Interspecific hybridization, polyploidization, and backcross of Brassica oleracea var. alboglabra with B. rapa var. purpurea morphologically recapitulate the evolution of Brassica vegetables

Brassica oleracea and B. rapa are two important vegetable crops. Both are composed of dozens of subspecies encompassing hundreds of varieties and cultivars. Synthetic B. napus with these two plants has been used extensively as a research model for the investigation of allopolyploid evolution. However, the mechanism underlying the explosive evolution of hundreds of varieties of B. oleracea and B. rapa within a short period is poorly understood. In the present study, interspecific hybridization between B. oleracea var. alboglabra and B. rapa var. purpurea was performed. The backcross progeny displayed extensive morphological variation, including some individuals that phenocopied subspecies other than their progenitors. Numerous interesting novel phenotypes and mutants were identified among the backcross progeny. The chromosomal recombination between the A and C genomes and the chromosomal asymmetric segregation were revealed using Simple Sequence Repeats (SSR) markers. These findings provide direct evidence in support of the hypothesis that interspecific hybridization and backcrossing have played roles in the evolution of the vast variety of vegetables among these species and suggest that combination of interspecific hybridization and backcrossing may facilitate the development of new mutants and novel phenotypes for both basic research and the breeding of new vegetable crops.

Brassica oleracea and B. rapa are two closely related taxa that contain the most important cruciferous vege-  [23][24][25] . B. oleracea var. alboglabra has been suggested to be a primitive type of kale crop and a possible ancestor of various cultivated B. oleracea vegetables 23,24 . B. rapa contains an even greater number of subspecies and varieties, such as B. rapa ssp. chinensis (pakchoi), B. rapa ssp. pekinensis (Chinese cabbage), B. rapa var purpurea (purple Cai-tai, B. compestris L. var purpurea was used in the past), B. rapa ssp. parachinensis (Cai xin), B. rapa ssp. dichotoma, B. rapa ssp. japonica, B. rapa ssp. Narinosa (wutacai), B. rapa ssp. Nipposinica (mizuna), B. rapa ssp. oleifera, B. rapa ssp. rapa (turnip), and B. rapa ssp. trilocularis 23,24,26 . The oldest members of B. rapa were planted thousands of years ago. However, most varieties of B. compestris cultivars, the hundreds of landraces of pakchoi in southern China and Chinese cabbage in northern China, emerged between the 14 th and 17 th centuries 27 . Though this evolutionary boom has attracted much attention, the mechanisms underlying it remain unclear.
In the present study, an interspecific hybrid of B. oleracea var. alboglabra and B. rapa var. purpurea (the primitive type of cultivated B. oleracea and B. rapa) was generated. After two rounds of backcrossing of the synthetic allotetraploid plant with B. rapa L. var. purpurea, the morphological variation and genetic composition of the progeny were investigated. The stereotypical traits of B. oleracea var. gemmifera (Brussels sprouts), B. oleracea var. gongylodes (kohlrabi), B. chinensis (pakchoi), and B. pekinensis (Chinese cabbage) were phenocopied or partially phenocopied in the BC 2 plants, providing direct evidence in support of the hypothesis that interspecific hybridization and backcrossing played roles in the evolutionary expansion of vegetable varieties in B. oleracea and B. rapa. The intergenomic recombination and asymmetric segregation of chromosomes could be one of the principal mechanisms underlying this process. Many interesting mutants and previously unknown phenotypes were obtained, indicating that the combination of interspecific hybridization and backcrossing may facilitate the development of more mutants and novel phenotypes.

Creation of hybrid plants by embryo rescue.
A cytoplasmic male sterile line of B. oleracea var. alboglabra was hand-pollinated with pollen from an inbred line of B. rapa L. var. purpurea. On the 8 th day after pollination, the young siliques were harvested, the surfaces were sterilized, and then ovules were dissected out and grown in embryo culture media. Three out of the ≈ 400 ovules developed into seedlings. Morphologically, the hybrid plants fall between their parents (Fig. 1). Chromosome analysis from the root tip cells revealed that the hybrid somatic cells contained 19 chromosomes (Fig. 1b). A series of codominant Simple Sequence Repeats (SSR) markers indicated ten of which were derived from B. rapa L. var. purpurea, and nine of which were derived from B. oleracea var. alboglabra (Fig. 1e). The hybrid plants were male sterile and produced no pollen. On the main stalk, the embryos were lethal when hand-pollinated with pollen from either B. rapa L. var. purpurea or a fertile B. oleracea var. alboglabra plant. However, some adventitious stalks generated from the bases of the plants produced several seeds by backcrossing with B. rapa L. var. purpurea. These partially fertile stalks were stout with shortened internodes that distinguished them morphologically from the original plants, indicating that these branches may underwent a spontaneous chromosomal doubling.
