Plant materials

The sweetpotato SC1149-19 clone used as a hexaploid flow-cytometry reference, and for the 454-RNAseq and the Ion Torrent PGM sequencing of COSII-marker amplicons was obtained from the US-Vegetable Lab, SC. The sweetpotato ZhengHong 3 (ZH 3, PI 606266) and Jewel (PI 531122) clones used as hexaploid flow-cytometry references, Ipomoea littoralis Blume (PI 573335) for the flow-cytometry analysis and the 454-RNAseq, and Ipomoea tenuissima Choisy (PI 553012) as a diploid flow-cytometry reference, and for the 454-RNAseq and the Ion Torrent PGM sequencing of its COSII-marker amplicons were provided by the USDA-ARS Plant Genetic Resources Conservation Unit, Griffin, GA (https://npgsweb.ars-grin.gov/gringlobal/site.aspx?id=22). The sweetpotato breeding line LSP used for the Ion Torrent PGM sequencing of COSII-marker amplicons was an in-house hybrid mothered by the Jewel (PI 531122).

Flow cytometry analyses

Flow cytometry analyses of nuclear DNA contents in three Ipomoea species were performed essentially following the procedure and recommendations by Doležel [25]. Briefly, suspensions of intact nuclei were isolated by chopping fresh young leaf tissues in Galbraith's buffer following the one-step protocol, stained using DAPI (4’, 6-diamidino-2-phenylindole) and digested simultaneously with RNase as recommended. The relative fluorescence of the stained nuclei was measured by using a FACSCalibur^™ flow cytometer (BD Biosciences, CA, USA) following the protocols for measurement of relative nuclear DNA fluorescence intensity and ploidy analysis with an external standard. The stained nuclei of diploid I. tenuissima were used as an external reference standard, and its G1peak was positioned at channel 200. Three independent replica measurements were performed using tissues from different pot-grown plants. Prior to the measurements, the instrument was aligned and calibrated by following the manufacturer’s recommendations using the BD^™ DNA QC Particles (catalog No. 349523).

RNA extraction, transcriptome sequencing and datasets

Total RNA samples were prepared using a procedure optimized for high-starch and high-latex tissues of the tree Ipomoea species. Briefly, about 5 g of tissues were pulverized in liquid nitrogen, homogenized and lysed in a15 mL preheated (65°C) buffer [1M Tris-HCl (pH 7.4,10% SDS, 0.5 M EDTA, 20% PVP-40 and 0.01% β-Mercaptoethanol). After incubation at 65 ^oC for 15 min, the suspension was mixed with 0.1 volumes of 5 M potassium acetate and 0.25 volume of cold absolute ethanol, extracted with 1 volume of cold chloroform:isoamyl alcohol by vortexing for 2–3 min, and centrifuged at 7000g for 10 min at room temperature. The supernatants were extracted with equal-volume of Phenol: Chloroform:Isoamyl-Alcohol (50:50:1) once, cold Chloroform:Isoamyl-Alcohol (50:1) twice, and then mixed with concentrated LiCl solution to a final 3 M for overnight incubation at –20°C. Raw RNAs were recovered by centrifugation at 12,000g for 20 min, washed twice with 2 ml of 2 M LiCl solution, and dissolved in 1 ml RNase-free TE buffer [10 mM Tris-HCl (pH 7.5), 2 mM EDTA]. The RNAs were recovered using routine Sodium Acetate and Ethanol precipitation, and stored in TE buffer. Three RNA pools were made by combining equimolar amounts of 42 RNA samples isolated from leaf, vine, fibrous root and storage root tissues of sweetpotato SC1149-19 plants under various growth conditions (field, pots and hydroponic) and treatments (drought, cold & viral challenges), of 8 samples from leaf, vine and root tissues of I. littoralis plants at various growth stages, and of 9 RNA samples from leaf, vine, flower and root tissues of I. tenuissima plants. Custom transcriptome sequencing libraries from the three RNA pools were prepared, partially normalized and sequenced using 800 cycle runs on a Roche 454 GS-FLX Titanium instrument by the Center for Genomics and Bioinformatics of Indiana University. The sequencing reads were cleaned of adaptor sequences, and assembled into isotigs by the center. The numbers of assembled isotigs for the RNA pools from the sweetpotato SC1149-19, I. littoralis and I. tenuissima accessions were 53,468, 29,107 and 27,666, respectively.

A collection of 2152 tomato COSSII cDNAs was obtained from the Sol Genomics Network (https://solgenomics.net) based on the COSII member list [26] (https://www.sgn.cornell.edu/markers/cosii_markers.pl). The transcriptome dataset from the sweetpotato clone Tanzania [27] was downloaded from the International Potato Center (https://research.cip.cgiar.org/confluence/display/SPGI/Home). The two datasets of predicted cDNAs from two I. trifida lines (Mx23Hm and 0431–1) [28] were downloaded from the sweetpotato GARDEN (http://sweetpotato-garden.kazusa.or.jp). The dataset of ESTs from an Ipomoea nil variety was from Wei, et.al [29]. All the eight cDNA datasets were converted into searchable databases in Geneious (https://www.geneious.com) [30], and used for the phylogenetic analyses.

