Natural diversity in the asparagine synthetase gene family in wheat
Full-length coding sequences of ASN genes from the wheat landrace Chinese Spring were used as queries in BLASTn searches against the genome assemblies of 14 common wheat varieties and spelt wheat (T. aestivum ssp. spelta) to characterise natural allelic variation in the wheat asparagine synthetase gene family. The results are shown in Table 1, ordered by gene name [14, 15] and the corresponding annotated gene model ID from the Chinese Spring RefSeq v1.1 genome assembly . For each orthologous gene, Sorting Intolerant From Tolerant (SIFT) analysis was performed on the translated protein to predict whether the variation in amino acid sequences was likely to disrupt protein function (highlighted in yellow in Table 1) or to be tolerated (highlighted in green). Full details of specific amino acid changes for all wheat varieties are provided in Additional file 1, Table S1.
There were deletions, polymorphisms and presence/absence variation in several wheat ASN genes. For example, TaASN-B1 was deleted in SY Mattis, while eight other varieties carried an allelic variant with a 16 bp deletion in exon seven, introducing a frame shift and bringing a premature stop codon into frame. The presence of this deletion means that the gene is predicted to encode a 375 amino acid protein with a C-terminal truncation of 209 amino acids, including part of the asparagine synthetase domain, indicating that this protein is likely to be non-functional (Additional file 1, Fig. S1). TaASN-B3.2 was deleted in Norin 61, CDC Stanley and Lancer, whereas TaASN-A3.2 and TaASN-D3.2 were present in all analysed varieties, and showed no polymorphisms predicted to impact protein function (Table 1). In contrast, 12 wheat varieties carried TaASN-A3.1 alleles with polymorphisms predicted to disrupt protein function (Table 1). Some varieties carry a combination of alleles predicted to disrupt the function of multiple asparagine synthetase proteins. For example, CDC Stanley carries alleles predicted to affect the function of the enzymes encoded by TaASN-A1, TaASN-B1 and TaASN-A3.1, in addition to a deletion of TaASN-B3.2, while SY Mattis carries deletions of TaASN-B1 and TaASN-B2, and disruptive alleles of TaASN-A3.1 and TaASN-D2 (Table 1). The most common presence/absence variation was of TaASN-B2, which was deleted in eight of the 15 genotypes assayed, including Chinese Spring (Table 1).
Characterisation of the TaASN-B2 deletion
The deletion containing TaASN-B2 mapped to chromosome arm 3BS in the Chinese Spring RefSeq v1.1 genome assembly (Fig. 1a). Alignment of the surrounding region was performed between the annotated Chinese Spring genome and the corresponding region of the Svevo and Jagger genomes, both of which contain the ASN-B2 gene, to evaluate other features of this locus (Fig. 1a and b). The deletion in Chinese Spring was 12,752 bp with respect to the Svevo genome and 12,770 bp with respect to the Jagger genome. A putative open reading frame predicted to encode an F-box protein was detected upstream of TaASN-B2 in the deleted region (Fig. 1a). Directly downstream of the deletion in Chinese Spring and the corresponding region in Svevo and Jagger there is a large, long terminal repeat (LTR) retrotransposon, Inga, belonging to the Ty1-copia family  (Fig. 1a), the identity of which was confirmed using the TREP database . Analysis of the other genome assemblies for the genotypes shown in Table 1 revealed that all eight varieties lacking TaASN-B2 had identical breakpoints.
Wider screening for the presence/absence of ASN-B2
Because of their potential role in determining free asparagine concentrations in the wheat grain, allelic variation in ASN2 genes was explored in a broader set of wheat germplasm. Comparison of the three TaASN2 homeologues in the Cadenza genome revealed they share a common gene structure, each containing 11 exons (Fig. 2a). The encoded proteins shared > 99% identity at the amino acid level, with only eight polymorphic residues between homeologues (Fig. 2b). Although four of these polymorphisms fell in the glutamine amidotransferase (GATase) domain (Fig. 2b), none were predicted to affect protein function according to the SIFT analysis (Table 1).
The length of intron 1 varied between homeologues and was 1104 bp in TaASN-A2, 1411 bp in TaASN-D2, but only 175 bp in TaASN-B2 (Fig. 2a). A pair of redundant primers was designed to amplify a DNA fragment from the first intron of all three homeologues, allowing for the reaction products to be readily distinguished based on size and to detect the presence of TaASN-B2. A second pair of homeologue-specific primers was designed to anneal upstream and downstream of the deleted region containing TaASN-B2 to amplify a DNA fragment only in genotypes carrying this deletion. The presence of TaASN-B2 was, therefore, demonstrated by the amplification of a 434 bp product with the first primer pair and failure to amplify a PCR product using the second primer pair. The results of the analysis are shown in Fig. 3a-d and summarised in Additional file 1, Table S2a. Overall, TaASN-B2 was deleted in 52 of 63 UK winter wheat varieties assayed (82.5%) (Fig. 3e). The deletion was most common in the biscuit (G3) class (93.3%) and least common in the breadmaking (G1) (70%) class (Fig. 3e). An additional set of 24 global wheat varieties were analysed using a similar PCR assay (Additional file 1, Fig. S2a) and the results are shown in Additional file 1, Fig. S2b, and summarised in Additional file 1, Table S2b. The TaASN-B2 deletion was less common among these wheats than in the UK varieties, being present in just 50% of the genotypes (Additional file 1, Table S2b).
