Chromosome-based survey sequencing reveals the genome organization of wild wheat progenitor Triticum dicoccoides

Bi-parametric flow cytometry provides access to individual T. dicoccoides chromosomes

To circumvent genome complexity of tetraploid T. dicoccoides, all 14 chromosomes were individually isolated by flow cytometry from two genotypes, based on bi-parametric analysis of GAA microsatellite content and DAPI fluorescence intensity, as described earlier (Akpinar et al., 2015b). Chromosomal DNA amplified from flow-sorted chromosomes was sequenced to a depth of 39–58 × , assuming that chromosome sizes are similar to those estimated for Triticum durum cv. Timilia (Venora et al., 2002). After exclusion of ultra-short contigs of <200 bases that likely represent collapsed repeats (Mayer et al., 2014; Simpson et al., 2009), the de novo assembly built from >480 Gb sequence data was mapped to the recently published genome of T. dicoccoides genotype Zavitan (Avni et al., 2017) to eliminate any impurities and contamination from organellar genomes and/or rDNA sequences. Neighbouring scaffolds of the de novo assembly mapping to the appropriate Zavitan chromosomes closer than a distance of 50 nucleotides were merged with ‘N's to build chromosomal super-scaffolds. On average, chromosomal super-scaffolds represented ~32%–78% of their respective Zavitan chromosomes. These reference-guided final super-scaffolds, referred as ‘chromosome assemblies’ hereafter, ranged from 288.3 to 700.4 Mb in total length, with N50 values of 1092–4938 bp. (Table 1).

T. dicoccoides genome is swamped by repetitive elements

As a hallmark of Triticeae genomes, chromosome assemblies were also swamped by repetitive elements which indicated an overall repeat content of 78.2% for the entire genome, marginally underestimated due to the exclusion of ultra-short contigs. While Class I retroelements dominated the overall repetitive landscape, at the family level, the Jorge family of Class II DNA transposons (En-Spm superfamily) was prominent, reaching as many as 15% of all bases masked as repetitive elements in all chromosome assemblies. Among repeat families, the most striking difference was observed for undetermined LTR elements, which collectively comprised over one-fifth of all repeat annotations from A-genome chromosomes, in contrast to B-genome chromosomes, where these elements comprised only ~12.5% of all repeat annotations combined (Figure 1).

T. dicoccoides genome is highly populated with genes that are conserved among grasses

To identify genes encoded by the T. dicoccoides chromosomes and explore syntenic relationships, chromosome assemblies were compared against the fully annotated proteomes of the related grasses Brachypodium, rice and sorghum and high-confidence proteins of its close relative, barley, in addition to transcript assemblies and predicted proteins of T. dicoccoides (Akpinar et al., 2015a; Avni et al., 2017) (Table S1, Figure S1). To avoid redundancy in gene number estimations due to orthologous relationships, Brachypodium, rice, sorghum and barley genes were clustered into nonredundant orthologous groups based on sequence similarity using OrthoMCL (Li et al., 2003). In total, 60 392 contigs were associated with 14 882 nonredundant orthologous genes from related grasses, representing orthologous coding loci. Of these, loci found on only one chromosome, on homeologous group chromosomes or all chromosomes, were excluded. Remaining 6430 coding loci found on two or more chromosomes corresponding to the same orthologous grass gene may point to putative interchromosomal gene duplication or gene loss events. Overall, assuming an average coding sequence length of 2772 bases for wheat (Luo et al., 2013), these observations suggested that the entire T. dicoccoides genome may encode ~37 000 conserved genes that have orthologs in the four related grass genomes, with a peak of conserved gene fraction of 1.3% for chromosome Tdic4A. The estimation of total gene content of the genome should take functional intra-chromosomal gene duplicates that widely diverged from their grass orthologs (neo-functionalization) and genes that are specific to the wheat lineage into account, while putting special care on the well-known prevalence of nonfunctional pseudogenes in wheat genomes (Choulet et al., 2010).