Each of the 2,152 tomato COSII cDNAs was used to BLAST-search its homologs in the other above-mentioned seven transcriptome databases from the five Ipomoea species using the Blastn program (Low complexity filter, Max E-value 1e-5, Match-mismatch scoring 2–3, and Gap-cost 52) within the Geneious suit. The numbers of hits from each of the searches, even those from the two sweetpotato datasets, were usually less than 10. All the hits from the searches using one tomato query sequence were aligned by using one of four multiple alignment programs (Geneious, MUSCLE, ClustalW and MAFFT) depending on overlapping lengths and numbers of the COSII cDNA homologs, and then locally refined using the Realign Region program within the Geneious suit to yield a set of overlapping COSSII cDNA homologs from the eight datasets. In an alignment, the cDNA variants of a COSII gene from the two sweetpotato datasets were manually screened for apparent assembly chimeras by cross-referencing haplotypes (sets of sequence variations) between the COSII cDNA variants from the same sweetpotato clone, and between matching ones from the two different sweetpotato clones. A COSII cDNA variant from one sweetpotato dataset is much less likely chimeric if its haplotype could be mostly matched in one from the other independently sequenced and assembled dataset from a different sweetpotato clone. On the other hand, a sweetpotato COSII cDNA variant was regarded as an apparent chimera, and thus removed from the alignment, if the cDNA variant displayed cross-overs of parts of two or more haplotypes of other cDNA variants from one or both of the sweetpotato datasets. Additionally, paralogous cDNA variants from multiple-copy COSII genes in the initial sets of alignments were also screened. If more than six cDNA variants having two or more distinct sets of sequence variations for a COSII gene in sweetpotato, or more than two types of cDNA variants for its homologs in the diploid I. tenuissima and/or I. trifida were identified in an alignment, these cDNA variants of the COSII gene were regarded as expressed paralogs, and the alignment excluded from further analysis. For rest of the alignments, a sub-alignment block encompassing the maximal overlapping regions for all the cDNA homologs were extracted, and used for downstream partition analyses. The COSII cDNA homologs from tomato and I. nil in the alignments were used as a reference and an outgroup for the successive hybridization network analyses, respectively.

Phylogenetic partition analyses

Stephens’s phylogenetic partition [24] is a statistical test for identifying non-random local spatial clustering of two or more variable sites carrying mutations common to only a partition of intra-or interspecific homologous DNA sequences among many others. It calculates the probability P(d≤ d_o) for the distance (d) between two or more partition-associated variable sites, if originated from randomly distributed mutations, is smaller or equal to the observed distance (d_o) between the sites in a sample (n nucleotides) of homologous DNA sequences. A non-random spatial clustering, when P(d≤ d_o) at 5% or less, of two or more variable sites having mutations shared by a particular partition of interspecific homologous DNA sequences predicts a homology of these mutations (rather than homoplasy) and their flanking regions, and thus closer relatedness of the homologs within the same partition than those in a different partition.

In each of the extracted sub-alignments, two interspecific variable sites are regarded as being significantly clustered, or far apart at the 5% level if the two sites are separated by a distance d_o ≤ 0.025 (n—1), or d_o ≥ 0.776(n—1), respectively, where n should be the consensus lengths of full-length homologous cDNAs. Since the full-length cDNA homologs were unavailable for most of the alignment sets, the length of the longest partial cDNA homolog in the initial alignment was used as the “n” for the tests, which represented a much more stringent test for spatial clustering of two variable sites. For example, in an alignment block having the longest partial cDNA homolog of 1000 bp, two interspecific variable sites separated by less than 25bp, or by more than 775 bp are significantly clustered, or far apart at the 5% level, respectively. A multi-nucleotide indel is regarded as a significant spatial clustering of adjacent variable sites as it represents statistically improbable random distribution of adjacent clustered variable sites. But, an SSR site, regardless of the unit number, is treated as a single variable site. The COSII cDNA homologs from one or two I. trifida lines, I. littoralis and I. tenuissima in the extracted sub-alignments were first assessed for statistically non-random clusters of variations including at least one specific to the reference species at interspecific variable sites, which define three variation partitions, Itr/Itrk/, Ils/ and Itn/, respectively. The cDNA variants of about one third of the COSII genes from the two I. trifida lines displayed different species-specific mutations at the same or different interspecific variable sites, and were regarded as two sub-partitions (Itr/ or Itrk/) of the Itr/Itrk/ specific to I. trifida. The COSII cDNA variants from sweetpotato in the sub-alignments were then distinguished into the three interspecific variation partitions by their differential conservation of the statistically non-random clusters of interspecific variations in each of the reference cDNAs from the three relatives, or into an additional partition (Hp4/) by its sweetpotato-specific non-random cluster(s) of variations at the interspecific variable sites. The Circos online program (http://mkweb.bcgsc.ca/tableviewer/) was used to create the Circos graph summarizing the partition analyses.

Amplicon sequencing of selected COSII-marker regions

The genomic amplicon sequencing of COSII-marker regions from the screened single-copy COSII genes was part of an effort for development of genotyping-by-sequencing markers in sweetpotato, which will be reported in more details elsewhere. Briefly, a total of 864 PCR degenerate or non-degenerate primer pairs in nine 96-well plates were designed by modification of the tomato COSII-marker primer pairs, and were used to amplify corresponding homologous COSII-maker regions (short introns and flanking primer-anchoring exon borders) in two sweetpotato clones (SC1149 and LSP) and the I. tenuissima accession. Based on primer-site sequences in the sweetpotato and I. tenuissima cDNA homologs of the tomato COSII cDNAs, the primer modification included corrections of non-conserved bases, uses of degenerate bases, shifting 3–10 bases away from the exon-intron junctions, and redesigning one of the primers at nearby sites to avoid too many degenerate bases in the primers for amplifications from the corresponding COSII homoeolog triplet in sweetpotato and the COSII homolog in I. tenuissima. The 864 primer pairs and their amplification efficacy were summarized in a master list provided on a data webpage (detailed below).

Amplicons from the 864 PCR amplifications each for the three genomic DNA samples were sized and quantified using the LabChip^® GX Touch^™ Nucleic Acid Analyzer (PerkinElmer, San Jose, CA), and pooled in equimolar into three size fractions (60 to 250 bp, 250 to 400 bp and 400 to 1500 bp). The three amplicon pools each for the three samples were separately used to make sequencing libraries with (the 400 to 1500 bp pool) or without (the two pools of less than 400bp) further fractionation using the Ion Xpress^™ Plus Fragment Library Kit (ThermoFisher Scientific, San Francisco, CA). Equimolar aliquots of the two non-fractionated libraries of less than 400 bp were mixed and used for templating in a single emulsion PCR using the Ion PGM^™ Hi-Q^™ View OT2 Kit, and their enriched beads sequenced on a single chip following the manufacturer’s 400-bp sequencing protocol. The libraries made of fractionated amplicons were size selected for ~450 to 525 bp, templated using the Ion PGM^™ Template IA500 Kit, and their enriched beads sequenced on a single chip following the manufacturer’s 500-bp sequencing protocol. Each set of sequencing reads was quality trimmed using the sequencer’s default software, and de novo assembled into two sets of contigs separately using the CAP3 as a plugin and the Geneious assembler in the Geneious suit. All the 12 contig datasets (2 fractions x 2 assembly x 3 samples) were converted into searchable databases in the Geneious suit for Blast-searches of contigs for amplicon variants from each of the COSII genes in the two sweetpotato clones and I. tenuissima. The two independently assembled contig groups each for the amplicon variants of a COSII gene from the two sweetpotato clones were first aligned, and apparent assembly chimeras manually screened and removed by cross-referencing haplotypes between the contigs from the two assemblies and between matching ones from the two clones. The cleaned-up alignment was then aligned with the two contig groups for the homologous COSII amplicon(s) from I. tenuissima for further reduction of apparent assembly chimeras among the contigs from both species, and for variant calling and haplotype grouping. Both the haplotype tree method using the Geneious tree builder, FastTree and PHYML, and the Stephens’s phylogenetic partition were used for haplotype grouping of the aligned COSII-maker amplicon variants from sweetpotato with their homologous amplicon from I. tenuissima as a reference.

