Next-generation sequencing, FISH mapping and synteny-based modeling reveal mechanisms of decreasing dysploidy in Cucumis


  • Luming Yang,

    1. Horticulture Department, University of Wisconsin, Madison, WI, USA
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Dal-Hoe Koo,

    1. Horticulture Department, University of Wisconsin, Madison, WI, USA
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Dawei Li,

    1. Horticulture Department, University of Wisconsin, Madison, WI, USA
    2. Horticulture College, Northwest A & F University, Yangling, China
    Search for more papers by this author
  • Tao Zhang,

    1. Horticulture Department, University of Wisconsin, Madison, WI, USA
    Search for more papers by this author
  • Jiming Jiang,

    1. Horticulture Department, University of Wisconsin, Madison, WI, USA
    Search for more papers by this author
  • Feishi Luan,

    1. Horticulture College, Northeast Agricultural University, Harbin, China
    Search for more papers by this author
  • Susanne S. Renner,

    1. Department of Biology, University of Munich, Munich, Germany
    Search for more papers by this author
  • Elizabeth Hénaff,

    1. Centre for Research in Agricultural Genomics Consejo Superior de Investigaciones Científicas-Institut de Recerca i Tecnologia Agroalimentàries-Universitat Autònoma de Barcelona-Universitat de Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Walter Sanseverino,

    1. Institut de Recerca i Tecnologia Agroalimentàries, Centre for Research in Agricultural Genomics Consejo Superior de Investigaciones Científicas-Institut de Recerca i Tecnologia Agroalimentàries-Universitat Autònoma de Barcelona-Universitat de Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Jordi Garcia-Mas,

    1. Institut de Recerca i Tecnologia Agroalimentàries, Centre for Research in Agricultural Genomics Consejo Superior de Investigaciones Científicas-Institut de Recerca i Tecnologia Agroalimentàries-Universitat Autònoma de Barcelona-Universitat de Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Josep Casacuberta,

    1. Centre for Research in Agricultural Genomics Consejo Superior de Investigaciones Científicas-Institut de Recerca i Tecnologia Agroalimentàries-Universitat Autònoma de Barcelona-Universitat de Barcelona, Barcelona, Spain
    Search for more papers by this author
  • Douglas A. Senalik,

    1. Horticulture Department, University of Wisconsin, Madison, WI, USA
    2. US Department of Agriculture/Agricultural Research Service, Vegetable Crops Research Unit, Madison, WI, USA
    Search for more papers by this author
  • Philipp W. Simon,

    1. Horticulture Department, University of Wisconsin, Madison, WI, USA
    2. US Department of Agriculture/Agricultural Research Service, Vegetable Crops Research Unit, Madison, WI, USA
    Search for more papers by this author
  • Jinfeng Chen,

    1. College of Horticulture, Nanjing Agricultural University, Nanjing, China
    Search for more papers by this author
  • Yiqun Weng

    Corresponding author
    1. Horticulture Department, University of Wisconsin, Madison, WI, USA
    2. US Department of Agriculture/Agricultural Research Service, Vegetable Crops Research Unit, Madison, WI, USA
    Search for more papers by this author


In the large Cucurbitaceae genus Cucumis, cucumber (C. sativus) is the only species with 2= 2= 14 chromosomes. The majority of the remaining species, including melon (C. melo) and the sister species of cucumber, C. hystrix, have 2= 2= 24 chromosomes, implying a reduction from = 12 to = 7. To understand the underlying mechanisms, we investigated chromosome synteny among cucumber, C. hystrix and melon using integrated and complementary approaches. We identified 14 inversions and a C. hystrix lineage-specific reciprocal inversion between C. hystrix and melon. The results reveal the location and orientation of 53 C. hystrix syntenic blocks on the seven cucumber chromosomes, and allow us to infer at least 59 chromosome rearrangement events that led to the seven cucumber chromosomes, including five fusions, four translocations, and 50 inversions. The 12 inferred chromosomes (AK1–AK12) of an ancestor similar to melon and C. hystrix had strikingly different evolutionary fates, with cucumber chromosome C1 apparently resulting from insertion of chromosome AK12 into the centromeric region of translocated AK2/AK8, cucumber chromosome C3 originating from a Robertsonian-like translocation between AK4 and AK6, and cucumber chromosome C5 originating from fusion of AK9 and AK10. Chromosomes C2, C4 and C6 were the result of complex reshuffling of syntenic blocks from three (AK3, AK5 and AK11), three (AK5, AK7 and AK8) and five (AK2, AK3, AK5, AK8 and AK11) ancestral chromosomes, respectively, through 33 fusion, translocation and inversion events. Previous results (Huang, S., Li, R., Zhang, Z. et al., 2009, Nat. Genet. 41, 1275–1281; Li, D., Cuevas, H.E., Yang, L., Li, Y., Garcia-Mas, J., Zalapa, J., Staub, J.E., Luan, F., Reddy, U., He, X., Gong, Z., Weng, Y. 2011a, BMC Genomics, 12, 396) showing that cucumber C7 stayed largely intact during the entire evolution of Cucumis are supported. Results from this study allow a fine-scale understanding of the mechanisms of dysploid chromosome reduction that has not been achieved previously.


The genus Cucumis (family Cucurbitaceae) contains 52 species (Schaefer, 2007; Sebastian et al., 2010), including two important vegetable crops, cucumber (C. sativus L., 2= 2= 14) and melon (C. melo L., 2= 2= 24). Both species are of Asian origin, and diverged from a common ancestor approximately 10 million years ago (Sebastian et al., 2010) (Figure 1). Among the approximately 40 Cucumis species for which chromosome counts have been obtained, C. sativus is the only species with 2= 14 chromosomes; all other species have 2= 24, and occasionally 2= 48 or 72 chromosomes (Kirkbride, 1993). The origin of the seven cucumber chromosomes has been a long-standing question, with opposite hypotheses being proposed to explain it: a fragmentation hypothesis that postulated de novo regeneration of centromeres from = 7 to = 12 (Kozhukhow, 1930; Whitaker, 1933; Bhaduri and Bose, 1947) and a fusion hypothesis that postulated that = 7 was derived from = 12 via unequal translocation or fusion of non-homologous chromosomes (Trivedi and Roy, 1970). More recently, Han et al. (2009) proposed the involvement of centromere repositioning in evolution of cultivated cucumber chromosome C7 based on cucumber/melon comparative FISH mapping data. The results of molecular phylogenetic studies suggest that = 12 is ancestral in the genus Cucumis (Ghebretinsae et al., 2007; Renner et al., 2007; Sebastian et al., 2010).

