Oriental wheat sprout samples were prepared with four replicates from each source (IT308132, IT308447, IT311253, or IT330600), along with a blank sample. Metabolites in the sprouts of four Oriental wheat cultivars were tentatively characterized using UHPLC−QTOF MS in negative ion mode. The base peak ion (BPI) chromatograms of the 80% ethanol extracts from the Oriental wheat sprout samples are shown in Figure 1. Although the BPI chromatograms displayed a large number of peaks, fourteen metabolites were annotated based on their mass spectra. The annotation was achieved by analyzing retention times and fragmentation patterns of major molecular ions and comparing them with previously reported data. Table 1 presents the retention time (t_R), calculated and observed deprotonated molecular ion m/z values ([M–H]⁻), calculated mass error (ppm), proposed molecular formulas, and characteristic MS/MS fragment ions for each annotated compound.

Among the annotated metabolites, one phenolic acid and two benzoxazinone derivatives were tentatively characterized. These compounds and their glycosides have been previously reported in various cereal crops such as corn, rye, wheat, and adlay (34–36). Notably, DIMBOA [2,4-dihydroxy-7-methoxy-2H-1,4-benzoxazin-3(4H)-one] and its derivatives are known to contribute allelopathy, insect resistance, and herbicide tolerance (34). Their glycosylated forms are generally considered to be relatively non-toxic constituents in plants (35).

The MS analysis of peak 1 (t_R = 5.6 min, mass error = 0.6 ppm) revealed a molecular ion [M – H]⁻ at m/z 371.0974. The MS/MS spectrum displayed a fragment ion at m/z 193.0504, corresponding to the loss of a glucuronide moiety (−178 Da) ([M – H – 178]⁻) (Supplementary Figure 1). Based on comparison with previously reported MS data, this compound was annotated as dihydroferulic acid-4-O-glucuronide (36). This metabolite has been previously reported in wheat (T. aestivum) but has not been found in Oriental wheat (T. turgidum ssp. turanicum).

Peak 2 (t_R = 9.7 min, mass error = 0.5 ppm) produced a deprotonated molecular ion at m/z 356.0985 [M − H]⁻. The MS/MS spectrum showed a fragment ion at m/z 194.0461 [M – H – Glu]⁻, consistent with the loss of glucoside moiety (−162 Da) and indicative of HMBOA [2-hydroxy-7-methoxy-2H-1,4-benzoxazin-3(4H)one] (Supplementary Figure 2). Based on this fragmentation pattern and comparison with reported data, this peak was annotated as 2-(glucopyranosyloxy)-7-methoxy-2H-1,4-benzoxazin-3(4H)one (HMBOA-glucoside) (37), a compound previously detected in grains of durum wheat (T. durum, cv. Kamut).

Peak 3 (t_R = 10.1 min, mass error = 0.5 ppm) exhibited a major molecular ion at m/z 372.0934 [M − H]⁻ (Supplementary Figure 3). The MS/MS fragmentation pattern included: m/z 210.0417 [M – H – Glu]⁻, representing the loss of a glucoside moiety (−162 Da), producing DIMBOA [2,4-dihydroxy-7-methoxy-2H-1,4-benzoxazin-3(4H)-one]; m/z 164.0352 [M – H – Glu – OH]⁻, corresponding to an additional loss of a hydroxyl group; and, m/z 149.0135 [M – H – Glu – OH – HCOH]⁻, attributed to a further loss of a formyl group (HCOH), producing MBOA (6-methoxy-benzoxazolin-2(3H)-one) (34). Based on the these spectral features and comparison with published literature (37), peak 3 was tentatively assigned as 2-(glucopyranosyloxy)-4-hydroxy-7-methoxy-2H-1,4-benzoxazin-3(4H)-one (DIMBOA-glucoside), a metabolite also reported in grains of durum wheat (T. durum, cv. Kamut).