Due to their shared ancestry, cereal genomes exhibit widespread colinearity, forming large ‘syntenic’ regions on chromosomes that carry orthologous genes (The International Brachypodium Initiative, 2010). Comparison of the chromosome assemblies to the related grasses reproduced the well-known patterns of synteny at the chromosome level, providing evidence for the well-documented 4AL-5AL-7BS translocation, as well (Figure S2). While Tdic5A and Tdic5B had similar conservation patterns in general, a unique region of synteny to the distal regions of Bd1 (Bd1 for Brachypodium distachyon chromosome 1) and Sb1 (Sb1 for Sorghum bicolor chromosome 1) and to the proximal region of Os3 (Os3 for Oryza sativa chromosome 3), was observed only for Tdic5A (Figure 2).

Overview of the structural comparison of wild and modern wheat genomes

To explore the overall conservation between modern and wild wheat genomes, chromosome assemblies were compared against the genomes of wild emmer wheat Zavitan (Avni et al., 2017) and bread wheat cultivar Chinese Spring (IWGSC RefSeqv1). We individually mapped the chromosome assemblies to each genome and used any sequence mapping to specific locations on both the Zavitan and Chinese Spring genomes as links to connect the two respective positions. For each of the 14 chromosomes, approximately 400 000–500 000 such links demonstrated large-scale structural variations between these wild and domesticated genotypes. As expected from the close evolutionary relationship between these two species, extensive colinearity between orthologous chromosomes was evident (Figure S3). Despite the extensive conservation between the closely related tetraploid wild genomes, we identified over 45M sequence variations, including single nucleotide polymorphisms (SNPs) and InDels, between our chromosome sequences and the Zavitan genome that passed stringent filtering criteria (Akpinar et al., 2016) (Table S2). These SNPs could be useful in designing novel SNP markers to distinguish regions of interest within these highly similar genomes.

MiRNAs from the noncoding RNA pool hint at potential toolboxes of T. dicoccoides in response to stress

Chromosome assemblies contained 28 011 putative precursor loci for miRNA biogenesis, representing 59 miRNA families (Table S3). Of these, expression evidence was present for 49 families (83%) at the small RNA (sRNA) and/or pre-miRNA level, among previously published durum wheat sRNA sequences and T. dicoccoides transcripts (Akpinar et al., 2015a; Distelfeld, 2016; Liu et al., 2015). Predicted miRNAs with precursor sequences containing transposable elements (TEs) by more than 50% of their lengths were classified as ‘low confidence (LC)’, as TE-associated miRNAs are not easily distinguished from siRNAs, particularly from genome sequence data alone. The precursor sequences of these LC miRNA families predominantly contained Class I DNA transposons, abundantly from the Mariner superfamily (Table S3). Notably, 11 miRNA families were processed from both LC precursors and precursors that contained few to no TEs (classified as ‘high confidence (HC)’). Seventeen miRNA families contained loci on all chromosomes, suggesting functional redundancy and/or potential TE-mediated proliferation events and loci for 49 miRNA families were common to both subgenomes (Figure S4). On a broader scale, using the same identification procedure, precursor sequences for 26 HC miRNA families identified from chromosome assemblies were also present in the whole-genome sequences of T. dicoccoides genotype Zavitan (Avni et al., 2017) (27 HC miRNA families predicted in total) and the hexaploid T. aestivum cv. Chinese Spring (31 HC miRNA families predicted in total), suggesting a core collection of regulatory miRNAs in polyploid wheats. In all three species, the total number of miRNA encoding loci was the highest in the B subgenomes. When mapped back to the genome, majority of mature miRNAs predicted from chromosome assemblies aligned with intergenic regions (94.35% of all predicted mature miRNAs), pointing out to a vast genomic source for miRNA biogenesis. On average, only 0.11% and 5.54% of predicted miRNAs aligned to exons and introns, respectively, of gene models (IWGSC RefSeqv1), indicating potential autoregulatory circuits.