Figure 1.

Chronogram of selected Cucurbitaceae species used in the present study. The estimations of divergence (in millions of years ago) are from Sebastian et al. (2010).

Despite their distant phylogenetic relationship and sexual incompatibility, the genome sequences of melon and cucumber appear to be highly conserved. The cross-species transferability of molecular markers between cucumber and melon is approximately 50% (e.g. Neuhausen, 1992; Katzir et al., 1996; Danin-Poleg et al., 2000; Park et al., 2004; Gonzalez et al., 2010). The genome size of melon is approximately 425 Mb and that of cucumber is approximately 367 Mb, but the number of protein-coding genes in both genomes is similar (Li et al., 2011b; Garcia-Mas et al., 2012), and the size difference is believed to be mainly due to expansion of intergenic regions and proliferation of transposable elements in the melon genome (Gonzalez et al., 2010; Garcia-Mas et al., 2012). Comparative genetic mapping and whole-genome alignment have enabled establishment of syntenic relationships of 12 melon and seven cucumber chromosomes, showing that six of the seven chromosomes arose from fusions and intra-chromosome rearrangements, while one has remained largely unchanged between cucumber and melon (Huang et al., 2009; Li et al., 2011a; Garcia-Mas et al., 2012).

Inferring the history of chromosome reduction in Cucumis based on comparative studies between melon and cucumber only is insufficient for a detailed understanding because at least 24 other species fall between these two species in the most complete phylogeny of the genus (Sebastian et al., 2010). Of these 24 species, the one closest to C. sativus is C. hystrix (2= 2= 24), from which it diverged approximately 4.6 million years ago (Sebastian et al., 2010). Inter-specific hybrids between C. hystrix and C. sativus have been generated by embryo rescue (Chen et al., 1997). Its position as the sister species of C. sativus confers a critical role for C. hystrix in understanding the process of chromosome reduction in Cucumis, and the family-wide phylogenies now available also establish the distance of C. sativus and C. hystrix to C. melo and other species in Cucumis and related genera.

Studying chromosome synteny and colinearity among species for which a phylogeny establishes the direction of evolution allows reconstruction of the likely events that led to present karyotypes. Events of particular interest are reductions in chromosome number (dysploidy), which have been analyzed using this approach in Brassicaceae (Koch and Kiefer, 2005; Yogeeswaran et al., 2005; Lysak et al., 2006; Mandakova and Lysak, 2008; Cheng et al., 2013), Poaceae (Luo et al., 2009) and Rosaceae (Vilanova et al., 2008; Illa et al., 2011; Jung et al., 2012). Here we performed whole-genome sequencing and de novo genome assembly of two C. hystrix accessions to develop a high-density C. hystrix linkage map, which was then used to establish genome-wide chromosome synteny and marker colinearity by alignment of shared markers to the cucumber and melon draft genome assemblies. The melon and cucumber synteny was refined through whole-genome alignment. We also performed large-scale comparative fluorescence in situ hybridization (FISH) mapping among the melon, cucumber and C. hystrix genomes to validate or reveal syntenic blocks or structural arrangements. The previously inferred conservation and retention of cucumber chromosome 7 (C7) during evolution was validated among several Cucumis species, and micro-synteny analyses and annotation of an 8 Mb region in C7 was performed for cucumber, melon and watermelon (Citrullus lanatus, 2= 2= 22). The resulting data shed light on the mechanisms of dysploid chromosome reduction that led from an = 12 ancestor to the = 7 karyotype of cucumber.


Next-generation sequencing and de novo assembly of the C. hystrix genome

We sequenced two accessions of C. hystrix, WI7001 and WI7002, using both Roche/454 GS FLX Titanium and Illumina HiSeq 2000 technologies (Table 1). The 454 pyrosequencing generated 651.8 and 418.0 Mb high-quality reads from WI7001 and WI7002, respectively. From approximately 252 million raw HiSeq 2000 reads, we obtained 5218 Mb WI7001 and 4432 Mb WI7002 sequences. Hybrid assembly of Illumina contig sequences and Roche/454 paired-end reads resulted in 11 649 scaffolds containing total 209 Mb sequences, with an N50 scaffold size of 50 831 bp and the largest scaffold of 560 870 bp (N50 is a weighted median statistic such that 50% of the entire assembly is contained in a scaffolds equal to or larger than this value) (Table 2).

Table 1. Summary of data for whole-genome sequencing of the C. hystrix genome
C. hystrix accessionsNext-generation sequencing methodNumber of raw readsTotal length (bp)Mean length (bp)Total length of clean reads (Mbp)
WI7001Illumina HiSeq 2000135 870 7549 782 694 288725218
454 GS FLX Titanium2 034 004651 804 005320652
WI7002Illumina HiSeq 2000117 104 4648 431 521 408724432
454 GS FLX Titanium1 192 482418 001 825351418
Total 256 201 70419 284 021 526 10 720
Table 2. Summary statistics of C. hystrix genome assembly
AssemblyTotal numberTotal bases (bp)Mean size (bp)Maximum size (bp)Minimum size (bp)N50
Scaffolds11 649209 044 45217 945560 870182550 831
Large contigs124 535210 428 975168921 9325002249
All contigs200 296232 662 438116221 9321002012

To assess the quality of this C. hystrix draft assembly, the 40 largest scaffolds were blast-aligned against the Gy14 cucumber and DHL92 melon genome scaffold assemblies (Table S1). Of the 40 scaffolds, 38 showed a high degree of sequence homology with the scaffolds of cucumber or melon or both. This C. hystrix draft genome assembly provides insights into micro-synteny among the C. melo, C. hystrix and C. sativus genomes. Of the approximately 12.2 Mb sequences in the 40 scaffolds, 11.5 Mb aligned to 13.1 Mb of melon scaffold sequences, indicating that the melon genome is approximately 13.9% larger than that of C. hystrix in these regions. Meanwhile, the syntenic regions in the cucumber genome in the 11.7 Mb C. hystrix scaffold regions were slightly longer (by 543 kb) (Table S1), suggesting that the C. hystrix genome may not be significantly larger than the cucumber genome.