Morphological divergence among BC 1 (AAC) plants. A total of 72 seeds were harvested from the partially fertile F1 stalks that were hand pollinated with B. rapa var. purpurea. After sowing, 67 seedlings germinated and completed a life cycle. Within this population, morphological variations were observed in the size and shape of the leaves, the size and color of flowers and stalks, the structure of the inflorescence and the presence of surface wax (Fig. 2). White petals and surface wax were the dominant traits controlled by the genes from C genome, as indicated by the F1 plants (Figs 1 and 2). The occurrence of yellow petals and waxless surfaces suggests that some of the chromosome segments from the C genome had been lost from the BC 1 plants.
Morphological traits of BC 2 progeny. After hand pollination with B. rapa L. var. purpurea as the male parent, all of the 67 resulting BC 1 plants were fertile. These 67 BC 2 lines produced 935 seedlings. It was deduced that each plant in this generation contained a genome composed of AA′ C (0-9) (A′ probably contains segments from C due to chromosomal recombination; C (0-9) contains 0-9 chromosomes from the C genome that probably contain segments from A). Flow cytometry showed that the chromatin content diverged significantly (Fig. 3). Morphologically, the BC 2 plants bore a stronger resemblance to B. rapa L. var. purpurea, but the detailed traits diverged significantly among individuals within the population (Fig. 4). The plants were clustered based on 17 morphological traits, including plant height, architecture, stooling, stalk shape, stalk color, waxiness of the stalk, brawniness of the stalk, lateral stalk number, rosette leaf shape, rosette leaf color, rosette leaf crimp, waxiness of the leaf, petiole color, bolting time, flowering time, petal color and style structure. As shown in Fig. 5, 19 of the 274 (6.93%) BC 2 plants clustered with B. rapa L. var. purpurea (group I), indicating that these plants were of the AA′ genotype, from which the chromosomes derived from the C genome had been eliminated. However, the individuals within this group were found to differ from each other significantly, indicating that recombination between chromosomes from the A and C genome occurred frequently during two rounds of meiosis, resulting in the replacement of many segments of the A′ genome with fragments from the C genome. Group II comprised BC 1 plants. These plants were less diverse morphologically because they underwent only one round of recombination and contained an AA′ C′ genome. The other plants were clustered according to nine major terms in group III, indicating the wide range of segregation events and complicated genetic interactions.
Phenocopying of Brassica vegetables by BC 2 progeny. Some BC 2 plants displayed traits characteristic of vegetables produced by plants other than their progenitors. As shown in Fig. 6, one BC 2 plant generated enlarged corms that resemble kohlrabi (B. oleracea var. carlorapa) (Fig. 6a). Some BC 2 plants produced axillary bud balls, which are characteristic of Brussels sprouts (B. oleracea L. var. gemmifera) (Fig. 6b). A number of BC 2 plants displayed a phenotype nearly identical to several pakchoi landraces (B. rapa ssp. chinensis). For example, some BC 2 plants phenocopied the "Paopaoqing, " a pakchoi landrace that produces characteristic dark green wrinkled rosette leaves (Fig. 6c). The leaves of the BC 2 plants frequently folded inward (Fig. 6d), a fundamental developmental process that leads to the formation of the leafy head in cabbage and Chinese cabbage. This suggests that the leafy head formation if the petiole will be short and that the inner leaves will fall inward. These phenotypes morphologically recapitulate the evolutionary history of several Brassica vegetable varieties.
Interesting mutants and novel phenotypes. In addition to the evidence presented for morphological variation and evolutionary recurrence, many interesting mutants were generated at a relatively high frequency. For example, one mutant presented two cotyledons that merged into a monocotyledon (Fig. 7a). The backcross progeny segregated into mono and dicotyledonous individuals at a 1:1 ratio, indicating that the mutation is controlled by a single dominant gene. This mutant might be a good model for the study of cotyledon development and determination of the molecular basis of mono and dicotyledonous segregation, which is a fundamental question in the evolution and phylogeny of flowering plants. Another interesting mutant generated young plants from its underground root system (Fig. 7b). Similar schedules have been adopted as a primary method of reproduction by many plants. This mutant might be useful for studying mechanisms of this kind. Plants resembling bamboo shoots and Zingiber striolatum could be used as new vegetable cultivars.

