The three crop types used in this study are heading Chinese cabbage (B. rapa ssp. pekinensis), non-heading Chinese cabbage (B. rapa ssp. chinensis), and turnip (B. rapa ssp. rapifera). Heading Chinese cabbage include the crops: Sanye (ssp. pekinensis var. dissoluta Li), Banjieqiu (ssp. pekinensis var. infarcta Li), Huaxin (ssp. pekinensis var. laxa Tsen et Lee) and Jieqiu (ssp. pekinensis var. cephalata Tsen et Lee) while non-heading Chinese cabbage include the crops: Baicai (ssp. chinensis var. communis), Wutacai (ssp. chinensis var. rosularis), Caitai (ssp. chinensis var. utilis), Zicaitai (ssp. chinensis var. purpurea) and Taicai (ssp. chinensis var. taitsai) (Li, 1990). Each crop type consists of several genotypes with different types of leaf curvature. In total, 56 genotypes were used for characterization of leaf curvature. All of them are the inbred lines except for two turnip cultivars. The genotypes used in this study are presented in Supplemental Table S1.

For comparison with the genotypes of Chinese cabbage, Arabidopsis mutants deficient in miRNA-mediated pathways were also selected.

The seeds of Chinese cabbage and Arabidopsis were sown in petri dishes with moistened filter paper and sealed with parafilm, moved to a growth chamber and grown at 22°C with 16 h of light. Four days later, they were transplanted to a growth chamber in the SIPPE phytotron and transferred to the field at the experimental station of SIPPE, Songjiang, Shanghai. More than 20 individual plants for each crop were sampled for various measurements.

The leaves were labeled at different developmental stages for quantitative measurement of leaf curvature. Before leaf flattening, we defined the type and direction of leaf curvature. To measure the global transverse curvature, we fixed two points, a and b, on the two lateral margins of the leaf at the widest point (Liu et al., 2011). Curvature indexes (CIs) for downward and upward curvature were calculated using the formulae CI = (a′b′ − ab)/a′b′ and CI = (ab − a′b′)/a′b′, respectively, where ab is the distance between points a an b on the two margins before leaf flattening and a′b′ is the distance between a and b on the two margins after flattening. The number of leaves for each measurement was more than 20. The significant differences were referred to P < 0.001 (Student's t-test) between genotypes.

The leaf samples were harvested from the 3-week-old seedlings of different genotypes and frozen immediately in liquid nitrogen. Then, the materials were sent to ShanghaiBio Corporation where the RNAs were isolated for miRNA microarray. Isolation of total RNA is done by the phenolic separation of DNA and RNA using Trizol (Invitrogen, Carlsbad, CA). The purification of miRNA samples, labeling, and hybridization analysis were performed following the instructions (Liu et al., 2008). Each sample was assayed in duplicate. The data were extracted using LuxScan (CapitalBio), and the differential miRNAs were selected using SAM (Significance Analysis of Microarrays ver. 3.0).

Total RNA was extracted from seedlings using TRIzol (Invitrogen) and treated with DNase I (TaKaRa) to remove DNA contamination. Four micrograms of RNA was used for reverse transcription with oligo(dT) primers. PCR was performed with the Rotor-Gene 3000 system (Corbett Research) using SYBR Premix Ex Taq (TaKaRa) according to the manufacturer's instructions. BrpACTIN mRNA was used as an internal control. The comparative threshold cycle (Ct) method was employed to determine relative transcript levels (MyiQ2 two-color real-time PCR detection system; Bio-Rad). Expression was normalized relative to that of BrpACTIN. Three biological replicates and three technical replicates were performed. The significant differences were referred to P < 0.001 (Student's t-test) between genotypes.

The Brp-MIR166g-1 gene was cloned by PCR from seedlings of heading Chinese cabbage (Bre) and inserted into pCAMBIA1300 binary vectors under the control of the AA6 promoter. The binary constructs were delivered into Agrobacterium tumefaciens strain GV3101(pMP90RK) (Weigel and Glazebrook, 2006). The Bre plants were transformed using a vernalization-infiltration method as described previously (Bai et al., 2013). The transgenic plants of the T1 generation were pollinated, and homozygotes were selected from the T2 and T3 populations.