A selection of other wheat species was also screened for the presence of an ASN-B2 gene (Fig. 4). An ASN2 gene was identified in Aegilops speltoides (genome BB); however, while an ASN-B2 gene was present in some tetraploid wheat genotypes (genomes AABB) it was absent in others (Fig. 4). Both pasta wheat (T. turgidum ssp. durum) varieties assayed in the study, Svevo and Kronos, were shown to have an ASN-B2 gene, as was Polish wheat (T. turgidum ssp. Polonicum), but the gene was absent in rivet wheat (T. turgidum ssp. turgidum). There was some ambiguity in the result for makha wheat (T. macha) in that there was a clear positive result for the presence of the ASN-B2 gene but a faint band amplified in the assay for the deletion (Fig. 4). This band was still present when the experiment was repeated (data not shown) and is likely due to genetic heterogeneity in the sample.
Expression profiles of wheat ASN genes during development
Raw sequencing reads from public RNA-seq datasets were mapped to the IWGSC RefSeq v1.1 genome assembly to provide a comprehensive overview of the expression profile of each TaASN gene. The expression values for each dataset are provided in Additional file 2 as mean Transcripts Per Million (TPM) values. In a dataset encompassing roots, leaves, stems, spike and grain, each sampled at three developmental stages , TaASN1 transcript levels were highest in young roots and leaves, whereas the three homeologues of TaASN3.1 and TaASN3.2 showed a broader expression profile, with transcripts detected in all assayed tissue types across different stages of development (Fig. 5a). TaASN4 homeologues were also broadly expressed, with TaASN-A4 transcript levels highest in root and spike tissues, and TaASN-B4 and TaASN-D4 more highly expressed during stem development (Fig. 5a). As shown previously [18, 19], TaASN2 showed a grain-specific expression profile, with transcript levels highest at Zadoks stage 85 (Z85), which corresponds to the soft dough stage  (Fig. 5a). Furthermore, TaASN-A2 accounted for 83.3% of all TaASN transcripts in grain tissues at Z85, while TaASN-D2 contributed just 3.0%, consistent with previous results .
To further explore the expression of TaASN genes in the grain, an expression dataset from six stages of grain development in the variety Azhurnaya was analysed . As expected [18, 19], high transcript levels of TaASN-A2 and TaASN-D2 were found in five developmental stages which, when combined, accounted for between 69 and 86% of all TaASN transcripts in these tissues (Fig. 5b). The exception was the grain milk stage, where TaASN3.1 transcript levels were higher than TaASN2 (Fig. 5b), as shown previously . TaASN-B2 transcripts were detected at negligible levels in this dataset, suggesting its deletion in the Azhurnaya genome.
Analysis of expression data from an embryo development timecourse in the common wheat variety AC Barrie  revealed that TaASN2 transcript levels were highest in the mature embryo stage (Fig. 5c). Among TaASN2 homeologues, TaASN-A2 was again the most highly expressed gene, while TaASN-B2 and TaASN-D2 were expressed at similar levels (Fig. 5c). However, TaASN2 transcripts were detected only at negligible levels at all other developmental timepoints, including earlier stages of embryo development and in endosperm and pericarp tissues, where TaASN3.1 transcripts were more abundant (Fig. 5c). Taken together, these data confirm the specific activity of TaASN2 in grain tissues and the mature embryo, and indicate a broader role for TaASN3.1 genes across development, including the early stages of embryo development.
Inter-varietal variation in TaASN expression profiles during grain development
To analyse variation in ASN transcript levels in wheat grain, RNA-seq reads were mapped from grain samples at 14 days post anthesis (DPA) and 30 DPA taken from 27 worldwide wheat varieties . At 14 DPA, total TaASN3.1 transcript levels ranged from 2 to 32 TPM and were greater than TaASN2 in 22 of the 27 varieties assayed (Fig. 6a), consistent with previous results . At 30 DPA, TaASN2 homeologues were the most highly expressed asparagine synthetase genes in all varieties assayed (Fig. 6b). At this latter timepoint, total TaASN2 transcript levels showed large variation between genotypes, ranging from 28 to 242 TPM (Additional file 2). Several lines exhibited very low TaASN-B2 transcript levels and the deletion of this gene was confirmed in five of these lines using the PCR assay (Additional file 1, Fig. S2b; Additional file 1, Table S2b). There were also lines with readily detectable TaASN-B2 transcripts (>five TPM), and the presence of this gene was confirmed for four of these lines (Additional file 1, Fig. S2b; Additional file 1, Table S2b).