Eleven flavonoids were annotated from Oriental wheat sprout samples. In plants, flavonoids typically accumulate as O-glycosylated derivatives; however, in Triticum species, the accumulation of flavonoid C-glycosides has also been reported (38). In MS/MS fragmentation, O-glycosylated flavonoids characteristically lose the sugar moiety via cleavage of the O-glycosidic bond, resulting in fragment ions corresponding to aglycones. In contrast, C-glycosylated flavonoids exhibit distinct fragmentation patterns due to the breakage of the carbon–carbon (C–C) bond between the aglycon and the sugar moiety (39). Mono-C-glucosyl flavonoids (6-C- or 8-C-isomers) typically show characteristic ions such as [Ag + 41]⁻ and [Ag + 71]⁻; additionally, the [M – H – H₂O]⁻ ion is exclusively detected in 6-C-isomers under negative ion mode. Di-C-glycosyl flavonoids display the characteristic ions at [Ag+83]⁻ and [Ag+113]⁻. For di-C-hexosyl flavonoids, fragment ions at [M – H – 90]⁻ and [M – H – 120]⁻ are commonly observed, whereas 6-C-hexosyl-8-C-pentosyl-flavonoids typically exhibit fragment ions at [M –H – 60]⁻, [M – H – 90]⁻, and [M–H–120]⁻ (40). It has been suggested that 6-C-hexosyl-8-C-pentosyl-flavonoids tend to show a strong [M – H – 120]⁻ ion, whereas 6-C-pentosyl-8-C-hexosyl flavonoids exhibit a higher intensity for the [M – H – 90]⁻ ion (41). For O-glycosyl-C-glycosyl flavonoids, fragment ions at [M – H − 164]⁻ and [M – H − 150]⁻ indicate the loss of hexose and pentose residues, respectively, characteristic of O-glycosylation at phenolic hydroxyl positions or 2″-O-glycosylated-C-glycosyl derivatives. Moreover, fragment ions at [M–H–hexose–120]⁻ and [M – H – pentose − 120]⁻ are indicated of 6″-O-glycosylated-C-glycosyl flavonoids (or those glycosylated at the 5″–3″ carbon positions) (42).

The MS analysis of peak 4 (t_R = 13.1 min, mass error = 0.2 ppm) showed a major molecular ion at m/z 579.1351 [M–H]⁻ (Supplementary Figure 4). The MS/MS spectrum exhibited fragment ions at m/z 519.1137 [M – H – 60]⁻, 489.1030 [M – H – 90]⁻, 459.0921 [M – H –120]⁻, 399.0715 [aglycone + 113]⁻, and 369.0607 [aglycone + 83]⁻, consistent with a C-hexosyl-C-pentosyl-luteolin structure containing both hexose and pentose moieties. The relatively higher intensity of the [M – H – 90]⁻ ion compared to [M – H – 120]⁻ suggested a 6-C-pentosyl-8-C-hexosyl substitution pattern. Based on fragmentation features and comparison with reported literature (38), peak 4 was annotated as luteolin-6-C-arabinoside-8-C-glucoside (isocarlinoside), a flavonoid previously reported in T. aestivum but not in Oriental wheat (T. turgidum ssp. turanicum).

Peak 5 (t_R = 13.3 min, mass error = 0.3 ppm) exhibited the same molecular ion at m/z 579.1352 [M − H]⁻ and a fragmentation pattern similar to that of peak 4 (Supplementary Figure 5). However, the relative intensity of the [M – H – 120]⁻ ion was greater than that of the [M – H – 90]⁻ ion, indicating a 6-C-hexosyl-8-C-pentosyl substitution pattern. Based on these observations and comparison with reference data, this peak was tentatively assigned as luteolin-6-C-glucoside-8-C-arabinoside (carlinoside) (38), which has also been reported in T. aestivum but not in Oriental wheat (T. turgidum ssp. turanicum).

The MS analysis of peak 6 (t_R = 14.3 min, mass error = 2.5 ppm) showed a major molecular ion at m/z 447.0938 [M − H]⁻ (Supplementary Figure 6). The MS/MS spectrum displayed characteristic fragment ions at: m/z 429.0830 [M – H – H₂O]⁻, m/z 357.0615 ([M – H – 90]⁻ or [aglycone + 71]⁻), m/z 327.0509 ([M – H – 120]⁻ or [aglycone + 41]⁻), and m/z 285.0404 [aglycone – H]⁻, indicating a mono-C-glycosylated luteolin. The loss of 18 Da corresponding to [M – H – H₂O]⁻ is typically observed in 6-C-hexosyl flavones but is rarely observed in 8-C-hexosyl flavones. Therefore, based on these findings and literature comparisons (40), peak 6 was annotated as luteolin-6-C-glucoside (isoorientin), a compound reported in the commercial Kamut® (T. turgidum ssp. turanicum) as well as various cultivars of T. turgidum ssp. durum (5).

