A spatiotemporally regulated transcriptional complex underlies heteroblastic development of leaf hairs in Arabidopsis thaliana

Abstract

Heteroblasty refers to a phenomenon that a plant produces morphologically or functionally different lateral organs in an age‐dependent manner. In the model plant Arabidopsis thaliana, the production of trichomes (epidermal leaf hairs) on the abaxial (lower) side of leaves is a heteroblastic mark for the juvenile‐to‐adult transition. Here, we show that the heteroblastic development of abaxial trichomes is regulated by a spatiotemporally regulated complex comprising the leaf abaxial fate determinant (KAN1) and the developmental timer (miR172‐targeted AP2‐like proteins). We provide evidence that a short‐distance chromatin loop brings the downstream enhancer element into close association with the promoter elements of GL1, which encodes a MYB transcription factor essential for trichome initiation. During juvenile phase, the KAN1‐AP2 repressive complex binds to the downstream sequence of GL1 and represses its expression through chromatin looping. As plants age, the gradual reduction in AP2‐like protein levels leads to decreased amount of the KAN1‐AP2 complex, thereby licensing GL1 expression and the abaxial trichome initiation. Our results thus reveal a novel molecular mechanism by which a heteroblastic trait is governed by integrating age and leaf polarity cue in plants.

Introduction

Plant accumulates physiological and morphological changes with age (Bäurle & Dean, 2006). The gradual transitions in leaf morphology and size (allometry) that occur along the shoot of a plant are termed as heteroblasty (Goebel, 1900; Telfer et al, 1997; Tsukaya et al, 2000; Costa et al, 2012). Based on the heteroblastic traits, the vegetative phase of plant development can be divided into juvenile and adult phases. In the model plant Arabidopsis thaliana, the transition from juvenile to adult phase is marked by changes in the initiation of trichomes (leaf hairs) on the abaxial (lower) surface of the leaf, an increase in the length/width ratio of the leaf blade, and the production of serrations on the leaf margin (Poethig, 2003, 2013; Huijser & Schmid, 2011; Yu et al, 2015). Physiological studies reveal that the cues that guide these developmental transitions include the conserved and intrinsic mechanisms as well as extrinsic inputs, such as nutrients, light, and temperature. However, the underlying molecular mechanism is poorly understood.

Trichome initiation is mediated by a trimeric transcriptional activation complex containing GLABRA1 (GL1), GL3, and TRANSPARENT TESTA GLABRA1 (TTG1) (Schiefelbein, 2003; Guimil & Dunand, 2006; Ishida et al, 2008; Balkunde et al, 2010; Tominaga‐Wada et al, 2011; Grebe, 2012; Yang & Ye, 2013; Robinson & Roeder, 2015). Once a threshold level reaches, this activator complex activates expression of downstream regulators including the homeodomain transcription factor GLABRA2 (GL2) and promotes trichome cell fate and trichome differentiation (Rerie et al, 1994; Zhao et al, 2008; Pattanaik et al, 2014). Despite the adaxial (upper) side of all the leaves is covered with trichomes, the abaxial trichomes only appear after the 6^th leaf is produced under long‐day conditions (Fig 1A), indicating that the initiation of abaxial trichomes is controlled by both plant age and leaf polarity.

Figure 1. KAN1‐TOE1 repressive complex regulates abaxial trichome initiation.

• A
  Leaf trichomes on the abaxial (ab) and adaxial (ad) side of the wild‐type Arabidopsis rosette leaves. Note that the abaxial side of the 1^st leaf is glabrous. Scale bar indicates 1 mm.
• B
  Schematic drawing summarizing the miR156‐SPL‐miR172‐AP2 signaling cascade.
• C
  The abaxial trichome phenotype. The 1^st leaf with abaxial trichome was scored (n = 16). Error bars represent SEM. Two asterisks above the bars indicate significant differences from WT (Student's t‐test, P < 0.01)
• D
  TOE1 bound to KAN1 in yeast. Transformed yeasts were spotted on selective medium. AD, GAL4 activation domain. BD, GAL4 DNA‐binding domain.
• E
  BiLC analyses of KAN1‐TOE1 interactions in N. benthamiana leaves. KAN1 and TOE1 were translationally fused to HA‐LUCn or LUCc‐Myc, respectively. Red and white means the high signals and blue and dark means the low signals.
• F
  Physical interactions between TOE1 and KAN1 in planta. 6xMyc‐rTOE1 and KAN1‐3xHA were stably expressed in Arabidopsis thaliana. Immunoprecipitation was performed using anti‐HA antibody. 6xMyc‐rTOE1 fusion proteins were detected by Western blot using anti‐Myc antibody. IB: immunoblot, IP: immunoprecipitation.
• G, H
  The effect of pSPL9::KAN1 on leaf polarity and abaxial trichome initiation in the wild‐type (G) and toe1 toe2 (H) background. Scale bar indicates 1 mm.

Source data are available online for this figure.

In many plant species, the upper surface of the leaf is covered with a thicker cuticle and contains a densely packed layer of palisade mesophyll cells to optimize the capture of light, whereas the stomata and spongy mesophyll cells in the underside of the leaf function in gas exchange and transpiration (Chitwood et al, 2007; Bowman & Floyd, 2008). Genetic studies have demonstrated an important role of transcription factors, microRNA (miRNA), and trans‐acting short interfering RNAs (ta‐siRNAs) in specifying abaxial fates (Chitwood et al, 2007; Husbands et al, 2009; Braybrook & Kuhlemeier, 2010; Skopelitis et al, 2017). KANADI genes (KAN1‐KAN4), members of the GARP family of transcription factors, are the key regulators of abaxial identity in leaf morphogenesis. KAN1 is expressed abaxially. The mutation in KAN1 leads to adaxialized leaves, whereas overexpression of KAN1 causes abaxialized leaves and meristem arrest (Eshed et al, 2001; Kerstetter et al, 2001; Emery et al, 2003). Another abaxial fate determinant belongs to the YABBY (YAB) family, which consists of six members in A. thaliana. At least three YAB genes are expressed in the abaxial side within leaf primordia (Sawa et al, 1999; Siegfried et al, 1999). Expression analyses indicate that the YAB genes act partially downstream of the KANs and ectopic YAB gene expression is sufficient to specify abaxial fate in Arabidopsis (Eshed et al, 2001). AUXIN RESPONSE FACTOR 3 (ARF3) and ARF4, which are targeted by TAS3‐derived ta‐siRNAs, ta‐siR2141, and ta‐siR2142, act cooperatively with KANs to specify abaxial fate in Arabidopsis (Pekker et al, 2005; Adenot et al, 2006; Fahlgren et al, 2006; Garcia et al, 2006; Kelley et al, 2012). Although ARF4 transcripts are localized in the abaxial domain of within primordia, ARF3 is expressed more ubiquitously throughout the meristem and primordia.

