Plant materials

Seeds of ‘TB2018’ were derived from about 50 almost ripe ears of wheat found by chance in a private property located in the territory of Matino (Lecce province, Apulia region, Southern Italy), on a private property road, and at the following coordinates: 40° 00’N, 18° 11’ E, 75 m a.s.l. In living memory, there is no evidence that wheat was produced on this land. After a year of growth in the same place, the seeds were sown in an experimental field at the CREA Research Center for Cereal and Industrial Crops (CREA-CI), S.S. 673, km 25,200, 71122 Foggia, Apulia region, Italy.

Triticum turgidum ssp. durum cultivars ‘Marco Aurelio’ and ‘Senatore Cappelli’ were obtained from “Società Italiana Sementi” (SIS, San Lazzaro di Savena (Bologna), Italy, www.sisonweb.com).

The pedigree/genealogy of ‘Marco Aurelio’ is ‘Orobel’//’Arcobaleno’/’Svevo’. ‘Senatore Cappelli’ is an old cultivar of durum wheat, dating back to 1915, obtained by the geneticist Nazareno Strampelli at the Research Center for Cereal Growing in Foggia, Italy, through genealogical selection (nr. 231/1915) carried out in Foggia on the North African population ‘Jenah Rhetifah’.

The set of Italian durum wheat cultivars used in the morpho-physiological characterization is provided in S1 Table.

Plant growth conditions

A quality testing experiment, programmed for a two-year period, was started in autumn 2018 at the CREA (Consiglio per la Ricerca in Agricoltura e l’analisi dell’Economia Agraria) Research center for cereals and industrial crops (Foggia, Italy; 41° 27′ 36” N, 15° 30′ 05” E; 75 m a.s.l.) on clayey soil (Typic Chromoxerert). The main characteristics of the soil were 30% clay, 25% sand; pH 7.5; 12.5 g kg^-1 total C.

The climate is typical of a Mediterranean environment characterized by the presence of a dry season between May and September. Based on 62 years of climatic data (starting 1955), the annual mean temperature is 15.7°C (range 9.7–21.8°C), with a precipitation of 529 mm, and with a high variability, especially for rainfall (range 272–803 mm).

The experiment was established with 11 wheat accessions and 2 fertilization levels in a randomized block design with 2 replicates and elementary plots of 10 m², repeated for two years for a total of 44 plots per year. Within this experiment, we set an a priori contrast between the main effect on ‘TB2018’ and ‘Senatore Cappelli’ for the analyzed variable.

Normality of residual distribution and homogeneity of variances was checked, respectively, with the Shapiro-Wilk and the Levene tests [14]. The sowing was carried out with a plot seeder on 7 December 2018. The germinability was 95% and 350 germinable seeds were distributed per m². In pre-sowing, 92 kg/ha^-1 of phosphorus (as P₂O₅) were distributed; in the growing period, three fractional fertilizations were carried out with 120 kg/ha^-1 of nitrogen (ammonium nitrate at 26%): i) tillering (BBCH 24) with 60 kg/ha^-1 (15 March 2019), ii) lifting II node (BBCH 32) with 30 kg/ha^-1 (05 April 2019) and iii) barrel (BBCH 41) with the last 30 kg/ha^-1 (23 April 2019).

The control of wild herbs/weeds was carried out with a treatment (herbicide) based on an Atlantis + Biopower + Zenit formulation. The earing date was about 34 days from 1 April and the average height was 110 cm.

The harvest was carried out at the physiological maturation stage corresponding to grade 92 of the Zadoks scale [15], with a plot combine. The average yield was 4.11 t/ha.

