BcpLH organizes a specific subset of microRNAs to form a leafy head in Chinese cabbage (Brassica rapa ssp. pekinensis)

HYL1 (HYPONASTIC LEAVES 1) in Arabidopsis thaliana encodes a double-stranded RNA-binding protein needed for proper miRNA maturation, and its null mutant hyl1 shows a typical leaf-incurvature phenotype. In Chinese cabbage, BcpLH (Brassica rapa ssp. pekinensis LEAFY HEADS), a close homolog of HYL1, is differentially expressed in juvenile leaves, which are flat, and in adult leaves, which display extreme incurvature. BcpLH lacks protein–protein interaction domains and is much shorter than HYL1. To test whether BcpLH is associated with defects in microRNA (miRNA) biogenesis and leaf flatness, we enhanced and repressed the activity of BcpLH by transgenics and investigated BcpLH-dependent miRNAs and plant morphology. BcpLH promoted miRNA biogenesis by the proper processing of primary miRNAs. BcpLH downregulation via antisense decreased a specific subset of miRNAs and increased the activities of their target genes, causing upward curvature of rosette leaves and early leaf incurvature, concurrent with the enlargement, earliness, and round-to-oval shape transition of leafy heads. Moreover, BcpLH-dependent miRNAs in Chinese cabbage are not the same as HYL1-dependent miRNAs in Arabidopsis. We suggest that BcpLH controls a specific subset of miRNAs in Chinese cabbage and coordinates the direction, extent, and timing of leaf curvature during head formation in Brassica rapa.


Introduction
Leafy heads are types of agricultural product composed of numerous incurved leaves. Crop species with leafy heads include Chinese cabbage (Brassica rapa ssp. pekinensis, syn. Brassica campestris ssp. pekinensis), cabbage (B. oleracea var. capitata), brussels sprouts (B. oleracea var. gemmifera), and lettuce (Lactuca sativa). Unlike the grains of corn, rice, and wheat, which provide starch and proteins for food, leafy heads supply mineral nutrients, crude fiber, and vitamins for health. The vegetative development of these crop species is divided into seedling, rosette, folding, and heading stages. The seedling and rosette leaves perform normal photosynthesis, whereas the head leaves serve as nutrient storage organs. The flatness of rosette leaves and the proper incurvature of heading leaves are essential for the high yield and quality of leafy heads. However, the genetic basis underlying leaf incurvature and head formation is unclear.
Leaf curvature is determined by leaf morphogenesis. In particular, the roles of adaxial-abaxial polarity, cell division, phase transition, and the genesis of leaf formation have long been a focus of studies 1,2 . The flatness of leaves can be described in terms of Gaussian curvature, in which a flat surface grows isotropically; for example, a uniformly expanding disc maintains zero Gaussian curvature 3 . Although leaves of many plant species have approximately zero Gaussian curvature, there are many more ways for a leaf to adopt negative or positive curvature than zero curvature for natural variation. Several transcription factors responsible for leaf adaxial-abaxial polarity have been shown to participate in the establishment of leaf curvature. The adaxial side is specified by the activity of members of the class III Homeodomain Leucine Zipper (HD-ZIP III) family of transcription factors. HD-ZIP III genes are targets of miR165/6. Of these, PHABULOSA (PHB), PHAVOLUTA (PHV), and REVOLUTA (REV) act redundantly to promote the adaxial cell fates of leaf primordia [4][5][6][7] . Dominant gain-of-function mutations in these transcription factors have been characterized that cause an expanded expression domain, promoting the adaxial growth of leaves 4,7 . ATHB8 (HB-8) and CORONA (CNA) play antagonistic roles against REV in certain tissues while performing overlapping functions with those of REV in other tissues 4 . Loss of function of the HD-ZIPIII gene results in abaxialized organs 6,[8][9][10] . Members of the miR319a-targeted TEOSINTE BRANCHED1/CYCLOI-DIA/PCF (TCP) gene family function in the maintenance of the normal shape and flatness of leaves via arrested cell division at the front of leaves 3 . miR156-targeted SPL genes control the transition of leaves from the juvenile to adult stage by the mediation of morphological and physiological changes 11,12 . In Chinese cabbage, miR319a modulates the head shape of Chinese cabbage by differentially arresting cell division in leaf regions 13 . The silencing of the miR156-targeted SPL genes promotes early leaf incurvature and heading 14 .
MiRNAs and their targets have been shown to function in many plant development processes and to be involved in protein processing. HYL1 has been verified to participate in the biogenesis of miRNAs in combination with DICER-LIKE1 (DCL1) and SERRATE (SE) 15,16 . As a result of a reduction in miRNAs, plants with the hyl1 null allele exhibit multiple phenotypic abnormalities, such as leaf hyponasty, delayed flowering, altered root gravity responses, and altered responses to hormones 17,18 . The N-terminal region of HYL1, which has two tandem dsRBD domains alone, is adequate to completely rescue the phenotype of hyl1 mutant 19 .