Recent studies have revealed that the temporal cue guiding abaxial trichome initiation is controlled by a miRNA signaling cascade (Fig 1B). miR156, whose level is gradually decreased as plant ages, maintains juvenile traits by repression a group of transcription factors, named SQUAMOSA PROMOTER‐BINDING PROTEIN‐LIKEs (SPLs) (Chen et al, 2010; Wang & Wang, 2015; Yu et al, 2015). miR172 acts downstream of miR156 to promote adult epidermal identity. miR156 regulates the expression of miR172 via SPL9 which directly promotes the transcription of MIR172B (Aukerman & Sakai, 2003; Wu et al, 2009; Hyun et al, 2016). miR172 regulates abaxial trichome formation through its target genes, APETALA2 (AP2)‐like transcription factors including AP2, TARGET OF EAT1 (TOE1), TOE2, TOE3, SCHLAFMUTZE (SMZ), and SCHNARCHZAPFEN (SNZ) (Aukerman & Sakai, 2003; Chen, 2004; Mathieu et al, 2009; Yant et al, 2010; Zhang et al, 2015). The increased level of miR172 (p35S::MIR172) induces abaxial trichome in the first pair of leaves (Wu et al, 2009), whereas overexpression of target mimicry of miR172 (p35S::MIM172) (Todesco et al, 2010), which attenuates miR172 activity, delays the appearance of abaxial trichomes (Fig 1C). However, how miR172‐targeted AP2‐like transcription factors repress trichome initiation is largely unknown.

In this manuscript, we show that the production of abaxial leaf trichomes is determined by a positional cue executed by the leaf polarity regulator KAN1 and by a temporal cue executed by the miR172‐AP2 timing module. The integration of temporal and spatial cues is achieved by the formation of the KAN1‐TOE1 complex. Moreover, we found that the KAN1‐TOE1 complex binds to the downstream sequence of GL1 and represses its expression through enhancer‐facilitated chromatin loop.

Results

miR172‐targeted AP2‐like transcription factors repress abaxial trichome initiation

We scored the first leaf with abaxial trichomes in the single and high‐order mutants of miR172‐targeted AP2‐like gene family. Compared to wild type, toe1 and toe2 mutants developed early abaxial trichomes. Although the timing of abaxial trichomes in smz snz double mutant was normal, toe1 toe2 smz snz quadruple mutant produced abaxial trichomes on the first leaf (Fig EV1A and B). In addition, toe1 toe2 ap2 triple mutant exhibited enhanced early abaxial trichome phenotype than toe1 toe2 double mutant (Fig EV1B). Thus, these results indicate that these four AP2‐like transcription factors are functionally redundant in regulating trichome initiation.

Figure EV1. miR172‐targeted AP2‐like genes play redundant role in repressing abaxial trichome initiation.

1. Leaf trichomes on the abaxial (ab) and adaxial (ad) side of wild type and mutants. Note that the abaxial side of the 1^st leaf in toe1 toe2 smz snz quadruple mutant produces trichomes. Scale bar indicates 1 mm.
2. The abaxial trichome phenotype. The 1^st leaf with abaxial trichome was scored. Error bars represent SEM (n > 15).
3. Expression of AP2 and SMZ in shoot apices. Scale bar indicates 50 μm.
4. The expression pattern of TOE1‐sGFP reporter in young leaf primordium. Scale bar indicates 100 μm.
5. Complementation of toe1 toe2 by pTOE1::TOE1‐HA. Shown here is the flowering time of a representative T3 plant. Plants were grown in long days. The total number of leaves after bolting was scored. Error bars represent SEM (n = 17).

While p35S::MIM172 or toe1 toe2 smz snz quadruple mutant exhibited late abaxial trichome phenotype, the timing of adaxial trichome production was not affected (Fig EV1A). These results suggest two molecular mechanisms underpinning the repression of abaxial trichomes by miR172‐targeted AP2‐like transcription factors: First, the expression of AP2‐like genes is confined to the abaxial domain of leaf. As a result, the downregulation of AP2‐like genes with age promotes the production of abaxial trichomes. Alternatively, the repression of trichome initiation by AP2‐like genes is dependent on leaf polarity factor(s). The AP2‐like genes only repress trichome initiation at the abaxial side of leaf. To distinguish between these two possibilities, we analyzed the expression of TOE1, AP2, and SMZ in the wild‐type shoot apices. Using in situ hybridization assays, we found that AP2 and SMZ were expressed ubiquitously in the leaf primordium with the highest transcript levels in the vascular tissues (Fig EV1C). Because TOE1 mRNAs were not detectable for an unknown reason, we generated the transgenic plants expressing green fluorescent protein (GFP) reporter for TOE1 (pTOE1::sGFP‐N7). The TOE1 promoter was functional because the relevant genomic construct fully complemented the toe1 toe2 mutant phenotype (Fig EV1E). TOE1‐GFP reporter gave strong and even GFP signals in leaf primordium (Fig EV1D). Thus, these results, together with the previous findings that miR172 is expressed widely in the leaf primordium (Wollmann et al, 2010), led us to conclude that miR172‐targeted AP2‐like genes repress trichome initiation with the help of leaf polarity factor(s).

miR172‐targeted AP2 transcription factors bind to KAN1

To gain insights into how miR172‐targeted AP2‐like genes control abaxial trichome initiation, we performed a yeast two‐hybrid screen using Arabidopsis transcription factor library and TOE1 as a bait (Pruneda‐Paz et al, 2014). Interestingly, KAN1, one of the leaf abaxial fate determinants, was identified as one of the potential interactors (Fig 1D). Similar interaction was observed between KAN1 and AP2, TOE2, TOE3, or SNZ (Fig EV2A and B). The yeast assays with truncated proteins further revealed that both KAN1 and KAN2 bound with AP2‐like proteins through their N‐terminal domain and that there were no direction interactions between KAN3/4 and AP2‐like proteins (Fig EV2C–G). To confirm KAN1‐TOE1 interaction in planta, we performed a bimolecular luminescence complementation (BiLC) assay. KAN1 and TOE1 were translationally fused to the hemagglutinin (HA)‐tagged amino‐terminal half (HA‐LUCn) or to the Myc‐tagged carboxyl‐terminal half (LUCc‐Myc) of firefly LUCIFERASE (LUC), respectively. A strong luminescence signal was observed in Nicotiana benthamiana leaves where KAN1‐HA‐LUCn and TOE1‐Myc‐LUCc were coexpressed (Figs 1E and EV2H). Next, we analyzed whether KAN1 interacts with TOE1 in Arabidopsis. To this end, we generated pKAN1::KAN1‐3xHA plants in which 3xHA‐tagged KAN1 (KAN1‐3xHA) was expressed from its own promoter. The encoded KAN1‐3xHA fusion protein was functional because it was able to rescue the leaf polarity defect of kan1‐13. We crossed pKAN1::KAN1‐3xHA to 6xMyc‐tagged TOE1 overexpression line (p35S::6xMyc‐rTOE1, miR172‐resistant form of TOE1). When the protein extract was immunoprecipitated with anti‐HA antibody, 6xMyc‐TOE1 fusion proteins were readily detected (Fig 1F), indicating that TOE1 interacts with KAN1 in vivo.