B. rapa shows a broad spectrum of leaf curvature and morphological variation. To characterize the direction, degree and position of leaf curvature in B. rapa, we selected the 14th leaves (rosette leaves) from a collection of 56 accessions for observation and measurement (Supplemental Table S1). The accessions of heading (rapa ssp. pekinensis), non-heading (rapa ssp. chinensis), and turnip (rapa ssp. rapa) were named with the initials rp, rc, and rr, respectively. Most accessions showed global curvature in the transverse or longitudinal axis of the leaves, as the entire leaves curved downward or upward (Figure 1). However, some accessions displayed local leaf curvature as local regions of the leaf such as the blade tip, margin and petiole displayed certain types of curvature.

During plant growth, the direction, degree, and position of leaf curvature and the inclination angle are changed at different development stages. For convenience of comparison, we chose the 14th leaf of plants for observation of leaf curvature. Among 56 accessions, 16 representative genotypes with different leaf curvature were given in Table 1 and Figure 1.

According to the direction, degree and position of curvature and the inclination angle, we classified the 56 accessions into five types of leaf curvature: downward curving leaves, inward curving petioles, serrated blades, wavy margins, and upward curving leaves. Upward curving leaves (rosette leaves along both the transverse and longitudinal axes) were observed in rp-2, rc-2, and rc-5 plants; upward curving petioles (small inclination angle of petioles) in rc-3 plants; serrated blades in rc-6 plants; wavy margins in rp-6 plants; and downward curving leaves (rosette leaves along both the transverse and longitudinal axes) in rr-1 plants.

In many cases, global curvature was mixed with local curvature in a single leaf and wavy margins, deep serration, upward curvature, and inward curvature appeared together. For example, rc-3 leaves curved downward as a whole; their apical regions curved downward, but their central regions curved upward. In such cases, global and local curvature were measured separately. Accordingly, some genotypes were categorized into two or three types of leaf curvature.

Heading Chinese cabbage shows different head shapes. The head tops of heading Chinese cabbage vary due to the diversity of leaf curvature. As shown in Figure 2, the rp-1 heads were oval with a cone-shaped top, as the leaf tips curved upwardly. The rp-2 heads were round as the leaf tips curved inwardly and thus overlapped. The rp-7 heads were cylindrical as the leaf incline angle became small or near zero. The rp-10 heads were upright with overturned leaf tips, due to downward curvature of the leaf tips. The rp-11 heads were elliptical as the head leaves were twisted in a counter-clockwise direction. Inside the leafy heads of heading Chinese cabbage, the head leaves were extremely incurved compared with those in the rosette. To compare the phenotypes of the different head leaves among crops, we harvested the leaves of leafy heads progressively from the outside to the inside and observed the direction and degree of leaf curvature in order. As shown in Figure 3, the leaves in different types of leafy heads were distinct in the direction and degree of curvature. For rp-1 leaves, the outermost leaf (leaf 6) of the leafy head curved inward transversely while the inner leaf (leaf 1) curved inward longitudinally and transversely. For rp-2 leaves, both the outermost and innermost leaves curved inward along both the longitudinal and transverse axes, and the degree of leaf curvature increased progressively from leaf 6 to leaf 1. For rp-5 leaves, the outermost leaves of the leafy head curved inward transversely but outward longitudinally while the inner leaves curved inward only transversely. For rp-9 leaves, both the outermost and innermost leaves curved inward transversely but outward longitudinally, even though all the leaves had wavy margins.

The shapes of the leaves and leafy heads of B. rapa plants are not only related to the curvature of leaves, but also to leaf incline angles. All leaves in B. rapa plants are inclined at various angles to the vertical axis from 0° to 180°. In Figures 4A,B, the leaf angle, blade angle and petiole angle of a rc-2 plant represented by α, β, and γ, respectively. If α = 0, the leaf was upright, and if α = 90°, the leaf was flat to the horizontal plane. Interestingly, the leaf angle, blade angle and petiole angle of this plant were distinct. The blade angle β was more than 90° while the leaf angle α and petiole angle γ were less than 90°. From the outer to the inner leaves, the leaf angle, blade angle and petiole angle became progressively smaller. We noticed that the petiole angles of rc-1 plants were so small that the petioles were bundled into a cluster.

Leaf incline angles varied by accessions and not by the crop types. At the rosette stage, the leaf angles were ranked rp-1>rc-3>rc-4>rp-9>rr-1>rc-2>rc-1 (Figure 4). The petiole angle was greatest in rp-1 and smallest in rc-1. The leafy head of rp-7 was special in that the leaf, blade and petiole angles were almost the same and were all nearly zero, meaning that the leaves were erect. We noticed that the outer leaves around the rp-7 head remained inclined at various angles, distinct from the internal leaves in the leafy head. Importantly, a small leaf incline angle or leaf erectness is necessary for the formation of leaf heads.