Morpho-physiological analysis

For the morphological analysis the descriptors employed were those described in the CPVO TP/120/2 protocol for Distinctness, Uniformity and Stability (DUS) [16, 17] and the corresponding UPOV web site [18]. The growth stages were referred to by the decimal codes of [15]. A principal component analysis (PCA), multivariate approach was followed employing the above-mentioned descriptors, to describe and classify the accession under study. A set of Southern Italian durum wheat cultivars and the corresponding descriptors available online, following the same coding scheme, were chosen as reference. The numerical values corresponding to indicators listed in the CPVO/UPOV identification protocols were used to compose a spreadsheet file (S1 Table) that was used as input for the R-based software ‘prcomp’ [19]. The resulting PCA data object was plotted using the R-based software ‘ggbiplot’ v.0.55 [20].

Total protein extraction and SDS-PAGE analysis

Whole mature and dry seeds were ground to a meal and flours were suspended in an extraction buffer (Urea 6 M, SDS 2%, DTT 5 mM; 1:10 w/v) and stirred for 2 hours at room temperature. The suspension was then centrifuged at 10,000 x g at 4°C for 30 min to separate solubilized proteins. Protein quantification was performed with the Bradford method [21].

SDS-PAGE analysis was carried out according to [22] on 12% polyacrylamide gel under reducing conditions. Gels were stained with Coomassie Blue G-250 (Bio-Rad, Hercules, CA, USA) and the relative molecular mass of polypeptides was determined by comparison with standard proteins (GE Healthcare, Chicago, IL, USA) [23].

In vitro growth conditions and phenotypic characterization

Seeds were sterilized by soaking in a solution of 50% bleach and sterile water for 15 minutes followed by a couple of quick washes in sterile water and then five washes for 5 minutes each in sterile water. Seeds were germinated and grown in 50 mm diameter plates at 22°C for 3 days in the dark on 3M filter paper soaked in sterile water or on an alternative substrate composed of inert vegetable fiber soaked in sterile water.

For the phenotypic observations, images of seedlings were taken. The total root and lateral root number were counted. The longest primary seminal root length and the lateral root length of each seedling were also measured using the ImageJ software [24].

Genomic DNA extraction, library preparation and sequencing

Total genomic DNA was extracted from 10 g of ground seeds using the Qiagen Plant DNA kit (Qiagen). DNA quality was assessed by conventional 0.8% (w/v) agarose gel electrophoresis [25] and spectrophotometry using a Nanodrop (ThermoFisher) [26].

The Celero TM DNA-Seq kit (NuGEN, San Carlos, CA, USA) was used for library preparation following the manufacturer’s instructions. Both input and final library were quantified by a Qubit 2.0 Fluorometer (Invitrogen, Carlsbad, CA, USA) and quality tested by an Agilent 2100 Bioanalyzer High-Sensitivity DNA assay (Agilent Technologies, Santa Clara, CA, USA). Libraries were then prepared and sequenced on a NovaSeq 6000 (Illumina) instrument in paired-end 150 mode (IGATech, Udine, Italy).

Bioinformatic analysis of the ‘TB2018’ genome sequence

A primary bioinformatic analysis included the following steps:

• Base calling and demultiplexing. Processing raw data for both format conversion and de-multiplexing by Bcl2Fastq v.2.0.2 of the Illumina pipeline [27].
• Adapter sequences were masked with Cutadapt v1.11 [28].

A low (5X) coverage of the ‘TB2018’ genome, yielding 552.82 million reads, was estimated sufficient for the scope of the analysis.

Quality control and mapping of reads against the reference genome of Triticum turgidum L. var. durum cv. ‘Svevo’ (genome sequence: GCA_900231445.1.fasta; [13]) was done through the NGSEP (v.4.01) software [29]. For mapping reads an index of the ‘Svevo’ reference genome was first obtained through the Bowtie v.2.4.3 software [30].

Quality metrics and statistics of reads were obtained through the Picard v.2.26.3 software [31].

The identification of sequence variants (SNPs, InDels) was performed through the NGSEP software using the VCFannotate command of the NGSEP (v.4.01) software [29]. Ploidy was addressed using the appropriate flags as described in the NGSEP manual [29, 32].