Peak 7 (t_R = 14.7 min, mass error = 1.0 ppm) exhibited a deprotonated molecular ion at m/z 579.1356 [M − H]⁻ (Supplementary Figure 7). The MS/MS spectrum showed fragment ions at m/z 459.0934 [M – H – 120]⁻, 429.0823 [M – H – 150]⁻, and 357.0609 [aglycone + 71]⁻, suggesting O-pentosylation at the 2″ position of the C-glucosyl-luteolin backbone. The absence of [M – H – H₂O]⁻ indicated that the C-glucoside moiety was located at the 8 position. The fragment ion at [M – H − pentose]⁻ supports the assignment of O-glycosylation to the 2″ position (41). Based on features and comparison with published data, this compound was annotated as luteolin-8-C-glucoside-2″-O-arabinoside (adonivernith) (38), which was has been previously reported in T. aestivum leaves, but not in Oriental wheat (T. turgidum ssp. turanicum).

Peak 8 (t_R = 15.0 min, mass error = 2.3 ppm) showed a major molecular ion at m/z 563.1414 [M − H]⁻ (Supplementary Figure 8). MS/MS fragmentation revealed five major ions: m/z 503.1180 [M – H – 60]⁻, m/z 473.1084 [M – H – 90]⁻, m/z 443.0977 [M – H – 120]⁻, m/z 383.0759 [aglycone + 113]⁻, and m/z 353.0661 [aglycone + 83]⁻. These fragments are consistent with a C-hexosyl-C-pentosyl-apigennin structure. The relatively higher intensity of the [M – H – 120]⁻ ion suggested a 6-C-hexosyl-8-C-pentosyl substitution pattern. Accordingly, the compound was tentatively assigned as apigenin-6-C-glucoside-8-C-arabinoside (schaftoside) (43), which has been previously detected in Kamut® (T. turgidum ssp. turanicum) and other T. turgidum ssp. durum cultivars (5).

Peak 9 (t_R = 15.2 min, mass error = 1.3 ppm) had a deprotonated molecular ion at m/z 593.1514 [M − H]⁻ (Supplementary Figure 9). Fragment ions observed in the MS/MS spectrum included: m/z 503.1178 [M – H – 90]⁻, m/z 473.1085 [M – H – 120]⁻, m/z 383.0763 [aglycone + 113]⁻, m/z 353.0662 [aglycone + 83]⁻. These finding indicate a di-C-glucosyl-apigenin, annotated as apigenin 6,8-di-C-glucoside (vicenin II) (43). This compound has been reported in commercial Kamut® (T. turgidum ssp. turanicum) and other cultivars of T. turgidum ssp. durum (5).

Peak 10 (t_R = 17.2 min, mass error = 1.6 ppm) showed a major molecular ion at m/z 431.0985 [M − H]⁻ (Supplementary Figure 10). The MS/MS fragmentation pattern included: m/z 341.0663 ([M – H – 90]⁻ or [aglycone + 71]⁻), m/z 311.0558 ([M – H – 120]⁻ or [aglycone + 41]⁻). These fragments suggest a mono-C-hexosyl-apigenin. The absence of [M – H – H₂O]⁻ supported the assignment of the glucosyl group to the 8-position. Based on comparison with previous literature, this compound was annotated as apigenin-8-C-glucoside (vitexin) (43), also detected in Kamut® (T. turgidum ssp. turanicum) and various cultivars of T. turgidum ssp. durum (5).

Peak 11 (t_R = 17.8 min, mass error = 2.4 ppm) showed a molecular ion at m/z 577.1571 [M − H]⁻ (Supplementary Figure 11). The MS/MS fragmentation patterns include m/z 457.1138 [M – H – 120]⁻, 413.0876 [M – H – 146]⁻, and 341.0663 [aglycone + 71]⁻, indicating O-deoxyhexocylation at the 2″ position of a C-glucosyl-apigenin backbone. The absence of the [M – H − H₂O]⁻ ion supports a C-8-glucosyl substitution. Based on these results and literature comparison, the compound was annotated as apigenin 8-C-glucoside-2″-O-rhamnoside (43), previously reported in wheat (T. aestivum) (44), but not in Oriental wheat (T. turgidum ssp. turanicum).

