AIL and HDG proteins act antagonistically to control cell proliferation

Plant growth is driven by stem cells within the meristems, which are maintained throughout the lifespan of a plant to ensure continued growth. At the same time, stem cell proliferation has to be kept in balance and contained within the meristem to prevent neoplastic growth. The AINTEGUMENTA-LIKE (AIL) subfamily of the APETALA2/ETHYLENE RESPONSE ELEMENT-BINDING FACTOR (AP2/ERF) family of transcription factors play an important role in defining the meristematic competence of plant cells (Horstman et al., 2014b). The Arabidopsis thaliana (L.) Heyhn (Arabidopsis) AIL clade comprises AINTEGUMENTA (ANT), AIL1, PLETHORA1 (PLT1), PLT2, AIL6/PLT3, PLT7, BABY BOOM (BBM) and AIL5/PLT5 (Nole-Wilson et al., 2005). Arabidopsis AIL genes are expressed in the embryo, the flower and the root and shoot meristems, where they act redundantly to define and/or maintain the stem cell niches (Galinha et al., 2007; Mudunkothge and Krizek, 2012; Horstman et al., 2014b). Although mutations in single AIL genes only lead to minor developmental defects, double and triple ail mutants exhibit more severe phenotypes, such as rootlessness (plt1;plt2;plt3), embryo lethality (bbm;plt2) (Galinha et al., 2007) or shoot meristem arrest (ant;ail6/plt3;plt7) (Mudunkothge and Krizek, 2012). The overexpression phenotypes of AIL proteins also support the notion of a role for these proteins in promoting meristematic competence. Overexpression of Brassica napus BBM or Arabidopsis AIL5/PLT5 induces formation of somatic embryos (Boutilier et al., 2002; Tsuwamoto et al., 2010), whereas overexpression of PLT1 and PLT2 induces ectopic root identity (Aida et al., 2004). In addition, the increased cell divisions due to AIL overexpression can also lead to increased floral organ size, as shown for both AIL5/PLT5 and ANT overexpression (Krizek, 1999; Nole-Wilson et al., 2005).

Although it is clear that AILs are key regulators of meristem function and cell proliferation, how AIL overexpression can trigger ectopic organ formation or embryogenesis is poorly understood. Transcription factor function is mediated in the context of multi-protein complexes. To provide insight into the mode of action of BBM and the signalling network in which it functions during cell proliferation, we identified and characterized BBM-interacting proteins. Here, we show that BBM and other AIL proteins interact with members of the HD-ZIP class IV/HOMEODOMAIN GLABROUS (HDG) transcription factor family. Sixteen HD-ZIP IV proteins have been identified in Arabidopsis, including MERISTEM LAYER1 (ATML1), GLABRA2 (GL2), ANTHOCYANINLESS2 (ANL2) and PROTODERMAL FACTOR2 (PDF2) (Nakamura et al., 2006). HD-ZIP IV/HDG genes are expressed in the L1 layer throughout the plant, where they function to specify epidermis identity and control development of its associated structures, such as trichomes, stomata or giant cells (Abe et al., 2003; Nakamura et al., 2006; Roeder et al., 2012; Peterson et al., 2013; Takada et al., 2013). Our results show that BBM and HDG proteins have antagonistic functions, with BBM stimulating cell proliferation and HDGs stimulating cell differentiation. In addition, we found evidence for transcriptional cross-regulation between BBM/AIL and HDG genes, suggesting a complex regulatory network for cell proliferation control.

We employed the yeast two-hybrid system to identify BBM-interacting proteins. Due to the strong and extensive transcriptional autoactivation activity of B. napus BBM1 (supplementary material Figs S1 and S2, Table S1 and see supplementary methods), we used the CytoTrap system (Aronheim, 1997) to screen a library created from B. napus embryos for interactions with BBM. We identified ten HDG transcription factors as interacting partners (supplementary material Table S2), nine of which are most similar to Arabidopsis HDG11 and one to HDG1 (supplementary material Table S2). Our subsequent studies focussed on the Arabidopsis BBM and HDG orthologs.

The interaction between BBM and HDG1 and HDG11 proteins was verified in planta, using Förster Resonance Energy Transfer detected via Fluorescence Lifetime Imaging Microscopy (FRET-FLIM; Fig. 1A) (Bücherl et al., 2014), as was the interaction with HDG12, which functions redundantly with HDG11 in trichome development (Nakamura et al., 2006). These results indicate that BBM also interacts with HDG proteins in planta. Using the Gal4 yeast two-hybrid system we found that BBM also interacts with HDG2, HDG3, HDG10, ANL2, ATML1 and PDF2 (Fig. 1B). Next, we determined whether these interactions also extend beyond BBM, by testing for HDG interactions with AIL proteins from the two major AIL clades, the ANT clade (ANT and AIL1) and the BBM/PLT clade (PLT7) (Horstman et al., 2014b). All three AIL proteins interacted with multiple HDG proteins (Fig. 1B). Our results show that BBM and other AIL proteins interact with phylogenetically distinct members of the HDG family (Nakamura et al., 2006).