For the annotation to the Triticum durum reference genome, the GFF3 file version 1.51 was used [33]. For all the analyses the high-performance computing (HPC) cluster at the University of Milano [34] was employed running under a CentOS 7 operating system with a minimum of 16 cores and 64 GB RAM for the most demanding runs of index building and read mapping.

The effect of any structural and single nucleotide variant detected by the comparison of the ‘TB2018’ genome against the ‘Svevo’ reference genome was estimated through the Variant Effect Predictor (VEP) v.104 software [35, 36] hosted at the Ensembl web site [37]. The list of VEP parameters and values used for the analysis is provided (S2 Table).

Variants were classified according to the VEP scheme [38], and further crossed with biological descriptors of gene function retrieved through the Gene Ontology [39] and the BioMart [40] online bioinformatic resources. The output files following VEP analysis at the single chromosome level were used to filter genes whose mutation(s) are expected to produce a relevant effect on protein coding (“IMPACT = HIGH” flag).

The analysis of repetitive elements was done on the GenSAS server [41] using the RepeatMasker v.4.1.1 program [42] that screens DNA sequences for interspersed repeats and low complexity DNA sequences. The output of the program is a detailed annotation of the repeats that are present in the query sequence. The details of parameters followed is reported (S1 Text).

The results of the repetitive element analysis are made available through the GenSAS server via the Apollo [43] genomic annotation platform which is coupled with JBrowse [44] for viewing feature alignments and for manual curation. Raw output data in GFF3 format can be accessed (https://doi.org/10.13130/RD_UNIMI/SLKJLY).

Genome-wide comparison of ‘TB2018’ against durum wheat germplasm

The methodology followed was based on the derivation, from the whole-genome resequencing data of ‘TB2018’, of genotypes at selected SNP loci to compare with a collection of 259 durum wheat cultivars already genotyped for the same loci [45]. The whole-genome sequence of ‘TB2018’ described in the above section was used to derive the SNP positions corresponding to the 3,541 SNP loci selected in [45].

The list of 3,541 probes spanning the SNP loci was first provided (Dr. Francesca Taranto, personal communication) and represents a subset of the probes of the Illumina iSelect 15k wheat SNP array aligned to ‘Durum’ and ‘Zavitan’ genomes. The chip contains 13,600 gene-associated SNP markers [46] and is an optimized and reduced version of the 90k iSELECT SNP-chip described by [47].

The iSelect 15k array chip sequences for both Durum and Zavitan wheat species are now available on the web page of the GrainGenes Database of Triticeae and Avena [48].

The following analytical pipeline was followed:

1. the list of probes was used to run a multiple BLAST v. 2.13.0+ [49] search against a local database generated from the ‘TB2018’ resequenced consensus genome with the parameters listed in S1 Text.
2. the output of the BLAST search was formatted to correct the positions of the SNPs relative to the coordinates of the ‘TB2018’ resequenced genome. Such coordinates were used as an input file in the next step to obtain the corresponding genotypes through the ‘getfasta’ command of the ‘Bedtools’ v. 2.25.0 software [50].
3. the output of ‘bedtools getfasta’ was used to derive the complementary sequence at each SNP position and the list of separate SNPs was joined to form a continuous nucleotide string (FASTA pseudomolecule) that was used for comparison to the corresponding sequences of the 259 reference wheat cultivars of [45];
4. the multiple sequence alignment and the phylogenetic tree were generated using the ClustalW-based algorithm and pipeline provided by the EMBL-EBI facility [51].

For sake of reproducibility of the analysis and comparison with data shown in [45] a neighbor-joining tree was also generated using MEGA v.11 [52] and a total of 1,000 replicates were used to generate bootstrap values. The FigTree v.1.4.4 software [53] was used for the graphical visualization of the tree.

To further investigate the placing of ‘TB2018’ against the five “Cappelli” entries from [45], the G-DIRT software [54] was run using the set of 3,509 SNP loci retained from the above analysis and formatted according to the instructions (S2 Text) with the parameters shown in S3 Text.

