Promoter Variations Strongly Enhance the Transcription Level of the BoLMI1 Gene, Causing Lobed Leaves in Ornamental Kale (Brassica Oleracea L. Var. Acephala)

Bin Zhang Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Wendi Chen Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Xing Li Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Wenjing Ren Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Li Chen Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Fengqing Han Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Zhiyuan Fang Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Limei Yang Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Mu Zhuang Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Honghao Lv Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Yong Wang Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers Yangyong Zhang (  zhangyangyong@caas.cn ) Chinese Academy of Agricultural Sciences Institute of Vegetables and Flowers


Background
Leaves are essential organs that play an important role in plants, including carbon assimilation, gas exchange, water transport and nutrient distribution [1]. Leaf shape can signi cantly affect both leaf function and plant architecture [2,3]. A typical variation in leaf shape involves the leaf margin, which can be unlobed, serrated or lobed [4]. Lobed leaves can be easily visualized even in the primary leaf stage, which can be used as an indicator trait for hybrid production [5,6]. Compared to unlobed-or serrated-leaf lines, plants with lobed leaves are better adapted to environmental stresses [7,8]. With improved heat transfer and light energy absorption, lobed leaves are advantageous for high-density planting and mechanized production [9]. Additionally, lobed leaves are also a graceful decorative trait for ornamental plants such as kale [10].
Ornamental kale (Brassica oleracea L. var. acephala) is an attractive ornamental crop owing to its polymorphic, colorful leaves [11]. Lobed-leaf genes have been genetically analyzed and mapped in some Brassica species. For example, the lobed-leaf trait in B. rapa is controlled by major gene or polygenic effects [12][13][14]. In B. napus, the incomplete dominant lobed-leaf gene BnLL1 was mapped to the distal end of chromosome A10 [15]. In ornamental kale, some studies have shown that the lobed-leaf trait exhibits incomplete dominance over the unlobed-leaf trait [10,16,17]. Genetic analysis of an interspeci c hybrid between B. napus and Rorippa indica (L.) Hiern revealed that the lobed-leaf trait is controlled by a dominant gene [9]. Moreover, Ren et al. mapped a quantitative trait locus (QTL) associated with lobed leaves to chromosome 9 of ornamental kale anked by insertion-deletion (InDel) markers LYIn39 and LYIn40, with genetic distances of 0.17 cM and 0.11 cM, respectively [4].
With the development of high-throughput sequencing technology and the release of B. oleracea draft genomes [18,19], a growing number of genes that govern important traits have been mapped in this species. Bulk segregant analysis (BSA) is a rapid and accurate gene mapping method that was rst developed and performed in plants [20]. This method is characterized by bulk genotyping of a pool of segregants that share the same phenotype. InDel has been considered as an ideal source for marker design due to its high-density distribution and genotyping e ciency. Using InDel markers, many genes/QTLs have been mapped in B. oleracea, including the yellow-green leaf gene ygl-1 [21], the purple leaf gene BoPr [22], QTLs associated with heading traits [23], male sterility genes [24,25] and the petal color gene BoCCD4 [26].
In the present study, we developed F 1 , F 2 and BC 1 populations descended from the ornamental kale inbred line 18Q2513 (with lobed leaves) and 18Q2515 (with unlobed leaves). These populations were used for inheritance analysis, ne mapping and cloning of the leaf shape-related gene BoLl-1. The ndings provide new insight into the molecular mechanism underlying leaf shape formation in ornamental kale.