Hybridization network construction

For hybridization network analyses, two series of alignment blocks each containing eight homologous partial cDNAs from 16 and 4 COSII genes, respectively, in five Ipomoea species were concatenated using the Geneious software. In the two concatenations, each of the 8 rows of 16 or 4 concatenated partial cDNAs from various COSII genes, respectively represents a taxon of sampled expressed sequences in the same variation partition from one species, and was labeled by its partitions and source species. In less than 1% of the alignment blocks, the homologous partial COSII cDNA from I. nil was shorter than the consensus length or not identified in the published dataset so that they were replaced with the homologous regions of corresponding full-length cDNAs from GenBank. Additionally, columns of “-N-” with the numbers of “N”s equal to the position number of an alignment block in the concatenations were inserted in front of the alignment block starting from the second one, which serve as identifiers of the alignment blocks, but do not interfere the network analyses. The SplitsTree4 program (http://www.splitstree.org/) was used to construct a hybridization network from each of the two concatenations. The two hybridization networks were resolved by using hybridization transformations of two split networks that were built by the split decomposition from the pairwise LogDet distances between the 8 taxa in each alignment matrix of the two concatenations. The least square fit (LSFit) measuring the degree of fit between the pairwise distances in the hybridization-network graphs and the pairwise distances in the alignment matrix were used for evaluating robustness of the hybridization networks.

DNA ploidy of Ipomoea littoralis

For references used in comparative phylogenetic analyses of sweetpotato transcriptomes, we obtained deep-coverage transcriptome datasets of partial cDNAs (>800 bp in average) from I. littoralis and I. tenuissima using the long-read 454-NGS. Initial comparative analyses of COSII cDNAs from I. littoralis, I. tenuissima and I. trifida revealed two distinct lineages between cDNA variants of many COSII genes in I. littoralis, one extremely trifida or tenuissima-like and the other littoralis-specific (examples given below), which seemed to be more consistent with them being expressed from two homoeologous loci in two diploid subgenomes than from paralogous ones in a diploid genome.

To ascertain ploidy of the I. littoralis accession, its nuclear content relative to the diploid I. tenuissima and three hexaploid sweetpotato accessions was assessed by using flow-cytometry analyses of freshly isolated intact nuclei from young leaf tissues. Fig 1 shows the results from one of three independent analyses. The nuclear DNA content of the I. littoralis accession was an average 1.355-fold that of I. tenuissima, and average 0.506 to 0.552-fold that of the three sweetpotato clones so that it should be a 2.7x and 3 to 3.4x DNA ploidy relative to I. tenuissima and sweetpotato, respectively. Considering the presence of two distinct lineages of some COSII variants, this DNA ploidy level of the I. littoralis accession could be more reasonably explained by a massive genome reduction (or diploidization) in a tetraploid after speciation rather than by a higher nuclear DNA content in a closely related diploid, i.e. a probable paleotetraploid. The results also ruled out the suggestion that the I. littoralis accession might be a feral form of the hexaploid sweetpotato [14]. Notably, the nuclear DNA contents of the three sweetpotato accessions were average 2.29 to 2.50-fold that of I. tenuissima, equivalent to a 4.6x or 5x DNA ploidy level, which displayed a similar degree (about 1x DNA ploidy) of genome reduction after its speciation as that in I. littoralis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow-cytometry analyses of relative nuclear DNA contents in the Ipomoea littoralis Blume accession.

Suspensions of intact nuclei were prepared from young leaf tissues, stained using DAPI and analyzed using a FACSCalibur^™ flow cytometer. The diploid I. tenuissima was used as an external reference standard and its G1peak positioned at channel 200. The three sweetpotato clones were used as external hexaploid references.

https://doi.org/10.1371/journal.pone.0229624.g001

Two distinguishable types and clustered distribution of variations in cDNA variants of many sweetpotato COSII genes

We first sought to assess phylogenetic differentiation among homoeolog triplets of screened single-copy COSII genes in sweetpotato through detailed comparisons of variations among their cDNA variants with those in respective homologs from three of sweetpotato’s closest relatives. For the analyses, we obtained 811 alignments of equal-consensus-length homologous regions from partial cDNA variants of 811 single-copy COSII genes in two sweetpotato clones, and from their corresponding cDNA homologs from two I. trifida, I. littoralis and I. tenuissima, I. nil and tomato lines, by screening alignments of overlapping homologous cDNAs from 2152 COSII genes in these species and lines. The rest of 1341 alignments were not selected due to presence of apparent expressed paralogs or too many apparent assembly chimeras of sweetpotato COSII genes, or lack of one or more partition reference cDNAs. In the 811 alignments, the pair-wise sequence variations between COSII cDNA variants from sweetpotato, and between COSII cDNA homologs from sweetpotato and its three closest relatives, were mostly less than 3% and 5%, respectively. However, variations between cDNA variants of many sweetpotato COSII genes, albeit small, displayed two important patterns, as illustrated in the variation-highlighting views of two concatenations of homologous cDNA regions from 16 and 4 COSII genes, respectively, in the five Ipomoea species (Fig 2). The first concatenation (variation highlighting view in Fig 2A, and sequence in Supplemental S2 Fig) comprised three taxa (IB_Itr/Itrk_C1, IB_Itn_C1 and IB_Ils_C1) of partial cDNA variants of 16 sweetpotato COSSII genes in the Itr/Itrk/, Itn/ and Ils/ partitions (detailed below), and five taxa of corresponding partial COSII cDNA homologs from the two I. trifida lines (Itr_C1 and Itrk_C1), I. tenuissima (It_C1), I. littoralis (IL-C1), and I. nil (Inil_C1, outgroup) lines. The second concatenation (variation highlighting view in Fig 2B, and sequence in Supplemental S3 Fig) comprised three taxa (IB_Itr/Itrk_C2, IB_Hp4_C2 and IB_Ils_C2) of partial cDNA variants of 4 sweetpotato COSSII genes in the Itr/Itrk/, Hp4/ and Ils/ partitions (detailed below), and five taxa of corresponding partial cDNA homologs from the two I. trifida lines (Itr_C2 and Itrk_C2), I. tenuissima (IT_C2) I. littoralis (IL-C2) and I. nil (Inil_C2, as an outgroup) lines.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Illustration of two distinguishable forms of variations and four phylogenetic partitions among cDNA variants of some COSII genes in sweetpotato.