Figure EV2. Interaction between KANs and miR172‐targeted AP2‐like proteins.

1. KAN1 was fused to GAL4 BD domain and AP2‐like proteins fused to GAL4 AD domain. Please note that KAN1 did not interact with SMZ.
2. Western blot analysis showing that SMZ‐AD was not properly expressed in yeast. TOE3‐AD and SMZ‐AD fusion proteins were probed with anti‐HA antibody.
3. Diagrams of KAN1 and KAN2 constructs for interaction studies. The green boxes indicate the Myb domain.
4. KAN1 bound to TOE1 through its N‐terminal domain and KAN2/3/4‐BD did not interact with TOE1 in yeast.
5. KAN2/3/4‐AD did not interact with TOE1 in yeast.
6. KAN2‐AD was not expressed in yeast. KAN2/3/4‐AD fusion proteins were probed with anti‐HA antibody. Please note that KAN3/4‐AD were expressed.
7. The N‐terminal domain of KAN2 bound to AP2‐like proteins.
8. Western blot analyses showing that LUCn and LUCc fusion proteins were properly expressed.

Source data are available online for this figure.

Genetic interaction between KAN1 and miR172‐AP2 timing module

In A. thaliana, there are four KAN1‐like proteins (KAN1‐KAN4), with KAN1 and KAN2 playing a predominant role (Eshed et al, 2001; Izhaki & Bowman, 2007; Reinhart et al, 2013; Huang et al, 2014). Compared to wild type (Columbia ecotype), kan1‐13 mutant exhibits leaf polarity defects (Eshed et al, 2001). The appearance of abaxial trichomes was accelerated in kan1‐13 mutants (Fig 1C). Conversely, expression of KAN1 under the control of a leaf primordia‐specific promoter (SPL9) results in the radialized leaf devoid of trichomes (Fig 1G) (Eshed et al, 2001; Wang et al, 2008), indicating that KAN1 acts as a repressor for trichome initiation. kan1 kan2 double mutant produces meristematic outgrowth on abaxial side of leaf (Eshed et al, 2001), which hinders us to investigate the functional redundancy of KAN genes on abaxial trichome initiation.

To understand the genetic interaction between miR172‐AP2 timing module and KAN1, we introduced kan1‐13 mutation into p35S::MIM172. p35S::MIM172 kan1‐13 plants developed abaxial trichomes earlier than p35S::MIM172 (Fig 1C), suggesting that the repression of abaxial trichome initiation by miR172‐targeted AP2s is dependent on KAN1 activity. It is unlikely that AP2s regulate trichome production through repressing KAN1 and KAN2 because the transcript level of both genes was not changed in either p35S::MIR172 or p35S::MIM172 plants (Appendix Fig S1). Moreover, pSPL9::KAN1 toe1 toe2 developed trichomes on radialized leaves (Fig 1H). Taken together, the physical interaction between KAN1 and miR172‐targeted AP2‐like transcription factors and genetic analyses indicate that the temporal repression of abaxial trichomes is mediated by the KAN1‐AP2 spatiotemporal complex.

Identification of GL1 as the downstream target of the KAN1‐TOE1 complex

To understand how KAN1‐TOE1 complex represses abaxial trichome initiation, we surveyed the published chromatin immunoprecipitation‐sequencing (ChIP‐seq) data for KAN1 (Merelo et al, 2013; Xie et al, 2015). Interestingly, GL1 was identified as a potential target for KAN1 (Fig 2A). It has been shown that the downstream enhancer element (+2,464 to 2,752 base pair, bp) is required for proper expression of GL1 and trichome initiation (Larkin et al, 1993). The potential binding site of KAN1 was located in the 3′ downstream of GL1, adjacent to enhancer (Fig EV3A).

Figure 2. GL1 is a direct target of KAN1.

1. ChIP‐seq reveals GL1 as a direct target of KAN1. The gray and green areas indicate the relative enrichment in wild type and p35S::FLAG‐GR‐KAN1, respectively. The coding region of GL1 is shown.
2. ChIP‐PCR analyses using the DEX‐treated p35S::FLAG‐GR‐KAN1 seedlings. The mock plants were used as controls. The chromatin extract was immunoprecipitated with anti‐FLAG antibody. Error bars represent SEM (n = 3). The enrichment in the control samples was set to 1.0.
3. Competitive EMSA showing binding of KAN1‐BD (DNA‐binding domain of KAN1) to DNA fragment dI. The relative amount (labeled oligonucleotide was set to 1.0) of unlabeled competitive oligonucleotide is indicated on the top. The position of the probe (dI) is shown in (Fig EV3A). The assay without KAN1‐BD protein (–) was used as a control.
4. ChIP‐PCR analyses using the p35S::6xMyc‐rTOE1 seedlings. The wild‐type (WT) plants were used as controls. The chromatin extract was immunoprecipitated with anti‐Myc antibody. Error bars represent SEM (n = 3). The enrichment in the control samples was set to 1.0.

Source data are available online for this figure.

Figure EV3. Validation of binding of KAN1 and TOE1 to REAT at the GL1 locus.

1. Diagrams of pGL1::GL1 T0 and pGL1::GL1 T6‐T7 constructs. The black, magenta, and blue boxes indicate three exons of GL1, REAT, and enhancer, respectively. The probes (dI, mI, mllI, and al) used for EMSA experiments. The position of REAT, M1, and M3 regions is shown in Fig 3A. The putative TOE1‐binding site is indicated by dashed line box. P stands for the GL1 promoter region for ChIP experiment (Fig 5B).
2. ChIP‐PCR analysis using 10‐day‐old wild type (WT) and pKAN1::KAN1‐3xHA seedlings. The chromatin extract was immunoprecipitated with anti‐HA. Error bars represent SD (n = 3). The enrichment in the control samples was set to 1.0.
3. EMSA showing that KAN1‐BD bound weakly to M1 (ml), but not to M3 (mllI). The REAT probe (dl) was used as positive control.
4. Competitive EMSA showing binding of TOE1 to DNA fragment dI. The relative amount (labeled oligonucleotide was set to 1.0) of unlabeled competitive oligonucleotide is indicated on the top. The position of the probe (dI) is shown in (A). The assay without TOE1 protein (‐) or probe al was used as a control. The position of al probe is shown in (A).
5. The abaxial trichome phenotype in different genotypes. T1 transgenic plants were screened with Basta. The 1^st leaf with abaxial trichome was scored. Error bars represent SEM (n = 20).

Source data are available online for this figure.