Results
Genetic analysis of leaf shape in ornamental kale The leaf shape throughout all the F 1 plants (comprising 16 individuals) generated by crossing 18Q2513 with 18Q2515 was lobed; thus, the lobed-leaf trait is dominant over the unlobed-leaf trait in these two ornamental kale lines. The F 2 population comprised 120 individuals, with 92 displaying lobed leaves and 28 unlobed leaves. According to a chi-square test, the segregation ratio is 3:1. The BC 1 P 1 population contained 850 individuals, with 429 lobed-leaf individuals and 421 unlobed-leaf individuals, and the segregation ratio was con rmed to be 1:1 by a chi-square test. The 200 BC 1 P 2 individuals all had lobed leaves (Table 1). These results indicate that the lobed-leaf trait is controlled by a single dominant gene, which was named BoLl-1.
Fine mapping of the BoLl-1 gene by BSA-seq and linkage analyses To identify markers associated with lobed leaves, the SNP index and Δ(SNP index) between the two bulks were calculated using high-quality SNPs. The average SNP-index and Δ(SNP-index) of the two bulks across a 1-Mb genomic interval were measured using a 10-kb sliding window and plotted against the genome position. The highest peak region, which was considered to be the candidate interval associated with BoLl-1, contains approximately 1.33 Mb (53.34-54.67 Mb) on chromosome 9 according to the 'TO1000' reference genome (Fig. 2a). For the candidate region of BoLl-1, 3280 SNPs between parental lines were identi ed, 410 of which are effective; 593 InDels were identi ed, 35 of which are effective (Table S1).
To further delineate the location of BoLl-1, 16 InDel and seven SNP markers (by comparing resequencing data of the parents with the sequence of the TO1000 reference genome) within the 1.33-Mb candidate region and its anking regions (600 kb on each side) were designed. Ultimately, ve InDel and three SNP markers showed polymorphisms between the two parents. A total of 429 recessive individuals of the BC 1 P 1 population were subsequently used for BoLl-1 ne mapping.
A linkage map consisting of ve InDel and three SNP markers was constructed using MapDraw (Fig. 2b). The SNP markers SL4 and SL6 were found to be tightly linked to BoLl-1, with genetic distances of 0.6 cM and 0.6 cM, respectively. Based on the marker locations in the reference genome, BoLl-1 was ultimately delimited to a 127-kb region (53680797-53808289 bp) on chromosome C09.
Prediction and expression analysis of the candidate genes Based on the 'TO1000' reference genome [19], 21 genes were identi ed within the 127-kb interval ( Table  2). According to annotations from the Brassica oleracea genome and BLASTX (best hit) to A. thaliana, only two genes Bo9g181710 and Bo9g181720 are related to the formation of leaf shape. These two genes are homologues of the LATE MERISTEM IDENTITY 1 (LMI1) gene in Arabidopsis, which encode a homeodomain leucine zipper class I (HD-Zip I) meristem identity regulator that plays an important role in leaf morphogenesis and bract formation. Thus, we designated that Bo9g181710 and Bo9g181720 were candidate genes controlling lobed leaf shape in ornamental kale.
To analyze the expression patterns of Bo9g181710 and Bo9g181720, qRT-PCR was performed using young leaves of the parents. The expression level of Bo9g181710 in lobed-leaf 18Q2513 was signi cantly higher than that in unlobed-leaf 18Q2515, whereas no signi cant difference in Bo9g181720 expression between the parental lines was detected (Fig. 3).

Sequence analysis of the candidate genes
To determine the causal relationship between the candidate genes and leaf shape formation, a comparative sequence analysis of the Bo9g181710 and Bo9g181720 genes body and ~ 3-kb promoter region was performed using genomic DNA from 18Q2513 and 18Q2515. No sequence variations (between the parental lines) in the coding sequences of Bo9g181710 were detected, while a 1737-bp deletion (1466 bp upstream of the transcription start site), a 92-bp insertion (1466 bp upstream of the transcription start site) and an SNP (765 bp upstream of the transcription start site) were identi ed within the Bo9g181710 promoter region of 18Q2513 (Fig. 4a). Conversely, no variation was detected in either the promoter or coding regions of Bo9g181720 between the 18Q2513 and 18Q2515. Combined with expression analysis, we speculated that the Bo9g181710 may control the formation of leaf shape in ornamental kale, and renamed it BoLMI1.
Sequence analysis further revealed that BoLMI1, which consists of three exons and two introns, encodes a putative 219-amino acid protein containing a homeobox domain and a leucine zipper domain (Fig. 4). Sequence alignment of the BoLMI1 protein and its seven homologues from other cruciferous species revealed that BoLMI1 shared a high degree of identity with its homologues in B. napus (98.17%) and B. rapa (91.32%) but a relatively lower degree of identity with Camelina sativa (54.92%) (Fig. 4b).
Furthermore, a phylogenetic analysis of the BoLMI1 protein and its close homologues was carried out to evaluate their evolutionary relatedness. The results showed that BoLMI1 is closely related to B. napusATHB-51 and is located in the same clade as other cruciferous plants, indicating that they may be derived from the same ancestor gene (Fig. 5).
Veri cation of BoLMI1-speci c markers Using the co-dominant marker CMLMI1 and the dCAPS marker DMLMI1, we determined whether the variations the BoLMI1 promoter are also present in 118 different cabbage inbred lines (with unlobed leaves) and another ornamental kale inbred line 2523 (with lobed leaves). The results indicated that the insertion, deletion (detected by co-dominant marker CMLMI1) and the SNP (detected by dCAPS marker DMLMI1) were present only in the lobed-leaf ornamental kale inbred line 2523 ( Fig. 6; Figure S1). These markers exhibited 100% accuracy which can be used for marker-assisted selection. Overall, the analyses strongly indicated that the variations in the promoter of BoLMI1 exist only in lobed-leaf ornamental kale inbred lines and are responsible for the change in leaf shape from unlobed to lobed.

