Analysis of color differences in hulless barley
The color differences in the seed coat of hulless barley are shown as *L*a*b (L* represents the brightness, a* represents the red-green axis, and b* represents the blue-yellow axis) in Table 1, which reflects the significant differences in the color characteristics of the five hulless barley cultivars (Leónet al, 2006). Analysis of the color value showed that the brightness factor L*, red factor a*, and blue factor b* decreased from white hulless barley to colored hulless barley. In the cluster analysis (Fig. 1B), the white hulless barley could be clearly distinguished from the colored hulless barley varieties.
Table 1
Color values of the seed coats of five hulless barley cultivars
Number
|
Cultivar
|
Seed coat color
|
L
|
a
|
b
|
1
|
1127
|
white
|
56.27 ± 0.13
|
8.52 ± 0.26
|
24.83 ± 0.88
|
2
|
BQ-2
|
purple
|
47.49 ± 0.62
|
3.61 ± 0.13
|
15.32 ± 0.12
|
3
|
GZL
|
blue
|
49.14 ± 0.72
|
2.27 ± 0.15
|
12.94 ± 0.27
|
4
|
XLQK
|
blue
|
46.17 ± 0.33
|
5.81 ± 0.28
|
19.34 ± 0.12
|
5
|
GZ-160505
|
black-purple
|
36.61 ± 0.45
|
5.03 ± 0.25
|
8.37 ± 0.45
|
Metabolic profiling
In total, 608 metabolites were detected (Supplementary Table S1), including 147 flavonoids, 87 amino acids and derivatives, 78 phenolic acids, 74 lipids, 47 organic acids, 46 nucleotides and derivatives, 42 alkaloids, 10 lignans and coumarins, eight tannins, four terpenoids, and 71 others. The clustering heatmap of the metabolites (Fig. 2) showed there were significant differences in the metabolite levels of the five cultivars, with the colored hulless barley varieties being clearly distinguished from the white cultivar. This finding was generally consistent with the dendrogram based on the color difference values (Fig. 1b). GZL and BQ_2, although cultivated in different regions, grouped into one category according to seed coat color and possessed similar metabolite contents. The white cultivar 1127 could be clearly distinguished from the other cultivars. This finding was demonstrated by clustering analysis of the samples and showed that the metabolites of hulless barley were strongly related to seed coat color.
Differential metabolite analysis using PCA
The first two principal components of the PCA score plot were responsible for 51.31% (28.68% for PC1 and 22.63% for PC2) of the variation in the metabolite profiles. There was a clear separation of these five cultivars, suggesting that each group had a relatively distinct metabolic profile. The distinct separation of 1127 (white hulless barley) on the left side of the plot from the other four varieties, which clustered on the right side of the plot, confirmed the distinction of white hulless barley from the colored cultivars. The HCA and PCA results suggested that the differences in the metabolite contents might be responsible for the differences in the seed coat color of hulless barley.
Differential metabolite analysis using OPLS-DA
OPLS-DA is a multivariate statistical analysis method for supervised pattern recognition that can maximize the differentiation between groups and identify differential metabolites. In this study, OPLS-DA was carried out to further verify the significantly different metabolites between the five hulless barley cultivars.
High predictability (Q2) and strong goodness-of-fit (R2X, R2Y) of the OPLS-DA models were observed for comparisons between 1127 and XLQK (R2X = 0.771, R2Y = 1, Q2 = 0.989; Fig. 4A), 1127 and GZ_160505 (R2X = 0.776, R2Y = 1, Q2 = 0.989; Fig. 4B), 1127 and GZL (R2X = 0.773, R2Y = 1, Q2 = 0.987; Fig. 4C), 1127 and BQ_2 (R2X = 0.768, R2Y = 1, Q2 = 0.986; Fig. 4D), XLQK and BQ_2 (R2X = 0.744, R2Y = 1, Q2 = 0.979; Fig. 4E), XLQK and GZL (R2X = 0.763, R2Y = 1, Q2 = 0.987; Fig. 4F), XLQK and GZ_160505 (R2X = 0.748, R2Y = 1, Q2 = 0.982; Fig. 4G), GZL and BQ_2 (R2X = 0.715, R2Y = 1, Q2 = 0.976; Fig. 4H), GZL and GZ_160505 (R2X = 0.748, R2Y = 1, Q2 = 0.982; Figure; 4I), and BQ_2 and GZ_160505 (R2X = 0.748, R2Y = 1, Q2 = 0.982; Figure; 4J). The validation of the OPLS-DA models presented a high interpretation rate (R2Y) and prediction degree (Q2 > 0.9), indicating that the models were appropriate.