To confirm the ChIP‐seq result, we performed ChIP‐PCR analyses using the wild‐type and p35S::FLAG‐GR‐KAN1 plants, in which KAN1 fused with the rat glucocorticoid receptor (GR) and FLAG epitope was expressed from 35S promoter (Merelo et al, 2013). Four regions (a to d) near the enhancer of GL1 were analyzed (Fig EV3A). While there was no enrichment in the wild‐type samples, three fragments (b to d) which overlapped with the enrichment peak in ChIP‐seq data were readily amplified in the dexamethasone (DEX, the ligand for GR)‐treated p35S::FLAG‐GR‐KAN1 samples (Fig 2B). The direct interaction between KAN1 and fragment d was further confirmed by ChIP‐PCR analyses using pKAN1::KAN1‐3xHA plants and electrophoretic mobility shift assay (EMSA) (Figs 2C and EV3A–C). Because TOE1 and KAN1 form a complex, we asked whether TOE1 binds to the same region of GL1 as KAN1. EMSA assay demonstrated that TOE1 bound to dI (Fig EV3A and D). This result was further validated by ChIP‐PCR analysis with p35S::6xMyc‐rTOE1 plants (Fig 2D). Taken together, these results indicate that GL1 is a direct target of KAN1‐TOE1 complex.

The KAN1‐TOE1 complex represses GL1 expression through REAT

The finding that KAN1 binds to the downstream sequence of GL1 encourages us to test whether this regulatory region is responsible for abaxial trichome repression. To this end, we generated GL1 genomic constructs with or without the region between 3′ enhancer and stop codon. We named these two constructs as pGL1::GL1 T0 and pGL1::GL1 T1 (Fig 3A). pGL1::GL1 T0 was able to rescue the glabrous phenotype of gl1 mutant (Fig 3B). Interestingly, pGL1::GL1 T1 construct not only rescued the glabrous phenotype of gl1 but also caused the appearance of abaxial trichome at the first pair of leaves (Fig 3B and C). The removal of 3′ downstream sequence of GL1 did not interfere with the GL1 function because the density of adaxial trichomes was comparable between pGL1::GL1 T0 and pGL1::GL1 T1 (Fig 3D).

Figure 3. KAN1‐TOE1 complex represses GL1 through REAT.

• A
  Diagrams of pGL1::GL1 series constructs and generation of gl1 mutants by CRISPR/Cas9. The black and blue boxes indicate the last exon of GL1 and enhancer, respectively. The deleted regions in four gl1 mutants (M1‐M4) are shown. Red bases, insertions, or mutations.
• B
  Leaf trichomes on the abaxial (ab) and adaxial (ad) side of pGL1::GL1 T0 and pGL1::GL1 T1 plants. Note that the abaxial side of the 1^st leaf in pGL1::GL1 T1 produced trichomes (arrowhead). Scale bar indicates 1 mm.
• C
  The abaxial trichome phenotype in pGL1::GL1 T0 and pGL1::GL1 T1 plants. T1 transgenic plants were screened with Basta. Error bars represent SEM (n = 20).
• D
  The number of adaxial trichomes on the 1^st rosette leaf of pGL1::GL1 T0 and pGL1::GL1 T1 plants. Error bars represent SEM (n = 13).
• E, F
  The effect of pSPL9::KAN1 on leaf polarity and abaxial trichome production in pGL1::GL1 T0 (D) or pGL1::GL1 T1 (E) background. Scale bar indicates 1 mm.
• G
  TOE1 represses abaxial trichome production through the downstream sequence of GL1. The late abaxial trichome phenotype caused by TOE1 overexpression was suppressed in pGL1::GL1 T1 background. Error bars represent SEM (n = 20).
• H
  Identification of REAT. T1 transgenic plants were screened with Basta. Error bars represent SEM (n = 20).
• I
  The abaxial trichome phenotype. Error bars represent SEM (n = 10).

Next, we introduced pSPL9::KAN1 and p35S::6xMyc‐rTOE1 into GL1::GL1 T1 gl1. While pSPL9::KAN1 produced glabrous abaxialized leaf, pSPL9::KAN1 pGL1::GL1T1 gl1 gave rise to the abaxialized leaf with trichomes (Fig 3E and F). Similarly, the late abaxial trichome phenotype of p35S::6xMyc‐rTOE1 was suppressed by pGL1::GL1 T1 gl1 (Fig 3G). These results collectively indicate that KAN1‐TOE1 complex binds to the downstream sequence of GL1 and represses its expression and that the role of KAN1 in leaf polarity and trichome initiation can be separated.

Progressive truncations (each 200 bp) from the stop codon to enhancer further revealed that the region (+2,356 to 2,465 bp) was absolutely required for abaxial trichome repression (Figs 3A and H, and EV3A and 3E). We thus named this region as REAT (Repressive cis‐Element for Abaxial Trichome). Notably, REAT fell within the KAN1‐binding sequence revealed by ChIP‐seq and ChIP‐PCR analyses (Fig 3A). To provide direct in vivo evidence that REAT is the key regulatory sequence in regulating abaxial trichome initiation, we tried to knock out REAT region in the wild‐type genome using the clustered regularly interspaced short palindromic repeats (CRISPR)–CRISPR‐associated protein 9 (Cas9) system (Yin et al, 2017). Two guide RNAs (sgRNAs) flanking the REAT region were designed. Among 65 T1 plants, we identified four transgenic lines harboring the deletion in different length within the REAT region (Fig 3A). Of note, the phenotypic analyses of homozygous mutants in T2 generation revealed that the truncation of REAT led to early abaxial trichome phenotype (Fig 3I). The CRISPR/Cas9‐mediated REAT‐knockout plants never developed trichomes on the first leaf because the removal of REAT in these plants was not complete. In agreement of these findings, EMSA experiment revealed that KAN1 did not bind to the truncated REAT probe (M3) (Fig EV3C).

Chromatin looping at the GL1 locus is required for the repression of abaxial trichome initiation

Having confirmed that KAN1 binds to REAT and that REAT is required for GL1 repression in abaxial leaf domain, it remains unclear how the 3′ localized KAN1‐TOE1 complex represses GL1 expression. Recent evidence suggests that long‐range enhancers and gene promoter are in close proximity, which might reflect the formation of chromatin loops (Krivega & Dean, 2012; Liu et al, 2013; Ariel et al, 2014; Pombo & Dillon, 2015; Weber et al, 2016). To reveal whether the GL1 enhancer is required for the repression of abaxial trichome, we generated two genomic complementation constructs in which the enhancer region was completely removed or truncated (pGL1::GL1 T8 and T9, Fig 4A). Consistent with previous report (Larkin et al, 1993), pGL1::GL1 T9 was unable to rescue glabrous phenotype of gl1 mutant, whereas pGL1::GL1 T8 partially restored trichome initiation (Figs EV4A and 4B). Next, we examined whether a chromatin looping exists at the GL1 locus by capturing chromosome conformation (3C) experiment (Dekker et al, 2002). The nuclei of cross‐linked wild‐type seedlings were digested with restriction enzyme MfeI or SpeI/XbaI (Figs 4B and EV4C). Quantitative real‐time PCR (qRT–PCR) with two divergent pair of primers revealed that there was an interaction between the promoter region close to the GL1 transcription start site and the 3′ enhancer region (Figs 4B and C, and EV4D). This result was further verified by Sanger sequencing of the chimeric 3C‐PCR products (Fig EV4E and F). Moreover, 3C experiments revealed that the gene loop was evident in pGL1::GL1 T0, but greatly disrupted in pGL1::GL1 T8 and T9 (Figs 4D and EV4A and G), suggesting that enhancer mediates gene loop formation. We generated another construct with 8.0‐kb fragment insertion immediately after stop codon of GL1 (pGL1::GL1 T11, Fig 4A). If the GL1 promoter recruits distal enhancer by chromatin looping, the distance between enhancer and promoter is insignificant. Indeed, we found that the pGL1::GL1 T11 construct sufficiently rescued the gl1 phenotype (Fig 4E).