The two alignments A-1 and B in zoom-out variation-highlighting views were two concatenations (_C1 and _C2) of 16 and 4 alignment blocks, respectively, which included three cDNA variants of the 16 and 4 COSII genes from the sweetpotato SC1149-19 or Tanzania clones (combination of IB_ with_C1 or _C2), and their homologs from I. tenuissima (IT_), I. trifida (Itr_ and Itrk_ for two lines), I. littoralis (IL_) and I. nil (Inil_) in matching column-orders. Black bars highlight all sequence disagreements at various positions against the consensus in individual alignment blocks. The green and maroon triangles indicate sites of two distinguishable forms of variations in among cDNA variants of some COSII genes in sweetpotato. The three cDNA variants each of the 4 and 16 sweetpotato COSII genes carried differential non-random clusters of variations that are either well conserved in one of the reference homologs from I. trifida, I. tenuissima and I. littoralis (as in A), or unshared (_Hp4_ in B) at the common interspecific variable sites, and could be thus distinguished into four species-specified phylogenetic partitions, which were marked with _Itr/Itrk_, _Itn_, _Ils_ and _Hp4_, and shaded with blue, purple, red and pink colors, respectively. The A-2 and A-3 alignments are zoom-in views of the first four concatenated blocks, and the first 132-bp (consensus length) of the fourth block in the first concatenation, respectively. The A-3 alignment also illustrates detailed phylogenetic variation partitions of three cDNA variants of the sweetpotato COSII gene (IB_ and CIP_ for isotigs from the SC1149-19 and Tanzania clones, respectively) with corresponding cDNA homologs from the three relatives.

https://doi.org/10.1371/journal.pone.0229624.g002

First, these local intragenic variations could be usually distinguished into two contrasting types. A more predominant type of variations, including SNPs, MNPs (Multinucleotide Polymorphisms) and Indels, among cDNA variants of many sweetpotato COSII genes were at the sites (e.g. those labeled by green-triangles in Fig 2A-2, 2A-3 and 2B) that were interspecifically variable among respective COSII cDNA homologs from sweetpotato and its three closest relatives, and were either differentially conserved in the reference homolog(s) from one or two of the three relatives (Fig 2A-3) or unique to the COSII cDNA variant(s) from sweetpotato (more details below). These differentially interspecifically conserved variations in cDNA variants of a sweetpotato COSII gene most likely were of distinct origins before speciation of sweetpotato, and were thus homoeoalleles among homoeologous loci of the COSII gene. Some additional variations (e.g. those labeled by maroon triangles in Fig 2A-2 and 2B) among cDNA variants of these sweetpotato COSII genes were not shared interspecifically in any of the reference cDNA homologs, and not at interspecifically variable sites. These intraspecific intragenic variations in a sweetpotato COSII cDNA variant were thus most likely alleles at one of the three homoeologous loci. Furthermore, some variations at interspecifically variable sites or other sites in cDNA variants of less than 5% of the COSII genes were not conserved in the same reference homolog as their neighboring interspecific variations, but shared in a second reference homolog(s), which suggested that they should be most likely derived from parallel mutations from the original residues as those at the same sites in the first reference homolog(s) before or after speciation of sweetpotato. These variations could not be easily determined to be homoeoalleles or alleles as they can be either type or both, but did not impair phylogenetic partition of the cDNA variants. An example of such variations can be seen in Fig 2A-3. The fifth residue “C” of the IB_isotig45341 was most likely a mutation from “A” as judged by the flanking differential variations homologous with those in IT_isotig18823 from I. tenuissima, and thus homoplasious to the “C” at the interspecifically variable site in the cDNA Itrk_sc0007727.1_g000001.1 from one of the two I. trifida lines. Thus, this SNP might be homoeoallelic to those in IB_isotig47343 and CIP_SP_Con_17460 from the two sweetpotato clones, and conceivably allelic to an unidentified one having the same “A” residue at the site as that in the IT_isotig18823.

Secondly, the interspecifically variable sites among cDNA homologs of a COSII gene from sweetpotato and its three closest relatives were present in statistically non-random clusters as determined by the Stephen’s statistical phylogenetic partition. These non-random clusters of variations among the COSII cDNA homologs in the alignments were usually spaced by nearly identical sequences of several tens to hundreds of base pairs for all the cDNA homologs from sweetpotato and its three closest relatives, as can be visually noticed in (Fig 2A-2, 2A-3 and 2B).