Discussion
In previous studies, the lobed-leaf trait was reported to be controlled by an incomplete gene or a QTL in ornamental kale [4,10,16,17,28]. In the present study, we analyzed the inheritance of leaf shape using F 2 and BC populations derived from a cross of lobed-leaf ornamental kale with unlobed-leaf ornamental kale, showing that the lobed-leaf trait is controlled by a single dominant nuclear-encoded gene. These ndings suggest that the molecular mechanism underlying leaf shape formation in ornamental kale is complex.
Ren et al. mapped the lobed-leaf gene BoLl to chromosome 9 of ornamental kale anked by InDel markers LYIn39 and LYIn40, with genetic distances of 0.17 cM and 0.11 cM, respectively [4]. Two candidate genes, Bol010029/Bo9g181710 and Bol010030/Bo9g1181720, were revealed, but no sequence variations were found in their promoter and coding regions according to the B. oleracea '02-12' (cabbage) [18] and 'TO1000' (Chinese kale like) genomes [19]. Therefore, the authors did not conclude which gene controlled the formation of leaf shape in ornamental kale. In our study, based on the 'TO1000' genome, the BoLl-1 gene was nely mapped to a 127-kb (53680797-53808289 bp) interval on chromosome 9. SNP markers SL4 and SL6 were tightly linked to BoLl-1, anking the gene at genetic distances of 0.6 cM and 0.6 cM, respectively. Sequence analysis of the parental alleles revealed no sequence variations in the coding sequence of Bo9g181710, whereas three variations were identi ed in the promoter region. In contrast, no sequence variations were detected in the promoter and coding regions of Bo9g181720. The expression level of Bo9g181710 in lobed-leaf 18Q2513 was signi cantly higher compared with unlobed-leaf 18Q2515, though the expression level of Bo9g181720 was similar between the parental lines. Thus, we further con rmed that the Bo9g181710 may control the formation of leaf shape in ornamental kale.
In B. napus, Hu et al. reported that a 2624-bp insertion (317 bp upstream of the transcription start site) and three SNPs were identi ed in the BnA10.LMI1 promoter sequence, along with 12 SNPs in the 3′ anking sequence, which were considered to be the cause of the lobed-leaf formation [29]. In ornamental kale, the genes that determine leaf shape are not fully understood. Ren et al. mapped the BoLl gene and found no sequence variations in the promoter and coding regions of candidate [4]. In our study, three variations, including an SNP, a 1737-bp deletion, and a 92-bp insertion (765 bp, 1466 bp, and 1466 bp upstream of the transcription start site, respectively) were identi ed in the BoLMI1 promoter region compared with the 'TO1000' reference genome. Through veri cation analyses of BoLMI1-speci c markers corresponding to the promoter variations revealed that the variations existed only in lobed-leaf ornamental kale inbred lines. These variations strongly enhance the transcription levels of BoLMI1, thus changing leaf shape from unlobed to lobed.
Leaf shape plays an important role in the reproduction and evolution of plants. Increasing evidence indicates that lobed leaves can improve photosynthesis e ciency and agronomic pro tability [7,[30][31][32][33]. LMI1-like genes encoding an HD-Zip I transcription factor have been functionally identi ed in several plants, and they were reportedly involved in leaf shape formation [29,[34][35][36][37]. For example, Hu et al.
(2018) identi ed the BnA10.LMI1 gene, which was responsible for the lobed-leaf shape in Brassica napus. In addition, the BnA10.LMI1 knockout mutations in the HY (with lobed leaves) background were su cient to produce unlobed leaves. In this study, we identi ed an LMI1-like gene, BoLMI1, which was the causal gene underlying the lobed-leaf trait in ornamental kale. Thus, our ndings further strengthen the potential for revealing the molecular mechanism underlying leaf shape formation, and we showed that BoLMI1-speci c markers (CMLMI1 and DMLMI1) can be used for marker-assisted selection in ornamental kale breeding.

Conclusions
In this study, the lobed-leaf trait is shown to be controlled by a single dominant gene, BoLl-1, in ornamental kale. The BoLl-1 gene was ne-mapped to a 127-kb fragment. A homologue of Arabidopsis LMI1, BoLMI1 was identi ed as a strong candidate gene. Three variations were identi ed in the promoter region of BoLMI1. These variations strongly enhance the expression of BoLMI1, causing the leaf shape change from unlobed to lobed. This study lays a foundation for cloning BoLMI1 and provides new insight into the formation of leaf shape in ornamental kale.