Functional annotation of the putative targets of T. dicoccoides miRNAs identified from the chromosome assemblies revealed an array of biological processes, molecular functions and cellular components, with subtle differences across subgenomes and homeologous groups. These potential miRNA-target pairs hint at an intriguing toolbox that can be used by T. dicoccoides to cope with stress conditions, including nuances for fine-tuning. For instance, Tsn1, a biotic stress-related gene involved in the response to tan spot disease in durum wheat, was a target of the miR2118 family members from multiple chromosomes with slightly different mature miRNA sequences (Table S3). Whole-genome network of potential miRNA-target pairs can be of substantial importance to understand the regulation of a given phenotype with its full dimensions.

In addition to miRNAs, precursors for two other classes of noncoding RNAs, tRNAs and long noncoding RNAs (lncRNAs) were identified from the chromosome assemblies. Consistent with previous observations, a marked abundance of tRNA^Lys genes, speculatively resulting from co-proliferation following an ancient TE capture, was observed for all chromosomes (Figure S5) (Akpinar et al., 2014, 2015b; Lucas et al., 2014; Tanaka et al., 2013). As lncRNAs are typically mined from transcriptomics data, lncRNA candidates on chromosome assemblies were explored using a comparative approach. Potential lncRNA sequences were extracted from a combined set of RNA-sequencing data from Triticum turgidum ssp. dicoccoides genotype Zavitan, six different tissues of T. aestivum, and also drought stressed and irrigated samples of T. aestivum with a strict set of elimination steps (Cagirici et al., 2017) (Experimental Procedures). A total of 89,623 lncRNAs passing all elimination steps were then mapped to the T. dicoccoides chromosome assemblies and genome sequences of wild emmer wheat Zavitan (Avni et al., 2017) and bread wheat cultivar Chinese Spring (IWGSC RefSeqv1) to confidently identify a core set of lncRNAs. Importantly, of 83 822 lncRNAs common to all three genomic resources, 23 713 were potential targets of mature miRNAs from chromosome assemblies (Table S4). These miRNA-lncRNA associations may provoke target mimicry, within stress response pathways as well. For instance, a member of the miR437 family (mature miRNA sequence: AAAGUUAGAGAAGUUUGACUU) targeted an aquaporin PIP1-5 transcript in T. dicoccoides, along with by 8, 24 and 89 lncRNAs in identified from leaf, seed and root tissue transcriptomes of T. aestivum, respectively. It is tempting to speculate that under drought stress, target mimicry imposed by lncRNAs can alter miRNA function, resulting in reduced down-regulation of aquaporin proteins and improved water uptake.

T. dicoccoides chromosome assemblies cover several agronomically important loci

Chromosome assemblies contained several genomic loci coding for agronomically relevant traits, which not only demonstrated the utility of the assemblies but, also in some cases, provided novel insights (Data S1). Domestication is a key event shaping the wheat genome. In barley, nonbrittle rachis 1 (Btr1) and nonbrittle rachis 2 (Btr2) control an important domestication-related trait, ‘Brittle Rachis (BR)’. Loss-of-function mutations in either of these genes disrupt the BR phenotype that enables free dispersal of grains at maturity (Pourkheirandish et al., 2015). In tetraploid wheat, homeologous copies of both Btr1 and Btr2 genes were found on chromosomes 3A and 3B (Avni et al., 2017). In this study, two contigs, Tdic3A-contig10048876 and Tdic3B-contig9087392, contained coding regions for Btr1-A and Btr2-B (designated as TdicBtr1-A and TdicBtr2-B), respectively, as indicated by their sequence similarities to barley Btr genes. Notably, TdicBtr1-A, with 93% sequence identity to the wild barley (Hordeum vulgare subsp. spontaneum) Btr1 gene (KR813340.1), had a 2-bp deletion at position 290 that resulted in a premature stop codon. Therefore, the predicted translated product of TdicBtr1-A was a 97-amino acid peptide that was likely nonfunctional. By contrast, TdicBtr2-B was 88% identical to its wild barley ortholog (KR813339.1) and was translated into a full-length 198 amino acid protein product (Table 2).