Four differentiable variation partitions of sweetpotato COSII genes

By using the Stephen’s phylogenetic partition, cDNA variants of various sweetpotato COSII genes in the 811 alignments could be distinguished into three variation partitions, marked as the Itr/Itrk/, Itn/ and Ils/, based on their differential conservation of non-random clusters of interspecific variations with the corresponding reference homologs from I. trifida, I. tenuissima and I. littoralis, respectively, and into an additional Hp4/ partition by their sweetpotato-specific non-random clusters of interspecific variations. Additionally, one or multiple sweetpotato COSII cDNA variants (counted as a single different one) in one third of the alignments were of multiple mixed partition-types (Mix) of interspecific variations in different regions, implying homoeologous recombinations or chimeric assembly, or could not be clearly partitioned due to insufficient differential clusters of interspecific variations (Ind), indicating lack of phylogenetic differentiation. A total of 372 different COSII cDNAs of such types were included in the Mix/Ind group (Fig 3A). Please also note that since minor allelic variations between two sweetpotato COS II cDNA variants of the same interspecific variation partition are of no direct relevance to the phylogenetic study on the origin of their residing subgenome, the two COSII cDNAs were counted as one homoeoallele type in the summary Circos graph in Fig 3A.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Phylogenetic partitions of the cDNA variants of 811 COSII genes in sweetpotato.

Alignments of equal-consensus-length homologous regions from cDNA variants of 811 COSII genes from two sweetpotato clones (IB_ and CIP_ for those isotigs from the SC1149-19 and Tanzania) and from homologous COSII cDNAs from two I. trifida lines (Itr_ and Itrk_), one I. littoralis (IL_) and one I. tenuissima (IT_) accessions, one I. nil line (Inil_), and tomato (Solyc_) were used for the phylogenetic partitions of cDNA variants of the sweetpotato COSII genes. The Circos plot (A) summarized numbers and types of partitions or group that could be distinguished among cDNA variants of the COSII genes in sweetpotato, and total numbers of different COSII cDNAs that were identified in each of the four partitions and the Mix/Ind group. In 505 of the 811 alignments, single or multiple cDNA variants of a COSII gene from the two sweetpotato clones could be distinguished into only one (the orange-colored arcs) of the four variation partitions, the Itr/Itrk/(blue-colored), Itn/ (purple-colored) or Ils/(red-colored) partitions by their differentially conserved non-random clusters of interspecific variations with the corresponding reference homologs from the two I. trifida lines, I. tenuissima and I. littoralis accessions, respectively, and the Hp4/ (pink-colored) by their unshared non-random clusters of variations at the interspecific variable sites; or were placed into the Mix/Ind group (gray-colored) as their variations at the interspecific variable sites were of mixed (Mix/) or indistinguishable (Ind/) relatedness with those in the corresponding three reference homologs. In the rest of the 811 alignments, the cDNA variants of a COSII gene from the two sweetpotato clones were of two (brown-colored arcs) or three (green-colored arcs) types of interspecific variations, which could be differentiated into various combinations of partitions and the Mix/Ind group. The sub-alignments shown in B to G were two examples each for one, two and three partitions of cDNA variants of different sweetpotato COSII genes. In the alignments D, E and F, the gray-shaded variations between the two COSII cDNA variants from the same or different sweetpotato clones exemplified the intraspecific variations (alleles) between cDNA variants of the COSII gene in the same partition.

https://doi.org/10.1371/journal.pone.0229624.g003

As summarized in the Circos graph (Fig 3A), among the 811 sweetpotato COSII genes, a majority of them (505) were of identifiable cDNA(s) of a single variation type in one of the four partitions or the Mix/Ind group. This indicated that a large percentage of the 505 COSII genes in sweetpotato, excluding some with unidentified variants due to coverage and analyses errors (e.g. out of variable regions), may not be differentiated at their respective homoeologous loci, but were of different lineages. The other two groups of 238 and 68 sweetpotato COSII genes were of two or three identifiable types of cDNA variants in different double or triple combinations of variation partitions or the Mix/Ind group, which demonstrated that at least ~38% of the 811 sampled COSII genes in sweetpotato should be partly or completely differentiated in two or three lineages among their respective three homoeolog triplets. Examples of single, double and triple identifiable variants of various COSII genes that could be differentiated into one, two and three partitions were illustrated with colored-block highlighting in Figs 3B–3G and 4A–4D. Please note that various gene tree methods were attempted, but usually failed to yield consistent results on the relatedness of partial cDNA homologs of a COSII gene form the five Ipomoea species in these alignments, indicating lack of robustness of the tree methods in dealing with the small-scale local variations in the relatively short (<1 kbp) sequence alignments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Illustration of the Hp4/ partition of cDNA variants from four sweetpotato COSII genes.

In the zoom-out variation-highlighting views of four sub-alignments, an Hp4/ partition of the cDNA variant(s) from one or both of the two sweetpotato lines, which displayed distinct clusters of variations at interspecific variable sites as compared to those in all the homologs from the other four Ipomoea species and tomato, could be the only identifiable partition of cDNAs for a COSII gene (A), one of the two partitions of cDNA variants in combination with the Itn/ (B) or Itr/Itrk/ partition (C) for two COSII genes, and one of the three partitions of cDNA variants in combination with the Ils/ and Itr/Itrk/ partition for another COSII gene (D). All the prefixes marking the species source of cDNAs and color-coding of the partitions were the same as in previous figures.

https://doi.org/10.1371/journal.pone.0229624.g004

The largest Itr/Itrk/ partition

The Itr/Itrk/ partition included 376 different sweetpotato COSII cDNAs from single-partition ones of 189 COSII genes, and one of two or three cDNA variants of 147 and 40 COSII genes, respectively (Fig 3A). This disproportionally larger number of COSII cDNAs in the Itr/Itrk/ than those in each of the other three partitions implied that the expressed homoeologs from two or all three of the homoeolog triplets of some of these COSII genes should be undifferentiable so as to be all included in the same Itr/Itrk/ partition. Examples of cDNA variants of sweetpotato COSII genes that can be differentiated into double combinations of Itr/Itrk/ with Itn/, Ils/ or Hp4/ partitions, and triple combination of Itr/Itrk/ with Itn/ and Ils/, with Ils/ and Hp4/, were illustrated in Figs 3D, 3F and 4C and Figs 3G, 3E and 4D, respectively. More than half of the reference COSII cDNA homologs from the two I. trifida lines displayed partly different sets of allelic variations (Itr/ and Itrk/) at clustered interspecific variable sites, indicating two distinct genetic lineages, one or both of which may be shared in one or two cDNA variants of a sweetpotato COSII gene as illustrated with the blue highlighting in Figs 3E and 4C.

