransferase that recruits TBP and RNA Pol II (Ding et al., 2011) to promoters for transcription initiation (Ding et al., 2012). However, little is known about how the trxG factors themselves become targeted to specific genes. The KAN proteins are members of the GARP family of transcription factors (Kerstetter et al., 2001), which contain a Myb-like domain that can bind DNA (Hosoda et al., 2002). Here, KAN1 and KAN2 emerge as putative components of an ATX1/ULT1/ULT2 multiprotein complex, with the ULT proteins acting as molecular links between the chromatin-modifying enzymes and the sequence-specific transcription factors. The transcription factors can provide target gene specificity to the complex by binding to precise regions of the DNA, facilitating chromatin modification by the trxG enzymes to maintain the active transcriptional state of the target locus (Narlikar et al., 2002).
Previous work has shown a role for KAN1 in the direct transcriptional repression of the ASYMMETRIC LEAVES2 gene (Wu et al., 2008). Recent transcriptomics studies of Arabidopsis seedlings have identified large numbers of downregulated KAN1 target genes as well as a smaller number of upregulated target genes (Merelo et al., 2013;Reinhart et al., 2013;Huang et al., 2014). These data indicate that KAN1 functions primarily as a transcriptional repressor but can also promote active transcription. (A) Early during gynoecium initiation, ULT1/2 and KAN1 are expressed in mutually exclusive domains. ULT and KAN act antagonistically to pattern the adaxial-abaxial polarity axis, with KAN1 and KAN2 conferring abaxial identity and ULT1 and ULT2 counterbalancing their function in the adaxial domain.
(B) Later, ULT1/2 and KAN1 are expressed in coincident domains, and their protein products associate to pattern the apical-basal axis. A plausible scenario is that the KAN transcription factors bind to specific downstream target loci and the associated ULT trxG factors promote their active transcriptional status. This complex may activate a specific target gene or genes (X) that repress SPT expression in the basal, abaxial region of the gynoecium via a separate genetic pathway from that of ETT.
Thus, the KAN1 Myb domain protein, like a number of other transcription factors (Fujimoto et al., 2000;Miao et al., 2004;Ikeda et al., 2009), may have a dual capacity to act as an activator or a repressor depending on the developmental context.
This study envisions a scenario where, in the leaves and in the early stages of gynoecium development, the trxG factor ULT1 maintains high transcription levels of one or more downstream adaxial genes that are negatively regulated by KAN1 in abaxial cells ( Figure 10A). At these stages, ULT1/2 and KAN1 are expressed in exclusive domains in the gynoecium and thus do not associate physically. Later, ULT1/2 and KAN1 become coexpressed and act in a dose-dependent manner to pattern the gynoecium along the apical-basal axis ( Figure 10B). The ULT and KAN proteins promote basal gynoecium identity by associating together to maintain the active transcriptional status of its target genes. These proteins would be responsible for activating specific target genes that would in turn non-cell-autonomously repress SPT expression in the basal region of the gynoecium via a separate genetic pathway from ETT. The Arabidopsis gynoecium is a highly complex organ in which patterning events are tightly regulated. Our work provides insights into this key developmental process, as it uncovers a link between chromatin regulatory proteins and transcription factors in the control of organ patterning. It will be important to characterize additional elements that associate with the ULT and KAN proteins to understand how chromatinremodeling and transcription processes influence each other to regulate organ formation.

Plant Materials and Growth Conditions
All Arabidopsis thaliana genotypes were in the Ler accession except for the ProASA1:GUS lines, which were in a mixed Ler/Columbia-0 background. The kan1-13 line was isolated from a 0.2% ethyl methanesulfonatemutagenized population of ult1-2 seeds in the Ler accession and was introgressed into Ler three times prior to analysis. Plants were grown on either Murashige and Skoog medium or soil (1:1:1 mixture of perlite: vermiculite:topsoil) under continuous light (120 mmol m 22 s 21 ) at 22°C. Seeds were stratified at 4°C for 4 d before exposure to light. Plants were watered every 2 d with a 1:1500 dilution of Miracle-Gro 20-20-20 fertilizer for the first week. Before in vitro culture, seeds were surfaced-sterilized for 45 min in a vacuum chamber containing 1.5% (v/v) HCl in a bleach solution. Mutant plants were genotyped by PCR prior to analysis (primers are listed in Supplemental Table 1).