Previous studies showed that BBM, HDG1, HDG11 and HDG12 are expressed in embryos and roots (Boutilier et al., 2002; Nakamura et al., 2006; Galinha et al., 2007). We examined their expression patterns in more detail using translational GFP fusion reporters. BBM expression was observed throughout the embryo from the 4-celled stage until the globular stage (Fig. 2A,E,I) and became basally localized at the heart stage, as previously reported (Fig. 2M) (Galinha et al., 2007). HDG11 and HDG12 expression was observed in all cells at the 4- and 16-cell embryo stages (Fig. 2C,D,G,H), and became restricted to the protoderm from the globular stage onward (Fig. 2K,L,O,P). HDG1 was weakly expressed and its expression was first observed in the embryo protoderm starting at the late globular stage (Fig. 2J).

BBM was expressed in the stem cell niche and the provascular tissue of mature roots (Fig. 2Q). Expression of all three HDG genes was observed in the epidermis, the outer layer of columella cells and lateral root cap of mature primary and lateral roots (Fig. 2R-T). These HDG genes were also expressed in the L1 layer of the floral meristems (Fig. 2U-W), shoot apical meristem (SAM) and leaf primordia (Fig. 2X-Z). HDG1 expression was also observed in the subepidermal layers of the flower meristem, SAM and leaf primordia (Fig. 2U,X). BBM expression was not observed in the shoot or flower. In summary, BBM expression overlapped with HDG11 and HDG12 expression during early embryo development, and later with all three HDG genes in progressively smaller regions of the protoderm. Post-embryonically, there was only a small overlap in expression of BBM and the three HDG genes, in a few epidermal cells close to the root stem cell niche.

To determine the functions of HDG proteins, we generated Arabidopsis HDG1, HDG11 and HDG12 overexpression lines (p35S::HDG). Approximately 10% of the primary transformants (n≥200 per construct) showed (similar) mutant phenotypes, with HDG1 resulting in the most severely altered phenotypes. Most of the affected seedlings were small and showed increased anthocyanin production (Fig. 3B). Seedlings of the most severe lines had a short primary root lacking lateral roots and stopped growing after producing a few leaves that were narrow and curled upward (Fig. 3B). Leaf fusions and leaves with holes were occasionally observed (Fig. 3C,D). The majority of seedlings with these phenotypes either died or was sterile, complicating further analysis of the lines. Therefore, we created dexamethasone (DEX)-inducible GR-HDG1, GR-HDG11 and GR-HDG12 proteins (p35S::GR-HDG) and selected the primary transformants directly on DEX-containing medium. We observed the same mutant phenotypes as described above. Again, the p35S::GR-HDG1 (n=468) mutant phenotypes were most pronounced, and in some cases more severe than the p35S::HDG1 mutant phenotypes. The most severely affected p35S::GR-HDG1 seedlings developed very narrow, gutter-shaped or radialized leaves (Fig. 3E-G, inset K), with occasional leaf ruptures (Fig. 3H). No aberrant phenotypes were observed when p35S::GR-HDG1 seedlings were grown without DEX (Fig. 3E, left) and overexpression of the HDG1 transgene was confirmed (supplementary material Fig. S3), indicating that the observed phenotypes were due to ectopic expression of HDG1.

The phenotypes of p35S::HDG1 and DEX-induced p35S::GR-HDG1 seedlings were examined in more detail using scanning electron microscopy (SEM). SEM analysis showed that the leaves of some HDG1 overexpression lines were radialized, but without an obvious abaxial or adaxial identity (Fig. 3N-P). The leaf surface comprised large numbers of smooth, elongated cells (Fig. 3O,P), reminiscent of the cells found on sepals and leaf margins (Fig. 3J) (Roeder et al., 2012), rather than the puzzle-shaped cells found in wild-type leaves (Fig. 3J). Some of the HDG1 overexpression seedlings developed a first set of radialized leaves, but did not grow further (Fig. 3M,N), whereas other HDG1 overexpression seedlings developed a radialized first leaf pair and a second leaf pair (Fig. 3O) with normal adaxial/abaxial patterning, although with a larger number of elongated cells (Fig. 3O). In addition, we observed holes (Fig. 3K) and large ruptures in the leaf epidermis (Fig. 3L). The altered leaf shape of the HDG1 overexpression seedlings complicated a general comparison of epidermal characteristics in these lines with those of wild-type seedlings (Fig. 3J). We did not observe any changes in trichome morphology; however, the radialized leaves of HDG1 overexpression seedlings contained less trichomes (Fig. 3M-P), and these were often positioned on the distal end of the leaf (Fig. 3O).