Partly and completely differentiated homoeologous COSII-marker regions of COSII genes

To further confirm differentiation (partly or completely) of homoeolog triplets of many COSII genes, we sought to sequence and compare genomic amplicon variants of the COSII-marker regions from two sweetpotato clones (SC1149-19 and LSP breeding line) and I. tenuissima. Since the original tomato COSII-marker primers anchored in exons were designed to amplify short introns of high allelic heterogeneity, the corresponding homologous COSII-marker regions in sweetpotato should likewise include more variable introns of homoeologs of various COSII genes, which has been sample-confirmed in our subsequent effort in developing genotype-by-sequencing markers using the sweetpotato COSII-marker amplicons (data not shown). Using 864 pairs of modified COSII-marker primers, we obtained 729 sets of genomic amplicon variants of COSII-maker regions of 729 out of the 811 COSII genes from the two sweetpotato clones and their homologs from I. tenuissima. Using both haplotype tree and phylogenetic partition methods, analyses of variations among COSII-marker amplicon variants of the 729 COSII genes from the two sweetpotato clones with the homologous partial cDNAs and COSII-marker genomic amplicons of corresponding COSII genes from I. tenuissima as references in alignments revealed four different levels of differentiation of the COSII-maker amplicon variants from these sweetpotato COSII genes.

For 170 COSII genes, their COSII-marker amplicon variants displayed so small number of intragenic variations that no or two dubious major clades could be distinguished by corresponding haplotype trees, but without clear phylogenetic partition as illustrated in Fig 5A. For another 140 COSII genes, their COSII-marker amplicon variants could be clearly distinguished into two major clades on the haplotype trees, and correspondingly the tenuissima-like and -unlike partitions by shared or different clusters of interspecific variations between the amplicon variants and the reference homolog, as illustrated in Fig 5B. For the rest of 419 COSII genes, their COSII-marker amplicon variants could be either distinguished into two major clades, one of which, usually the tenuissima-unlike one, could be further distinguished into two partitions by additional interspecific variations shared by a subgroup of the amplicon variants as in Fig 5C, or three major clades corresponding to three interspecific variation partitions as in Fig 5D. Thus, the number of COSII genes that were partly or completely differentiated at homoeologous loci accounted for about 78% of the 729 sampled COSII gene in sweetpotato. The consistency of the variant grouping by the haplotype tree method and the phylogenetic partition with a higher resolution for the latter, as demonstrated in Fig 5B to 5D, also validate the phylogenetic partition method for grouping COSII cDNA variants. In addition, the intraspecific (brown-triangle labeled) variations and the clustered interspecific (green-triangle labeled) ones, as illustrated in Fig 5, could be also clearly distinguished in introns of at least one of COSII-marker amplicon variants from 559 sweetpotato COSII genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Illustration of four levels of variation differentiation of COSII-marker genomic amplicon variants among 729 sweetpotato COSII genes.

The panels show variation-highlighting views of alignments of, and haplotype trees for the four sets of COSII-marker amplicon variants (_a1 to _a5) from four COSII genes in two sweetpotato clones (IB-SC1149- and IB-LSP-), as compared to their respective reference cDNA (_amp) and genomic amplicons (_G-amp) from corresponding homologous COSII-marker regions from I. tenuissima (IT_). The color-coded highlighting bars represent sequence variation(s) against consensus at the positions in each alignment. The black lines (gaps) connecting two horizontal gray bars for representing reference cDNA amplicons in each alignment delineate introns in respective COSII-marker genomic amplicons. In the panel B, C and D, distinct cluster(s) of variations that were exclusively shared by some amplicon variants in each set, and those that were shared with or different from those in the reference COSII-marker amplicons, distinguished the three sets of COSII-marker amplicon variants into two or three phylogenetic variation partitions, which are separated by wider gaps in the alignments. Those variations between partitions of amplicon variants were most likely homoeoalleles, and labeled by green triangles. Additional variations between the amplicons variants in the same partition were most likely true alleles, and were marked by brown triangles.

https://doi.org/10.1371/journal.pone.0229624.g005

What is noteworthy in the above comparative analyses of cDNA and genomic amplicon variants of COSII genes is the predominance of interspecific (i.e. homoeoallelic) over intraspecific (i.e. allelic) variations among the partly and completely differentiated cDNA and genomic amplicon variants of more than two third of sampled COSII genes. This strongly argued against the frequently claimed high heterozygosity in sweetpotato in literature, and is more consistent with a very low heterozygosity due to very limited hybridizations between the vegetatively propagated genotypes due to widespread incompatibility and flowering unsynchronization among various sweetpotato genotypes, as compared to other crops. The observed genetic heterogeneity in sweetpotato should be mostly homoeoallelic, rather than allelic.

Tracing ancestries of the four COSII genes lineages

To distinguish ancestries of the four differentiated lineages of COSII genes in sweetpotato, we constructed hybridization networks from the two above-mentioned alignment concatenations (variation highlighting view in Fig 2, and sequences in Supplemental S2 and S3 Figs). The two concatenations included two sets of fully differentiated expressed homoeolog triplets in two triple-partition combinations (Itr/Itrk/, Itn/ and Ils/, and Itr/Itrk/, Ils/ and Hp4/) from16 and 4 sweetpotato COSII genes, which were aligned with corresponding reference homologs from the three closes-relative species and one outgroup homolog from I. nil. Please note that these 16 and 4 COSII genes out of the 811 analyzed ones were all that could be identified to have three fully differentiated expressed homoeolog triplets in matching triple-partition combinations, and corresponding reference homologs of equal-consensus-length from all the other four Ipomoea species for proper alignment concatenations. The two hybridization networks (Fig 6) were surprisingly simple, and much more “tree” like. They are robust, albeit a much lower resolution in the second network, as phylogenies of the five taxa of the COSII cDNA homologs from I. trifida, I. tenuissima, I. littoralis and I. nil each in the two networks are consistent between the two networks, and more importantly with the taxonomy of the four species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6.