Among the 40 C. hystrix scaffolds, ten had one melon/two cucumber or one cucumber/two melon alignment patterns (Table S1). The sequences of these ten C. hystrix scaffolds could be used to improve the cucumber or melon draft genome assemblies by estimating the gap size between two adjacent scaffolds or anchoring scaffolds that had not been placed to the cucumber or melon draft genome assembly. Conversely, the melon and cucumber scaffold assemblies can be used to anchor more C. hystrix scaffolds for draft genome assembly.

Development of an SSR-based C. hystrix genetic map

For genetic map construction, we screened 2826 cucumber and 1004 melon microsatellite markers (SSRs), and identified 185 (7%) and 63 (6%) polymorphic SSRs, respectively, between WI7001 and WI7002. Polymorphic SSRs were also identified in silico between the two C. hystrix contig assemblies. We first performed genome-wide identification of microsatellite sequences in WI7001 and WI7002 contig assemblies. In 128 257 and 117 711 SSRs identified, respectively, from WI7001 and WI7002, 373 were selected based on in silico polymorphism between the two parents and maximum sequence alignment with the Gy14 cucumber genome. The polymorphism of 312 (84%) SSRs was empirically validated. In contrast, among 89 randomly selected SSRs, only 28 (33%) were empirically polymorphic, suggesting that our high-throughput whole-genome screening strategy based on in silico polymorphisms was highly efficient with respect to time and cost for linkage map development.

The resulting C. hystrix genetic map contained 416 markers (see Table S2 for details), of which 215, 151 and 50 were derived from C. hystrix, cucumber and melon, respectively. Using a logarithm of odds (LOD) score of 7.0, the 416 markers were grouped into 12 linkage groups, designated H01–H12 based on their synteny with the respective melon chromosomes (I to XII). Key statistics for this genetic map are presented in Table 3 and Figure S1. The 1001.5 cM map length was comparable to two previous melon genetic maps (Diaz et al., 2011; Li et al., 2011a), suggesting near-complete coverage of the C. hystrix genome.

Table 3. Information for the C. hystrix linkage map and the syntenic relationships with cucumber and melon
Linkage group (chromosome)Number of lociMap length (cM)Map interval (cM)Number of markers anchoredSynteny to melonSynteny to cucumber
H024797.52.13633II + VIIIC1 + C6
H033584.92.43020IIIC2 + C6
H054780.41.73734VC2 + C4 + C6
H082456.12.3178II + VIIIC1 + C4 + C6
H113989.22.33021XIC2 + C6

Synteny and chromosome rearrangements between the C. hystrix and C. melo genomes

Primer sequences of the 416 markers on the C. hystrix genetic map were used as queries for in silico PCR or blast searches with the DHL92 melon assembly as the template, and 261 (75%) had in silico PCR amplicons or blast alignments covering approximately 294 Mb (93%) of the DHL92 melon genome. The alignment of the 261 shared markers between the 12 C. hystrix and corresponding melon chromosomes is shown in Figure 2(a). The physical locations of these markers in the DHL92 melon assembly are shown in Table S3.

Figure 2.

Syntenic relationships of Cucumis hystrix (H01–H12) with melon (I–XII) (a) and cucumber (C1–C7) (b) chromosomes. Synteny between C. hystrix and melon is based on alignment of 262 shared markers on the C. hystrix genetic map and the DHL92 melon draft genome assembly (a); synteny between C. hystrix and cucumber is based on alignment of 348 shared markers on the C. hystrix genetic map and the Gy14 cucumber draft genome assembly. Syntenic blocks are connected by lines (shared markers) with the same color.

Ten of the 12 C. hystrix chromosomes were highly syntenic with melon, thus retaining the ancestral condition. In C. hystrix chromosome H02, the region from 53.1–97.5 cM (12 marker loci) was syntenic to melon chromosome II, whereas the 0–42.8 cM block (21 loci) was syntenic to approximately 7.2 Mb in the distal (lower) end of melon chromosome VIII (Table S3). Seven of the eight markers in H08 were shared with chromosome VIII, but one (NR39) was located in a syntenic region of melon chromosome II, indicating a reciprocal translocation between H02 and H08, as supported by sequence alignment of C. hystrix scaffolds anchored to H02 and H08 with the DHL92 assembly (Figure 3a).

Figure 3.

Cucumis hystrix-specific reciprocal translocation. (a) The H02–H08 translocation in relation to melon chromosome II (MII) and chromosome VIII (M VIII) revealed from sequence alignment of C. hystrix scaffolds and the DHL92 melon genome assembly. (b) Comparative genetic mapping between C. hystrix and melon identified the H02–H08 translocation (A). Two fosmid clones 11.9 (green) and 12.3 (red) were used to probe mitotic (B–E) or meiotic pachytene (F) chromosomes of five Cucumis species (B–F).

Marker colinearity between the C. hystrix and C. melo genomes was inconsistent in many syntenic blocks (Table S3), which may be due to chromosome rearrangements between the two species or to errors in genetic mapping or whole-genome assembly. We performed large-scale comparative pachytene FISH mapping among C. sativus, C. hystrix and C. melo to assess the structural rearrangements. Molecular markers mapped to seven cucumber chromosomes were used to screen a cucumber fosmid library to identify positive clones as FISH probes. We identified 128 fosmid clones (Table S4), all of which were unambiguously mapped to the seven cucumber chromosomes. Among 122 clones tested in C. hystrix, all resulted in single and excellent FISH signals; of 119 tested in C. melo, all but three gave good FISH signals (see Figure S2 for examples). The results from comparative FISH mapping of 12 C. hystrix and melon chromosomes are shown in Figure 4, and revealed at least 14 inversions between the two genomes, five of which appear to be nested within additional inversions. All fosmid loci in chromosomes H09 and H10 were colinear with those in melon chromosomes IX and X, respectively, suggesting that the two chromosomes were conserved during evolution.

Figure 4.

Syntenic relationships between Cucumis hystrix and melon based on comparative genetic and pachytene FISH mapping. Syntenic chromosomes are arranged side by side. Black ovals indicate putative centromere locations.