A closer look at the protein level revealed candidate sequences on two contigs on chromosome 3A and three contigs on chromosome 3B for Btr2-A and Btr1-B, respectively. Potential coding regions on contigs from chromosome 3B shared relatively low sequence similarity with Btr1 sequences, which is consistent with the presence of Btr-like genes along the genome (Pourkheirandish et al., 2015). However, Tdic3A-Cluster30852 contained 494 bases from the 3ʹ end of a potential Btr2-A gene separated from its 61-base 5′ region by a 3.7-kb unrelated sequence. Together, these segments of the Btr2-A gene shared high sequence identity with its barley ortholog (90%), and its sequence was identical to that of Zavitan TtBtr2-A (Avni et al., 2017). In addition, the unrelated sequence separating the 3ʹ and 5ʹ ends of Btr2-A carried TE elements including both DNA transposons and retroelements. This organization suggested that a TE insertion within the coding sequence of TdicBtr2-A could have relocated the first 61 bases at the 5ʹ end to a downstream position, splitting this sequence into two segments and resulting in the loss of a genomic sequence encoding 15 amino acids in the translated product. Notably, Cluster30852 originated from the merging of two Tdic3A contigs that mapped to the Zavitan 3A chromosome at positions 41 bases apart. Such an organization could have been the result of ambiguous (and erroneous) mapping due to the presence of repetitive sequences, although the presence of flanking nonrepetitive sequences would reduce this possibility. This observation could also be explained by a small-scale genome rearrangement on Tdic3A in our assembly with respect to the Zavitan genotype. In this case, even though the 5ʹ and 3ʹ fragments were identical to the BTR2 protein, such an organization should abolish its function completely (Table 2, Figure S6).

The chromosome assemblies also covered two vernalization genes that are important regulators of flowering in wheat. Tdic5A-contig17747970 and Tdic5B-contig15503489 contained homeologous coding regions for Vrn1, while Tdic7A-contig20604678 potentially contained coding sequences for Vrn3. Intriguingly, two contigs from Tdic4B might harbour coding sequences for the Vrn2 locus, along with TE-related sequences. The Vrn2 locus, which was mapped to the 5A chromosome, was organized as two tandemly duplicated genes, ZCCT1 and ZCCT2. The protein products of both genes were composed of a putative zinc finger domain and a CCT domain, and they shared 76% sequence identity (Distelfeld et al., 2009). The 11-kb Tdic4B-contig10186201 contained a two-exon ORF that was translated into a 217-residue peptide. This translated product shared 93% similarity with T. turgidum ZCCT2-B2a (ACI00359.1). Importantly, the product encoded by contig10186201 did not exhibit any of the three mutations at positions 16, 35 and 39 of the CCT domain that are predicted to disrupt protein function (Distelfeld et al., 2009). The translation product of a second contig on Tdic4B, contig10211027, consisted of 209 amino acids and was 97% similar to T. turgidum ZCCT1-B1 (ACI00355.1). Both of these ORFs were surrounded by TEs, including the highly promiscuous Gypsy family of retrotransposons. Whether these ORFs represent a fully functional Vrn2-like locus that was translocated into Tdic4B awaits further study.

Leaf rust caused by Puccinia triticina is an important disease that affects wheat production worldwide(Sela et al., 2012; Xie and Nevo, 2008). A leaf rust resistance gene, Lr10, was located inside the 13-kb contig6213438 from Tdic1A, which may potentially represent a new allele (Data S1). The Lr10 locus resided within in a gene-rich segment on the distal end in close proximity to another resistance gene, Pm3, conferring resistance against powdery mildew disease. In our chromosome assemblies, the 17-kb contig, Tdic1A-contig6388763, carried a coding region for the Pm3 locus. Importantly, a number of resistance genes against Hessian fly have also been mapped to the same gene-rich region on 1AS, a few of which have already been introgressed from close and distant relatives into hexaploid wheat (Data S1).