Two hybridization networks constructed from two concatenations of 16 (A) and 4 (B) individual alignments of partial cDNA homologs of 16 and 4 COSII genes from five Ipomoea species. The first concatenation (_C1) comprised three taxa (IB_Itr/Itrk_C1, IB_Itn_C1 and IB_Ils_C1) corresponding to 16 expressed COSII homoeolog triplets in the Itr/Itrk/, Itn/ and Ils/ partitions, and five taxa of corresponding COSII cDNA homologs from the two I. trifida lines (Itr_C1 and Itrk_C1), I. tenuissima (It_C1), I. littoralis (IL_C1), and I. nil (Inil_C1 as an outgroup). The second concatenation comprised three taxa (IB_Itr/Itrk_C2, IB_Hp4_C2 and IB_Ils_C2) corresponding to 4 expressed COSII homoeolog triplets in the Itr/Itrk/, Hp4/ and Ils/ partitions, and 5 taxa of corresponding COSII cDNA homologs from the two I. trifida lines (Itr_C2 and Itrk_C2), I. tenuissima (IT_C2) I. littoralis (IL_C2) and I. nil (Inil_C2, as an outgroup). The two hybridization networks were the HybridizationNetwork transformations of two split networks that were resolved from the LogDet distances between the 8 taxa each in the two concatenated alignment matrices by the SplitDecomposition method. The LSFit was the least squares fit between the pairwise distances of splits in the two hybridization networks and the taxa in the alignment matrices. The internal nodes that explicitly predict ancestors in the two networks were highlighted with red dots, and number labeled except the earliest one in the networks. The blue-colored splits indicated four detected hybridization events (H1 to H4).

https://doi.org/10.1371/journal.pone.0229624.g006

As the hybridization network is explicit in predicting ancestors [22], the nodes (red-dots) on the two network trees predict ancestors of the COSII multi-gene lineages in the Ipomoea species. The two hybridization networks revealed that two direct branching and one hybridization event (H2 or H3) could account for phylogenies of the set of expressed COSII homoeolog triplets of the particular triple-partition combination from sweetpotato in each network. First, the IB_Itn_C1 and the IB_Hp4_C2 taxa from sweetpotato branched in parallel, relative to all the taxa from three reference species and to the two matching hybrid nodes ancestral to the two taxa of the same Ils/ partition from sweetpotato, in the two networks. The IB_Itn_C1 taxon, representing the COSII homoeologs of a tenuissima-like lineage in sweetpotato, and the IT-C1 taxon from I. tenuissima, were sisters branched from the node N3, which was in turn a sister node of the N4’ leading to the Itrk_C1 and Itr_C1 taxa representing two genetic lineages in I. trifida. Thus, the IB_Itn_C1 taxon from sweetpotato should be an evolutionary cousin to the Itr_C1 and Itrk_C1 taxa from I. trifida, and probably originated from an ancestral species in a sibling relationship to I. trifida. Moreover, the IB_Hp4_C2 taxon, representing the COSII homoeologs of the sweetpotato-specific lineage in sweetpotato, descended from the incompletely resolved node M, which was also a progenitor to the other four taxa from three Ipomoea species. The parallel phylogenies of the IB_Itn_C1 and the IB_Hp4_C2 taxa in the two networks predicted that the two taxa from sweetpotato probably descended from one progenitor species hosting the N3 node. Secondly, the IB_Itr/Itrk_C1 and the IB_Itr/Itrk_C2 taxa representing the COSII homoeologs of trifida-like genetic lineages in sweetpotato also branched in parallel from the nodes N5 and M leading to the Itr_C1, Itr_C2 and Itrk_C2 taxa from I. trifida in the two networks. The IB_Itr/Itrk_C1 taxon from sweetpotato branched from the Itr_C1 lineage even more recently than the other genetic lineage, Itrk_C1, in I. trifida in the first network, and thus was probably derived directly from an I. trifida genotype. However, the IB_Itr/Itrk_C2 taxon from sweetpotato branched from a distant common ancestral node M to the Itr_C2 and Itrk_C2 taxa from I. trifida, and the IB_Hp4_C2 taxon from sweetpotato, displaying a deep coalescence pattern. The IB_Itr/Itrk_C2 taxon may thus represent those sweetpotato COSSII homoeologs that may have retained ancient variations of the corresponding homologs in a distant conman ancestral species to I. trifida and to the one that contributed the IB_Hp4_C2 taxon in sweetpotato due to incomplete lineage sorting (ILS) so that the taxon could still have been contributed from the same I. trifida genotype as the IB_Itr/Itrk_C1 taxon. Lastly, the IB_Ils_C1 and IB_Ils_C2 taxa, representing the COSII homoeologs of the littoralis-like genetic lineages in sweetpotato, were descendants of hybrid nodes N8 and N11 between an ancestor (N7 or N10) leading to the IB_Itn_C1 or IB_Hp4_C2 taxon and an ancestor (N9 or N9’) lineal to the IL_C1 or IL_C2 from I. littoralis, respectively. The two hybrid parents N9 and N9’that branched in parallel from the two most distant ancestral nodes (N1 and N1’) leading to all the other 6 taxa in the two networks were most likely hosted in one ancestral species. The other two hybrid parents N7 and N10 were also in parallel in the two networks, and would be causally in a second ancestral species. However, both pairs of hybrid parents were also direct ancestors to two additional interspecific taxa (IB_Itn_C1 and IL_C1, IB_Hp4_C2 and IL_C2) each in the two networks. This topology implied that the two parallel hybridizations in the two networks should be originated only from a hybridization of two autotetraploids in which the two pairs of hybrid parental nodes were duplicated. Additionally, the hybrid node N4 leading to the Itrk_C1 in the first network should be derived from an intraspecific introgression event, so as the N12 lineal to the Itr_C2 and Itrk_C2 in the second network if the IB_Itr/Itrk_C2 was contributed from an I. trifida genotype.

A species tree embodying the two hybridization networks

To reconstruct species phylogenies accounting for the ancestries of the four lineages of sweetpotato COSII genes, we attempted the most parsimonious inference of a species tree that could be reconciled with the two hybridization networks minus the outgroup splits. The species tree shown in Fig 7 was mainly delineated by the two parallel H2 and H3 hybridizations in the two networks, and by the homoeologous relationships between the IB_Itn_C1 and IB_Ils_C1, and between the IB_Hp4_C2 and IB_Ils_C2 taxa in sweetpotato. A hypothesis of requiring the fewest evolutionary steps to cause the two hybridizations in the two networks would be two synchronous hybridizations of the two pairs of COSII lineages of different ancestries during an interspecific hybridization between their host tetraploids. Furthermore, the homoeologous relationships between the two taxa of each pair that was coupled by a hybridization in each network implied that the closest ancestors to each pair of the taxa should be hosted in two diploid subgenomes of a hybrid tetraploid.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. The most-parsimonious species tree embodying the two hybridization networks.