The shoot and root meristem were also affected by ectopic HDG1 overexpression. In the most severe cases, the shoot meristem was absent (Fig. 3L). Small leaves were observed occasionally in meristem-arrested seedlings (Fig. 3K). These were visible at a later stage, when the surrounding leaves were fully developed. It was not clear whether these leaves developed from axillary meristems or through adventitious growth, but they never developed further. When these seedlings were transferred to medium lacking DEX prior to complete meristem arrest, they recovered and developed into wild-type-looking seedlings.

The loss of root meristem function due to HDG1 overexpression was confirmed by the reduced growth rate and shortened root meristem of DEX-induced p35S::GR-HDG1 seedlings compared with wild-type seedlings (Fig. 4). Despite these root growth defects, DEX-induced p35S::GR-HDG1 seedlings were able to produce lateral roots (Fig. 4A). The root meristem defect observed here resembles the phenotype of bbm;plt loss-of-function mutants (Galinha et al., 2007), suggesting opposite roles for BBM and HDG1 proteins, with BBM promoting root meristem activity and HDG1 stimulating meristem differentiation.

The elongated cells found in HDG1 overexpression lines are reminiscent of giant cells, which are differentiated, endoreduplicated cells found in the sepal epidermis (Roeder et al., 2012). Similarly elongated cells are found along the margin of cotyledons and leaves (Fig. 5A,B) and in the root. We used the enhancer trap line YJ158, which reports GUS activity in giant/elongated cells (Eshed et al., 2004; Roeder et al., 2012) (Fig. 5C), to determine whether HDG1 overexpression seedlings show enhanced giant/leaf margin cell production. In DEX-induced p35S::GR-HDG1 seedlings with the narrow leaf phenotype (Fig. 3B), GUS staining was more intense, but still restricted to the margins, as in the control (data not shown), whereas p35S::GR-HDG1 seedlings with gutter and/or pin-shaped leaves (Fig. 3G and Fig. 5D,E) showed GUS expression throughout the leaf surface, and also in the cotyledon blade and petiole (Fig. 5F). In contrast with HDG1 overexpression lines, cotyledons of p35S::BBM seedlings consist of small, undifferentiated cells (Fig. 5G,H): the cotyledons lacked the jigsaw-shaped cells, stomata and the elongated margin cells of wild-type leaves (Fig. 5A,B). As expected based on cell morphology, YJ158 marker expression was weak (Fig. 5I) or completely absent in p35S::BBM seedlings.

Sepal giant cells are highly endoreduplicated, a differentiation process that occurs after the establishment of giant cell identity (Roeder et al., 2012). Endoreduplication is also considered a sign of advanced cell differentiation in leaf pavement cells (Melaragno et al., 1993). In accordance with our observations on changes in epidermal differentiation, we observed a shift toward higher ploidy levels in p35S::GR-HDG1 seedlings and toward lower ploidy levels in p35S::BBM seedlings (Fig. 5J). Taken together, our results suggest that HDG1 and BBM promote and inhibit epidermal cell differentiation, respectively.

To further investigate the roles of HDG proteins during development, we examined hdg1, hdg11 and hdg12 mutant phenotypes during embryogenesis and root development, the developmental stages in which BBM functions. However, none of the single, double or triple hdg1, hdg11 or hdg12 lines showed mutant phenotypes in these tissues (supplementary material Fig. S4 and see supplementary methods). The lack of embryo and root phenotypes might reflect redundancy between members of the HDG family (Nakamura et al., 2006).

We observed that ∼1% of p35S::HDG primary transformants formed ectopic shoots (Fig. 6A,B) or embryo-like tissue. None of the transformants with these phenotypes showed the overexpression phenotypes described above. Notably, these proliferating seedlings resemble BBM overexpression seedlings, which also show ectopic organ formation and somatic embryogenesis (SE) (Boutilier et al., 2002). Cell proliferation phenotypes were also observed at a similar frequency in p35S::GR-HDG1 primary transformants that were grown on DEX-containing medium, but that failed to recover the wild-type phenotype after transfer to DEX-free medium (Fig. 6C-E). These data, together with the lack of similar phenotypes in hdg1 T-DNA insertion mutants, suggest that co-suppression of multiple HDG genes underlies this cell proliferation phenotype.