The formation of a leafy head is a multitrait. The size, shape, weight, and compactness of leafy head and heading time are under the control of different genetic loci 20 . All plants with leafy heads undergo leaf curvature transitions from downward to inward. Leaf incurvature at late developmental stages is essential for the high yield and quality of leafy heads. In 2000, we reported that the gene Brassica rapa ssp. pekinensis LEAFY HEADS (BcpLH) was isolated by differential hybridization of cDNA libraries using flat rosette and upwardly curved folding leaves of Chinese cabbage 21 . To determine whether BcpLH functions in leaf curvature, we investigated BcpLH-regulated miRNAs and miRNA-targeted genes through the overexpression or silencing of BcpLH. We found that BcpLH controlled the timing of leaf curvature and leafy head formation through integration of some important miRNAs.

Results
BcpLH is downregulated at the folding stage of Chinese cabbage Chinese cabbage plants are characterized by downwardcurving leaves at the seedling stage, flat leaves at the rosette stage, upward-curving leaves at the folding stage and inward-curving leaves at the heading stage (Fig. 1a). BcpLH isolated by differential hybridization between rosette and folding leaves was considered to contribute to heading. Genomic sequencing of B. rapa revealed another copy of BcpLH, which we named BcpLH2. The amino acid sequences of the two BcpLH proteins are highly identical (96%), which indicated that BcpLH and BcpLH2 may function redundantly in B. rapa.
To examine the contribution of BcpLHs in Chinese cabbage, we measured the temporal and spatial expression patterns of the BcpLH genes. We isolated RNA samples from the shoot tips of plants at the four developmental stages. Real-time PCR showed that the expression levels of both BcpLH or BcpLH2 increased progressively from the seedling stage, during the rosette stage and to the heading stage, while they were downregulated at the folding stage (Fig. 1c). This result was consistent with that of the differential hybridization, which indicated that BcpLH functions at the key folding stage. In situ hybridization demonstrated that both BcpLH and BcpLH2 were expressed mainly in the shoot apical meristems and developing leaves. The difference is that, compared with BcpLH2 expression, BcpLH expression in developing leaves was more preferential in the adaxial region than in the abaxial regions (Fig. 1b). BcpLH2 was expressed mainly in the shoot apical meristem and tips of developing leaves, whereas BcpLH was expressed preferentially in the adaxial regions of developing leaves (Fig. 1b).
To investigate the subcellular localization of BcpLH, we fused GFP with BcpLH and performed a transient expression of p35S:BcpLH-GFP and p35S:BcpLH2-GFP in leaf protoplasts of Chinese cabbage. Subcellular fluorescence showed that BcpLH and BcpLH2 were localized simultaneously in the nucleus and cytoplasm (Fig. 1d).
The knockdown of BcpLH altered the timing of leaf curvature and leafy head formation Considering the special expression pattern of BcpLH, we hypothesized that BcpLH plays a key role in the heading of Chinese cabbage. First, we overexpressed BcpLH in Chinese cabbage under the control of the AA6 promoter using in planta transformation via the vernalizationinfiltration method 22 . The phenotype of the transgenic plants did not differ from that of the wild type, even though BcpLH mRNA and protein levels markedly increased ( Supplementary Fig. 1). To determine the physiological roles of BcpLH and BcpLH2, we cloned Bre antisense sequences of BcpLH and BcpLH2 and inserted them into binary vectors under the control of the AA6 promoter and then transferred those constructs into Bre plants of Chinese cabbage plants that produced a round head. We named the two pAA6:BcpLH antisense transgenic lines LHas-1 and LHas-2 and the two pAA6:BcpLH2 antisense lines LH2as-1 and LH2as-2 (Fig. 2a). To confirm the knockdown of BcpLH in the transgenic plants with antisense BcpLH, we isolated protein samples from developing leaves (1 cm long) of the LHas-1 plants at the rosette stage. Western blotting showed that the amount of BcpLH protein was reduced in the LHas-1 cells (Fig. 2b), revealing that antisense BcpLH specifically reduced the accumulation of BcpLH proteins. In the field, compared with the wild-type plants, which had flat leaves and a round head, the four transgenic lines showed more crinkly leaves, an earlier heading time and a head shape that transitioned from round to cylindrical (Fig. 2a). The 13th leaf of the Bre plants was flat at the rosette stage, and the 16th-17th leaves began to fold (Fig. 2a, d). By contrast, the 13th leaves of the LHas-1 lines were curved inward, with crinkles and more bulges, exhibiting the properties of leaves at the folding and heading stages (Fig. 2a, d). Compared to that of the wild type, the first day with upward curvature in the LHas-1 plants was 10 days earlier (Table 1), and the first leaf with upward curvature was earlier (by 3 leaves); in addition, in the latter, the juvenile phase (seedling stage) was 4 days shorter, and there was 1 fewer leaf in the juvenile phase (Fig. 2e). As a result, the timing of the upward and inward curvature of the LHas-1 leaves occurred much earlier than did that of the wild types, leading to early heading, and the number of incurved leaves increased, resulting in taller, heavier, and larger heads.