Intergenomic recombination and asymmetric segregation as indicated by SSR analysis.
To investigate the genetic mechanism underlying the aforementioned morphological observations, 30 SSR markers (2-5 per chromosome, details in Table 1) were designed and used for genetic analysis of 93 BC 2 plants, the F 1 , and their initial progenitors. As shown in Fig. 8, the BC 2 plants contained 0-9 whole or partial C chromosomes from B. oleracea var. alboglabra. Intergenomic recombination between A and C chromosomes was detected between each pair of adjacent markers. Because both A and C genomes have been extensively rearranged since their divergence, the homeologous recombinations can result in asymmetric segregations with chromosome deletion and duplication, as demonstrated in Fig. 9.
A neighbor-joining (NJ) tree was generated based on this set of SSR data (Fig. 10). The BC 2 plants also clustered in nine major branches, which are in accordance with the tree constructed with morphological traits (Fig. 5). Due to two generations of backcrossing with B. rapa L. var. purpurea, it is not surprising that the B. oleracea var. alboglabra outbranched far from the BC 2 groups. The F 1 and B. rapa L. var. purpurea clustered separately in two outside branches, which were again consistent with the morphological tree. One of these two branches contained the BC 2 plants harboring nearly the whole set of C chromosomes and the other those that had lost nearly all of the C chromosomes. The inner branches were composed of those harboring many different numbers of C chromosomes. These molecular data confirmed the morphological observations. InDel mutations detected in BC 2 plants. The SSR analysis also detected four novel bands in the BC 2 plants (Fig. 11). All of these four mutations are short stretch insertions. The mutation rate was 0.15%, much higher than the spontaneous mutation rate, indicating the minor genomic alteration such as short stretch insertion was induced by interspecific hybridization.

Discussion
Many interspecific hybrids have been produced from various types of wild and cultivated A and C genome plants [15][16][17][18][19][20][21][22] . However, this is the first report of the hybridization of the two flower stalk vegetables, B. oleracea var. alboglabra and B. rapa L. var. purpurea. Morphological analysis, chromosome counting, flow cytometry and SSR analysis demonstrated that hybridization was successful. The hybrid plants showed a relatively high degree of spontaneous chromosome doubling. Because the initial progenitors were highly homologous, the morphological divergence among the BC 1 plants could not have been derived from genetic segregation between sister chromosomes. Mechanisms such as recombination between homoeologous chromosomes and alterations in the expression pattern and DNA methylation status of genes have been proven to lead to rapid phenotypic changes in newly synthesized allopolyploids [16][17][18][19][20][21][22][28][29][30][31] . The DNA methylation and expression level were not analyzed in these newly synthesized hybrids. However, homoeologous chromosome recombination was found to be active in this system. The changes in qualitative traits, such as flower color, indicate the non-homologous chromosomal recombination and gene loss took place during meiosis in the F 1 plants 32,33 . The relatively low fertility rate of the first round of backcrosses indicated that non-homologous chromosomal recombination occurred extensively, resulting in female gamete lethality due to chromosome disorder during meiosis. This is consistent with the observation that newly synthesized allopolyploids undergo consistent chromosome crossover and recombination 17,34,35 . Chromosome recombination requires homologous sequences for DNA pairing. Sequencing of cabbages and Chinese cabbages has shown that the A and C genomes share many homologous blocks 10,13 , presenting opportunities for recombination. Both the A and C genomes underwent two rounds of genome duplication and produced three subgenomes 9,36 . This type of genome redundancy provides considerable plasticity for non-homologous recombination between A and C chromosomes. These facts may be the reason for the extensive homeologous recombination displayed in the hybrid plants, which leaded to chromosome deletion, duplication and rearrangement in high frequency and finally resulted in the numerous phenotypes observed in the BC 2 progenies. On the other hand, our backcrossing strategy also contributing to the high frequency homeologous recombination observed, because a complete single set of A chromosomes always obtained from the recurrent parent, which will rescue the lethal chromosome rearrangements in the maternal plants.
