LEUNIG and SEUSS co-repressors regulate miR172 expression in Arabidopsis flowers

The ABCE model of flower development (Coen and Meyerowitz, 1991; Krizek and Fletcher, 2005; Theissen and Saedler, 2001) successfully explains how floral organ identity is established through the combinatorial action of A, B, C and E classes of genes, all of which, except APETALA 2 (AP2), encode MADS box transcription factors. Central to the model is the antagonistic interaction between the A and C class genes. Whereas class A activity is repressed by class C genes in the inner two whorls, class C activity is repressed by class A genes in the outer two whorls (Bowman et al., 1991; Coen and Meyerowitz, 1991; Drews et al., 1991).

The class C gene AGAMOUS (AG) encodes a transcriptional repressor of the class A gene APETALA 1 (AP1) in the inner two whorls (Gustafson-Brown et al., 1994; Mandel et al., 1992). Two interacting transcriptional co-repressors, LEUNIG (LUG) and SEUSS (SEU), have been shown to repress AG transcription in the outer two whorls (Conner and Liu, 2000; Franks et al., 2002; Liu and Meyerowitz, 1995; Sridhar et al., 2004). Neither LUG nor SEU contains any known DNA-binding domains; their binding to the AG cis-regulatory elements is dependent on a direct physical interaction between SEU and the two MADS box proteins AP1 and SEPALLATA 3 (SEP3) (Gregis et al., 2006; Gregis et al., 2009; Sridhar et al., 2006).

AP2 is another transcriptional repressor of AG (Bowman et al., 1991; Drews et al., 1991). Using chromatin immunoprecipitation, AP2 was shown to directly bind within the AG second intron (Yant et al., 2010). Ectopic AP2 activity via the pAP3:AP2r transgene resulted in the repression of AG mRNA, supporting a direct repression of AG by AP2 (Wollmann et al., 2010). It is not yet known whether the repressive activity of AP2 depends on its recruitment of transcriptional co-repressors such as LUG/SEU.

The outer whorl-specific activity of AP2 has been largely attributed to a microRNA (miRNA), miR172, that specifically blocks AP2 mRNA translation as well as causing AP2 mRNA cleavage in the inner two whorls (Aukerman and Sakai, 2003; Chen, 2004; Schwab et al., 2005). In situ hybridization has revealed that miR172 is expressed most strongly in the inner whorls (Chen, 2004). A recent re-examination by in situ hybridization revealed that the spatial expression patterns of AP2 mRNA and miR172 are largely complementary, although there is a transient overlap in the second and possibly third floral whorls (Wollmann et al., 2010). Nevertheless, what determines the inner whorl-specific expression of miR172 remains unknown. A genome-wide analysis of AP2 binding sites using ChIP-seq indicated that miR172b is a target of AP2 (Yant et al., 2010), suggesting a possible role of AP2 in miR172 regulation.

Here, we show that LUG directly and negatively regulates miR172c and miR172e expression in sepals. This direct repression of miR172 by LUG also requires SEU and AP2, suggesting that AP2 might recruit the SEU-LUG co-repressor complex to repress miR172 transcription in the outer floral whorls. Our study provides important insights into miRNA regulation and suggests a positive-feedback loop, in which AP2 maintains its own outer whorl-specific activity by repressing the expression of its cognate miRNA.

lug-3, seu-1 and ap2-2 mutants and 35S:GFP-LUG; lug-16, ProSEU:GFP-SEU; seu-1, ProSEU:GFP-SEU; ap2-2 and 35S:AP2m3 transgenic lines, all in the Landsberg erecta (Ler) background, were grown at 20°C under long-day conditions (16 hours light/8 hours dark). 35S:GFP-LUG in pAVA393 (Conner and Liu, 2000) was excised with HindIII and SacI and cloned into pCAMBIA2300 (Cambia, Brisbane, Australia), which was used to transform and rescue lug-16. ProSEU:GFP-SEU, previously shown to rescue seu-1 (Azhakanandam et al., 2008), was crossed into ap2-2+/+ant-9 to yield ProSEU:GFP-SEU (WT) and ProSEU:GFP-SEU (ap2-2). 35S:AP2m3 was described previously (Chen, 2004; Zhao et al., 2007).