Figure 4. DNA looping is required for the repression of abaxial trichome.

1. Diagrams of pGL1::GL1 T0 and pGL1::GL1 T8‐T11 constructs. The black, magenta, and blue boxes indicate three exons of GL1, REAT, and enhancer, respectively.
2. Strategy for mapping juxtaposed regions in chromosome loops at the GL1 locus. Green and red lines indicate introns and 5′ or 3′ regulatory sequences, respectively. Gray bars depict MfeI sites. Green and red arrows depict the positions of the primers used for 3C‐PCR.
3. qRT–PCR analyses of 3C samples from the 4‐day‐old or 8‐day‐old wild‐type seedlings using the primers shown in (B). Unfixed seedlings were used as control. Two biological replicates were performed. Error bars represent SEM.
4. qRT–PCR analyses of 3C samples from pGL1::GL1 T0, T8, and T9 gl1 seedlings using the primers shown in (Fig EV4A). Two biological replicates were performed. Error bars represent SEM.
5. The abaxial trichome phenotype in different genotypes. T1 transgenic plants were screened with Basta. The 1^st leaf with abaxial trichome was scored. Error bars represent SEM (n = 20).

Figure EV4. Validation of chromatin loop at the GL1 locus.

• A
  Strategy for mapping juxtaposed regions in chromosome loops at the transgenic GL1 locus. Orange line and box indicate on nopaline synthase (NOS) terminator and coding sequence for Basta on the binary vector, respectively. Gray bars depict XhoI sites. J and K are two primers used for 3C‐PCR. Please note that K primer is specific for pGL1::GL1 T0, T8, or T9 transgene.
• B
  Phenotype of transgenic plants. Representative T1 plants were shown. Please note that pGL1::GL1 T0 fully rescued the gl1 phenotype, whereas pGL1::GL1T8 or T9 partially or hardly rescued gl1. Scale bar indicates 5 mm.
• C
  Strategy for mapping juxtaposed regions in chromosome loops at the GL1 locus. Green and red lines indicate introns and 5′ or 3′ regulatory sequences, respectively. Gray bars depict SpeI/XbaI sites. Green and red arrows depict the positions of the primers used for 3C‐PCR.
• D
  qRT–PCR analyses of 3C samples from the 10‐day‐old wild‐type seedlings using the primers shown in (A). Unfixed seedlings were used as control. Error bars represent SEM (n = 2).
• E–G
  Sequencing chromatogram showing 3C‐PCR products. The MfeI, SpeI/XbaI, and XhoI sites are underlined. The 5′ promoter and 3′ downstream sequence of GL1 are shown in green and red, respectively (C and D). The 5′ promoter region of GL1 and vector sequence are shown in black and orange, respectively (E).
• H
  Adaxial trichome phenotype of pGL1::GL1 T0 and T10. Please note that pGL1::GL1 T10 construct fully rescued the adaxial glabrous phenotype of gl1. Scale bar indicates 1 mm.
• I
  qRT–PCR analyses of 3C samples from the seedlings of different genotypes using the primers shown in (Fig 4B). Unfixed seedlings were used for data normalization. Error bars represent SEM (n = 2).

To examine whether the enhancer is required for the repression of GL1 by KAN1‐TOE1 complex, we generated the construct in which the enhancer was moved to the 5′ end of the GL1 promoter, pGL1::GL1 T10 (Fig 4A). Interestingly, pGL1::GL1 T10 not only restored adaxial trichome production of gl1 mutant (Fig EV4H), but also produced the abaxial trichomes on the first leaf (Fig 4E), indicating that REAT was no longer functional without closely linked enhancer. Thus, these results demonstrate that the repression of GL1 by KAN1‐TOE1 complex is dependent on enhancer‐mediated chromatin looping.

KAN1‐TOE1 complex causes transcriptional repression of GL1 probably through histone deacetylation

It has been shown that TOE1 represses gene expression by recruiting TOPLESS (TPL), which encodes a general transcriptional corepressor interacting with transcription factor through EAR motif (Long et al, 2006; Kagale & Rozwadowski, 2011; Causier et al, 2012; Krogan et al, 2012). tpl‐1 mutant was identified in Landsberg erecta (Ler) accession background (Long et al, 2002). Wild‐type Ler plants started producing abaxial trichomes from the 3^rd or 4^th leaf. In contrast, tpl‐1 mutant developed the abaxial trichomes on the first pair of leaves (Fig 5A). In addition, the late trichome phenotype in p35S::MIM172 plants was suppressed in the tpl‐1 mutant background (Fig 5A). Therefore, these results suggest that KAN1‐TOE1 complex inhibits abaxial trichome initiation with the help of TPL.

Figure 5. KAN1‐TOE1 complex may trigger histone deacetylation at the GL1 locus.

1. The abaxial trichome phenotype in different genotypes. The 1^st leaf with abaxial trichomes was scored. Error bars represent SEM (n = 20). Two asterisks above the bars indicate significant differences from toe1 or toe1 toe2 mutant (Student's t‐test, P < 0.01).
2. ChIP analyses using 8‐day‐old (juvenile) or 18‐day‐old (adult) pMIR165A::NTF plants. The nuclei were insolated by INTACT method. The chromatin extract was immunoprecipitated with anti‐H3K9Ac or anti‐H3K14Ac antibody. Error bars represent SD (n = 3). The enrichment in the 8‐day‐old seedlings sample was set to 1.0.
3. Proposed model for the initiation of abaxial trichome by an age‐regulated repressive complex. The question marker indicates a putative protein facilitating the formation of a chromatin looping at the GL1 locus. In juvenile phase at the abaxial side of the leaf, the KAN1‐TOE1 complex represses GL1 expression through chromatin looping. The complex no longer exists in adult phase.