Peaks 12 and 13 appeared as nearly co-eluting peaks in the BPI chromatogram, with molecular ions at m/z 461.1088 [M − H]⁻ and m/z 593.1503 [M − H]⁻, respectively (Supplementary Figures 12, 13). In the extracted ion chromatogram for m/z 461.1088, fragment ions at m/z 371.0768 ([M − H − 90]⁻ or [aglycone + 71]⁻) and m/z 341.0665 ([M − H − 120]⁻ or [aglycone + 41]⁻) were detected, indicating a mono-C-hexosyl-chrysoeriol structure. The absence of [M – H – H₂O]⁻ indicated an 8-C-glucosyl substitution. Based on spectral data and literature comparison, peak 12 was annotated as chrysoeriol-8-C-glucoside (scoparin) (45), previously found in wheat (T. aestivum) but not in Oriental wheat (T. turgidum ssp. turanicum). For peak 13, the extracted ion chromatogram at m/z 593.1503 revealed fragment ions at m/z 443.0977 [M – H – 150]⁻, 371.0663 [aglycone + 71]⁻, and 341.0665 [aglycone + 41]⁻, indicating O-pentosylation at the 2″ position of the C-glucosyl-chrysoeriol. The absence of [M – H − H₂O]⁻ ion supports a C-8-glucosyl substitution. This compound was annotated as chrysoeriol-8-C-glucoside-2′′-O-arabinoside (45), previously isolated from Setaria italica (46) and detected in Dendrobium offificinale (47), and is here reported in Triticum species for the first time.

Peak 14 (t_R = 19.4 min, mass error = 0.7 ppm) showed a major molecular ion at m/z 607.1667 [M − H]⁻ (Supplementary Figure 14). The MS/MS fragmentation include m/z 487.1241 [M – H − 120]⁻, 443.0987 [M – H – 146 – 18]⁻, and 371.0767 [aglycone + 71]⁻, consistent with O-deoxyhexosylation at C-2″ of a C-glucosyl-chrysoeriol. The absence of [M – H − 18]⁻ confirmed the presence of C-8-glucosyl substitution. Based on literature comparison, this compound was annotated as chrysoeriol-8-C-glucoside-2″-O-rhamnoside (48), previously detected in Cecropia hololeucaa (48), and is now reported in Triticum species for the first time.

To clarify the chemical differences among the sprouts of Oriental wheat cultivars, the metabolite profiles of ethanol extracts from four Oriental wheat sprouts grown under the same environmental conditions were analyzed by UHPLC−QTOF MS. PCA, OPLS-DA, and S-plot, which have been widely used in metabolome analysis of highly complex samples in recent years, are tools to identify significant differences between BPI chromatograms. This serves as an effective approach to compare differences between experimental settings in untargeted and targeted metabolome studies. The four samples were clustered into three groups according to PCA (Figure 2A). Groups I and II were each grouped separately with IT308132 and IT311253 from Afghanistan, while group III consisted of two plants collected in Türkiye, IT308447 and IT330600. The corresponding PCA loading plot showed that ten markers were responsible for group separation (Figure 2B). Four markers (peaks 6, 9, and 14 and an unidentified peak with a molecular ion at m/z 661.1376, t_R = 15.09 min) that shifted in the same direction as group I were suggested as distinguishable markers for group I; one marker (peak 4) was distinctive for group II. Five makers aligned with group III, including peaks 2 and 3 and unidentified peaks with molecular ions at m/z 306.1191 (t_R = 3.57 min), m/z 432.1141 (t_R = 15.42 min), and m/z 440.0797 (t_R = 10.03 min). Since the molecular ions of these markers are shown to be even-numbered, they are presumed to contain nitrogen atoms. PCA can identify the overall direction of variation, while OPLS-DA can distinguish variation between specific groups. The OPLS-DA score plot showed that the Oriental wheat sprout samples were clearly separated according to the countries from which they were collected (Figure 2C). The OPLS-DA model produced one predictive and one orthogonal (1 + 3) components, showing a cross-validated predictive ability (Q²) of 0.988 and a total explained variance (R²Y) of 0.991. In most cases, a Q² value greater than 0.5 is considered adequate, and the difference between R² and Q² should be less than 0.3. The S-plot (point, t_R–m/z pair) from the OPLS-DA model, which is a useful tool for comparing the magnitude and reliability of a variables, was also analyzed. The markers, based on the country of collection, were examined according to their distribution in the S-plot (Figure 2D). As with the PCA results, the differential markers of samples collected in Türkiye showed even molecular ions due to the inclusion of nitrogen atoms, as seen in peaks 2 and 3. Conversely, the markers specific to the Afghan samples mainly comprised flavonoid C-glycosides, including peaks 4, 6, 9, and 14. Therefore, these results could serve as important references for identifying quality markers in Oriental wheat breeding programs based on country of cultivation.