RNA was isolated from inflorescences using TRI reagent (Sigma-Aldrich, St Louis, MO, USA). For small RNA blots, 30 μg total RNA was separated on a 15% acrylamide gel, transferred, and cross-linked onto Hybond-N nylon membrane using EDC [N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride; Sigma-Aldrich] (Pall and Hamilton, 2008). [γ³²P]ATP and the mirVana Probe & Marker Kit (Ambion, Austin, TX, USA) were used to label the miR172 oligo (see Table S1 in the supplementary material). The small RNA blot was hybridized and washed as previously described (Lau et al., 2001; Lee and Ambros, 2001).

For northern blots, PCR-amplified AP2 first exon or 5S RNA was labeled with [α³²P]dCTP (Perkin-Elmer, Waltham, MA, USA) with Ready-To-Go DNA Labeling Beads (Amersham, Piscataway, NJ, USA). Then 15 μg total RNA was separated on a 1% agarose gel, transferred onto BrightStar-Plus membrane (Ambion), hybridized and washed using the Northern Max-Gly Kit (Ambion). The signal intensity was measured using Image J (NIH).

For AP2 quantitative (q) RT-PCR (see Table S1 in the supplementary material), the QuantiSure First-Strand cDNA Kit (Accugen Bioscience, Rockville, MD, USA) was used, followed by qRT-PCR with SsoFast EvaGreen Supermix on a CFX96 real-time PCR machine (BioRad, Hercules, CA, USA). The Ct^AP2 was subtracted from the Ct^GAPC to yield the ΔCt^WT and ΔCt^lug3. The Pfaffl formula 2^−ΔΔCt (Livak and Schmittgen, 2001), where −ΔΔCt=ΔCt^WT–ΔCt^lug3, gave the expression difference. Error bars represent the s.d. of two biological replicates (three technical repeats each).

For Fig. S1 in the supplementary material, RNAs isolated from wild-type (Ler) and lug-3 inflorescence tissues were enriched for small RNAs, labeled with Cy5, and hybridized to a custom-designed miRNA chip as described (Pang et al., 2009). Average expression values were first normalized against internal positive controls (tRNAs) before they were compared between wild type and lug-3.

A direct tandem oligonucleotide concatamer (4×) of the sense miR172 strand sequence 5′-AGAATCTTGATGATGCTGCAG-3′, with a T7 RNA polymerase-binding site (5′-CCCTATAGTGAGTCGTATTA-3′) at the 3′ end, served as the template for T7 transcription with the digoxygenin (DIG) RNA Labeling Mix (Roche, Basel, Switzerland). The miR172 sense probe was made by a similar strategy. In situ hybridization was carried out based on Carr and Irish (Carr and Irish, 1997); slides were hybridized at 42°C overnight and were washed three times at 42°C for 30 minutes with 0.2× SSC.

Two to three grams of inflorescences with stage 1-11 flowers were used in ChIP based on Sridhar et al. (Sridhar et al., 2006). Protein A Dynabeads (Dynal/Invitrogen, Carlsbad, CA, USA) and a polyclonal anti-GFP antibody (AB290-50, Abcam, Cambridge, MA, USA) were used. qPCR with the QuantiTect SYBR Green PCR Kit (Qiagen, Valencia, CA, USA) was performed on the CFX96 qPCR machine. The percent efficiency for each primer is shown in Table S4 in the supplementary material.

Based on Mukhopadhyay et al. (Mukhopadhyay et al., 2008), the Ct value for the plus antibody sample (+AB) and for −AB was independently subtracted from the Ct value of the input to yield ΔCT. Then ΔCT^+AB was subtracted from ΔCT^−AB to yield the ΔΔCT for each sample. The Pfaffl formula was used to calculate the fold enrichment.