In Arabidopsis, leaf curvature was quantified using the formula CI = (a′b′ – ab)/a′b′, where ab is the distance between points a and b on the two margins of a leaf before flattening, and a′b′ is the distance between a and b on the two margins after flattening (Liu et al., 2010). To determine whether this formula was applicable to B. rapa, we calculated CI values for several representative types of leaf curvature (Figure 5). In rp-1 heading leaf, the straight-line distance between a and b along longitudinal axis before flattening is shorter than the one between a′ and b′ after flattening. In rc-1 petiole, the straight-line distance between a and b along transverse axis before flattening is shorter than the one between a′ and b′ after flattening. rc-2, a genotype of non-heading Chinese cabbage, had the lowest (−0.63) transverse CI while rc-3, another genotype of non-heading Chinese cabbage, had the highest (0.19) transverse CI for blades (Figure 6A). rp-2, a genotype of heading Chinese cabbage, had the lowest (−0.37) longitudinal CI for blades while rr-1, a genotype of turnip, had the highest (0.14) (Figure 6B). Although the leaves of these genotypes showed global curvatures, the directions of leaf curvature along the transverse and longitudinal axes were different.

The CIs of leaves were related to the order of the leaves on the shortened stem. In rp-2 plants, either the transverse or longitudinal CI increased progressively from the innermost to outermost leaves in the leafy head (Figures 6C,D), in which the transverse and longitudinal CIs of the innermost blades were −1.03 and −0.51 while those of the outermost blades were −0.11 and −0.03. In rc-1 plants, both the transverse and longitudinal CIs of the petioles decreased progressively from leaves 1 to 4 (Figures 6E,F), indicating that the severity of petiole inward curvature increased from the outer to inner leaves. Specifically, the transverse and longitudinal CIs of the innermost petioles were −0.15 and −0.16 while those of the outermost blades were −0.24 and −0.18.

The various types of leaf curvature in different genotypes of B. rapa were similar to many mutants of Arabidopsis deficient in miRNA-directed pathways. In total, 22 Arabidopsis lines including mutants deficient in miRNA biogenesis, MIRNA genes and miRNA-targeted genes, and transgenic lines related to leaf curvature were selected. To examine the possible relationship between miRNAs and leaf curvature in B. rapa, we compared the leaf shapes of the different genotypes with those of previously reported miRNA-deficient Arabidopsis mutants. The downward curvature of rc-3 leaves resembled that of the jba-1D mutant in which the MIR166g gene is activated (Williams et al., 2005). The wavy margins of rp-3, rp-9, rc-4, and rc-5 leaves (Figure 1) were similar to those of the jaw-1D mutant in which the MIR319a gene is activated (Palatnik et al., 2003; Table 1). The upward curvature of rp-2, rc-2, and rc-5 plants resembled that of 35S::mCNA plants in which the miR165/6-targeted CNA gene is overexpressed (Liu et al., 2010), and the downward curvature of rc-3 leaves looked like that of the rev-6 mutant (Prigge et al., 2005). The blades that curved upward transversely and downward longitudinally in rp-3 and rp-5 leaves were similar to those of the hyl1 mutant in which the function of HYL1 (responsible for miRNA biogenesis) is lost (Yu et al., 2005). The upward curling of rp-5 was the same as that of mir159ab mutants (loss-of function mutants of MIR159a and MIR159b genes) (Allen et al., 2007). The rp-7 leaves curved inward transversely. The leaflets in the rc-5 blades were equivalent to those of plants overexpressing the miR164–targeted CUC1 gene (Hasson et al., 2011).

To examine whether the leaf curvature of B. rapa was associated with miRNA accumulation, we chose the accessions rc-1, rc-6, rp-2, rp-9, and rr-1 as representatives of the five types of leaf curvature (upward curving petioles, serrated blades, upward curving leaves, wavy margins, and downward curving leaves), respectively, for miRNA microarray analysis. There was great difference in the accumulation of miRNAs among the different genotypes (Table 2). Compared with rp-2, the accumulation of all types of miR169 in rc-1, rc-6 and rr-1 seedlings was increased by more than 2 times, while the accumulation of all types of miR172 and miR162 was decreased by 2 times. In rp-9 seedlings, the accumulation of miR319ab was increased 2 times compared with rc-6 plants, indicating that higher accumulation of miR319a is associated with the wavy leaf margins of rp-9. Compared with rc-6 plants, which have flat leaves, miR166 accumulation in rr-1 plants was increased by more than 1.5 times, suggesting that a relatively high level of miR166 is associated with the leaf downward curvature of the rr-1 genotype.