Feature-based molecular networking has become a widely used method for analyzing MS/MS data due to its capacity to visualize large datasets based on fragmentation ma similarity and to suggest possible structures using extensive cloud-based libraries. Additionally, it facilitates semi-quantitative comparisons between groups by evaluating differences in MS1 feature intensities (32). In the present study, feature-based molecular networking analysis was employed to reinforce the annotation of chemical markers that distinguish Oriental wheat samples based on their countries of origin, as suggested by multivariate analysis. This method also aimed to uncover additional candidate markers. Feature-based molecular networking of the Oriental wheat sprout samples was conducted with grouping based on collection origin: Türkiye (orange) and Afghanistan (green), as indicated in the pie charts. The results closely aligned with those from the multivariate analysis (Figures 3, 4). Analog-based library searches further enhanced the reliability of metabolite annotation, providing plausible matches for many of the features (Table 2).

Among the fourteen key peaks annotated from the BPI chromatograms, flavonoid C-glycosides constituted the majority of metabolites in the Oriental wheat sprouts. The multivariate analysis identified peaks 4, 6, 9, and 14 as differential markers for Afghanistan samples. In the molecular network, clusters B − D were specifically enriched with flavonoid C-glycosides, and the aforementioned peaks corresponded to the following features: MF8 in cluster B, MF11 in cluster C, and MFs 14 and 22 in cluster D (Figure 3). The pie charts visualizations clearly indicated the predominance of these components in the Afghanistan samples, as the green segments were significantly larger than the orange segments. Notably, MF11 (peak 6, cluster C) had a high cosine score (0.93) in the GNPS library, consistent with its annotation as luteolin-6-C-glucoside. Other plausible analogs supported additional annotations: MF8 (peak 4, cluster B) matched apigenin-6-C-hexoside-8-C-pentoside, which is consistent with the annotation of luteolin-6-C-arabinoside-8-C-glucoside; MFs 19 and 22 (peaks 11 and 14, respectively, in cluster D) matched apigenin-8-C-glucoside-2″-O-rhamnoside and chrysoeriol-8-C-glucoside-2″-O-rhamnoside. Furthermore, three previously unidentified features were assigned as flavonoid C-glycosides based on spectral library searches with cosine scores greater than 0.8: MF7 (apigenin 6,8-C-digalactoside), MF10 (apigenin-6-C-glucoside-7-O-glucoside), and MF17 (apigenin-8-C-glucoside). However, MF15, despite being a distinguishing feature of the Afghan samples, could not be structurally annotated using the current methods.

The molecular networks also included clusters of feature nodes that were quantitatively more abundant in the Türkiye samples. In the pie charts within these clusters, the orange segments—representing Türkiye—were larger than the green segments—representing Afghanistan. Clusters F, G, and H were interpreted as potentially associated with nitrogen-containing compounds, which were previously suggested as distinguishing markers for the Türkiye samples in the multivariate analysis (Figure 4). The GNPS library search proposed HMBOA-O-hexoside and DIMBOA-O-hexoside as likely identities for the features in clusters F and G, respectively. MF6, a marker that remained unidentified the multivariate analysis, was found to be clustered with MF5 (corresponding to peak 2 in cluster F, annotated as HMBOA-O-glucoside) and also yielded an analog library hit as DIMBOA-O-hexoside. Furthermore, MF1, a feature in cluster H, had a library match with another nitrogen-containing compound, DIBOA-O-hexoside, with a cosine score of 0.87. However, for MFs 2 and 16, located in clusters I and J, respectively, no plausible structural annotations were retrieved from the GNPS library search.