A ternary plot showing the relative contributions of each homeologue to overall TaASN2 transcript levels in the grain at 30 DPA in different wheat varieties (Fig. 6c) revealed that TaASN-A2 transcript levels were generally greater than TaASN-D2 in varieties with very low TaASN-B2 transcript levels (likely associated with the deletion of this gene in these varieties). By contrast, in varieties with relatively high TaASN-B2 expression, TaASN-A2 and TaASN-D2 were generally more evenly expressed, and in four varieties, TaASN-D2 transcript levels were higher than TaASN-A2 (Fig. 6c). Nevertheless, overall TPM values for TaASN-A2 and TaASN-D2 were not higher in varieties that lacked TaASN-B2 compared with those in which TaASN-B2 was present (Additional file 2), so there was no evidence of increased expression of these genes to compensate for the lack of TaASN-B2 transcripts.
To investigate the expression dynamics further, the expression of the TaASN2 homeologues was also analysed by RT-qPCR in two wheat varieties possessing TaASN-B2 (Cadenza and Duxford) and two varieties lacking it (Spark and Claire) (Fig. 7). The results of the analysis of variance for this experiment are shown in Additional file 1, Table S3a, revealing significant effects (p < 0.001) of variety, timepoint, and homeologue, and the interactions between these factors, on relative expression levels. In Cadenza and Duxford, mean TaASN-A2 expression was the highest of the three homeologues across all timepoints (14, 21 and 28 DPA), whereas mean TaASN-D2 expression was the lowest (Fig. 7a and b). In both varieties, mean TaASN-B2 expression was greater than TaASN-D2 expression at all timepoints, and in Cadenza at 21 DPA, matched the levels of TaASN-A2 expression (Fig. 7a). In Claire and Spark, mean TaASN-A2 expression was greater than TaASN-D2 expression in all samples, and showed similar expression dynamics across timepoints to Cadenza and Duxford (Fig. 7c and d). Notably, there was no evidence of higher expression of TaASN-A2 or TaASN-D2 in Claire and Spark compared with Cadenza and Duxford.
Contribution of TaASN-B2 to free asparagine concentration in the grain
The screen of varieties for the presence or absence of TaASN-B2 (Fig. 3, Additional file 1, Additional file 1, Table S2) included 63 UK varieties for which free asparagine concentration in the grain had been determined in field trials grown in the UK over two growing seasons (2011–2012 and 2012–2013) . This meant that an assessment could be made of the effect of the TaASN-B2 deletion on free asparagine concentrations in the grain. Of the 63 varieties in the field trials, eleven possessed TaASN-B2 while 52 did not.
The grain from these field trials had been produced in plots in which sulphur was either supplied or withheld . We analysed the effect of TaASN-B2 alongside the other variables in these trials by ANOVA (Additional file 1, Table S3b), which revealed a significant (p < 0.001) effect of the TaASN-B2 deletion in the 2011–2012 field trial: varieties without TaASN-B2 had 13.18% less free asparagine relative to those with TaASN-B2 (Fig. 8a). There was no significant effect (p > 0.05) of the deletion by itself in the 2012–2013 field trial, but there was a significant (p < 0.001) interaction between TaASN-B2 presence/absence and sulphur treatment: there was no significant difference (p > 0.05) in free asparagine concentrations between varieties with and without TaASN-B2 under sulphur deficiency, but varieties without TaASN-B2 had 32.60% less free asparagine (p < 0.01) than those with TaASN-B2 under sulphur sufficiency (Fig. 8c).
The REML analysis based on the complete combined datasets also identified a significant effect of TaASN-B2 as well as an interaction between TaASN-B2 and sulphur treatment (Additional file 1, Table S3b). This analysis also further demonstrated that the interaction between TaASN-B2 and sulphur treatment changed by year, as is suggested by Fig. 8b and c, and that there is a significant interaction between year and sulphur treatment. Prediction and analysis of means from the REML analysis for the 11 varieties common to both trials (of the 63 that were analysed in total) suggests that the TaASN-B2 deletion effect is greatest under sulphur sufficiency across different growing seasons (Fig. 8d). The effect of sulphur deficiency differed greatly between the trials though, so it is difficult to reach a definitive conclusion regarding the effect of TaASN-B2 under sulphur deficiency based on the predicted means of the REML analysis.