While the rosette leaves of the wild-type plants were flat, those of the LHas-1 and LHas-2 plants wrinkled, with bugles and wavy margins. Compared to that in the wildtype leaves, the longitudinal curvature of the top regions in LHas-1 leaves became weaker, causing the head shape to transition from the round to oval (Fig. 2a, Table 1). While the leafy heads of the LHas-1 and LHas-2 plants were round shaped, those of the LHas-1 and LHas-2 plants were oval, apparently due to the constriction of the top regions of head leaves.
BcpLH is the homologous gene of HYL1 and rescues the phenotype of hyl1 plants Knockdown of BcpLH and BcpLH2 affected the heading of Chinese cabbage, so the next problem was to determine how BcpLH and BcpLH2 regulate heading. First, the amino acid sequences composing BcpLH and BcpLH2 were queried via BLAST. We found that BcpLH and BcpLH shared high identity, approximately 78%, with the two dsRNA-binding domains of HYL1 in Arabidopsis (Supplementary Fig. 2A). Compared with HYL1, both the BcpLH and BcpLH2 proteins have two conserved dsRNAbinding domains but lack the long C-terminal fragments containing a putative protein-protein interaction (PPI) domain. A phylogenetic tree of AtDRBs and BcpDRBs was constructed and showed that BcpLH and BcpLH2 were the definite orthologous genes of HYL1 in Chinese cabbage ( Supplementary Fig. 2B).
In Arabidopsis, HYL1 is responsible for miRNA biogenesis. In hyl1 mutants, the downregulation of a subset of miRNAs causes pleiotropic phenotypes, including phenotypes associated with leaf curvature, small stature and delayed phase transition 17,18 . Considering that, unlike HYL1, the BcpLH and BcpLH2 proteins lack the PPI domain, it was unclear whether BcpLH functions in miRNA biogenesis. Therefore, constructs of BcpLH and BcpLH2 with GFP at the C-terminus under the control of pBcpLH and pBcpLH2, respectively, were introduced into the Arabidopsis hyl1 mutant (Fig. 3a). Western blotting showed that the BcpLH and BcpLH2 proteins effectively accumulated in the transgenic lines compared with the hyl1 mutants (Fig. 3b). The hyl1 phenotypes were mostly rescued by BcpLH or BcpLH2 in the pBcpLH:BcpLH-GFP/ hyl1 and pBcpLH2:BcpLH2-GFP/hyl1 plants, and the degree of rescuing by BcpLH and BcpLH2 was 91.7% and 89.5%, respectively (Fig. 3c, Table 2). To determine whether BcpLH and BcpLH2 contribute to miRNA biogenesis in Arabidopsis, northern blotting was performed, and the accumulation of miRNAs was increased in the transgenic plants compared with the hyl1 mutants (Fig. 3d). Concomitantly, the expression of their pri-miRNAs was downregulated (Fig. 3e), and that of the corresponding miRNA-targeted genes was downregulated (Fig. 3f). We then used the HYL1 native promoter driving BcpLH, pHYL1:BcpLH-GFP, which was subsequently transformed into the hyl1 mutant ( Supplementary Fig. 3A). The pHYL1:BcpLH construct expectedly rescued the phenotype of hyl1 (Table 2, Supplementary Fig. 3B). The miRNA accumulation in the transgenic plants increased, while the expression of the corresponding target genes decreased ( Supplementary Fig. 3C, D). These results suggest that BcpLH, with only two dsRNA-binding domains, is able to rescue the hyl1 phenotype, and thus, its role in miRNA biogenesis is similar to that of HYL1.