Autophagy is a process in which cellular contents are degraded, resulting in either recycling of cytoplasmic components or the elimination of damaged or toxic molecules inside the cell. This process plays an important role in both abiotic and biotic stress responses (Klionsky et al., 2016). For instance, wheat homeologs of ATG6 play a role in responses to fungal pathogens in addition to abiotic stress factors (Yue et al., 2015). Additionally, the autophagy-related gene, TdATG8, which was recently cloned from wild emmer wheat, is responsive to water deficit (Kuzuoglu-Ozturk et al., 2012). The T. dicoccoides chromosome assemblies obtained in this study covered several autophagy-related genes, including previously cloned ATG4, ATG8 and ATG6 (Nagy et al., 2014; Pei et al., 2014; Yue et al., 2015). Considering the importance of autophagy-related genes in plant metabolism and stress responses which can be modulated for the improvement, all autophagy genes were mined and comparatively analysed in tetraploid Zavitan and hexaploid Chinese Spring genomes. In total, 44 coding regions corresponding to highly conserved 18 isoforms of ATG genes were identified across all wheat chromosomes (Data S2). Atg8 isoforms were found in three separate locations on homeologous groups 2 and 5 chromosomes, consistent with previous findings (Kuzuoglu-Ozturk et al., 2012). ATG genes were also conserved at the protein level in close relatives rice and Brachypodium, as well as in Arabidopsis thaliana albeit to a lesser extent (Table S5).

To gain deeper insight into the functional roles of ATG genes, we assessed the differential expression patterns of these genes including stress responsive pathways and cis regulatory elements. The promoter regions of ATG genes hosted recognition sites for many different transcription factor (TF) families such as WRKY, GATA, TCR and ERF, suggesting a complex interplay of multiple molecular pathways (Figure S7). Importantly, recognition sites for AP2, B3, TCP, NAC-NAM, bZIP, NF-YB TF families were detected in the promoter regions of all ATG genes, which indicates autophagy is commonly recruited in an array of developmental and stress response-related processes, under different genetic and/or environmental cues. Consistent with these scheme, ATG genes were differentially expressed in different tissues, developmental stages and environmental conditions (Figure 3). A hallmark of autophagy, Atg8 expression, was prominent in all tissues tested, potentially indicating a basal level of autophagy as a normal developmental process (Figure 3a). Furthermore, ATG genes were also differentially expressed in different layers of grain, with a relatively lower expression profile in endosperm compared to inner and outer pericarp tissues. ATG expression appeared to be high in transfer cells in grain, pointing out an increased autophagic activity in these cells (Figure 3b). Finally, ATG genes were also responsive to abiotic stress factors, such as drought and heat (Figure 3c), together with developmental changes such as senescence and photomorphogenesis.

Chromosome sorting, sequencing, assembly and filtering

Bivariate flow karyotyping and chromosome sorting were carried out as previously described (Akpinar et al., 2015b) using Triticum dicoccoides (2n = 4x = 28), genotypes TR191 (Karacadag region of Turkey) and MvGB436 (provided by Dr. István Molnár, Centre for Agricultural Research, Hungarian Academy of Sciences, Martonvásár, Hungary). Only chromosomes 2A and 4A were derived from genotype MvGB436, as flow karyogram did not allow separation of these chromosomes to an acceptable purity in the other genotype. Chromosomal DNA was amplified by isothermal multiple displacement amplification (MDA) (Data S2).