One well-known example of interspecific hybrid evolution is the U's triangle model, in which three tetraploids, B. napus, B. juncea, and B. carinata, were proposed to have evolved from hybrids of any two of the three diploids B. rapa, B. nigra, and B. oleracea 14 . Each of the three major Brassica vegetable species, B. rapa, B. oleracea, and B. juncea, have a large number of subspecies and varieties and various crop cultivars that produce a diverse variety of edible organs. Previous studies have suggested that these plants were domesticated independently in China, Europe, and the Middle East [37][38][39] . Primitive types, such as kale and turnip, may have been cultivated more than 5,000 years. However, historical records indicate that most other types emerged suddenly between 800-400 years before present 27 . The evolutionary history of the expansion of this cultivar is still a mystery; the natural variation and selection model cannot explain this rapid speciation. Recent re-sequencing and comparison between three subspecies of B. apa, a turnip, a rapid cycling, and the reference genome of Chinese cabbage estimated the date of divergence among the three morphotypes at approximately 250,000 YA, long predating the date of domestication 40 . One explanation for this paradox is that the genetic variation may have come from interspecific hybridization. The present study provides solid evidence that interspecific hybridization and backcrossing have the potential to generate multiple crop types within the B. rapa and B. oleracea clans. It is here proposed that the primitive crop types of B. rapa and B. oleracea were domesticated independently in China and Europe for more than two thousand years and evolved slowly and smoothly until the 12-14 th centuries, when the crops were introduced to each other between China and the Mediterranean via the Maritime Silk Road. These interspecific crossings were spontaneous and occurred in quotidian settings such as kitchen gardens. Backcrossing to B. rapa and B. oleracea crops took place in China and the Mediterranean, respectively, due to the population proportion effect. Genetic rearrangement and morphological variation manifested in the progeny of these crosses. Then, various traits were selected according to the selector's preferences and were consequently refined and stabilized in later generations to form the diverse varieties and cultivars known today. This scenario is also supported by historical biogeographical data that all of the varieties of B. rapa were first generated in southern China, where only one Chinese native kale crop, B. oleracea var. alboglabra, coexists with pakchoi cultivars, including B. rapa L. var. purpurea 27,41 . Even the Chinese cabbage, which is currently planted through much of northern China, Korea, and Japan, originated in southern China and was later distributed elsewhere 27 . In the current model, it was not possible confirm that the two plants used in this study were the initial ancestors of current cultivars, but a trend toward a polychronism was observed, suggesting that multiple interspecific hybridization events took place and that many Brassica plants, including allotetraploid crops, could have participated in them.
Mutants are important resources for genetic research and for plant breeding. Natural mutation is rare and the mutants are difficult to obtain. Several methods, including EMS and γ -radiation, have been used to induce mutations 42 . The majority of these methods induce point mutations, creating loss-of-function mutants. The results of the present work showed distant crossing to be a powerful method for producing mutations, including valuable gain-of-function mutants. In distant hybrids, mutants can be generated by chromosome fragment deletion and gene conjunction, which is called "genome reshuffling, " and by activated transposable elements, which induce mutations termed "genome shock" 16,17,43 . It is not easy to use linkage-map-based cloning on these types of mutants, so they have not been widely used for functional genetic analysis. Currently, with the rapid development of sequencing technology and the accompanying dramatic reduction in cost, these mutants may be more frequently utilized in genetic studies.
Beside the large chromosome fragment rearrangements, minor genomic alteration such as short stretch elimination and insertion also play important roles in the evolution of interspecific hybrid plants 30 . Though only four InDel mutations have been detected in this study, it is far from reflecting the true mutation level in these plants. Because SSR is not an effective mutation detecting tool and it cannot detect the point mutations.
The use of distant cross technology in the plant breeding industry have a long and glorious history. Many important agronomic traits, such as male sterility, disease resistance, and stress tolerance have been introduced to cultivars from wild and interspecific relatives by distant crossing [44][45][46][47][48] . However, crop diversity is currently in sharp decline, and current vegetable innovation cannot match that of the pre-industrial age. One reason for this trend is that commercial seeds are highly homogeneous; thus, new traits are not likely to occur nor to be adopted to form the new varieties. The present study shows that artificial distant hybridization and backcrossing can generate new traits at a relatively high frequency. Some of the new traits have the potential to breed new vegetable cultivars. For this reason, it is here proposed that breeders use distant hybridization and backcrossing strategies to generate new types of crops for the enrichment of the human diet rather than limiting themselves to the transfer of existing traits from one plant to another for the minor modification of commercial cultivars.