As mentioned above, the downward leaf curvature of some B. rapa accessions was comparable to that of the jba-1D mutant of Arabidopsis, while the wavy margins of some B. rapa accessions were similar to the jaw-1D mutant of Arabidopsis. We wondered whether high levels of miR166 were correlated with downward curvature in B. rapa. To address this question, we performed qPCR of miR166 using leaf samples of all 31 accessions showing downward curvature. Not all of the accessions with this type of leaf curvature showed higher levels of miR166. Among the accessions of the downward curving type in transverse direction, 84% showed upregulation of miR166 compared with rp-9 (with upward curvature; Table 3). Similarly, among the accessions with wavy margins, 89% exhibited upregulation of miR319a compared with rp-2 (without wavy margins). These results showed that high levels of miR319a were correlated with wavy margins.

The genome of Chinese cabbage is mesohexaploid and thus most of the genes in Chinese cabbage have 2–3 copies, showing duplication or triplication. However, the Brp-MIR165/6 gene is not in accordance with this rule. There are only 3 Brp-MIR165 loci and 7 Brp-MIR166 loci in Chinese cabbage, compared with 2 and 7 in Arabidopsis, respectively. By contrast, the miR165/6-targeted genes are duplicated or triplicated. This suggests that miR165/6-guided gene silencing in B. rapa is more complex than in Arabidopsis.

To verify the relationship between miRNA and leaf curvature in B. rapa, we constructed binary vectors of some MIRNA genes under the control of the AA6 promoter and overexpressed them in Chinese cabbage (Figure 7A). Brp-MIR166g is the most homologous to MIR166g within the MIR165/6 gene family of Arabidopsis, but has a “T”-to-“C” substitution in the miR166 region and generates a mature miR165 rather than a miR166 (Supplemental Figure S1). Brp-MIR166g was placed under the control of the AA6 promoter and transferred into heading Chinese cabbage Bre. Through PCR of the AA6 promoter, nine transgenic lines (166g-1 through 166g-9) were identified (Supplemental Figure S2). Among them, the 166g-2 line showed a 2.5-fold increase in miR165 accumulation compared with the wild-type (Figure 7B). In the 4th leaves of this transgenic line, BrpREV-1, BrpREV-2, and BrpPHB-1 were downregulated more than 2-fold (Figure 7C). This result indicated that overexpression of Brp-MIR166g caused the upregulation of miR165 and downregulation of its target genes.

In the field, the seedling leaves (8th leaves) of wild-type plants and 166g-2 plants carrying pAA6::Brp-MIR166g were downwardly curved (Figures 7D, 8A,D), and the rosette leaves (12th leaves) of the wild-type were basically flat (Figures 8B,C) whereas the rosette leaves of 166g-2 and 166g-3 plants were downwardly curved (Figures 7D, 8F,G). This demonstrated that higher accumulation of miR165 changed the direction and degree of leaf curvature at the rosette stage. When the young leaves on the shoot tip of the wild-type were upwardly and inwardly curved, those of 166g-2 plants were downwardly curved (Figures 8B,F). Because of the delay of leaf incurvature, the head size of 166g-2 plants was decreased and the heading time was delayed.

We noticed that the transverse CI of the 12th leaves in 166g-3 plants was higher than that of the 166g-2 plants (Figure 7D). Considering that the expression level of miR165 in 166g-3 plants was lower than in 166g-2 plants, this suggested that the expression levels of miR165 affected the degree of downward curvature of the rosette leaves in the transgenic plants carrying pAA6::Brp-MIR166g.

Overexpression of the Brp-MIR166g gene altered the adaxial–abaxial polarity of the rosette leaves. Compared with the wild-type, the adaxial regions in 166g-2 leaves contained many long cells similar to those in the abaxial regions of the wild-type (Figures 8D,H), and the intercellular spaces between the palisade mesophyll cells were larger than those of the spongy mesophyll cells of the wild-type. This result suggested that the adaxial regions in 166g-2 leaves had abaxial characteristics.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.