The H02–H08 reciprocal translocation between C. hystrix and C. melo was cross-validated with comparative FISH in five additional species representing the Cucumis phylogenetic tree, namely the Asian species C. debilis, the African species C. metuliferus, C. ficifolius and C. zeyheri, and the African/Pakistani species C. prophetarum, as well as C. melo and C. sativus. Two fosmid probes, 11.9 and 12.3, which mapped on cucumber chromosome C1 and melon chromosome II (Figure 4), were used in FISH of somatic chromosomes of these species. We found that both fosmid loci were located on the same chromosome in all species except C. hystrix (Figure 3b), suggesting that this reciprocal translocation is specific to C. hystrix. Both fosmid probes gave signals in cucumber C1, which appears to be the result of a fusion between C. hystrix chromosomes H02/H08 and H12 (see below).

Synteny and chromosome rearrangements between the C. hystrix and C. sativus genomes

Of the 416 markers on the C. hystrix genetic map, 348 (84%) aligned to the C. sativus genome, covering 95% (181/191 Mb) of the Gy14 draft genome assembly. Based on the chromosome synteny inferred from this alignment (Figure 2b) and the physical locations of these markers in the Gy14 genome (Table S5), the syntenic relationships between C. hystrix and cucumber chromosomes may be expressed as follows: H01 = C7, H02 = C1 + C6, H03 = C2 + C6, H04 = C3, H05 = C2 + C4 + C6, H06 = C3, H07 =C4, H08 = C1 + C4 + C6, H09 = C5, H10 = C5, H11 = C2 + C6 and H12 = C1 (Table 3).

To characterize the syntenic relationships between the two species in more detail, we divided the 12 C. hystrix chromosomes into 53 syntenic blocks. Each block was defined as a region on the C. hystrix genetic map that aligned with a continuous stretch of DNA sequences in the Gy14 genome and was anchored by at least one shared marker or fosmid position. The orientation of each syntenic block in relation to the Gy14 genome assembly was determined by the order of shared markers on the C. hystrix genetic map and in the Gy14 assembly, and further verified by pachytene FISH. Of the 53 syntenic blocks, 20 were colinear, and 25 contained inversions between C. hystrix and cucumber. The orientation of eight blocks remains unknown (Table S5). The resulting accurate view of the seven cucumber chromosomes in terms of their synteny with C. hystrix (Figure 5) may be simplified as follows: C1 = H02/H08 + H12, C2 = H03 + H05 + H11, C3 = H04 + H06, C4 = H7 + H8 + H05, C5 = H09 + H10, C6 = H03 + H11 + H08/H02 + H05 and C7 = H01. Chromosome C7 shows complete synteny with H01, while C1, C3 and C5 each appear to be the result of fusion of two C. hystrix chromosomes, and C2, C4 and C6 each contain syntenic blocks from more than two C. hystrix chromosomes. The arrangement of C. hystrix syntenic blocks in all but the C7 cucumber chromosome is clearly far more evolutionarily complicated than a simple fusion of two C. hystrix chromosomes.

Figure 5.

Cucumis hystrix (left) and melon (right) syntenic block views of cucumber chromosomes C1– C7 (center). Syntenic blocks that originated from the same C. hystrix chromosomes (1–12) or melon chromosomes (I–XII) have the same color. C. hystrix syntenic block boundaries are indicated by white lines or centromeres, and anchored by fosmid loci. The numbers to the left of each syntenic block are fosmid clones.

Synteny and chromosome rearrangements between the C. sativus and C. melo genomes

To refine the cucumber/melon synteny, we divided the DHL92 melon genome assembly into 91 syntenic blocks (Table S6), each >500 kb in size and exhibiting sequence homology with the Gy14 cucumber draft genome. Among the 91 blocks, 35 were anchored with two or more fosmid loci, providing additional evidence on the orientation of these syntenic blocks. Of the 91 syntenic blocks, 45 had at least one inversion between cucumber and melon. Melon chromosome V is known to be syntenic to cucumber chromosome C2 (Li et al., 2011a; Garcia-Mas et al., 2012). Sequence alignment identified two new syntenic blocks of melon chromosome V in cucumber chromosomes C4 and C6 (Figure S3), which was confirmed with pachytene FISH (Figure S2). A melon syntenic block view of cucumber chromosomes C1–C7 is shown in Figure 5. Except for the 14 C. hystrix -specific inversions and the H02/H08 translocation, the arrangements of C. melo and C. hystrix syntenic blocks in the cucumber genome were essentially the same.

Conservation of cucumber chromosome C7 with C. hystrix and C. melo

Comparative mapping and sequence alignment studies revealed a one-to-one whole chromosome synteny of cucumber chromosome C7 with melon chromosome I (MI) and C. hystrix H01. To verify this conserved synteny, we performed comparative FISH mapping using 14 fosmid probes located in C7 (71.1–72.4, Figure 5) in cultivated cucumber (C. sativus var. sativus), wild cucumber (C. sativus var. hardwickii, 2= 2= 14, the progenitor species of cultivated cucumber; Yang et al., 2012), C. hystrix (Figure 6a) and melon (C. melo). All 14 probes detected single hybridization signals in these species except for probe 72.0, which gave no signal in melon. The large paracentric inversion spanning six fosmid loci (71.1–71.6) in C7 is specific to C. sativus (Yang et al., 2012). The short arms of wild cucumber, C. hystrix and C. melo (defined by fosmid loci 71.1–71.8) were differentiated by one or two inversions.

Figure 6.

Synteny of cucumber chromosome 7 (C7) with other cucurbit species. (a) Pachytene FISH analysis was performed with 14 fosmid probes in three Cucumis species (CSS, C. sativus var. sativus; CSH, C. sativus var. hardwickii). CEN indicates the putative centromere location. (b) Alignment of draft genome sequences of C7, melon chromosome I and watermelon chromosomes W2 and W9 showing a high degree of sequence homology in this region.

The six fosmid loci (71.9–72.4) on the long arm showed complete colinearity across the three species (Figure 6a), suggesting a high level of conservation of this region in Cucumis. The region comprised approximately 8.0 and 13.0 Mb in the Gy14 cucumber (Yang et al., 2012) and DHL92 melon (Garcia-Mas et al., 2012) draft genomes, respectively. To investigate the scope of conservation of this region in more distant cucurbit species, we extracted sequences in the syntenic block of this region from the watermelon draft genome (Guo et al., 2013). This block is approximately 8.4 Mb in watermelon chromosome W2. Sequence alignment confirmed a high level of micro-synteny across the three species (Figure 6b and Figure S4). We also annotated genes, DNA transposons and retrotransposons of this region in the three assembled genomes as well as the C. hystrix scaffolds anchored to the region. Although 70% of genes in this region were conserved among the three species, the number of transposons in melon compared with cucumber and watermelon has increased 3.6 or 2.5 times, respectively (Table 4). Consistent with this, the gene density in melon was lower than that in cucumber, C. hystrix or watermelon. However, the distribution of mobile elements across the region was not uniform. While the distal telomeric region appeared to be conserved, with approximately the same size in the three genomes, there was notable expansion in the proximal centromeric region in melon, with a higher frequency of mobile elements (Figure S4). In watermelon, distribution of genes and transposons or retrotransposons was relatively uniform across this region, while the highly conserved regions were gene-dense in both melon and cucumber, with the region corresponding to the expansion in melon also being transposon-dense (Figure S4).