BcpLH is a direct component of miRNA processing
BcpLH rescued the miRNA levels of hyl1 in Arabidopsis, so we wanted to determine whether BcpLH participates in miRNA processing directly. In Arabidopsis, HYL1 colocalizes with AtSE and AtDCL1 in the nucleus 17 . The fulllength CDS regions of BrpDCL1 and BrpSE were amplified using cDNA synthesized from Chinese cabbage and cloned into a pEASY/blunt vector. The cloned gene The seeds were sown in pots and grown at 22°C in a SIPPE phytotron. The seedlings were transplanted into the field on August 24, 2015, at a SIPPE farm station. More than 20 plants were used for each measurement. The number of days was recorded from the first day after germination, while the number of leaves from the first primary leaf was recorded. The data are presented as the mean of 20 plants ND not detected fragments (BcpLH, BrpDCL1, BrpSE1) were sequenced and inserted into pSAT4-nYFP and pSAT4-cEYFP. Each of the two cloned constructs was transformed concurrently into the protoplasm of Chinese cabbage leaves. The results of bimolecular fluorescence complementation (BiFC) assays showed that BcpLH colocalized with BrpDCL1, BrpSE1 and BrpSE2 in the nucleus (Fig. 4a), indicating that BcpLH localizes in the D-body and is possibly associated with pri-miRNA processing in Chinese cabbage. Considering the conserved dsRNA-binding domains in BcpLH, we investigated the binding of pri-miRNAs to BcpLH by RNA electrophoresis mobility shift assays (EMSAs). Recombinant BcpLH-GST and BcpLH2-GST were expressed in Escherichia coli and purified by glutathione Sepharose resin. Moreover, pri-miR168a was transcribed by the T7 promoter as substrates in vitro. The purified proteins and pri-miRNAs were then incubated in mobility shift buffer at 4°C for 2 h. Northern blotting detected the mobility shift of pri-miR168a, thus showing that pri-miRNA168a did not bind to GST but instead bound specifically to the BcpLH-GST and BcpLH2-GST proteins (Fig. 4b). Equal amounts of purified BcpLH-GST and BcpLH2-GST proteins were used for the RNA EMSA assay. As expected, the bands of the BcpLH-and BcpLH2pri-miRNA complexes migrated more slowly than did the free pri-miRNAs, indicating the direct binding of BcpLH with the pri-miRNAs in vitro. Furthermore, RNA immunoprecipitation (RIP) was performed to investigate the binding of BcpLH with pri-miRNAs in vivo using samples from the leaves of wild-type Chinese cabbage.  The pri-miRNAs loaded by the BcpLH complex were examined by RT-PCR (Fig. 4c). The tested pri-miRNAs were detected in the BcpLH complex immunoprecipitated by anti-BcpLH but not in the immunoprecipitation from the "no-antibody" controls. We concluded that pri-miRNAs bind to BcpLH proteins in Chinese cabbage plants.
It was previously reported that the efficiency of pri-miRNA processing by DCL1 was enhanced by HYL1 23,24 . In this study, we focused on the function of BcpLH in pri-miRNA processing. In conjunction with a BcpLH antibody, a co-IP complex from Chinese cabbage was used in pri-miRNA processing. The substrates for miRNA processing, pri-miR166a and pri-miR168a transcripts, were obtained in vitro under the T7 promoter. pri-miR168a was cleaved only when the BcpLH complex was added to the reaction, and mature miR166 was detected only in the presence of the BcpLH complex. In vitro miRNA processing confirmed that pri-miR168a and pri-miR166a were cleaved only in the presence of BcpLH (Fig. 4d, e). Taken together, these results suggest that BcpLH is an important and direct component in pri-miRNA processing in Chinese cabbage.