We could detect GR-HDG1 transgene expression in the p35S::GR-HDG1 plants that recovered their wild-type phenotype after transfer to DEX-free medium, but could not detect GR-HDG1 expression in the plants that failed to recover, i.e. continued to overproliferate, in the absence of DEX (supplementary material Fig. S5). Expression of seven selected HDG genes (ATML1, PDF2, HDG2, HDG11, HDG12, ANL2, HDG3) was also reduced in these proliferating lines compared with wild-type seedlings (supplementary material Fig. S5). However, endogenous HDG1 expression was variably up- or downregulated in these lines depending on the quantitative real-time RT-PCR (qPCR) primer set that was used (supplementary material Fig. S5). Similar variable qPCR results were previously reported for silenced genes and could be caused by incomplete degradation of mRNA fragments (Shepard et al., 2005; Holmes et al., 2010). Expression analysis in subsequent generations was further complicated by the limited survival and fertility of these seedlings, and by the reversion of the surviving lines to the characteristic DEX-dependent overexpression phenotypes in the subsequent generation. Although indirect, these results imply that the cell proliferation phenotypes observed in a subset of the p35S::GR-HDG1 plants are not due to ectopic HDG1 overexpression, but rather to HDG gene silencing.

To determine whether downregulation of multiple HDG genes could cause these overproliferation phenotypes, we developed an artificial microRNA (amiRNA) construct (Schwab et al., 2006) that is predicted to target HDG3, HDG7, HDG11, PDF2 and ATML1. We observed the same cell proliferation phenotypes in four primary transformants after transformation of this amiRNA construct to the hdg1;anl2 double mutant (2/119 lines; Fig. 6F,G), and wild-type Col-0 (2/467 lines; Fig. 6H,I). These HDG amiRNA lines could not be propagated via seed. Together, the GR-HDG1 and amiRNA data provide support to the hypothesis that downregulation of multiple HDG genes leads to ectopic cell proliferation.

If HDG1 and BBM function antagonistically in the regulation of cell proliferation, then co-overexpression of HDG1 and BBM should mitigate the overexpression phenotype of the other protein. To test this hypothesis, we transformed a p35S::BBM construct to a characterized p35S::GR-HDG1 line (>90% penetrance of the phenotype) and examined the phenotypes of the double-transgenic seedlings before and after DEX-activation of the GR-HDG1 protein. The expression of both transgenes was verified using qPCR (supplementary material Fig. S6). The effect of HDG1 overexpression on the BBM phenotype was dependent on the penetrance of the BBM SE overexpression phenotype (Table 1). A line that showed a high penetrance of BBM-mediated SE was unaffected by co-overexpression of HDG1 (Table 1, line 1), whereas SE was reduced in double-transgenic lines with a lower penetrance of this BBM overexpression phenotype (Table 1, lines 2-5). In addition, we observed that the HDG1 overexpression phenotype was also compromised in the co-overexpression lines compared with the phenotype of the parental p35S::GR-HDG1 line, even in lines with a mild BBM overexpression phenotype (supplementary material Fig. S7). This suggests that BBM and HDG1 function antagonistically and that the balance between cell proliferation and differentiation depends on their relative concentrations.

We performed microarray experiments to understand how HDG1 controls cell proliferation, as well as its functional relationship with BBM. Direct HDG targets were identified using 5-day-old p35S::GR-HDG1 seedlings treated with DEX for 8 h in the presence of the translational inhibitor cycloheximide (CHX). HDG1 expression was significantly upregulated in CHX-treated p35S::GR-HDG1 seedlings compared with CHX-treated wild-type seedlings, but was not increased in DEX+CHX-treated p35S::GR-HDG1 seedlings compared with CHX-treated p35S::GR-HDG1 seedlings (supplementary material Table S4; DEX+CHX experiment). These data indicate that HDG1 is overexpressed, but that HDG1 does not regulate its own expression.

Statistical analysis identified 26 genes that were significantly differentially expressed at least twofold in DEX+CHX-treated p35S::GR-HDG1 seedlings compared with the controls (wild-type DEX+CHX-treated and p35S::GR-HDG1 CHX-treated; supplementary material Table S4; DEX+CHX experiment). The expression level of the GR-HDG1 fusion protein did not change after CHX treatment compared with untreated samples (supplementary material Fig. S8 and see supplementary methods), suggesting that the minimal change in gene expression is not due to a CHX-mediated reduction in HDG1. As the dataset did not provide clear links with the observed HDG1 overexpression phenotypes, we performed another microarray experiment in which p35S::GR-HDG1 and wild-type seedlings were treated for 8 h with only DEX. Using this approach, we identified 63 differentially expressed genes, including the HDG1 transgene. In contrast to the DEX+CHX experiments, most of the genes that were differentially expressed in response to GR-HDG1 activation were downregulated (79%; supplementary material Table S4; DEX experiment). The differential expression of a selection of these genes was validated by qPCR (supplementary material Fig. S9).