Agronomic parameters

The ‘TB2018’ landrace was included in an agronomic field trial for a set of reference Southern Italian durum wheat cultivars, in which, as part of the morpho-physiological characterization, a principal component analysis investigation was performed, considering 24 CPVO plant descriptors encompassing the major morphological descriptors of the species. The results show the relative contribution of each morphological trait (arrows in Fig 1) in the separation of the analyzed cultivars in the first two principal components, contributing to 25.2% and 20.6% of the total variance, respectively.

Download:

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

Fig 1. Principal Component Analysis (PCA) of durum wheat accessions based on morphological descriptors.

The first two components are shown in the X and Y axes, respectively. Arrows show how each descriptor contributes to the two principal components. The red oval highlights the proximity of ‘TB2018’ to ‘Senatore Cappelli’.

https://doi.org/10.1371/journal.pone.0291430.g001

Interestingly, the analysis placed the ‘TB2018’ landrace in proximity to the ‘Senatore Cappelli’ cultivar (oval in Fig 1), further supporting the hypothesis that the two are somehow related.

The comparison of ‘Senatore Cappelli’ and ‘TB2018’, was achieved by cultivation under standard agronomic conditions. The Shapiro-Wilk and the Levene tests excluded a significant deviation from normality and homogeneity of variances. No interaction between year, fertilization level and accessions resulted significant. The main effect of accession resulted significant (p< 0.01) as well as the contrast between the two accessions. Data analysis confirmed the striking evidence that ‘TB2018’ featured a shorter stem (average 110 cm) relative to ‘Senatore Cappelli’ (average 135 cm) which could provide an increased resistance to lodging.

Seed protein analysis confirms the similarity between ‘TB2018’ and ‘Senatore Cappelli’

A comparison between ‘TB2018’ and ‘Senatore Cappelli’ was performed through the analysis of soluble proteins extracted from flour obtained from mature kernels. The two cultivars were also compared to the ‘Marco Aurelio’ cv. used as control. Proteins extracted from ‘TB2018’ and ‘Senatore Cappelli’ had comparable yields of 19.3 ± 0.4 and 21.2 ± 0.5 mg of proteins/g of meal, while in ‘Marco Aurelio’ a slightly higher protein solubilization was obtained, yielding up to 26.6 ± 0.9 mg of proteins/g of meal.

The electrophoretic separation performed under reducing conditions allowed the qualitative comparison of total proteins among the three cultivars. Protein distribution patterns of ‘Senatore Cappelli’ (Fig 2 lane A) and ‘TB2018’ (Fig 2 lane B) displayed high similarity, while both patterns clearly differed from that of ‘Marco Aurelio’(Fig 2 lane C).

Download:

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

Fig 2. SDS-PAGE analysis of total protein extracts from seeds of the three durum wheat cultivars examined, performed under reducing conditions.

A: ‘TB2018’; B: ‘Senatore Cappelli’; C: ‘Marco Aurelio’; MW: molecular weight standard.

https://doi.org/10.1371/journal.pone.0291430.g002

‘Senatore Cappelli’ and ‘TB2018’ presented a higher content of proteins with molecular weights ranging from about 100 to 65 kDa, among which High-Molecular Weight (HMW) glutenins can be found (Fig 2 lanes A and B). These samples also differed from ‘Marco Aurelio’ for the electrophoretic pattern of proteins between 60 and 30 kDa where gliadins can be found [55, 56]. In addition, the Marco Aurelio pattern reported in lane C showed bands in the molecular range of about 30–21 kDa that could belong to Low-Molecular Weight (LMW) glutenins, while lanes A and B presented a lower content of these proteins.

Analysis of early developmental stages of plant growth revealed distinct root traits for the ‘TB2018’ landrace

To analyze their seedling phenotype, seeds of ‘TB2018’, ‘Senatore Cappelli’ and ‘Marco Aurelio’ were germinated in controlled conditions. After three days, germination rate and shoot elongation did not reveal any significant difference among the three accessions.