Table 4. Summary of genes, DNA transposons and retotransposons annotated in a highly conserved syntenic region across four cucurbit species
 MelonCucumber C. hystrix Watermelon
  1. ‘–’ indicates that no estimation was made because the contiguous C. hystrix genome sequences were too short.

Length of syntenic block (Mbp)
Mean (number of genes/Mb)83.3105.495.5106.9
Mean (kb/gene)12.09.510.59.5
DNA transposons
Mean (number/Mb)44.822.324.0
gypsy 5907935
copia 4035657
Mean (number/Mb)95.240.460.7
Grand total1821501712
Overall mean (number/Mb)140.162.684.8


Major chromosome rearrangements among C. melo, C. hystrix and C. sativus

The present study highlights the synergy from integrated use of high-throughput genome sequencing and assembly, comparative genetic and physical mapping, and whole-genome alignment to understand karyotype evolution in non-model plant species (Rocchi et al., 2006; Schubert and Lysak, 2011). In particular, using next-generation sequencing data, we developed a quick method of in silico polymorphism screening for development of a genetic map in C. hystrix, for which no available genetic or genomics resource was available. This cost-effective and highly efficient approach should have wide applicability in genetic mapping studies.

Comparative genetic and physical mapping revealed a high level of synteny among chromosomes of cucumber, C. hystrix and melon. However, structural rearrangements appear to be common within syntenic blocks among the three species, with inversions as the predominant event. For example, whole-genome pachytene FISH revealed 14 inversions between C. hystrix and melon (Figure 4), and 20 of the 53 syntenic blocks (47%) between C. hystrix and cucumber also contained inversions (Table S5). We previously identified six inversions differentiating wild (C. sativus var. hardwickii) and cultivated (C. sativus var. sativus) cucumbers (Yang et al., 2012), and these appear to correspond well with C. hystrix syntenic blocks 7B1 and 7B5 (C4), 9B2, 9B4 and 10B2 (C5), and 1B1 (C7) in the present study (Figure 5). Comparison of the orientations of these six syntenic blocks in C. melo, C. hystrix and C. sativus suggests that four inversions, 7B5 in C4, 9B2 and 9B4 in C5, and 1B1 in C7, may have occurred during domestication of cucumber (Figure S5). Furthermore, genome alignment and comparative FISH identified 45 of the 91 melon syntenic blocks (50%) as having at least one inversion in each syntenic block compared with cucumber, covering almost half of the cucumber and melon genomes (Table S6). The predominance of inversions found here matches findings from Rosaceae (Vilanova et al., 2008), Solanaceae (Livingstone et al., 1999; Wu and Tanksley, 2010) and Brassicaceae (Lagercrantz, 1998; Yogeeswaran et al., 2005). Inversions are believed to play an important role in karyotype evolution, speciation and local adaptation by reducing recombination and thereby protecting genomic regions from introgression (Hoffmann and Rieseberg, 2008; Kirkpatrick, 2010; Lowry and Willis, 2010; Ruiz-Herrera et al., 2012).

We also identified a reciprocal translocation involving C. hystrix chromosomes H02–H08 that sets this species and C. sativus apart from the other species analyzed (Figure 3). Comparative FISH in C. debilis, C. prophetarum, C. metuliferus, C. zeyheri and C. ficifolius suggests that this translocation may have arisen in the common ancestor of C. hystrix and C. sativus (Figure 1).

Retention of cucumber chromosome C7 during Cucumis evolution

We confirmed the conservation of cucumber chromosome C7 with melon chromosome I and extended it to C. hystrix H01 (Figure 6a). Chromosome C7 is also syntenic to chromosomes W2 and W9 of watermelon (Huang et al., 2009; Guo et al., 2013) (Figure 6b). However, we found that the degree of conservation varied in the long and short arms of C7 among C. sativus, C. hystrix, C. melo and watermelon. Their short arms were differentiated by one or two inversions, probably due to their high content of heterochromatin as indicated by DAPI-stained dark blue blocks in pachytene chromosomes (Figure 6a). By contrast, the long arms in these species showed complete colinearity that spanned 8.0, 8.4 and 13.0 Mb in the cucumber, watermelon, and melon draft genomes, with high levels of micro-synteny and colinearity of genes despite the genome size differences (Figure S4 and Table 4). In the two Cucumis species, genes and transposable elements were clustered, with higher transposable element density in the pericentromeric region and higher gene density in the distal telomeric region; by contrast, in watermelon, the distribution of transposable elements and genes was relatively uniform (Figure S4).

Inferred mechanisms of dysploid reduction from n = 12 to n = 7 in Cucumis

The evolution from an ancestral karyotype may be modeled through synteny-based comparative analysis in extant species (Abrouk et al., 2010; Schubert and Lysak, 2011; Salse, 2012). Geographically, the range of C. hystrix [found in Myanmar (Burma), North and West Thailand and Southwest China] overlaps that of wild cucumber (C. sativus var. hardwickii) (Sebastian et al., 2010), and phylogenetic studies have shown that C. hystrix is the sister species to C. sativus (Ghebretinsae et al., 2007; Renner et al., 2007; Sebastian et al., 2010). It is reasonable to assume that the common ancestor of C. sativus and C. hystrix had 12 chromosomes, AK1–12, with a structure that is similar, but not identical, to chromosomes H01–H12 of C. hystrix. Therefore, we modeled the ancestral karyotype of cucumber using both C. melo and C. hystrix, and reconstructed the history of karyotype changes from this = 12 ancestor to = 7 in cucumber. The main events during this process are summarized in Figure 7, and details are presented in Figure S5.

Figure 7.