Among the BC 3 plants, typical traits such as enlarged corms, bud balls, wrinkled rosette leaves, leaf folding, mono-cotyledon, and root-born seedlings were also observed, indicating that these traits are heritable rather than epigenetic or due to phenotypic plasticity. However, because the maternal progenitor B. oleracea var. alboglabra is a cytoplasmic male sterile line, their progeny cannot self, so the traits cannot be reproduced stably. After generations of backcrossing, the traits are gradually lost and the phenotype finally approaches that of the recurrent parent B. rapa L. var. purpurea. This study is to be repeated using a fertile system, and further stabilize and manifest the traits by selfing and sister crossing, which will provide stronger support for the hypothesis raised in this study.

Material and Methods
Plant material and growth conditions. B. oleracea var. alboglabra 4E286 is a cytoplasmic male sterile line. The B. rapa L. var. purpurea HCT3 is an inbred line. The seeds of these plants and their progeny were surface-sterilized in 1% NaClO and germinated on glass Petri dishes at 25 °C. After germination, seedlings were transplanted into a mixture of peat soil (peat:moss:perlite:vermiculite soil = 3:2:1:1). The seedlings were transplanted to a greenhouse after 2 weeks and were watered and fertilized regularly.
Distant hybridization and embryo rescue. The B. oleracea var. alboglabra cytoplasmic male sterile line 4E286 served as the maternal parent. The B. rapaL. var. purpurea inbred line HCT3 served as the paternal parent. The inflorescence was bagged and hand-pollinated one day later. Eight days after pollination, the siliques were harvested and surface-sterilized in 70% ethanol for 30 s followed by 1% NaClO for 15 min and washed three times in purified, sterile water. The siliques were dissected under a stereomicroscope on a superclean bench. The embryos were cultivated in a solidified MS medium containing 2 mg/L 6-benzyladenine (6-BA) and 0.2 mg/L 1-naphthaleneacetic acid (NAA) and then placed in a tissue culture room at 25 °C with a 16:8 light-dark cycle 49 . The germinated seedlings were propagated on the same medium and then transplanted to a root-induction medium composed of solidified MS + 0.2 mg/L NAA. The rooted plants were transplanted to soil, kept humid for a week, and then cultivated in a greenhouse.

Morphological observation and cluster analysis. The morphological traits recorded include
shape, color, the crimp and wax of the cotyledons, rosette-leaves, stem-leaves, petioles, stalks, flowers, and siliques. The plant height, architecture, bolting time, stooling and brawniness of stalk, number of lateral stalks, and flowering time were observed and measured at the seedling, vegetative, bolting, flowering, SSR analysis. The SSR markers were designed based on comparison between the genomes of B. oleracea 10 and B. rapa 13 . The genome sequences were downloaded from (ftp://bradata:zhl410ivf@brassicadb.org/Brassica_ oleracea/Bol_Chromosome_V1.1/BOL.seq.lst.new.chr20110802_check.fa.gz) and (ftp://bradata:zhl410ivf@ brassicadb.org/Brassica_rapa/Bra_Chromosome_V1.5/Brapa_sequence_v1.5.fa.gz). The SSRs were firstly identified in B. rapa genome using MIcroSAtellite identification tool-MISA (http://pgrc.ipk-gatersleben.de/misa/). Then the primers were design on primer 3 52 . Subsequently, the primers were mapped on B. oleracea and B. rapa genome using Electronic PCR (e-PCR) 53 . The SSR primers which produced fragments of more than six nucleotides difference between the two plants and also have only one local in each genome were retained. From which, 66 primer pairs equidistributed on the A genome were manual selected and experimentally tested via PCR analysis on B. oleracea var. alboglabra and B. rapa L. var. purpurea. The 30 SSR primers (Table 1)  clear distinguishable bands were used for further analysis. Genomic DNA was extracted from leaf tissues using the cetyl trimethylammonium bromide method. PCR amplifications were performed in a 20 μ L volume containing 20-50 ng template DNA, 0.5 pmol primers, 0.5 U Taq enzyme and 1× PCR reaction buffer. Reactions were performed with an initial denaturation step of 3 min at 95 °C; followed by 35 cycles of 95 °C for 30 s, 56 °C for