Differences in root development were instead observed (Fig 3A–3C).

Download:

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

Fig 3. Differences in root length at the seedling stage.

Representative images of the three wheat genotypes germinated on paper: ‘Marco Aurelio’ (A), ‘Senatore Cappelli’ (B) ‘TB2018’(C). Plants were grown in vitro on paper and parameters were measured at 3 days after sowing (DAS). Length of primary root, lateral root, and total roots (D) as well as number of lateral roots (E) were measured for each plant. Values are means of 32 independent biological replicates. Asterisks indicate statistical significance using Student’s t test (*P < 0.05, ns = not significant). Error bar = standard error.

https://doi.org/10.1371/journal.pone.0291430.g003

A reduction in root elongation was detected for ‘TB2018’ if compared to ‘Senatore Cappelli’. The reduction was found for the primary root length as well as for the sum of primary and lateral roots lengths measured at the time of sampling (Fig 3D). However, root length of ‘TB2018’ was like that observed for ‘Marco Aurelio’ and the total number of lateral roots was similar among the three accessions (Fig 3E).

Genome-wide characterization of the ‘TB2018’ landrace

Overall, data on protein profile and plant development provided evidence supporting the hypothesis that the ‘TB2018’ landrace and the registered ‘Senatore Cappelli’ durum wheat cultivars are somehow related. It also highlighted the presence in ‘TB2018’ of distinct traits of agronomic importance. Therefore, to gain insight into the relationship between TB2018 and other wheat genotypes and in view of its future exploitation as a source of interesting genetic variants, for the ‘TB2018’ landrace a low-coverage resequencing of its genome was performed and data were compared against the available ‘Svevo’ reference genome [13]. The bioinformatic pipeline comprised a first quality control of reads to discard those that did not pass a minimum threshold value [29]. The remaining reads were mapped against the ‘Svevo’ genome and yielded the Variant Call Format (VCF) files used to feed the Variant Effect Predictor (VEP) software. Generated data were analyzed at the single chromosome level yielding 14 sets of data corresponding to the 14 chromosomes of the reference genome covering the 7 A-series and 7 B-series haploids (S1–S14 Files).

The number of processed variants per chromosome varied from 95,986 (chromosome 1A) to 437,679 (chromosome 2B) with a range of putatively affected genes between 390 (chromosome 4B) and 971 (chromosome 2B) (S2 Fig, S3 Table). Interestingly, all the chromosomes of the B series featured a higher number of variants than their equivalents of the A series, with a notable 4.5-fold difference between pair 1B and 1A.

Besides several expected low-quality variants that were discarded from the analysis, most variants detected fall in the large intergenic non-coding regions and their information value is expected to be negligible in comparison to that of coding genes. The results of the chromosome-wide analysis are shown in S3 Fig.

Several of the hundreds of genes putatively affected by variants possess Gene Ontology descriptors that are linked to the morphological traits that distinguish ’TB2018’ from ’Senatore Cappelli’ (S4 Fig., S4 Table).

The most represented descriptor in the Gene Ontology “Biological process” domain is “Defense response” (GO:0006952), followed by “Protein phosphorylation” (GO:0006468). The fifth position is occupied by “Regulation of DNA-templated transcription” (GO:0006355) which includes transcription factors.

In particular, the analysis detected several variants in genes and gene families associated with known QTL loci for various physiological and agronomic traits. A special case is the occurrence of a mutation (a stop gained at protein position 438: Y/*) affecting a low-molecular-weight (LMW) glutenin subunit gene (TRITD1Av1G002790), featuring 4 transcripts (splice variants), 16 orthologues and 5 paralogues in the ‘Svevo’ reference genome.

The search for variants in regulatory genes involved in root development has highlighted a mutation (a stop gained at protein position 88: Q/*) in a member (TRITD1Av1G207410) of the WRKY family, one of the largest TF families in plants whose members, in addition to stress response and defense regulation, significantly determine plant development and growth [57].