The gray-shaded outline represents the most parsimonious species tree that could be reconciled with the two hybridization networks. The two hybridization networks minus their outgroups were fitted in the gray-shaded species tree trunks, having matching partition-color-coded splits, numbered nodes and hybridization events (H1 to H4 in blue) as those in the two redrawn color-coded hybridization networks minus their outgroups at the bottom left and right for comparison. The three pairs of sweetpotato COSII lineages fitted in the species tree and their corresponding taxa in the two hybridization networks were labeled as br, bt, bl and b4 for trifida-like, tenuissima-like, littoralis-like and sweetpotato-specific COSII lineages, respectively, in black (first network) and green (second network) color circles. The reference COSII lineages in the three sweetpotato relatives and their corresponding taxa in the two hybridization networks were labeled as tr or tk, t, and L for those from the two I. trifida lines, I. tenuissima, and I. littoralis, respectively, with subscripted number 1 or 2 indicating the first and second network in respective partition-color circles. The tree sections predicting ancestral species (diploid Ip-N1, -N2, -N3 and -N7, tetraploid Ip-Tn, -Lt and -Hb) to sweetpotato were delimited by dashed lines. The H1 and H4 hybridizations were modeled as intraspecific introgression (IS) in I. trifida. The br (IB_Itr/Itrk_C2) and b4 (IB_Hp4_C2) lineages (light-blue and pink colored, respectively) in the second network were modeled as outcomes of incomplete lineage sorting (ILS) in the tree. The dotted light-purple lines mark an additional predicted COSII lineage (btr) in sweetpotato and the trifida and tenuissima-like one (Ltr) in I. littoralis, which would have descended from an ancestral linage branched at the node N1/1’. The three color-gradient blocks underneath the subgenome designations were used to emphasize that only parts of these diploid subgenomes were differentiated as represented by the differentiated multi-gene COSII lineages.

https://doi.org/10.1371/journal.pone.0229624.g007

The species tree predicted that the hexaploid sweetpotato was originated from a hybridization between a I. trifida genotype carrying the two trifida-like COSII genetic lineages (br in black and green circles) and a tetraploid hybrid (Ip-Hb) hosting two interspecific (inter-subgenomic) pairs of COSII lineages, i.e. the tenuissima-like (bt) and sweetpotato-specific (b4), and the two littoralis-like hybrid (bl in blue and green circles) lineages. The inferred tetraploid hybrid Ip-Hb could have been derived from a hybridization between an autotetraploid (Ip-Tn), which was from doubling of an earlier ancestral diploid (Ip-N7) genome in a sister and a niece kinship to those of the extant I. tenuissima and I. trifida, respectively, and another tetraploid (Ip-Lt), which was from doubling of an earlier ancestral diploid genome (Ip-NL) in a grand-aunt and an aunt kinship to those of the same two Ipomoea species. In others words, I. trifida would be a direct progenitor to only one (B_s1’B_s1’) of the constituent diploid subgenomes in sweetpotato. The other two subgenomes in sweetpotato would have been derived from hybrids of two pairs of duplicated diploid genomes (in Ip-Tn and Ip-Lt) in a respective niece and aunt kinship to that of I. trifida. They would thus be extremely closely related to, but not the same origin as the first sweetpotato subgenome from I. trifida. Additionally, I. tenuissima and I. trifida would be in a niece and aunt kinship, derived from the predicted ancestor Ip-N2 hosting the network nodes N2 and M, and in a third and second cousin kinship to I. littoralis, respectively. By the parsimony principle, I. littoralis was modeled as a paleo-tetraploid derived from the predicted Ip-Lt tetraploid. Since the three diploid subgenomes in sweetpotato were only partly differentiated as represented by the three inter-subgenomic pairs of differentiable lineages of some COSII genes, they were thus noted using the three color-gradients blocks under the subgenome designations.

In the species tree, the sweetpotato-specific COSII lineage (b4) was fitted as one that retained some ancient polymorphisms from the most distant ancestor Ip-N1 (hosting network nodes 1/1’), i.e. resulting from an ILS, which was consistent with the fact that the expressed homoeologs of COSII genes in this lineage (Hp4/) carried unique ancient variations that could not be traced in homologs from the three closest relatives, nor in most cases from I. nil. Furthermore, based on the observed Itr-Itn-like second lineage (Ltr) for variants of some COSII genes in I. littoralis, the second subgenome in sweetpotato was predicted to also carry additional Itr-Itn-like btr lineages (dotted light-purple lines) in place of the Itn-like bt or the sweetpotato-specific b4 lineage for some COSII genes, which could have been carried over from the tree-predicted earlier ancestor (Ip_Lt) to I. littoralis during its hybridization with the predicted autotetraploid Ip-Tn. Just as expressed homoeologs of the Ltr lineage from some COSII genes in I. littoralis were almost indistinguishable from their homologs from I. trifida or I. tenuissima, the expressed homoeologs of the btr lineage from some sweetpotato COSII genes would have been too close to those of the trifida-like (Itr/Itrk/) or tenuissima-like lineage (Itn/) from the same COSII genes to be distinguished by the partition method, and thus hybridization not detectable. This predicted btr lineage of COSII genes in sweetpotato could have contributed to the disproportionally higher number of COSII cDNAs in the Itr/Itrk/ and Itn/ partitions, as these cDNAs could have been placed in the Itr/Itrk/ or Itn/ partition. The ancestral lineage to the predicted btr and the Ltr lineages was simplistically modeled as a distinct genetic lineage in an autotetraploid Ip_Lt and an earlier diploid Ip_NL ancestor due to lack of information. Additionally, the H1 and H4 hybridizations detected by the two networks were modeled on the species tree to be derived from intraspecific introgressions (IS) between genotypes of I. trifida, as explained in the network reconstruction.