Sequencing of amplified chromosomal DNA was carried out by Centrillion Genomic Services, Centrillion Technologies (Palo Alto, CA, http://www.centrilliontech.com/) on high output mode HiSeq 2500 v3 at a read length of 2x100PE. The quality metrics of the raw reads were assessed using FASTQC software (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). The Illumina paired-end reads were de novo assembled using the ABySS tool (Simpson et al., 2009). The k-values for the sequence assemblies were determined using KmerGenie (Chikhi and Medvedev, 2014) (Table 1). ABySS scaffolds were then mapped to the chromosome assemblies of T. dicoccoides genotype Zavitan (Avni et al., 2017). Scaffolds mapping to the respective Zavitan chromosomes at a distance closer than 50 nucleotides were merged with ‘N's using custom Python scripts, thereby producing the final reference-guided assemblies. Repetitive elements were identified by comparing the chromosome assemblies against the MIPS Repeat Element Database v9.3 p for Poaceae (ftp://ftpmips.helmholtz-muenchen.de/plants/REdat/) using RepeatMasker v.3.3.0 software (http://www.repeatmasker.org/) (Nussbaumer et al., 2013).

Protein-coding genes and syntenic relationships

Protein-coding genes were identified through BLAST searches against the fully annotated proteomes of Brachypodium distachyon (v1.2, http://mips.helmholtz-muenchen.de/plant/brachypodium) (The International Brachypodium Initiative, 2010), Oryza sativa (IRGSP-1.0, http://rapdb.dna.affrc.go.jp/download/irgsp1.html) (Tanaka et al., 2008), Sorghum bicolor (v1.4, http://mips.helmholtz-muenchen.de/plant/sorghum/) (Paterson et al., 2009) (1E-6, -length 30, -ppos 75); high-confidence Hordeum vulgare proteins (http://mips.helmholtz-muenchen.de/plant/barley/) (Mayer et al., 2012) (1E-6, -length 30, -ppos 90) and Triticum dicoccoides transcripts assembled from RNA-sequencing data (Akpinar et al., 2015a; Avni et al., 2017) and predicted proteins (Avni et al., 2017) (1E-30, -length 100, -pident 99 for transcripts; 1E-6, -length 30, -ppos 99 for proteins). A ‘Best Reciprocal Hit’ approach was adopted, and only the best reciprocal hits in blastx and tblastn searches with the given parameters were accepted as significant matches.

To estimate gene number, a nonredundant list of orthologous grass genes was constructed from the B. distachyon, O. sativa, S. bicolor and H. vulgare proteins described above using the OrthoMCL pipeline (http://www.orthomcl.org/orthomcl/) (Li et al., 2003) (Data S2).

Syntenic relationships were visualized as circle plots and heatmaps generated using Circos software (Krzywinski et al., 2009) and in-house Python scripts. Ribbons on the circle plots were obtained by bundling >200 links along 1-Mb intervals. Gene densities for B. distachyon, O. sativa and S. bicolor were counted on 500-kb intervals (light blue and light grey). The genomic positions of annotated genes were obtained from the MIPS database of plants (http://mips.helmholtz-muenchen.de/plant/genomes.jsp). For all BLAST searches, BLAST+ stand-alone toolkit, version 2.2.31 was used (Camacho et al., 2009). All computational analyses were carried out on a High Performance Computing Cluster. All functional annotations were performed using BLAST2GO (Conesa and Götz, 2008); the initial BLAST steps were run locally against all Viridiplantae proteins (1E-6, -outfmt 5, -max_target_seq 1).