Reconstruction of karyotype evolution from the = 12 ancestor to the = 7 wild cucumber (CSH). The 12 ancestral chromosomes (AK1–AK12) are divided into 53 syntenic blocks that are the same as in Cucumis hystrix. The origin of each syntenic block may be tracked by its name and color. T, translocation; I, inversion; F, fusion. The + or − symbol within each block represents its orientation (+, colinear; −, inversion) between C. hystrix and cucumber. Black ovals indicate centromeres as inferred from pachytene FISH.

Cucumber chromosome C7

C7 is highly conserved with chromosome I of melon and H02 of C. hystrix (Figure 6). C7 differentiated from AK1 by only one pericentric inversion in syntenic block 1B2. Another paracentric inversion in syntenic block 1B1 occurred during cucumber domestication (Figure S5A).

Cucumber chromosomes C1, C3 and C5

A common theme in the evolution of these three cucumber chromosomes is that each derives from the fusion of two ancestral chromosomes. For C1, this process was probably initiated by a reciprocal translocation between AK2 and AK8 producing two translocated chromosomes (AK2/AK8 and AK8/AK2). Then, AK12 was inserted into the centromeric region of AK2/AK8 followed by two pericentric and two paracentric inversions to form chromosome C1 (Figure S5B). For C3, a pericentric inversion occurred in both AK4 and AK6, creating one telocentric and one acrocentric chromosome, which then underwent a fusion event (Robertsonian-like translocation). This translocation generated a product consisting of telomeric and AK6 centromeric fragments, which was probably unstable during meiosis and therefore eliminated. The fused chromosome went through two para- and two pericentric inversions leading to C3 (Figure S5C), which is the longest cucumber chromosome. Similar to C3, the origin of C5 involved fusion of AK9 and AK10, followed by nine inversions. C5 of cultivated and wild cucumbers then evolved via three more inversions during domestication (Figure S5D).

Cucumber chromosomes C2, C4 and C6

The evolution of these three chromosomes appears to be more complex because each contains blocks that are syntenic to at least three ancestral chromosomes. Evolution may have started with a reciprocal translocation between AK5 and AK8/AK2, resulting in two intermediates (AK8/AK5 and AK5/AK8/AK2) (Figure S5E). Two translocations between AK7 and AK8/AK5 and a series of inversions created C04 of wild cucumber (Figure S5F). After one paracentric inversion, a reciprocal translocation between AK5/AK8/AK2 and AK11 may have resulted in two new intermediates, each of which recombined with a segment from AK3 leading to C02 (by translocation) and C06, respectively (Figure S5E).

This model of dysploid reduction explains the location and orientation of the 53 C. hystrix syntenic blocks in the seven cucumber chromosomes identified herein (Figure 5). The 59 chromosome rearrangement events included five fusions, four translocations and 50 inversions (Table 5). Assuming all fusion and translocations were reciprocal translocations, up to 114 breakpoints may explain the 59 rearrangement events. The majority of these breakpoints were presumably located in heterochromatic regions including centromeric, telomeric, pericentromeric or nucleolus organizer (NOR) regions. Indeed, almost half of the breakpoints (53/114) involve centromeric or telomeric positions (Table 5) that are rich in repetitive sequences (e.g. Figure 6 and Figure S2). For example, C. hystrix chromosomes H08 (AK8) and H10 (AK10) are acrocentric, and H08, H10 and H12 (AK12) have NORs in the telomeric ends (Figure 4). It makes sense, therefore, that AK08, AK10 and AK12 were involved in three of the five fusion events forming cucumber chromosomes C01, C05 and C04 (Figure S5B,D,F).

Table 5. Summary of cytogenetic changes during dysploid reduction from the = 12 ancestor to = 7 wild cucumber
Cucumber chromosomeAncestor donor chromosomeaNumber of breakpointscNumber of translocationsdNumber of inversionsNumber of fusionsAK centromere inherited by cucumber chromosome
  1. a

    The ancestral karyotype is similar to or the same as the C. hystrix karyotype. AK1–AK12 correspond to H01–H12, respectively, in terms of chromosome structure and organization.

  2. b

    t(AK2/AK8) = result of the chromosome translocation between AK2 and AK8 that preceded all other events.

  3. c

    Breakpoint counts assume all that fusions are due to reciprocal translocations. The numbers in parentheses indicate breakpoints that occurred in centromeric/telomeric regions.

  4. d

    Reciprocal translocations excluding those mediating chromosome fusions.

 t(AK2/AK8)b2 (0)10  
C1t(AK2/AK8), AK1211 (6)041AK12
C3AK4, AK615 (8)071AK4
C5AK9, AK1019 (7)0101AK9
C7AK12 (1)010AK1
C2AK3, AK5, AK1165 (31)3280AK5
C4AK5, AK7, t(AK8/AK2)   1AK7 or AK8
C6AK3, AK5, t(AK2/AK8), AK11   1AK11
Sum 114 (53)4505 

The mechanism of formation of cucumber C1 through nested chromosome fusion revealed here resembles the situation in some grasses (Luo et al., 2009), in which centromeric or telomeric illegitimate recombination between non-homoeologous chromosomes has been demonstrated (Murat et al., 2010). In Arabidopsis, chromosome rearrangements leading to descending dysploidy from an ancestral = 8 species to A. thaliana (= 5) also involved inversions, fusions and translocations (Lysak et al., 2006). Given the constraints on chromosome structure, most or all chromosome fusions are probably mediated by reciprocal translocations with or without preceding para- and pericentric inversions, resulting in a fusion chromosome and a meiotically unstable (a)centric fragment that is eliminated (Schubert and Lysak, 2011). Therefore, in addition to the expansion of intergenic regions and proliferation of transposable elements in the melon genome (Gonzalez et al., 2010; Garcia-Mas et al., 2012), the size difference between the cucumber and melon genomes may be partly explained by loss of these (a)centric fragments.