The Matchmaker System (Clontech, Mountain View, CA, USA) was used. ANT-pGAD424 (Krizek and Sulli, 2006) and AP2delta-pGAD424 were gifts of B. A. Krizek. SEU (1-563)-pGBT9 and LUH-pGAD424 were reported previously (Sridhar et al., 2004; Sitaraman et al., 2008). To make AP2-pGADT7, AP2 was excised from pGG30 with NcoI and EcoRI and cloned into pGADT7. For P426-GAPD, the GAPD promoter was excised with BamHI and EcoRI, cloned into pCR2.1-TOPO, excised again with SacI and EcoRI and cloned into p426 GAL, replacing GAL. For P426-SEU, SEU was excised from pCRII-TOPO-SEU (Franks et al., 2002) with HindIII and XhoI and cloned into p426 GAPD at HindIII-SalI. Constructs were introduced one at a time into Saccharomyces cerevisiae strain PJ694A according to the Yeast Protocols Handbook (Clontech).

AP2, SEU and LUH were cloned into pUC-SPYNE and pUC-SPYCE (Walter et al., 2004). Specifically, full-length AP2, SEU or LUH was amplified (for primers, see Table S1 in the supplementary material), cloned into pGEM-T Easy (Promega, Fitchburg, WI, USA) for AP2 and SEU or into pCR8/GW/TOPO (Invitrogen) for LUH, excised with SalI and XmaI, and then cloned into pUC-SPYNE and pUC-SPYCE. Helios Gene Gun (BioRad) was used according to a published procedure (Hollender and Liu, 2010). Results were imaged under a Zeiss Axio Observer.Z1 inverted microscope. The experiment was performed twice.

lug mutant flowers exhibit partially homeotic transformation of sepals to carpelloid sepals, indicating an expansion of C function into the first whorl (Liu and Meyerowitz, 1995). To test whether miR172, which is expressed in the inner floral whorls (Chen, 2004; Wollmann et al., 2010), expands into lug outer whorls, small RNA blots and in situ hybridization were performed. Increased miR172 was detected in lug-3 flowers (Fig. 1A), consistent with preliminary microarray data showing increased expression of miR172a-e (see Fig. S1 in the supplementary material). This increased miR172 in lug-3 correlates with a decrease in AP2 mRNA (Fig. 1B).

In situ hybridization was used to examine miR172 expression in lug-3 as well as in seu-1 because SEU is a known partner of LUG (Sridhar et al., 2006). The miR172 antisense probe was made by T7 transcription from a direct tandem oligonucleotide concatamer (4×) of the sense miR172 strand. Using a similar probe, miR172 was previously shown to be expressed in the entire floral meristem, but the expression was abated from sepals in stage 7 and older flowers (Chen, 2004). We showed that miR172 RNA is no longer present in the sepals of stage 6 and older wild-type flowers (Fig. 1C and see Table S2 in the supplementary material). By contrast, in lug-3 and seu-1 mutant flowers, miR172 expression was detected in the sepals of stage 6 or older flowers (Fig. 1E,F). This ectopic miR172 expression was, however, sometimes found in only one of the two sepals (Fig. 1E,F). Further quantification revealed that 82% of the lug-3 (14 of 17) or seu-1 (18 of 22) sepals at stages 6 to 11 showed ectopic miR172 expression (see Table S2 in the supplementary material). This correlates well with the incomplete sepal-to-carpel transformations in lug-3 and seu-1 mutants (Franks et al., 2002; Liu and Meyerowitz, 1995). The in situ data indicate that LUG and SEU are each required to repress miR172 expression in the sepals starting no later than stage 6.

A recent publication using a locked nucleic acid (LNA)-based probe revealed inner whorl-specific miR172 expression as early as floral stage 4 (Wollmann et al., 2010). This difference in the timing (i.e. stage) of miR172 RNA clearing in wild-type sepals might reside in the greater specificity and sensitivity of the LNA probe, which consequently is likely to detect specific miR172 species. We thus focused our analyses of miR172 expression on stage 6 and older sepals, for which both probe types detect a clearing of the miR172 signal in wild type.

The increased and ectopic miR172 expression observed above could result from ectopic carpelloid organ development in the outer floral whorls of lug-3 as carpels are known to express high levels of miR172. We therefore conducted ChIP to test whether LUG directly associates with miR172 chromatin.