Transposable elements (TEs) are distinct genetic elements that move and spread within the host genomes in multiple copies. TEs can be major constituents in plant genomes and can drive genome evolution, plasticity, and expansion [58, 59].

The analysis of repetitive elements carried out on the ‘TB2018’ genome by RepeatMasker [42] revealed 9,778,301 features distributed along the 14 chromosomes with a maximum of 838,038 on chromosome 3B and a minimum of 558,432 on chromosome 1A (S5a Fig). An example of the formatted output of the GenSAS server is shown (S5b Fig). Moreover, a study of the chromosome-wide distribution of the three main long terminal repeat (LTR) retrotransposon superfamilies Gypsy, Copia and EnSpm has been carried out (S5c Fig). The complete dataset is accessible at https://doi.org/10.13130/RD_UNIMI/SLKJLY.

To investigate the origin of ‘TB2018’, a genome-wide comparison was run against a dataset of SNP genotypes covering 3,541 marker loci of 259 durum wheat accessions released since the early 1900s and representative of landraces, old cultivars, and modern varieties [45].

The full list of the 3,541 used SNP loci and corresponding genotypes is shown in the accompanying file (S15 File). The vast majority (99.10% of total) of the SNP loci were properly called, while only a small number, 32 SNP loci (0.90%) were not called by the analytical pipeline.

Notably, two groups of such loci fell in contiguous portions of the genome spanning positions 11 Mb to 18.5 Mb and 412 Mb to 420 Mb of chromosome 1A and chromosome 5A, respectively. The unrepresented regions might be due to insufficient coverage of the corresponding genomic sequences at low-level (5x) sequencing, although, due to their extension and contiguity, structural variations (SVs) are a more probable alternative explanation. Interestingly, SVs caused by tandem repeats occurring at chromosome 5A have already been recognized as a factor affecting recombination and major cytological gene-affecting rearrangements in wheat [60]. Further investigation of the nature of these regions will be addressed in future work.

The reconstructed phylogenetic tree placed ‘TB2018’ in proximity with a group of accessions sharing the designation ‘Senatore Cappelli’, among which ‘Senatore Cappelli V-OCs’, ‘Cappelli V-OCs’, ‘Cappelli-MP-OCs’, ‘Cappelli AG-OCs’, and ‘Senatore Cappelli UP-OCs’. Moreover, in the same group are also present ‘Margherito.UP-L’ and ‘Bidi.UP-L’ which are known to be close relatives of ‘Senatore Cappelli’ (Fig 4).

Download:

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

Fig 4. Cluster analysis of the 259 durum wheat accessions compared to ‘TB2018’.

The ‘Cappelli’ accessions are indicated by red arrows; ‘Marco Aurelio’ is indicated by a green arrow; ‘TB2018’ is included in a red box. The subcluster including the ‘Cappelli’-like accessions is highlighted in blue.

https://doi.org/10.1371/journal.pone.0291430.g004

The genome-based comparison of ‘TB2018’ to a large set of wheat germplasm resulted in agreement with data coming from the morpho-physiological characterization that placed ‘TB2018’ close to the ‘Cappelli’ reference (Fig 1).

To further investigate the placement of ‘TB2018’ among the ‘Cappelli’ accessions described in [45] we performed an analysis with the G-DIRT software [54], a tool for the identification and removal of duplicate germplasm based on identity-by-state analysis using SNP data. The results indicate, out of the 3,509 SNP loci left from the genotyping test described above, that 613 markers were retained after linkage disequilibrium (LD) pruning, and 511 markers were retained after applying the Hardy-Weinberg Equilibrium (HWE) data filtration threshold of 0.05 and Heterozygosity threshold of 0.1. All the six input genotypes were retained after missing data filtration of 10%. Four genotypes were retained after removing duplicates with less than 0.1% of Homozygous difference (S3 Text).