BcpLH regulates leaf curvature and leafy head formation by miRNAs Considering that BcpLH directly participates in miRNA processing, we suspected that the change in leaf curvature and leafy head formation in LHas-1 plants was caused by miRNAs. To examine whether knockdown of BcpLH affects miRNA biogenesis and the subsequent effects on Chinese cabbage, we isolated RNA samples from developing leaves (1 cm long) of the LHas-1 line at the heading stage and performed small RNA deep sequencing and RNA-seq. The abundance of a subset of miRNAs changed by more than 1.5-fold (Fig. 5a). Among the 10 miRNAs examined, miR156a-f, miR159a, miR164a, miR165a, miR165b, and miR166a-e were downregulated, while miR168a, miR172a and miR319a, and miR319b were upregulated (Supplementary Table 2). The abundance of miR156a-f decreased 1.6-fold, whereas that of miR319a and miR319b surprisingly increased 2.4-fold. Among the 21 miR156-targeted genes, 14 were upregulated, and among the 22 miR319-targeted genes, 20 were downregulated (Supplementary Table 3). The RNA-seq data showed that the expression levels of miR156-targeted BrpSPL9-2 and miR166-targeted BrpREV-1 were upregulated (Supplementary Table 3). Northern blotting was then performed for LHas-1 and LHas-2 to confirm the changes in miRNA accumulation in the transgenic lines (Fig. 5b). In accordance with the small RNA deep sequencing results, miR156, miR164, and miR166 decreased, while miR319 increased in LHas-1. Real-time PCR showed that the expression levels of miR156targeted BrpSPL9-2, miR164-targeted BrpCUC3-1, and miR166-targeted BrpREV-1 were upregulated in LHas-1 Fig. 4 BcpLH is involved in miRNA processing. a BiFC showing the interactions between BcpLH and BcpLH, BrpSE, and BrpDCL1 in protoplasts of Chinese cabbage. BcpLH-nYFP and BcpLH-cYFP, BrpSE-cYFP, and BrpDCL1-cYFP were transiently coexpressed in protoplast cells of Chinese cabbage leaves. Left, YFP fluorescence. Right, bright field. b RNA-EMSA assay by incubating pri-miR168a with no protein, GST, BcpLH-GST or BcpLH2-GST. c RNA immunoprecipitation by BcpLH antibodies for pri-miRNA binding detection using wild type. Input was collected before immunoprecipitation, and UBQ5 was used as an internal control. d, e In vitro pri-miRNA processing using pri-miR166a (d) and pri-miR168 (e). pri-miRNAs were incubated with complexes immunoprecipitated with BcpLH antibodies (anti-LH) or without antibodies (beads). and LHas-2, which corresponded with the northern blot results and was consistent with the RNA sequencing (Fig. 5c). Corresponding to the changes in leaf curvature and leaf head formation in LHas-1, the early heading time was reminiscent of the transgenic plants harboring pAA6: BrpSPL9-1 reported by Wang et al. 14 ; the round-to-oval transition of head shape and the more wrinkled and bulging leaves were consistent with the transgenic plants harboring pAA6:Brp-MIR319a reported by Mao et al. 13 ; and the upward curvature of the leaves is in agreement with the pAA6:BrpREV-1 plants reported by Ren et al. 25 . These results verified that BcpLH regulates leaf curvature and leafy head formation via miRNAs.
To further examine whether miRNA processing efficiency or pri-miRNA cleavage accuracy or both contribute to the change in miRNAs resulting in the phenotype of LHas-1, we first quantified pri-miRNAs by real-time PCR. The results showed that the abundance of pri-miR156a, pri-miR166a and pri-miR164a increased considerably, while that of pri-miR319a decreased (Fig. 5d), thus showing that the processing of these pri-miRNAs is abolished in LHas-1 plants. The 5′ RACE (rapid amplification of cDNA ends) procedure was then performed to detect the 5′ cleavage sites in pri-miRNAs in both wild type and LHas-1. The accuracy of pri-miRNA cleavage was affected by the antisense BcpLH. Compared to the percentage in the wild-type pri-miR166a, 5% (1 in 20) of the cleavage sites in pri-miR166a in LHas-1 plants shifted by 1 nucleotide, and 6.7% (1 in 15) of cleavage sites in pri-miR168a shifted far away from the stem loop. Most of the cleavage sites were 16 bp away from the ssRNA-dsRNA junction of pri-miRNA in Bre. Hence, BcpLH is important for correct selection of the cleavage sites in pri-miRNAs (Fig. 5e, f). We conclude that BcpLH coordinates microRNA accumulation for the timing of leaf curvature and leafy head formation by ensuring miRNA processing efficiency or pri-miRNA cleavage accuracy in Brassica rapa. Overexpression of miRNAs partially rescued the phenotype of LHas-1 miRNAs contribute to phase transition and leaf development, which cocontribute to the formation of leafy heads. As LHas-1 with altered miRNA abundances affected the timing of leaf curvature and leafy head formation, transgenic lines overexpressing miRNAs or target genes were grown to check the rescuing of LHas-1. In our study, eMIR156a miR156-overexpressing transgenic lines, with high expression of miR156 and downregulated expression of BrpSPL9, exhibited delayed phase transition, in which the heading time was delayed to 40 days, or no heading occurred. However, LHas-1 × eMIR156a, with upregulated miR156 and reduced expression of BrpSPL9, headed 35 days later than did LHas-1. eMIR166g miR166overexpressing transgenic lines showed more downwardcurved leaves, while LHas-1 × eMIR166g impaired the upward-curved leaves of LHas-1 by increased amounts of miR166 and downregulated BrpREV expression. In addition, 5mTCP4 transgenic plant developed more flat leaves, while LHas-1 × 5mTCP4 presented leaves with fewer wrinkles and bulges as a result of the increased expression of BrpTCP4 ( Fig. 6; Supplementary Table 4). These data showed that miRNA or target gene overexpression could partially rescue the phenotype of LHas-1, indicating the mutual effect between miRNA accumulation and mutual regulation between target genes. We concluded that BcpLH coordinates miR156, miR166, and miR319, causing their target genes to affect leaf development and heading characteristics. The balance among the accumulation of miRNAs and the mutual regulation of their targets thus contributes to plant development and production.