TPL represses gene expression by recruiting histone deacetylase including HISTONE DEACETYLASE 6 (HDA6) and HDA19 (Wang et al, 2013; Pi et al, 2015). Although the timing of abaxial trichome production in hda19 was normal, hda19 toe1 and hda19 toe1 toe2 plants developed earlier abaxial trichomes than toe1 and toe1 toe2, respectively (Fig 5A; Appendix Fig S2). The weak phenotype of hda19 is probably due to the functional redundancy with HDA6. However, the growth arrest phenotype of hda6 hda19 hinders us to test this hypothesis (Tanaka et al, 2008). Nevertheless, the enhancement of the early abaxial trichome phenotype of toe1 and toe1 toe2 mutant by hda19 suggests that TPL‐HDA19‐mediated histone deacetylation contributes to the heteroblastic development of abaxial trichomes.

To investigate the histone acetylation level at the GL1 locus in abaxial side of the leaf, we deployed the INTACT (isolation of nuclei tagged in specific cell types) method (Deal & Henikoff, 2010), which allows us to isolate the nuclei at different domains in a leaf. To this end, we generated an INTACT line, in which the coding sequence of the nuclear targeting fusion (NTF) protein was driven by the abaxial domain‐specific MIR165A promoter (Appendix Fig S3) (Benkovics & Timmermans, 2014; Skopelitis et al, 2017). We purified the nuclei of the young leaves from juvenile plants (8‐day‐old) and adult plants (18‐day‐old) and performed ChIP‐PCR analyses using anti‐H3K9Ac and anti‐H3K14Ac antibodies. Both H3K9Ac and H3K14Ac are epigenetic markers for gene activation and have been reported to be targeted by HDA19 in plants (Jang et al, 2011; Krogan et al, 2012). The accumulation of these markers at the promoter of GL1 was significantly increased as plant aged (Figs EV3A and 5B). 3C experiments further revealed that the GL1 gene loop was not affected in tpl‐1, kan1‐13, M4, p35S::6xMyc‐rTOE1, toe1 toe2 smz snz, and hda19 toe1 toe2 mutants (Fig EV4I), indicating that the change in histone acetylation status of the GL1 locus does not perturb gene loop formation. Taken together, the above results suggest that KAN1‐TOE1 complex represses GL1 expression in the abaxial side of leaf probably through TPL‐mediated histone deacetylation in the juvenile phase.

Discussion

Our results suggest a model in which heteroblastic development of leaf trichomes is determined by integration of positional and temporal cues. In the positional axis, leaf polarity regulator KAN1 is exclusively expressed in the abaxial side of leaf. In the temporal axis, the gradual increase in miR172 leads to the reduction in AP2‐like transcription factors with age (Fig 5C). The integration of temporal and spatial axes is achieved by the formation of KAN1‐AP2 repressive complex. In the juvenile phase, KAN1‐AP2 complex binds to the 3′ downstream of GL1 and inhibits its expression at abaxial side of leaf, whereas such repression is relieved in the adult phase when the AP2 proteins no longer exist.

The heteroblastic traits usually vary between and within species (Koenig & Weigel, 2015). One of the best examples comes from a recent study on natural variation in leaf shape of Cardamine hirsuta, a close relative of Arabidopsis. By quantitative trait locus analyses, Tsiantis and his colleagues identify cis‐regulatory variation in the flowering time repressor FLOWERING LOCUS C (FLC) as the key determinator for heterochronic variation in leaf shape. The C. hirsuta genotypes with low‐expressing FLC alleles show both early flowering and accelerated age‐dependent changes in leaf form (Cartolano et al, 2015). In Arabidopsis, despite the recent progress in understanding the natural variation for adaxial trichome density (Mauricio, 2005; Symonds et al, 2005, 2011; Hilscher et al, 2009; Bloomer et al, 2012, 2014), how the timing of abaxial trichome production is diversified among different accessions is poorly understood. Moreover, the interspecific variations in trichome density and patterning in crucifers are not yet explored. The accomplishment of the genome sequences of diverse crucifer species such as C. hirsuta, A. lyrata, Capsella rubella, and Arabis alpina may pave the way for us to dissect the genetic and molecular basis for the variations in heteroblastic traits at interspecific scale (Hu et al, 2011; Slotte et al, 2013; Willing et al, 2015; Gan et al, 2016).

The existence of multiple upstream regulators of trichome initiation may underlie the heterochronic variation in trichome patterning within the same species. For example, the production of abaxial trichomes is controlled not only by age but also by exogenous hormones including gibberellin (Matias‐Hernandez et al, 2016). In Arabidopsis, the gibberellin‐deficient mutant delays the appearance of abaxial trichome whereas gibberellin treatment accelerates (Telfer et al, 1997). Although it is well known that gibberellin functions through the degradation of DELLA proteins (Daviere & Achard, 2013; Xu et al, 2014), it remains unclear how DELLA represses trichome initiation. One plausible mechanism is that gibberellin induces trichome initiation through miR172. However, a recent study reveals that a rice mutant defective in gibberellin biosynthesis exhibits long juvenile phase but the expression level of miR172 is similar between wild type and mutant, suggesting that DELLA does not promote adult traits through modulating miR172 abundance (Tanaka, 2012). Because of lack of a canonical DNA‐binding domain, DELLA regulates different aspects of plant development by interacting with diverse transcription factors, such as PHYTOCHROME INTERACTING FACTORs (PIFs), MYC2, and JASMONATE ZIM‐DOMAIN (JAZ) (Daviere & Achard, 2013). Therefore, the second possibility is that DELLA functions through binding with SPL or AP2‐like proteins. Indeed, it has been shown that DELLA can bind to SPL9 and SPL15 and inhibits their transcriptional activities (Yu et al, 2012; Hyun et al, 2016). The third possible molecular mechanism is that gibberellin promotes abaxial trichome initiation through gibberellin‐responsive C2H2 zinc finger proteins including GLABROUS INFLORESCENCE STEMS (GIS), GIS2, ZINC FINGER PROTEIN 5 (ZFP5), ZFP6, and ZFP8 (Gan et al, 2006; Zhou et al, 2013; Sun et al, 2014). However, although gis and gis3 mutant exhibits glabrous inflorescence stem phenotype, it is currently unknown whether these mutants delay abaxial trichomes production.

Another interesting phenomenon is that abaxial trichome production is promoted by long days, but delayed in short days in Arabidopsis (Telfer et al, 1997). The analyses of late flowering mutants further suggest that the formation of abaxial trichome initiation is correlated with flowering time, which is consistent with the notion that juvenile‐to‐adult phase change contributes to the acquirement of reproductive competence. Hence, the dissection how photoperiod regulates abaxial trichome initiation will be an important future direction.