We chose miR172c and miR172e, which showed the highest increase in expression in lug-3 mutants (see Fig. S1 in the supplementary material). We searched miR172c and miR172e promoters for conserved binding sites for class A genes (i.e. MADS box and AP2 domain). No MADS box binding site (CArG box) was found, but four putative AP2 binding sites (TTTGTT; T. Dinha and X.C., unpublished) were found in the promoter of miR172c and one in the promoter of miR172e. PCR and qPCR primers were designed to flank these putative AP2 binding sites (see Tables S3 and S4 in the supplementary material).

Nuclear extracts of inflorescences from 35S:GFP-LUG; lug-16 and negative control (non-transgenic wild type) were immunoprecipitated with an anti-GFP antibody. The immunoprecipitated DNA was first quantified by semi-quantitative PCR, testing a larger number of primers (see Fig. S2 and Table S3 in the supplementary material). qPCR was subsequently used to test the most relevant promoter regions of target genes (Fig. 2A). For miR172c, only binding site miR172c-2 showed over 4-fold enrichment with the GFP-LUG; lug-16 sample (Fig. 2A). The other three potential AP2 binding sites, including miR172c-1, were not enriched (Fig. 2A and see Table S3 in the supplementary material). For miR172e, the single putative AP2 binding site was enriched over 7-fold with the GFP-LUG; lug-16 sample (Fig. 2A). Neither miR172c nor miR172e was enriched with the negative control sample (Fig. 2A). A non-target control, E1F4a1, showed no enrichment (see Fig. S2 and Table S3 in the supplementary material). These data suggest that AP2, or other AP2 family members, might be involved in recruiting LUG to the promoters of miR172c and miR172e.

If AP2 recruits LUG/SEU to repress miR172, ap2 mutants should exhibit expanded expression of miR172 similar to lug and seu mutants. In situ hybridization revealed ectopic miR172 expression in stage 6 and 7 ap2-2 sepals (Fig. 1G); 20 out of 20 ap2-2 sepals at floral stages 6-11 ectopically expressed miR172 (see Table S2 in the supplementary material). This was supported by northern blots (Fig. 1H) showing increased miR172 in ap2-2 flowers. By contrast, plants that express a miR172-resistant AP2 (35S:AP2m3) did not show reduced miR172 (Fig. 1H). Perhaps an overproliferation of stamens, a miR172-expressing organ, in these 35S:AP2m3 lines (Zhao et al., 2007) compensates for a reduction of miR172.

As GFP-LUG (ap2-2) plants are not currently available, GFP-SEU (ap2-2) plants were used for ChIP analysis to test the role of AP2 in mediating the association of the LUG/SEU co-repressors with miR172c and miR172e (Fig. 2B). GFP-SEU (WT) and GFP-SEU (ap2-2) inflorescences were analyzed by ChIP using the anti-GFP antibody. The same qPCR primer pairs as in Fig. 2A were used for quantifying fold enrichment. miR172c-2 and miR172e were enriched 4- and 4.8-fold, respectively, with GFP-SEU (WT) but not with GFP-SEU (ap2-2) (Fig. 2B). These data support the contention that AP2 is required for mediating the association of SEU with the miR172c-2 and miR172e promoters. Similar to GFP-LUG (Fig. 2A), GFP-SEU failed to associate with miR172c-1 (Fig. 2B). Since GFP-LUG and GFP-SEU are both associated with the same AP2 binding sites (miR172c-2 and miR172e) out of those tested (Fig. 2A,B and see Table S3 in the supplementary material), it suggests that LUG and SEU might function together in a complex, tethered by AP2, to regulate miR172. Using ChIP-seq, Yant et al. (Yant et al., 2010) reported the binding of AP2 to miR172b but not to other miR172 genes. This difference might reside in the different developmental stages at which the tissues were collected in the two studies.

The above results suggest that AP2 might directly interact with SEU and/or LUG. A yeast two-hybrid assay revealed a direct interaction of AP2-AD with SEU-BD (Fig. 3A, sector 2) but not with LUG-BD (Fig. 3A, sector 5). A truncated AP2 (AP2delta) containing the two AP2 domains (residues 124-394) failed to interact with SEU-BD (Fig. 3A, sector 3). Thus, the two AP2 domains are not sufficient for the interaction with SEU. Another AP2 domain-containing protein, AINTEGUMENTA (ANT), failed to interact with SEU (Fig. 3A, sector 4), indicating that SEU specifically interacts with AP2.