Putative noncoding RNA species

A homology-based miRNA prediction approach was utilized to annotate miRNA-encoding sequences as described previously, with slight modifications (Alptekin et al., 2017; Kurtoglu et al., 2013, 2014; Lucas and Budak, 2012) (Data S2). Expression analysis of the predicted miRNAs was performed at both the pre-miRNA and mature miRNA level, as described in Data S2. Target sequences for putative miRNAs were predicted with the web-tool psRNATarget (http://plantgrn.noble.org/psRNATarget/), using transcripts assembled from RNA-sequencing data from two T. dicoccoides genotypes (Akpinar et al., 2015a; Cagirici et al., 2017), along with the T. dicoccoides whole-transcriptome assembly (Avni et al., 2017). For comparative analyses, miRNA identification was carried out on the genome assembly of wild emmer genotype Zavitan (Avni et al., 2017) and hexaploid wheat cv. Chinese Spring (RefSeqv1, International Wheat Genome Sequencing Consortium, https://urgi.versailles.inra.fr) using the same approach and target transcripts were identified among whole-transcriptome assembly (Avni et al., 2017) for Zavitan and high-confidence coding sequences (ftp://ftpmips.helmholtz-muenchen.de/plants/wheat/IWGSC/) for Chinese Spring. Putative tRNA genes were identified using the tRNAscan-SE 1.21 program(Lowe and Eddy, 1997) with default parameters for eukaryotic genomes. Pseudogenes and other undetermined annotations were excluded.

Identification of lncRNAs was performed using three available transcriptome data sets from (i) wild emmer genotype ‘Zavitan’ (both high-confidence and low-confidence transcripts, (https://wheat.pw.usda.gov/graingenes_downloads/Zavitan/), (ii) hexaploid wheat Chinese Spring (http://www.earlham.ac.uk/organisms/wheat) and (iii) Drought-stressed and irrigated samples of Chinese Spring. The coding potentials of these transcriptome sequences were assessed with CNCI (https://github.com/www-bioinfo-org/CNCI) (Sun et al., 2013), and ORFs were identified using Transdecoder (https://transdecoder.github.io/). Transcripts with protein-coding potential and/or with potential ORFs encoding polypeptides >100 amino acids in length were discarded. Remaining transcripts were compared against known protein sequences of B. distachyon, O. sativa, S. bicolor and H. vulgare, as above, and against protein sequences of Triticum aestivum (UniProt) and Triticum turgidum (NCBI, retrieved from 7 February 2017), together with reviewed protein sequences from the Swiss-Prot database with both ‘blastx’ (1E-5, -length 30, -ppos 80) and ‘tblastn’ (1E-5, -length 30, -pident 80). BLAST searches were also performed against Unigene sequences of B. distachyon (build #2), O. sativa (build #86), S. bicolor (build #30), H. vulgare (build #59) and T. aestivum (build #63) and EST sequences of T. turgidum (NCBI, retrieved from 7 February 2017) (1E-30,-length 90, -identity 80). In addition, BLAST searches were performed against the organellar genomes of T. aestivum (gene bank ID: NC_007579.1 and NC_002762.1) and T. turgidum (KJ614397.1), together with noncoding RNA sequences (rRNA, tRNA, snRNA and snRNA). Transcript sequences showing any homology in these analyses were discarded and candidates longer than 200 nucleotides were accepted as putative lncRNAs. The lncRNAs were mapped onto the genome assemblies of wild emmer genotype Zavitan and Chinese Spring, as well as the T. dicoccoides chromosome assemblies, using GMAP (-cross-species, -f ‘2) (http://research-pub.gene.com/gmap/) (Wu and Watanabe, 2005). For positive alignments, only lncRNAs mapping through their entire lengths with an alignment score of at least 40 were accepted. Finally, the lncRNA transcripts that contain noncanonical splice sites were checked in depth to assess whether these transcripts may still have potential coding features that eluded detection simply due to strandedness. This final elimination discarded the lncRNAs with noncanonical splice sites if their reverse complemented sequences coded for canonical transcripts.

Identification of SSRs and TE junctions and SNPs

SSRs from chromosome assembly contigs fully covering selected genes from T. dicoccoides and T. turgidum were identified independently using the Gramene SSR tool (http://archive.gramene.org/db/markers/ssrtool) and the MISA-MIcroSAtellite identification tool (http://pgrc.ipk-gatersleben.de/misa/). TE junctions for which ISBP markers can be designed were identified using IsbpFinder (Paux et al., 2010). SNPs were identified as previously described (Akpinar et al., 2016).