The centromere is the point of spindle attachment in cell division, and is essential for faithful segregation of chromosomes. The insertion of an ancestral chromosome AK12 into the centromeric region of translocated AK2/AK8 (Figure S5B) probably involved inactivation of the AK8 centromere, and the eight additional translocations or fusions inferred here probably led to loss of centromeres from AK2, AK3, AK6, AK7 (or AK8) and AK10, with cucumber chromosomes C1–C7 instead inheriting their centromeres from the remaining seven ancestral chromosomes (Table 5). Such weak constraints on the centromeres fit with speculations that centromeres are activated and maintained primarily by epigenetic mechanisms (Karpen and Allshire, 1997). Cucumber chromosome C7 has remained largely intact in the evolution of Cucumis (Figure 6). The centromere positions in C7, melon chromosome MI and C. hystrix H01 (Figure 6) appear to be consistent with a centromere repositioning event as proposed by Han et al. (2009). On the other hand, the different centromere positions between cucumber chromosome C6 and melon chromosome VIII (MeI in Han et al., 2009) may be the result of multiple rearrangements that may have occurred during the complex evolutionary history of C6 (Figure S5E).

Among the three Cucumis species, C. hystrix and melon diverged approximately 10 million years ago, and are differentiated by 14 inversions and a reciprocal translocation (Figure 4); the subsequent dysploid reduction from = 12 to = 7 in the ancestor of C. sativus then occurred at some time after that species diverged from C. hystrix approximately 4.6 million years ago (Sebastian et al., 2010). The factors that caused the accelerated karyotype change in the C. sativus lineage remain unknown. It would also be interesting to know why cucumber C7 remained largely unchanged during at least 12 million years ago of evolution. However, the syntenic relationships revealed here provide a fresh perspective in understanding chromosome evolution in Cucumis, and the newly sequenced genome of C. hystrix adds further information on a wild relative of cucumber.

Experimental Procedures

Plant materials

An F2 mapping population of 91 F2 plants was developed for construction of a linkage map in C. hystrix, which was derived from a cross between two accessions WI7001and WI7002 originally collected from Thailand and Southwest China, respectively. Additional species in the genus Cucumis used in comparative cytological analysis included inbred or plant introduction (PI) lines of cultivated cucumber (C. sativus var. sativus L., PI 249561), wild cucumber (C. sativus var. hardwickii Alef., PI 183967), melon (C. melo L., Top Mark), C. debilis W.J. de Wilde & Duyfjes, C. prophetarum L. (PI 193967), C. metuliferus E. Mey. ex Naudin (PI 482443), C. zeyheri Sond. (PI 364472) and C. ficifolius A. Rich. (PI 299570). Seeds of all PI lines were obtained from the US National Plant Germplasm System (Ames, IA, USA).

C. hystrix whole-genome sequencing and assembly

Sequencing of the genomes of WI7001 and WI7002 using both the Roche/454 (Branford, CT, USA) and Illumina systems (San Diego, CA, USA) was performed in the Biotechnology Center of the University of Wisconsin at Madison. For the 454 GS FLX Titanium pyrosequencing, 8 kb paired-end libraries were used; whole-genome shotgun sequencing was performed on the Illumina's HiSeq 2000 system. For the HiSeq 2000 raw reads, sequences with quality scores <28 and length <41 were removed to obtain high-quality reads. A hybrid assembly strategy was used to integrate all sequences from both platforms. First, the trimmed Illumina reads of WI7001 and WI7002 were assembled into contigs using the ABySS assembler (version 1.3.0) (Simpson et al., 2009) with k–mer size = 51. Then, for all contigs larger than 1 kb in length, each contig was chopped into 1 kb fragments, with neighboring ones having an 800 bp overlap. These artificial reads were combined with approximately 3.2 million 454 pair-end reads using Newbler 2.6 in heterozygotic mode (Margulies et al., 2005), with a minimum overlap length of 50 bp and a minimum overlap identity of 95%. To assess the quality of the resulting genome assembly, the largest 40 C. hystrix scaffolds were aligned to the Gy14 cucumber and DHL92 melon draft genome assemblies using blastn (Altschul et al., 1990) (cut-off E value ≤10−100). Visual alignment of these scaffolds with the cucumber and melon draft genomes was performed using the progressiveMauve procedure in MAUVE version 2.3.1 (Darling et al., 2004).

Development of molecular markers and linkage map construction in C. hystrix

To construct a C. hystrix genetic map, we first performed polymorphism screening between WI7001 and WI7002 using 2826 cucumber and 1004 melon microsatellite (SSR) markers (Yang et al., 2012). We also performed in silico identification of polymorphic SSRs between the two parental lines from the Illumina genome sequences. First, all microsatellites with 2–8 bp repeat motifs were identified from both WI7001 and WI7002 HiSeq 2000 contig assemblies using MISA (Thiel et al., 2003). As SSRs with dinucleotide motifs and longer repeats of motifs tend to be more polymorphic (Cavagnaro et al., 2010), a subset of SSRs was selected from these identified SSRs, and primer pairs flanking these SSRs were designed using Primer3 software ( Next, virtual PCR was performed using WI7001-derived SSRs with the WI7002 contig assembly as the template, or vice versa, resulting in SSRs that were polymorphic between the two parental genomes. To facilitate comparative analysis, these selected SSRs were subjected to in silico PCR using the Gy14 cucumber draft genome assembly as the template. Finally, 373 SSRs (prefixed with CHXSSR in marker names) that were polymorphic between WI7001 and WI7002 and had amplicons in the cucumber genome were selected by empirical polymorphism screening and linkage mapping in the C. hystrix F2 population. However, several regions on the C. hystrix map lacked markers to detect or anchor syntenic blocks in the cucumber or melon draft genome. To fill these gaps, C. hystrix scaffolds or contigs that were orthologous to those melon or cucumber syntenic regions were identified by blast sequence alignment. SSRs were selected in these C. hystrix regions for in silico PCR between the WI7001 and WI7002 contig assemblies to assure that there were PCR products in both parents. Finally, 89 such SSRs were identified for empirical polymorphism screening.

For marker analysis, unexpanded young leaves from embryo culture-generated seedling plants were collected for DNA extraction using the CTAB (cetyl trimethylammonium bromide) method (Murray and Thompson, 1980). The PCR procedure and electrophoresis of the PCR products were performed as described by Yang et al. (2012). Linkage analysis was performed using JoinMap 3.0 software (van Ooijen and Voorrips, 2001). Linkage groups were determined using a minimum LOD score of 7.0. Genetic distance was calculated using the Kosambi mapping function (Kosambi, 1944). H01–H12 were assigned as the 12 C. hystrix linkage groups and chromosomes that were syntenic to melon chromosomes I–XII, respectively.