Plant materials
The 18Q2513 female parent (P 1 ) is an ornamental kale inbred line with lobed leaves (Fig. 1a); the 18Q2515 male parent (P 2 ) is an ornamental kale inbred line with unlobed leaves (Fig. 1b). 18Q2513 was crossed with 18Q2515 to generate an F 1 population. An F 2 population was generated from selfpollination of the F 1 plants; BC 1 P 1 and BC 1 P 2 were then generated by BCs of F 1 × 18Q2513, F 1 ×18Q2515, respectively.
Additionally, 118 different cabbage inbred lines (with unlobed leaves) and another ornamental kale inbred line 2523 (with lobed leaves), were screened for BoLl-1 promoter variations. All of the plant materials used in the present study were grown in a greenhouse under normal management. All the plant materials are from the Institute of Vegetables and Flowers, Chinese Academy of Agriculture Sciences (IVFCAAS, Beijing, China).

Genetic analysis and whole-genome resequencing
Leaf shape was investigated visually. Segregation ratios for the F 2 and BC 1 populations were analyzed by chi-square (χ 2 ) tests using SAS software.
Fifty lobed-leaf BC 1 and fty unlobed-leaf BC 1 individuals were selected to construct two bulks. The two bulks and two parental lines were then used to construct paired-end sequencing libraries, which were subsequently sequenced by the Beijing Genomics Institute (BGI) (Shenzhen, China). SNP-index and sliding-window analyses were performed as previously described [27].
Marker development and ne mapping of the BoLl-1 gene InDel and SNP markers were designed based on candidate region resequencing data for the two parents. Markers were designed with amplicon lengths of 100-180 bp, GC contents of 40-50% and Tm values of 52-58°C. The markers that were polymorphic between the parents were then used to analyze unlobed-leaf individuals in the BC 1 P 1 populations.
Genomic DNA was extracted from young leaves of the parents and BC 1 P 1 individuals using a modi ed cetyltrimethylammonium bromide (CTAB) protocol [38]. The DNA concentration was subsequently determined using a spectrophotometer (BioDrop, UK) and adjusted to 40-50 ng/μL. 72°C for 10 min. The amplicons were separated by 8% polyacrylamide gel electrophoresis (160 V for 1.2 h), and the gel was stained with silver nitrate.
For each marker, individuals consistent with the 18Q2513 (lobed-leaf) allele, the 18Q2515 (unlobed-leaf) allele, and the F 1 allele were categorized as 'a', 'b', and 'h', respectively. Genetic distances between markers were calculated by the Kosambi map function [39], and a genetic map was constructed using MapDraw [40].

Candidate gene analysis
To identify the lobed-leaf gene BoLl-1, genes located within the candidate interval were analyzed based on annotations for the B. oleracea 'TO1000' reference genome (http://plants.ensembl.org/Brassica_oleracea/Info/Index) [19]. The expression patterns of candidate genes Bo9g181710 and Bo9g181720 were investigated using quantitative real-time PCR (qRT-PCR). Total RNA was extracted from young leaves of the parents using TRIzol reagent (Invitrogen, United States) according to the manufacturer's protocol, and PrimeScript TM RT Reagent Kit (Takara, Japan) was used to reverse transcribe cDNA from the total RNA extracted. qRT-PCR was carried out using a CFX96 Real-Time System (Bio-Rad) with SYBR Premix Ex TaqII Reagent Kit (Takara, Japan). The relative expression level of each gene was calculated using the 2 −ΔΔCt method [41]. The qRT-PCR primers used are listed in Table S2, and B. oleraceaactin was employed as a control.
Gene-speci c markers GL10 (primers GL10-F and GL10-R) and GL20 (primers GL20-F and GL20-R) (Table  S2) were used to amplify the promoter and genomic sequences of Bo9g181710 and Bo9g181720, respectively. The resulting PCR products were analyzed by electrophoresis on 1% agarose gels, followed by sequencing and alignment. The co-dominant marker CMLMI1 (primers CMLMI1-F1, CMLMI1-F2 and CMLMI1-R) and the derived cleaved ampli ed polymorphic sequence (dCAPS) marker DMLMI1 (primers DMLMI1-F and DMLMI1-R) (  Table 2 The 21 putative gene models in the target mapping region Gene structure and protein alignment of BoLMI1. a The BoLMI1 gene structure as well as promoter variations between 18Q2513 and 18Q2515 are shown; horizontal blue arrows represent speci c primers for amplifying the promoter and genomic sequences and detecting the promoter variations of BoLMI1. b Sequence alignment of the BoLMI1 protein and its seven homologues from other cruciferous species.
The homeobox domain as well as the leucine zipper domain are indicated Figure 5 Phylogenetic analysis of BoLMI1 and its 23 homologues from other species