Discussion
BcpLH functions differently from HYL1 of Arabidopsis in the processing of some primary miRNAs Both BcpLH and HYL1 are required for the processing of primary miRNAs and act as functional partners of DCL1 among the miRNA biogenesis machinery 26 . Although, unlike HYL1, BcpLH lacks a long PPI domain in its C-terminal region, it shows its ability to form homodimerization and ensure the correct selection of cleavage sites in pri-miRNAs. This result supports the previous findings that the N-terminal double-stranded RNA-binding domains are sufficient for processing primary miRNAs. However, the processing of some primary miRNAs is different in Chinese cabbage and Arabidopsis. In hyl1 mutants of Arabidopsis, the miR156, miR165/6, and miR319 contents are lower than those in the wild type, concurrent with a relatively high accumulation of pri-miR156, pri-miR165/6 and pri-miR319; in Chinese cabbage, however, the miR319 content in transgenic plants with BcpLH antisense is much higher than that in the wild type, concurrent with a low accumulation of pri-miR319. The miR160 content is much lower in hyl1 mutants of Arabidopsis than in the wild type, but the content in the transgenic plants of Chinese cabbage with antisense BcpLH is not lower than that in the wild type. This suggests that the same pri-miRNA reacts differently to BcpLH in different genetic backgrounds of Chinese cabbage and HYL1 in different genetic backgrounds of Arabidopsis.
The rosette leaves of Chinese cabbage and Arabidopsis are essentially flat. However, the rosette leaves of hyl1 mutants and LHas-1 plants are upward curving in the transverse direction, possibly due to the reduced contents of miR156 and miR165/6. The difference is that the wrinkled, bulging and wavy margins typical of jaw-1 mutants (which present enhanced expression of miR319a) occur on the rosette leaves of LHas-1 plants but do not occur on the rosette leaves of hyl1 mutants. The enhanced expression of miR319a in the rosette leaves of LHas-1 plants causes the wrinkled, bulging and wavy margins of the leaves. We suggest that the wrinkled, bulging and wavy margin leaf phenotype caused by BcpLH occurs mainly through miR319a.
Recently, we found that several pri-miRNAs bind to BcpLH differentially, and they compete with each other for binding ability. In LHas-1 plants, the disruption of the original balance between these miRNAs may alter the competence of some pri-miRNAs, thereby altering the levels of the related miRNAs. An attempt has been made to identify whether the enhanced expression of miR319a in the rosette leaves of LHas-1 plants is caused by the reduced levels of miR165/6 and/or miR156.
BcpLH regulates the direction, degree, and timing of leaf curvature Plants have formed a set of mechanisms to coordinate the morphogenesis of organs and the timing of developmental events. The flat leaves of Chinese cabbage generated at the seedling and rosette stages produce enough photosynthetic products to support plant growth, whereas those at the folding and heading stages are upwardly and inwardly curved for nutrient storage. The coordination of morphological changes and phase transition ensures the formation of leaf curvature. This coordination is apparently disrupted by antisense BcpLH. In LHas-1 plants, silencing of BcpLH via antisense causes rosette leaves to transition from being flat to being upward and causes the folding-stage leaves to transition from being upward to curving inward; moreover, the degree of upward curvature increases, and the inward curvature occurs sooner compared with that of wild type. Moreover, the wrinkled, bulging and wavy margins occur on rosette leaves. These altered leaf characteristics are beneficial for early heading, as LHas-1 plants form leafy heads much sooner than do wild-type plants.
The changes in the direction, degree and timing of leaf curvature are attributable to a decrease in a subset of miRNAs. In LHas-1 plants, miR165/6, miR156, and miR164 accumulation decreases; miR168, miR172, miR319 accumulation increases; and the accumulation of many of the other miRNAs does not change considerably. miR165/6, miR156, and miR319 regulate the development of adaxial identity, phase transition and arrest cell division at the front of leaves, respectively. Therefore, it is important to analyze the relationships between these miRNAs and the timing of leaf curvature.