Establishment of chromosome synteny among the C. hystrix, C. sativus and C. melo genomes

The C. hystrix genetic map was used to infer syntenic relationships of C. hystrix with cucumber and melon chromosomes as described by Li et al. (2011a). Briefly, for each marker mapped on the C. hystrix linkage map, in silico PCR was implemented using the Gy14 cucumber and DHL92 melon draft genome assemblies as templates to align this marker and associated scaffold or contig on the respective cucumber and melon chromosome. This was performed using a custom Perl script that used the National Center for Biotechnology Information NCBI, ( blastn program as the search engine (Cavagnaro et al., 2010). The order and physical locations of shared markers were used to detect macrosynteny and rearrangements among chromosomes. The resulting syntenic relationships were visualized using the circos program (Krzywinski et al., 2009). For comparison of the synteny between C. hystrix and cucumber chromosomes, each C. hystrix chromosome was divided into 3–7 syntenic blocks, each of which was a discrete linkage block on the C. hystrix genetic map occupied by several marker loci that had continuous DNA sequence matches in the Gy14 cucumber genome. These syntenic regions were the building blocks for inferring the evolutionary history of dysploid reduction from the = 12 ancestor to = 7 cucumber genomes.

The Gy14 cucumber and DHL92 melon draft genome sequences were used to refine the syntenic relationships identified from comparative genetic mapping. Whole-genome comparison was performed with Symap version 3.5 (Soderlund et al., 2011) using the nucmer algorithm of the MUMer package ( with parameter identity ≥90% and match length ≥1 kb. For simplicity, the melon genome assembly was divided into 91 syntenic blocks, where each block was defined as a region of ≥500 kb with continuous sequence matches in the Gy14 cucumber genome, without considering small gaps or micro-inversions that were common in genome sequence alignment.

Micro-synteny in a syntenic region across cucumber, melon and watermelon genomes

Our data suggested that cucumber chromosome C7 was largely conserved during chromosome evolution in several Cucurbitaceae species, including C. sativus, C. hystrix, C. melo and watermelon. To obtain insights into micro-synteny at the DNA sequence level, we performed detailed comparative analysis in a highly conserved region across the four species. This region in the long arm of C7 spanned 8 Mb (11–19 Mb) in the Gy14 draft genome, and the corresponding syntenic region in the DHL92 melon chromosome I (MI) comprised 13 Mbp (22–35 Mb in version cm3.5) (Garcia-Mas et al., 2012), and that in watermelon chromosome 9 (W9) comprised 8.4 Mbp (0–8.4 Mb) (Guo et al., 2013). Four scaffolds of the C. hystrix assembly of approximately 1.5 Mb were anchored to this syntenic block, which included scaffold00001, scaffold00005, scaffold00011 and scaffold00041. As the four C. hystrix scaffolds only accounted for approximately 12.5% of the 8 Mb cucumber syntenic region of cucumber (excluding Ns from paired-end sequencing), the four C. hystrix scaffolds were merged with a 100 kb spacer of Ns to avoid syntenic redundancy. The 8.4 Mbp watermelon sequences were reverse-complemented for easy plotting of the syntenic blocks.

We first aligned the sequences of this syntenic block from the four species using muscle version 3 (Edgar,2004) and progressivemauve using the non-colinear option (Darling et al., 2010). We next performed gene annotation for these sequences. Genome annotations for melon and watermelon in this region were downloaded directly from (watermelon) and (melon), respectively. The four C. hystrix scaffolds and the cucumber genomic region were annotated using the Eugene tool (Foissac et al., 2008) with Arabidopsis thaliana as a training set. We further performed transposon annotation for sequences of this region in cucumber, melon and watermelon. The transposon annotations for the melon and cucumber genome sequences were performed as described by Garcia-Mas et al. (2012). The ltr_finder program (Xu and Wang, 2007) was used to identify LTR retrotransposons de novo in the watermelon genome. The predicted elements were clustered (complete linkage, 80% similarity along 50% of sequence length), and a representative was taken from each cluster. The resulting set of sequences were further classified as gypsy or copia according to their similarity (tblastx, >80% identity along >200 bp) (Altschul et al., 1990) to known LTR retrotransposons in Repbase ( Copies of these elements in watermelon sequences were estimated using a copy finding program based on mite-hunter (Han and Wessler, 2010) and modified as described by Garcia-Mas et al. (2012), which annotates similar sequences and joins fragmented copies. As this procedure may fail to identify some ancient copies, we performed blast (tblastn, e >1e−10) using transposase and reverse transcriptase protein queries to identify DNA transposon- and retrotransposon-related sequences, respectively. There were therefore two levels of resolution of the transposon annotation: copies that were classified into super-families and families, and copies that were identified as transposon-related or retrotransposon-related.

Comparative fluorescence in situ hybridization (FISH) for cytological investigation of chromosome rearrangements

We performed large-scale comparative FISH using dual color-labeled (red and green) fosmid probes to investigate genomic rearrangements in cucumber, melon and C. hystrix. We first anchored fosmid loci onto the seven cucumber chromosomes. In a previous study, we identified 76 fosmid clones that detect chromosome differentiation between cultivated and wild cucumber chromosomes (Yang et al., 2012), all of which were used in the present study. Furthermore, 207 SSR markers at strategic map locations or target syntenic blocks were selected to screen the Straight 8 cucumber fosmid library (Yang et al., 2012), and 52 additional fosmid clones were identified. Fifty-two of the 53 C. hystrix syntenic blocks were anchored by at least one fosmid clone, which greatly increased the accuracy of alignment of syntenic blocks among the three species.

All FISH experiments were performed on meiotic pachytene chromosomes that were prepared from pollen mother cells, except for screening of the H02–H08 reciprocal translocation in several Cucumis species, for which mitotic metaphase chromosomes prepared from root tips cells were used. The karyotype of the cucumber chromosomes was first established using centromere- and telomere-specific repeat probes (Yang et al., 2012). Then all fosmid clones were anchored to the seven cucumber chromosomes. The physical order of adjacent fosmid clones in each chromosome was determined by two-color FISH. To improve the efficiency of performing large-scale comparative FISH, two strategies were used in the present study: multi-fosmid cocktail for probe preparation and multi-probing of the same pachytene chromosome preparation (up to eight times) (Cheng et al., 2001). The FISH procedure was performed as described by Koo et al. (2010).


This research was supported by US Department of Agriculture Current Research Information System Project 3655-21000-048-00D and a US Department of Agriculture Specialty Crop Research Initiative grant (project number 2011-51181-30661) to Y.W.