HDG1 targeted a diverse group of genes, including those involved in transport (e.g. ZIP1, SUC1, AAP4) and hormone biosynthesis, transport or signalling (e.g. GA3OX1, PIN5, ENP/MAB4, ARR16), as well as genes involved in biosynthesis and transport of methionine-derived aliphatic glucosinolates (MYB29, MAM1, CYP79F2, CYP83A1/REF1, IPMI1, IPMI2 and BAT5). Notably, HDG1 downregulated the expression of five positive regulators of meristem development/cell proliferation: CYCD3;1, CLE41, DAR2, RUL1 and AIL5/PLT5.

The BBM and HDG1 transcription factors interact, suggesting that they might regulate a common set of target genes. We compared the list of HDG1 target genes with direct BBM target genes that were obtained by ChIP-seq analysis of BBM binding sites in somatic embryos (supplementary material Table S5, Fig. S10). We observed BBM binding to 17 of the genes that showed differential expression after DEX activation of the GR-HDG1 fusion protein, including CLE41 and AIL5/PLT5 (Table 2; supplementary material Fig. S11). We selected five genes that showed promoter binding close to the translational start site by BBM and that had a reasonable gene expression change upon GR-HDG1 activation in the microarray experiment. qPCR analysis of gene expression changes after BBM-GR or GR-HDG1 activation showed that CLE41, RanBP2 and TRM13 were antagonistically regulated by BBM (up) and HDG1 (down), whereas AIL5 and ATC were downregulated by both BBM and HDG1 (Fig. 7). This suggests that HDG1 and BBM have common target genes that might be antagonistically regulated or co-regulated.

Our previous observation that BBM transcriptionally activates the HDG gene PDF2 and the epidermally expressed GASSHO1 (GSO1) [see supplementary material table S2 of Passarinho et al. (2008)] prompted us to examine whether BBM binds other HDG and L1-expressed genes. BBM binding was observed at HDG1, HDG5, HDG7, HDG8, HDG11, ANL2, PDF2, ATML1 and a set of epidermis-expressed genes, including GSO1, GSO2, CRINKLY4 (ACR4) and WEREWOLF (WER) (supplementary material Fig. S11). BBM binding was mostly observed in the promoters of these genes; however, in some cases introns were bound (supplementary material Fig. S12). Increased expression of HDG12, PDF2, GSO1 and GSO2 was observed when 5-day-old p35S::BBM-GR seedlings were treated with DEX and CHX, showing that these HDG and L1 genes are direct transcriptional targets of BBM (supplementary material Fig. S12).

The combined microarray and ChIP-seq data analysis suggest that BBM and HDG1 regulate a common set of target genes, but this regulation appears to be complex, as both coordinately and oppositely regulated transcription, as well as a transcriptional feedback loop between AIL and HDG genes, were observed. Additionally, our results uncovered a role for BBM in the transcriptional control of additional epidermal regulatory genes.

Members of the AIL family have been well described with respect to their positive roles in stem cell maintenance. Here, we have shown that the BBM AIL protein interacts with and regulates the expression of L1-expressed HDG proteins. Analysis of gain- and loss-of-function HDG phenotypes suggests that HDG proteins function antagonistically to AIL proteins to keep cell proliferation processes in check.

We observed that HDG1, HDG11 and HDG12 overexpression seedlings were smaller compared with wild type, and that they developed narrow leaves and accumulated anthocyanins. The most extreme overexpression phenotypes were observed in HDG1 overexpression lines, which showed root and shoot meristem arrest. Another notable feature of HDG1 overexpression lines was the increased formation of narrow, elongated cells on the leaf surface. In Arabidopsis leaves, elongated cells are found on the abaxial leaf surface, on the petiole and along the leaf margin. The elongated cells formed in HDG1 overexpression lines resemble margin cells and showed increased expression of YJ158, which marks large/giant cells in the sepal, leaf margin and abaxial surface (Eshed et al., 2004; Roeder et al., 2012). Our data therefore suggest that HDG1 is able to specify the identity of margin cells. Recently, two other HDG proteins, ATML1 and HDG11, were shown to regulate sepal giant cell formation prior to endoreduplication (Roeder et al., 2012). The HDG1-mediated increase in cell ploidy in leaves might be a secondary consequence of the increase in margin cells. Alternatively, the increased ploidy levels might also reflect the increased differentiation/reduced meristematic growth that characterizes HDG1 overexpression lines.