Bimolecular fluorescence complementation (BiFC) was used to confirm the interaction between SEU and AP2 via particle bombardment in onion cells. SEU and AP2 were each fused to the N-terminal and C-terminal fragments of YFP using the pUC-SPYNE and pUC-SPYCE vectors, respectively (Walter et al., 2004). The fusion proteins SEU-N and AP2-C were observed to interact in nuclei, whereas the negative controls did not (Fig. 3B).

Since LUG is known to interact with SEU (Sridhar et al., 2004), SEU might bridge the interaction between LUG and AP2. A yeast three-hybrid assay was employed to test whether LUG-BD interacts with AP2-AD in the presence or absence of SEU (Fig. 3C). The ‘−TLUH’ and ‘−TLUHA’ media represent two different selection stringencies. AP2-AD interacted with LUG-BD only when SEU was present (Fig. 3C, sector 2) and only in the less stringent −TLUH medium. AP2-AD failed to interact with LUG-BD when SEU was absent (sector 1). AP2delta-AD weakly interacted with LUG-BD in the presence of SEU (sector 3). The data support the proposal that SEU is capable of bridging the interaction between AP2 and LUG. The weakness of the interaction might reflect that all three proteins have to be expressed at the same time and in the same cellular compartments. This result is consistent with synergistic and semi-dominant genetic interactions among lug, seu and ap2 mutants during floral organ development (Franks et al., 2002; Liu and Meyerowitz, 1995), indicating functional interactions in vivo.

miR172 regulates many important developmental processes in diverse plant species (Aukerman and Sakai, 2003; Chen, 2004; Chuck et al., 2007; Jung et al., 2007; Lauter et al., 2005; Martin et al., 2009; Wu et al., 2009). Whereas much attention has focused on finding the targets of miRNAs and on their biogenesis and metabolism, less is known about their transcriptional regulation. Our analysis provides important insights into the domain-specific transcriptional regulation of miR172.

Integral to the ABCE model is the mutual antagonism between the A and C class genes. The repression of A class activity in inner whorls is conferred by at least two mechanisms: the post-transcriptional downregulation of AP2 by miR172 (Fig. 4a) and the repression of AP1 transcription by AG (Fig. 4b) (Aukerman and Sakai, 2003; Chen, 2004; Gustafson-Brown et al., 1994; Schwab et al., 2005). However, AG and miR172 appear to act independently of each other (Wollmann et al., 2010; Zhao et al., 2007). The repression of class C genes in the outer whorls depends on the LUG/SEU co-repressors recruited by the AP1/SEP3 MADS box proteins (Sridhar et al., 2006) (Fig. 4c) as well as by AP2 (Fig. 4d). The mutual repression between AP2 and miR172 ensures positive-feedback loops that promote their own expression in the respective A and C domains (Fig. 4). It remains unresolved as to how the A-C boundary is initially established. A strict temporal order of gene activation might be one mechanism, so that earlier-acting class A genes are subsequently switched off in the inner whorls once the class C genes are switched on.

Transcriptional co-repressors are known to regulate diverse targets and developmental pathways and their target specificity is conferred by DNA-binding partners (Liu and Karmarkar, 2008). This work has reported a new regulatory target, miR172, of LUG/SEU and identified AP2 as the DNA-binding partner of LUG/SEU for miR172 regulation, providing novel insights into co-repressor function in plant development.

We thank Robert Franks for ProSEU:GFP-SEU; seu-1; Beth Krizek for ANT-pGAD424 and AP2delta-pGAD424; Detlef Weigel for pGG30; Jörg Kudla for BiFC vectors; Thanh Dinha for sharing unpublished results and the northern of 35S:AP2m3; Joann Conner for 35S:GFP-LUG; V. V. Sridhar for AP2-pGADT7 and p426-SEU; Vidyadhar Karmarkar for initiating the project; and Ranjani Iyer and Kristi Mullen for generating 35S:GFP-LUG; lug-16 lines. The work is supported by grants from the National Science Foundation (IOB0616096 and MCB0744752) to Z.L. Z.L. is partially supported by the University of Maryland Agricultural Experiment Station.