Cloning and Sequencing of the kan1-13 Allele
The two halves of the KAN1 gene were amplified from wild-type Ler and kan1-13 plants by PCR (primers are listed in Supplemental Table 1) with Phusion enzyme (Finnzymes), and each was cloned into the PCR4-topo vector (Invitrogen). Positive clones were screened by PCR, and plasmid extraction was performed using the AccuPrep Plasmid Extraction Kit (Bioneer). Plasmids were sequenced using M13F and M13R primers. DNA and protein sequence analysis was performed using the Sequence Manipulation Suite (Stothard, 2000).

RNA Extraction and Quantitative Real-Time RT-PCR
Total RNA was isolated from flower clusters, and cDNA was prepared as described (Monfared et al., 2013). Ten nanograms of cDNA was amplified according to the manufacturer's instructions with iQ SYBR Green Supermix on the CFX96 Real-Time PCR detection system (Bio-Rad) using a 60°C annealing temperature for each primer pair (primers are listed in Supplemental Table 1). Three technical replicates were performed for each of three biological replicates.

Histology and in Situ Hybridization
Vascular staining was performed as described (Sessions and Zambryski, 1995). For GUS staining, GUS activity was analyzed as described (Jefferson, 1989) with the modification that 1 mM potassium ferrocyanide and 1 mM potassium ferricyanide were used. Incubation staining time was 12 to 16 h following vacuum infiltration. Because ProASA1:GUS activity is fertilization dependent (Sessions and Zambryski, 1995), we manually fertilized each gynoecium with pollen from the same flower 6 to 7 h prior to fixation.
The SPT antisense probe was transcribed from an SPT cDNA plasmid using the T3 promoter after SalI digestion. The SPT sense probe was transcribed using the T7 promoter after BamHI digestion. The KAN1 antisense probe was transcribed from a KAN1 cDNA plasmid using the T7 promoter after HindIII digestion. Each probe was hydrolyzed for 37 min. Probe synthesis, tissue fixation, and in situ hybridization were performed as described  with an additional 10-min acetic anhydride treatment (0.1 M triethanolamine, 0.025 M acetic anhydride) after the formaldehyde fixation step. Slides were mounted using Aqua Polymount (PolySciences).

Y2H Assays
Full-length ULT1, ULT2, KAN1, and KAN2 cDNAs were PCR-amplified from Ler inflorescence tissue (primers are listed in Supplemental Table 1). Each cDNA was cloned into the pCD and pC-ACT plasmids (Durfee et al., 1999) using the NdeI and BamHI restriction sites. Yeast strain PG 694a (James et al., 1996) was transformed with the recombinant plasmids as described (Durfee et al., 1999). Yeast transformants were selected on medium lacking leucine and tryptophan, and selection for protein-protein interactions was performed on medium lacking leucine, tryptophan, and histidine, supplemented with 3 mM 3-amino-1,2,4-triazole. Three independent colonies were separately cultured and grown for analysis, and each culture was spotted on the plate in duplicate.

BiFC Assays
The full-length ULT1, ULT2, KAN1, and KAN2 coding sequences were PCR-amplified (primers are listed in Supplemental Table 1) and blunt-end cloned into the pENTR-TOPO vector (Invitrogen). The coding sequences were then transferred into the destination vectors pE-SPYNE-GW and pE-SPYCE-GW (Weltmeier et al., 2006) through an LR reaction (Invitrogen). Positive clones were screened by PCR, and plasmid extraction was performed using the AccuPrep Plasmid Extraction Kit (Bioneer). The plasmids were sequenced using primers listed in Supplemental Table 1. The constructs were transformed into onion (Allium cepa) epidermal cells by particle bombardment using a Biolistic PDS-1000/He unit (Bio-Rad), as described (Sanford et al., 1993), and visualized by epifluorescence microscropy. Cobombardments were processed with 4 mg of each YFP construct plasmid construct used per shot, along with 0.5 mg of the pRecA-RFP construct. Bombardments were repeated at least three times for each combination.

Microscopy
Plant material was visualized using a Zeiss Stemi SV11 microscope, and images were acquired using a Canon D-40 digital camera. Cleared

of 17
The Plant Cell gynoecia were imaged using a Zeiss Axiophot fluorescence microscope under dark-field conditions. Individual flowers were prepared for scanning electron microscopy observation as described . For YFP visualization in onion cells, epidermal peels were examined 24 h after bombardment using a Zeiss Axiophot microscope. YFP and RFP fluorescence was visualized with the fluorescein isothiocyanate channel, and images were acquired with a Zeiss camera. Images of in situ hybridization slides were taken with a Q imaging Retiga-2000R digital camera connected to a Leica M216F microscope.

Supplemental Data
The following materials are available in the online version of this article. Supplemental Table 1. Primer Sequences Used in the Experimental Procedures.