Whereas overexpression of HDG genes induces meristem arrest and epidermal margin cell formation, knockdown of HDG expression by co-suppression or amiRNAs induced ectopic cell proliferation, including formation of shoots and embryo-like tissue. The formation of ectopic shoots was also observed occasionally in pdf2;hdg3 or atml1;hdg3 mutants (Nakamura et al., 2006). In addition, post-embryonic expression of ATML1-SRDX induced callus-like protrusions on cotyledons and leaves (Takada, 2013). Our HDG knockdown phenotypes were observed in a small proportion of transgenic lines, but were more severe than previously reported HDG loss-of-function phenotypes. Downregulation of a larger number of HDG genes might have allowed us to overcome the high degree of functional redundancy within the HDG family (Nakamura et al., 2006). However, the low frequency of mutant phenotypes suggests that either HDG knockdown was inefficient or that it has negative impact on embryo viability (Abe et al., 2003; San-Bento et al., 2014).

Taken together, these data suggest that HDG genes, besides their roles in the differentiation of specific epidermal structures, also have a general role in repressing cell proliferation in the epidermis. This role does not appear to be restricted to HDG genes, as loss-of-function or knockdown mutants in other epidermal-expressed genes that control epidermal differentiation also show cell over-proliferation phenotypes (Jin et al., 2000; Becraft et al., 2002; Ahn et al., 2004).

We showed that HDG1 downregulates the expression of genes involved in cell proliferation, including the D-type cyclin CYCD3;1. CYCD3;1 overexpression leads to ectopic/increased cell divisions, reduced cell expansion and endoreduplication (Dewitte et al., 2003). Conversely, loss of CYCD3 genes reduces leaf cell numbers and SAM size and stimulates endoreduplication (Dewitte et al., 2007). HDG1 also inhibits the expression of CLE41, which encodes a B-type CLE signalling peptide. Overexpression of CLE41 promotes the formation of axillary buds (Yaginuma et al., 2011), and co-overexpression of CLE6 (A-type) and CLE41 peptides induced ectopic divisions in root, leaf and the hypocotyl vasculature (Whitford et al., 2008), indicating a role for CLE41 in cell proliferation. In addition, HDG1 represses the expression of REDUCED IN LATERAL GROWTH1 (RUL1), which encodes a receptor-like kinase that positively regulates cambium activity (Agusti et al., 2011), and of AIL5/PLT5, which controls lateral root primordia initiation in a redundant fashion with AIL6/PLT3 and PLT7 (Hofhuis et al., 2013) and can induce increased organ size or SE when overexpressed (Nole-Wilson et al., 2005; Tsuwamoto et al., 2010). HDG1 could also indirectly downregulate other AIL genes through DAR2, which was shown to act upstream of PLT1/PLT2 in the control of root meristem size (Peng et al., 2013).

We have shown that HDG and BBM/AIL proteins interact in vitro and in planta. The interaction between BBM and HDG proteins is limited mainly to embryo development, where the expression patterns of BBM, HDG1, HDG11 and HDG12 overlap extensively. However, as HDG proteins also interact with other AIL proteins, the expression patterns of the other AIL genes must be taken into account as well. For example, PLT2 is expressed in all epidermal cells of the root meristem (Galinha et al., 2007), and overlaps with HDG expression in these cells. Although BBM is not expressed in the SAM, other AILs are expressed here, e.g. PLT7 (Mudunkothge and Krizek, 2012), and could interact with HDG proteins in the L1/L2 layers.

Interestingly, our HDG1 overexpression phenotypes resemble ail loss-of-function phenotypes: plt1;plt2 mutant roots terminate soon after initiation, whereas plt1;plt2;ail6/plt3 mutants do not form any roots (Aida et al., 2004; Galinha et al., 2007), and ant;ail6/plt3;plt7 mutants produce only a few leaves before the SAM terminates (Mudunkothge and Krizek, 2012). In addition to root and shoot meristem differentiation, we also observed other differentiation phenotypes in HDG1 overexpression seedlings (ectopic formation of margin cells and higher ploidy levels) that were opposite to those observed in BBM overexpression seedlings (decreased cellular differentiation of cotyledons cells and reduced ploidy). Similarly, AIL6 overexpression lines lack sepal giant cells (Krizek and Eaddy, 2012). In line with this antagonistic HDG-AIL model, we found that downregulation of HDG expression by co-suppression or by an amiRNA leads to adventitious growth, similar to BBM or PLT5/AIL5 (AIL) overexpression (Boutilier et al., 2002; Tsuwamoto et al., 2010), and that co-overexpression of BBM and HDG1 reduces the overexpression phenotypes of both proteins. Taken together, our results suggest opposite roles for AIL and HDG genes, with AILs promoting meristem activity and HDGs stimulating meristem/cellular differentiation, and that these interactions are mediated at both transcriptional and protein-interaction level.