BcpLH coordinates miR165 with miR156 and miR319 for the timing of leaf curvature In Chinese cabbage, miR156 prolongs the juvenile phase and delays the adult phase by silencing BrpSPL9-2, leading to early leaf incurvature and heading 14 ; miR319a regulates differential cell division arrest in forward leaf regions by silencing the BrpTCP4 gene, resulting in wrinkled, bulging and wavy leaf margins and causing the head shape to transition from round to cylindrical 13 ; and miR166 regulates the direction of leaf incurvature, causing changes in head size and heading time 25 . BcpLH and BcpLH2 control the expression levels of the BrpSPL9-2, BrpREV, and BrpTCP4 genes via miR156, miR166 and miR319, respectively, and thus, their downregulation affects the phase transition and head shape simultaneously. As such, miR165/6 coordinates with miR156 and miR319 to determine the timing of leaf curvature. The balance between the relative abundance of these miRNAs is essential for the correct timing of leaf curvature during vegetative growth. In Arabidopsis, HYL1 regulates leaf flatness by modulating the ratio of the expression of genes involved in adaxial to abaxial characteristics (adaxial to abaxial ratio), which determines the direction and extent of leaf incurvature 27 . In Chinese cabbage, BcpLH regulates the balance among miR156, miR165/6 and miR319 and ensures that leaf curvature occurs at the proper time during vegetative growth.
BcpLH promotes the processing of pri-miR165/6 and pri-miR156 but inhibits the processing of pri-miR319a. The contents of miR156, miR165/6 and miR319 affect the expression levels of HD-ZIP III, SPL, and TCP genes, respectively, and influence the adaxial identity, phase transition and cell division arrest at the front of leaves, thus affecting the timing of leaf curvature. Furthermore, the balance among miR156, miR165/6 and miR319 under the control of BcpLH facilitates the formation of leafy heads through the correct timing of leaf curvature. Hence, BcpLH has the potential for genetic manipulation of agricultural products. In future research, we could alter the expression of BcpLH and BcpLH2 to generate plants with characteristics such as better or more effective storage, which would bring great benefits to agricultural production.

Plant materials and growth conditions
An inbred line of Chinese cabbage (B. rapa ssp. pekinensis cv. Bre) was used in this study. The seeds were sown in a greenhouse on August 8, 2014. Two weeks later, the seedlings were transplanted into the field at the Songjiang Farm Station of SIPPE in early September.
The in planta transformation procedures with Bre using the vernalization-infiltration method are described by Bai et al. 22 . Briefly, Brassica plants with small flower buds at the early bolting stage were used for transformation. The plants were placed upside down in a vacuum desiccator that contained both infiltration media and the engineered Agrobacterium for vacuum infiltration. The Agrobacterium-infected plants were then grown in a dark room and incubated at 22/18°C, after which they were transferred to a chamber room after 2 days. The pollen of the Bre plants was then used to pollinate the transformed flowers manually. The seeds of the transgenic plants were harvested after they plants grew for 1-2 months in a growth chamber.
Arabidopsis thaliana wild-type plants and hyl1 (Nossen ecotype) mutants were used in this study. The growth conditions and transgenic methods are described by Wu et al. 19 .

RNA analysis
For isolating total RNA from plant samples, 1 ml of TRIzol per 0.2 g of plant tissue was used for extraction, and phenol:chloroform:isoamyl alcohol and chloroform were added for phase separation, followed by ethanol precipitation.
Nothern blotting was performed as described previously by Wu et al. 19 . Briefly, 30 µg of total RNA was loaded onto a 19% PAGE gel for electrophoresis at 150 V for 4 h, after which the gels were transferred to a Hybond membrane (Amersham Biosciences, GE Healthcare) subjected to 200 mA for 2 h. After UV cross-linking was performed, the membrane was then hybridized in ULTRAhyb ® Ultrasensitive Hybridizaton Buffer (Ambion, Austin, TX, USA) with DNA oligo probes. The probes were in the antisense orientation to the mature miRNA or U6 transcripts, with biotin labels at their 3′ terminal (TaKaRa, Otsu, Japan). The northern blot results were generated with a Light Shift EMSA Kit (Thermo Scientific, Waltham, MA, USA) and imaged using a FLA-5000 Phosphor imager (Fujifilm).
Total RNA samples were extracted from plant leaves using TRIzol, extracted with phenol:chloroform:isoamyl alcohol and chloroform, and then precipitated with ethanol.
For northern blotting, 30 µg of total RNA was resolved by 19% PAGE electrophoresis in 1 × TBE buffer and then transferred to a Hybond membrane (Amersham Biosciences, GE Healthcare), which was subjected to 200 mA for 2 h. The UV cross-linked membrane was subsequently hybridized in ULTRAhyb ® Ultrasensitive Hybridizaton Buffer (Ambion, Austin, TX, USA) using antisense probes of 3′-biotin-labeled oligo DNA (TaKaRa, Otsu, Japan) to mature miRNA or U6 transcripts. The hybridization signals were developed with a Light Shift EMSA Kit (Thermo Scientific, Waltham, MA, USA) and imaged using a FLA-5000 Phosphor imager (Fujifilm).