Differential metabolite screening, functional annotation, and enrichment analysis
The differential metabolites among the five cultivars were screened using a fold-change (≥ 2 or ≤ 0) and variable importance in the projection (VIP) value (≥ 1) of the OPLS-DA model. The screening results are shown in Supplementary Table S2 and are illustrated using volcano plots (Fig. 5A–D) and Venn diagrams (Fig. 5E, F). There were 181 significantly different metabolites between 1127 and BQ_2 (134 down-regulated, 47 up-regulated), 219 between 1127 and GZ_160505 (115 down-regulated, 104 up-regulated), 207 between 1127 and GZL (131 down-regulated, 76 up-regulated), and 246 between 1127 and XLQK (146 down-regulated, 100 up-regulated). Compared with white hulless barley, there were more down-regulated metabolites in the colored hulless barley, and these metabolites could clearly distinguish the colored hulless barley from the white hulless barley. The screening results also showed that, in total, about one third of the differentially abundant metabolites were flavonoids (62/181 in 1127 vs. BQ_2; 84/219 in 1127 vs. GZ_160505; 65/207 in 1127 vs. GZL; and 76/246 in 1127 vs. XLQK). These metabolites were mainly flavonoids and may constitute the representative differential metabolites for the different colored varieties. A difference in phenolic compound composition with grain color was previously observed in both barley and hulless barley (Abdel et al. 2012), which corroborates our results.
Using the intersection of each comparison group in the Venn diagram (Fig. 5E), a total of 70 common differential metabolites were observed, and each comparison group possessed unique differential metabolites. The Venn diagram in Fig. 5F shows the overlapping and unique metabolites among the comparison group. There were 28 metabolites that were only present in the colored hulless barley, 19 of which were flavonoids and three of which were phenolic acids. These metabolites were primarily flavonoids and may be considered as representative differential metabolites for colored hulless barley.
To gain further insight into the differences in flavonoids between colored and white hulless barley, differential flavonoids among the comparison groups were compared (Supplementary Table 3). The results showed that, except for BQ_2 (26/62 in 1127 vs. BQ_2), most of the flavonoids (69/84 in 1127 vs. GZ_160505; 47/65 in 1127 vs. GZL; and 62/76 in 1127 vs. XLQK) were up-regulated as the color of the seed coat changed from white to purple-black. Many flavonoids were up-regulated in the colored hulless barley, indicating that colored hulless barley is rich in flavonoids.
Based on the intersection of each comparison group (Supplementary Table 3), 14 common differential flavonoid metabolites were up-regulated among the comparison groups 1127 vs. BQ_2, 1127 vs. GZ_160505, 1127 vs. GZL, and 1127 vs. XLQK. Rutin was detected in all the samples, though the differences among the colored hulless barley and white cultivar were nonsignificant. Except for cyanidin 3,5-O-diglucoside and peonidin 3-O-glucoside in BQ_2, most of the anthocyanins were up-regulated in the colored varieties compared with the control. Isoflavones are generally exclusively present in legumes, such as soybeans, and play important roles in plant defense and nodulation (Jung et al. 2000; Kim et al. 2007; Shoeva et al. 2016). In the present study, three isoflavones were detected in all hulless barley varieties, of which afzelechin(3,5,7,4'-tetrahydroxyflavan) was down-regulated and 2'-hydroxyisoflavone was up-regulated, whereas aracarpene 2 was down-regulated in BQ_2 and up-regulated in GZ_160505 and XLQK.
Differential metabolic pathways among the samples
The differentially abundant metabolites among the five hulless barley cultivars were annotated using the KEGG database (http://www.genome.jp/kegg/) in order to obtain the detailed pathway information (Fig. 6A–D, Supplementary Table S4). The KEGG classification results and enrichment analyses indicated that the majority of differential metabolites were involved in metabolic pathways, including metabolic pathways (44/84 [52.38%], 1127 vs. BQ_2; 45/86 [52.33%] in 1127 vs. GZ_160505; 57/95 [60%] in 1127 vs. GZL; 65/113 [57.52%] in 1127 vs. XLQK), biosynthesis of secondary metabolites (22/84 [26.19%] in 1127 vs. BQ_2; 23/86 [26.74%] in 1127 vs. GZ_160505; 27/95 [28.42%] in 1127 vs. GZL; 34/113 [30.09%] in 1127 vs. XLQK); ABC transporters (12/84 [14.29%] in 1127 vs. BQ_2; 8/86 [9.30%] in 1127 vs. GZ_160505; 14/95 [14.74%] in 1127 vs. GZL; 19/113 [16.81%] in 1127 vs. XLQK); flavonoid biosynthesis (9/84 [10.71%] in 1127 vs. BQ_2; 13/86 [15.12%] in 1127 vs. GZ_160505; 7/95 [7.37%] in 1127 vs. GZL; 13/113 [11.50%] in 1127 vs. XLQK); and anthocyanin biosynthesis (5/84 [5.95%] in 1127 vs. BQ_2; 6/86 [6.98%] in 1127 vs. GZ_160505; 6/95 [6.32%] in 1127 vs. GZL; 6/113 [5.31%] in 1127 vs. XLQK). Enrichment analysis (Fig. 6A–D) also indicated that the differential metabolites of the comparison groups were mainly involved in flavonoid biosynthesis, flavone and flavonol biosynthesis, and anthocyanin biosynthesis.