We have shown that AIL and HDG proteins interact in yeast and in planta and that they have antagonistic functions. A mechanism for interacting proteins to exert antagonistic functions is through competitive inhibition. In this scenario, interaction of HDG and AIL proteins would inhibit their respective abilities to act as transcriptional regulators, with the balance between the amount of free HDG or AIL determining the developmental outcome. In support of this ‘titration model’, we found that BBM and HDG1 can suppress the overexpression phenotypes of each other in a dose-dependent manner. In this model, AIL overexpression would lead to opposite regulation of HDG target gene expression and vice versa. We have identified overlapping BBM and HDG1 gene targets, some of which are oppositely regulated upon BBM and HDG1 overexpression. Notably, some of the common target genes also appear to be co-regulated, suggesting that in addition to antagonistic functions, BBM and HDG also have common functions. This raises the possibility that they perform these functions in the same protein complex. Finally, we observed transcriptional cross-regulation between AILs and HDGs, suggesting an additional level of interaction.

Our results suggest that HDG and AIL act in concert to control cell proliferation and differentiation processes. Whether AIL-HDG function is cell autonomous (Hacham et al., 2011; Knauer et al., 2013; Nobusawa et al., 2013) and how local BBM-HDG interactions regulate these processes at a molecular level are intriguing questions for further research.

The pBBM::BBM-YFP (Galinha et al., 2007), p35S::BBM (Boutilier et al., 2002), p35S::BBM-GR (Passarinho et al., 2008) lines, HDG T-DNA insertion alleles hdg1-2 (SALK_062171), hdg11-1 (SAIL_865 G09), hdg12-2 (SALK_127261) and anl2-t1 (SALK_000196) (Nakamura et al., 2006), and the giant cell marker line YJ158 (Eshed et al., 2004) have been previously described.

Plants were grown at 21°C (16/8 h light/dark regime) on rock wool plugs supplemented with 1 g/l Hyponex fertilizer or in Petri dishes on medium containing half-strength Murashige and Skoog salts and vitamins (MS medium, pH 5.8), containing 0.8% agar and 1% sucrose (0.5MS-10). The medium was supplemented with 10 µM DEX when appropriate. To obtain a fully-penetrant BBM-GR overexpression phenotype it is necessary to sterilize seeds with liquid bleach, rather than bleach vapour.

All used primers are shown in supplementary material Table S3.

For ectopic expression of HDG1, HDG11 and HDG12, the open reading frames were amplified from Arabidopsis Col-0 cDNA and cloned into the Gateway (GW) overexpression vector pGD625 (Immink et al., 2002).

For inducible activation of HDG1, HDG11 and HDG12 the HDG coding regions were fused in-frame to the ligand-binding domain of the rat glucocorticoid receptor (GR) coding region and then cloned into pGD625. For co-overexpression, a p35S::GR-HDG1 line was transformed with a construct overexpressing the genomic Arabidopsis BBM fragment (p35S::gAtBBM) in pB7WG2.0 (Karimi et al., 2002).

The BBM and the HDG translational eGFP reporter constructs were made by cloning Arabidopsis Col-0 genomic DNA into the GW-compatible pGreenII vector AM884381 (NASC) (Zhong et al., 2008). The promoters of BBM, HDG1, HDG11 and HDG12 comprised 4.2, 0.65, 2.7 and 1.2 kb upstream of the translational start site, respectively.

BBM and HDG cDNA entry clones were used to generate the BBM-CFP and YFP-HDG plasmids used for FRET-FLIM experiments, and cloned using recombination into GW-compatible sCFP3A and sYFP2 vectors (Karlova et al., 2011).

The HDG amiRNA construct was designed and generated according to WMD3 (Schwab et al., 2006) and cloned into pGD625.

For ChIP-seq experiments, a pBBM::NLS-GFP construct was generated in pGREEN using a 4.2 kb pBBM fragment. The p35S::BBM-GFP construct was using the BBM (At5g17430) Col-0 cDNA in pK7FWG2.0 (Karimi et al., 2002).

Arabidopsis Col-0 plants were transformed by the floral dip method (Clough and Bent, 1998) using Agrobacterium tumefaciens strain C58 carrying the pMP90 Ti plasmid.

Protoplasts were transfected as described (Horstman et al., 2014a) and incubated overnight. All plasmid combinations were tested in three independent transfections (26-30 cells per combination), except for YFP-AP1, which was tested twice (26 cells). FRET-FLIM measurements were performed as previously described (Russinova et al., 2004). Photons were collected by a Hamamatsu HPM-100-40 Hybrid detector (Becker & Hickl; resolution 120 ps). 64×64 pixel images were acquired (60-150 s, count rate ∼10⁴ photons per second) using the Becker and Hickl SPC 730 module. The data were analysed with SPCImage 3.10 software (Becker & Hickl) using a one- or two-component decay model for donor-alone and donor-plus-acceptor samples, respectively. The two-component analysis gives a slightly reduced fluorescence lifetime compared with the one-component analysis. Significant differences between samples were determined using a two-tailed Student's t-test.