To perform quantitative real-time PCR, 50 μg of RNA was treated with DNase I (TaKaRa) to remove DNA contamination, followed by RNA extraction with phenol: chloroform. Five micrograms of RNA was reverse transcribed to produce cDNAs with PrimeScript ® Reverse Transcriptase (TaKaRa) in conjunction with oligo(dT) primers. Real-time PCR was performed with specific primer pairs (Supplementary Table 5) in a MyiQ2 Two-color Real-time PCR Detection System (Bio-Rad, Richmond, CA, USA). At least 3 biological replicates of quantitative PCR were performed for each gene. The relative transcript level of each gene was normalized to that of ACTIN cDNA for quantitation.

BiFC assay
Paired constructs were coexpressed in Arabidopsis protoplasts for 12 h at 22.5°C in the dark and subjected to confocal microscopy (Zeiss LSM 510 Meta) for imaging. BiFC signals were excited at 658 nm and detected with a narrow barrier filter.

RIP
Leaf tissue from five-week-old transgenic Arabidopsis plants was ground under liquid nitrogen and homogenized in 5 mL/g lysis buffer [50 mM Tris-HCl (pH 7.4), 100 mM KCl, 2.5 mM MgCl 2 , 0.1% NP-40, and 2× complete protease inhibitor cocktail; Roche]. After centrifugation for 15 min at 9500 × g, the clarified lysate was precleared for 20 min at 4°C with 10 μL of bed volume protein A-agarose (30 μg protein A)/milliliter. The precleared lysates were reacted with 4 μg of anti-LH (NEB) or anti-GFP (Sigma-Aldrich)/milliliter for 1 h at 4°C and then with 50 μL of bed volume protein A-agarose (150 μg protein A)/milliliter for 3 h at 4°C. The precipitates were washed three times in lysis buffer and then divided for protein and RNA analyses. Nucleic acids were recovered by treatment with 3 volumes of proteinase K solution [100 mM Tris-HCl (pH 7.4), 10 mM EDTA, 150 mM NaCl, 2% SDS, and 0.2 μg/μL proteinase K] for 15 min at 65°C, extracted with saturated phenol and phenol: chloroform, and then precipitated with ethanol. Five micrograms of RNA from the input extract or from IP fractions representing 150 mg of tissue was used for qPCR analysis. UBQ5 was used as a control.

RNA-EMSAs
BcpLH and BcpLH2 were independently cloned in frame in a pGEX4T-1 vector for bacterial expression and purification of fusion proteins, which were then used in RNA-EMSA assays. Pre-miR168a stem-loop RNA was transcribed by the T7 promoter in vitro. One microgram of purified GST, BcpLH-GST and BcpLH2-GST protein each was then incubated with the pre-miRNAs (1.5 pmol). The reactions were incubated in binding buffer [20 mM HEPES-KOH (pH 7.5), 10 mM KCl, 20 mM MgCl 2 , 0.5 mM EDTA, and 0.5 mM DTT] at 4°C for 30 min and were subsequently resolved on a 4.5% nondenaturing glycerol polyacrylamide gel, after which they were exposed to a storage phosphor imager screen for the detection of biotin-labeled miRNA.
Pri-miRNA processing in vitro RNA substrates were transcribed under the T7 promoter in vitro using PCR-generated templates. The in vitro transcription of RNAs was carried out for 3 h or overnight at 37°C in one reaction containing 1 µL of DNA template (100 ng), 4 µL of 5 × transcription buffer [400 mM HEPES (pH 7.5), 10 mM spermidine, 200 mM DTT, 125 mM MgCl 2 and each dNTP at 20 mM], 1 µL of RNase inhibitor (Ambion), 2 µL of T7 RNA polymerase and 12 µL of water. DNase-treated RNA was fractionated on a 6% polyacrylamide and 8 M urea gel (denaturing gel) and eluted overnight from gel slices in RNA elution buffer [0.3 M NaAc (pH 5.5) and 2% SDS] using a Thermomixer R (Eppendorf) at 4°C under 1200 rpm; afterward, the RNA was precipitated with ethanol and stored in RNasefree water. Briefly, 10 µL of each RNA cleavage assay mixture contained 20 mM Tris-HCl (pH 7.0), 50 mM NaCl, 4 mM MgCl 2 , 5 mM ATP, 1 mM GTP, 2 units of RNase inhibitor (TaKaRa), RNA substrate, and the co-IP protein complex with anti-LH. After incubation at 37°C for 30 min, the products were extracted with phenol: chloroform and precipitated. The processed products were fractionated by PAGE in a 19% acrylamide urea gel and detected by northern blotting.