Finally, the molecular mechanism by which the chromatin loop at the GL1 locus is formed is still unknown. The study of Arabidopsis FLOWERING LOCUS T (FT) locus demonstrates that the NUCLEAR FACTOR Y (NF‐Y) complex that binds to distal CCAAT boxes and CONSTANS (CO) that interacts with proximal CORE elements facilitates the formation of a gene loop required for FT activation (Cao et al, 2014). Intriguingly, the transcription of TERMINAL FLOWER 1 (TFL1), a functional antagonist for FT, also requires chromatin looping in which its promoter at the 3′ site is brought close to the transcription start site (Liu et al, 2013). The high‐order chromatin structure also exists at the FLC locus. 3C experiment reveals the presence of a gene loop at FLC, involving a physical interaction between the flanking 5′ and 3′ regions. Interestingly, this loop is significantly disrupted upon cold exposure during vernalization, suggesting a potential role in early stages of transcriptional shutdown, prior to the switch to a silenced chromatin state (Crevillen et al, 2013). The formation of FLC gene loop may be regulated by ATP‐dependent chromatin remodeling. BAF60, a SWI/SNF subunit, accumulates in the nucleus and promotes looping by modulation of histone density, composition, and post‐translational modification (Jegu et al, 2014). Very recently, an FLC promoter‐derived long noncoding RNA (lncRNA), named COLDWRAP, that is induced by vernalization and functions with another lncRNA COLDAIR to retain Polycomb at the FLC promoter through the formation of a repressive intragenic chromatin loop (Kim & Sung, 2017). Therefore, dissection of the proteins or lncRNA involved in enhancer looping at GL1 locus is another interesting research topic in the future.

Materials and Methods

Plant materials and growth conditions

Arabidopsis thaliana (ecotype Col‐0 and Ler) and N. benthamiana were grown at 21°C (day)/19°C (night) in long days (16‐h light/8‐h dark). ap2‐12 (Col‐0 ecotype), kan1‐13 (Col‐0 ecotype), smz snz (Col‐0 ecotype), toe1 toe2 (Col‐0 ecotype), smz snz toe1 toe2 (Col‐0 ecotype), tpl‐1 (Ler ecotype), p35S::MIM172 (Col‐0 ecotype), and p35S::MIR172 (Col‐0 ecotype) were described (Schwab et al, 2005; Long et al, 2006; Mathieu et al, 2009; Todesco et al, 2010; Yant et al, 2010).

For transgenic A. thaliana plants, the binary constructs were delivered into Agrobacterium tumefaciens strain GV3101 (pMP90) by freeze–thaw method. Transgenic plants were generated by floral dipping method (Clough & Bent, 1998) and screened with 0.05% glufosinate (Basta) on soil, 40 μg/ml hygromycin or 50 μg/ml kanamycin on half‐strength Murashige & Skoog (MS) plate.

Constructs

The oligonucleotide primers for all the constructs are given in Appendix Table S1. The map and DNA sequence for each construct are available upon request.

For yeast two‐hybrid constructs, the cDNAs of AP2, TOE1, TOE2, TOE3, SMZ, SNZ, KAN1, KAN2, KAN3, KAN4, and TPL were amplified and cloned into pGBKT7 or pGADT7 vectors (Clontech).

To generate pGL1::GL1 series vector (pGL1::GL1 T0‐T11), the downstream fragments of GL1 were PCR amplified and inserted into the binary vector JW1307 which harbors the 2.9‐kb genomic fragment of GL1. pGL1::GL1 T11 was generated by insertion of an 8‐kb irrelevant DNA into pGL1::GL1 T0. To generate pSPL9::KAN1, the KAN1 cDNA fragment was inserted into the binary vector JW815 which harbors 3.5‐kb upstream regulatory fragment of SPL9.

To generate pMIR165A::NTF, a 2.2‐kb upstream fragment of MIR165A (At1g01183) was introduced into the binary vector containing the NTF fragment (Deal & Henikoff, 2010, 2011).

BiLC constructs were generated as described (Gou et al, 2011). The cDNA fragments of KAN1 and TPL were amplified and cloned into the vector LW662 behind LUCc‐Myc under the control of 35S promoter. TOE1 coding region was cloned into the vector LW631 in front of HA‐LUCn under the control of 35S promoter.

For CoIP constructs, KAN1 and TOE1 were cloned into the binary constructs with 3xHA (JW819 for 3xHA carboxyl‐terminal fusion) or 6xMyc (JW1016 for 6xMyc amino‐terminal fusion) tag behind KAN1 or 35S promoter.

Genome editing by CRISPR/Cas9

We used egg cell‐specific promoter‐controlled CRISPR/Cas9 to generate gl1 mutant (Wang et al, 2015). Two sgRNAs were designed and cloned into pHEE401E as described.

Phenotypic analyses

The number of abaxial and adaxial trichomes on fully expanded leaves was examined and scored under stereomicroscope. The 1^st leaf with abaxial trichome was labeled.

Expression analyses

Total RNA was extracted with TRIzol reagent (Invitrogen, 15596018). 1 μg of total RNA was DNase I‐treated and used for cDNA synthesis with oligo (dT) primer (Fermentas, S0132). The average expression levels were calculated from 2^−ΔΔCt values. Biological triplicates with technical triplicates were performed. The qRT–PCR primers for TUBULIN (TUB) have been described (Wang et al, 2009). The oligonucleotide primers for all the genes are given in Appendix Table S1.

In situ hybridization

RNA in situ hybridization was performed as described (Wang et al, 2009; Lian et al, 2013). cDNA fragments of AP2 and SMZ were amplified and cloned into T‐vector, respectively. In vitro transcription was performed with T3 or T7 RNA polymerase (Roche, 11031163001) in which linearized vectors were added as templates. All primers used for preparing probes are listed in Appendix Table S1.

Microscopy

For confocal imaging, the tissues were fixed in ice‐cold PBS containing 2.5% paraformaldehyde (PFA), pH 7.0, washed with sucrose gradient PBS‐PFA solution, embedded with 6% low melting gel, and sectioned into thin slices with a Leica Sliding Microtome SM‐3000 at the thickness 110 μm. Prepared specimen was observed and scanned with Olympus FV1000 confocal microscope. Proper filter sets and lasers were selected for fluorescence signal scanning. For GFP, excitation light wavelength was 488 nm, emission, 510–550 nm.

Yeast two‐hybrid assay

Plasmids were transformed into yeast strain AH109 (Clontech) by LiCl‐PEG method. The transformants were selected on SD‐Leu‐Trp plates. The interactions were tested on SD‐Leu‐Trp‐His (SD‐LWH) or SD‐Ade‐Leu‐Trp‐His (SD‐ALWH) plate with 3‐amino‐1,2,4,‐triazole (3‐AT). At least ten individual clones were analyzed. The screen of Arabidopsis transcription factor library with TOE1 as bait was performed according to the published protocol (Pruneda‐Paz et al, 2014). Yeast soluble proteins were extracted with the buffer (50 mM HEPES, 10 mM EDTA, 50 mM NaCl, 10% glycerol, 1% PVPP, 2 mM DTT, 1 mM PMSF, 10 μM MG‐132, 1× protease inhibitor cocktail, pH 7.5) and subjected to Western blot analyses with anti‐HA (Roche, 12013819001).