β-glucuronidase (GUS) activity assays on seedlings were performed overnight at 37°C as described (Soriano et al., 2014), using 2.5 mM potassium ferri- and ferrocyanide. Seedlings were cleared with 70% ethanol prior to imaging.

Ploidy measurements were performed (Iribov, The Netherlands) on whole 10-day-old seedlings (three to four replicates), using one seedling per replicate.

GFP was visualized with a Leica SPE DM5500 upright confocal laser scanning microscope using the LAS AF 1.8.2 software. Roots were counterstained with 10 µg/ml propidium iodide (PI), and embryos and shoot/flower meristems with 10 µg/ml FM4-64. GFP, YFP, PI and FM4-64 were excited with a 488-nm solid-state laser and emissions were detected at bandwidths of 500-530, 510-560, 670-800 and 600-800 nm, respectively.

Cryo-SEM was performed as in Fatouros et al. (2012), except that samples were sputter-coated with 10 nm tungsten and the analysis was performed with SE detection at 2 kV and 6.3 pA. Digital images were contrast-adjusted with Photoshop CS5.

ChIP-seq experiments were carried out using a GFP antibody on (1) 14- to 17-day-old 2,4-dichlorophenoxyacetic acid (2,4-D)-induced somatic embryo cultures (Mordhorst et al., 1998) (4.8 g, ectopic shoots and callus removed) carrying a pBBM::BBM-YFP construct (Galinha et al., 2007) in the bbm-1 (SALK_097021; NASC) mutant background, and on (2) embryogenic seedlings (1.75 g) derived from a p35S::BBM-GFP line. The pBBM::BBM-YFP construct complemented the embryo lethal phenotype of the bbm;plt2 double mutant (not shown). 2,4-D somatic embryo cultures from a pBBM::NLS-GFP line and p35S::BBM seedlings served as negative controls for experiment (1) and (2), respectively. ChIP samples were prepared as described previously (Kaufmann et al., 2010; Smaczniak et al., 2012).

ChIP-seq libraries were sequenced on the Illumina HiSeq 2000 platform. Sequence reads that failed the CASAVA quality filter were eliminated. Sequence reads were mapped to the unmasked Arabidopsis genome (TAIR10; ftp://ftp.arabidopsis.org) using the SOAPaligner (v2) program (Li et al., 2009). A maximum of two mismatches and no gaps were allowed. Reads mapping in multiple genomic locations or to the chloroplast or mitochondrial genomes were discarded. ChIP-seq peaks were detected using CSAR (Muiño et al., 2011) with default parameter values except for ‘backg’, which was set to 2. Enrichment was calculated as the ratio of normalized extended reads between the pBBM::BBM-YFP or p35S::BBM-GFP samples versus their corresponding controls. False discovery rate (FDR) thresholds were estimated by permutation of reads between IP and control sample using CSAR. ChIP-seq results were visualized using Integrated Genome Browser 8.1.2 (Nicol et al., 2009). The ChIP-seq data is made available via NCBI (GEO accession: GSE52400).

For each sample, ∼40 five-day old seedlings were treated with 10 µM DEX and/or 10 µM CHX for 8 h. All conditions were replicated in triplicate. Microarray analysis was performed by NASC using Affymetrix Arabidopsis Gene 1.0 ST Arrays. Arrays were normalized using the RMA algorithm and differential expression was assessed with the LIMMA package and the Benjamini and Hochberg multiple testing correction (Benjamini and Hochberg, 1995; Carvalho and Irizarry, 2010). HDG1 targets were defined as showing a fold change >2 and FDR threshold of 0.05. The data are available via NCBI (GEO accession: GSE54312).

For analysis of (GR-)HDG1, BBM and target gene expression, seedlings were grown and treated as described in the text (DEX and CHX, both at 10 µM). DNAse-treated RNA was used for cDNA synthesis. qPCR was either performed using the SYBR green mix from Bio-Rad or the Biomark HD System (effect of BBM on HDG/L1 gene expression) on a 96.96 dynamic array chip according to the manufacturer's instructions (Fluidigm) and analysed using the manufacturer's software. In all experiments, the relative expression levels were calculated according to the 2^−ΔΔCT method (Livak and Schmittgen, 2001) using wild-type Col-0 as the calibrator and the SAND gene (Czechowski et al., 2005) as the reference. All primers are listed in supplementary material Table S3.

We thank Taku Takahashi for providing the hdg mutants, Renze Heidstra for the pBBM::BBM-YFP line, John Bowman for the YJ158 line, Jan-Willem Borst for advice concerning FRET-FLIM, Adriaan van Aelst and Tiny Franssen-Verheijen for performing the cryo-SEM analyses, and Iris Heidmann and Mieke Weemen for assistance.