CoIP and immunoblot analyses

7‐day‐old wild‐type, p35S::6xMyc‐rTOE1, and pKAN1::KAN1‐3xHA p35S::6xMyc‐rTOE1 A. thaliana seedlings were used for CoIP analyses. The soluble proteins were extracted in extraction buffer (50 mM HEPES, 10 mM EDTA, 50 mM NaCl, 10% glycerol, 1% PVPP, 2 mM DTT, 1 mM PMSF, 10 μM MG‐132, 1× protease inhibitor cocktail, pH 7.5). The immunoprecipitation in Arabidopsis was performed with anti‐HA antibody (Santa Cruz, sc‐7392) and Dynabeads (Thermo Fisher Scientific, 1003D) at 4°C. The beads were washed three times with wash buffer (50 mM HEPES, 150 mM NaCl, 10 mM EDTA, 0.1% Triton X‐100, 10% glycerol, 1 mM PMSF, pH 7.5). 3xHA‐ or 6xMyc‐fusion proteins were detected by immunoblot with anti‐HA‐peroxidase (HRP) (Roche, 12013819001) or anti‐Myc (Millipore, 05‐724) antibody.

EMSA assay

To construct plasmids for the expression of TOE1 and the DNA‐binding domain of KAN1 (KAN1‐BD) in E. coli, the coding sequence of TOE1 and KAN1 was amplified and cloned into pRSF‐Duet. Oligonucleotide probes were synthesized and labeled with biotin at the 5′ end. EMSA was performed using a LightShift Chemiluminescent EMSA Kit (Thermo Fisher Scientific, 20148X). Briefly, biotin‐labeled probes were incubated in 1× binding buffer, 2.5% glycerol, 5 mM MgCl₂, and 50 ng/μl Poly (dI dC) with or without proteins at room temperature for 20 min. For unlabeled probe competition, unlabeled probes were added to the reactions. The probe sequences are listed in Appendix Table S1.

BiLC analysis

BiLC assay was performed as described (Gou et al, 2011). A. tumefaciens was resuspended in infiltration buffer (10 mM methyl ester sulfonate, 10 mM MgCl₂, 150 μM acetosyringone, pH 5.7) at OD₆₀₀ = 0.8. p35S::P19‐HA (Papp et al, 2003) was co‐infiltrated to inhibit gene silencing.

ChIP analysis

Briefly, the 7‐day‐old wild‐type, p35S::FLAG‐GR‐KAN1, pKAN1::KAN1‐3xHA, and p35S::6xMyc‐rTOE1 seedlings were fixed according to a published protocol (Yu et al, 2013). p35S::FLAG‐GR‐KAN1 seedlings were treated with 10 μM DEX for 24 h before harvest. The chromatin extract was immunoprecipitated with anti‐FLAG beads (Sigma, F2426), anti‐Myc beads (Sigma, E6654), and anti‐HA beads (E6779). ChIP DNAs were reverse cross‐linked and purified with PCR purification kit (Qiagen, 28206). 1 μl DNA was used for qRT–PCR analyses. The relative enrichment of FLAG‐GR‐KAN1, KAN1‐3xHA, and 6xMyc‐rTOE1 on the GL1 locus was calculated by normalizing the amount of each immunoprecipitated fragment to input DNA and then by normalizing the value for transgenic plants against the value for wild type as a negative control. The ChIP assays with pMIR165A::NTF INTACT line were performed using anti‐H3K9Ac (Millipore, 07‐352) and anti‐H3K14Ac (Millipore, 07‐353) antibody.

3C assay

Three grams of the wild‐type, mutant, or transgenic seedlings were used for 3C experiments. The samples were fixed in 50 ml ice‐cold isolation buffer supplemented with 4% formaldehyde and ground with liquid nitrogen (Louwers et al, 2009). The slurry was resuspended with ice‐cold nuclei isolation buffer, incubated for 10 min at 4°C. The following steps were carried out generally according to the 3C protocol established for maize tissues (Louwers et al, 2009), except that 100 U of MfeI, SpeI/XbaI, or XhoI (New England Biolabs, R3589, R0145/R3133, R0146) and 50 U of T4 DNA ligase (New England Biolabs, M0202) were used for genomic DNA digestion and ligation, respectively. The genomic fragment (−2,186 to −1,962 bp) of GL1 devoid of the MfeI, SpeI/XbaI, or XhoI sites was amplified for normalization. The samples without fixation were used as controls. All primer sequences used for 3C assay are listed in Appendix Table S1.

INTACT

In vivo tagged nuclei were purified from approximately 1 g of pMIR165A::NTF plants using a modified version of the INTACT procedure described previously (Deal & Henikoff, 2011). The shoot apices with small developing leaves (< 0.5 cm in length) were harvested from 8‐day‐old or 18‐day‐old plants. The samples were fixed in NPBf (20 mM MOPS, 40 mM NaCl, 90 mM KCl, 2 mM EDTA, 0.5 mM EGTA, 1% formaldehyde, pH 7.0) and ground in liquid nitrogen. All subsequent stages were performed at 4°C. The resulting fine powder was resuspended in 20 ml nuclei purification buffer (20 mM MOPS, 40 mM NaCl, 90 mM KCl, 2 mM EDTA, 0.5 mM EGTA, 0.5 mM spermidine, 0.2 mM spermine, pH 7.0) supplemented with Complete protease inhibitors (Roche, 04693132001). The obtained lysate was filtered through Miracloth twice and centrifuged for 10 min at 1,000 g to pellet nuclei. Nuclei were resuspended in 1 ml NPB with 4 μg/ml DAPI, incubated on ice for 3 min, centrifuged for 5 min at 1,000 g, and resuspended in 1 ml NPB. After estimating the number of nuclei using epifluorescence microscopy, nuclei were incubated for 30 min under gentle rotation with 25 μl Streptavidin M280 Dynabeads (Thermo Fisher Scientific, 11205D). Bead‐bound nuclei were selectively recovered using a DynaMag 15 magnet (Thermo Fisher Scientific, 12301D) by carefully removing the supernatant. Bead‐bound nuclei were gently resuspended in 14 ml NPB, 0.5% BSA, and 0.1% Triton X‐100 and magnetically re‐captured. This wash step was repeated twice. Bead‐bound nuclei were resuspended in 1 ml NPB, and their purity and number were measured by epifluorescence microscopy.

Author contributions

LW and J‐WW designed experiments. LW performed most of the experiments. C‐MZ and Y‐XM performed yeast two‐hybrid, BiLC, EMSA, and immunoblot experiments. G‐DS analyzed the ChIP‐seq data. L‐ZL, JG, HL, LH, T‐QZ, H‐BT, HR, F‐XW, L‐YW, X‐LL, C‐SW, E‐WC, and X‐NZ generated transgenic plants. CL contributed to the chromatin looping experiments. LW, C‐MZ, and J‐WW analyzed data. J‐WW wrote the manuscript. All authors commented on the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

Source Data for Figure 1

Source Data for Figure 2

The EMBO Journal (2019) 38: e100063