3.1 Effects on the types of VOCs in red clover flowers
A total of 170 volatile components from the process of flowering of red clover detected by GC–MS were analyzed via petal diagram and Venn diagram (Fig. 2A), which included 63 alkanes, 35 alcohols, 34 fatty acids and their derivatives, 12 aldehydes, 11 aromatic hydrocarbons, 4 olefins, 3 acids, 2 ketones and 6 others (see Supplementary Table S3). We found that 15 VOCs existed in five flower stages of all treatments. Among these 15 compounds, heptane, 1-octen-3-ol, phytone, phytol and heptadecane were reported to have various aromas, such as sweet, fruity and jasmine (Chung TY et al.1993; Nakaya S et al. 2015; Feng T et al. 2020). There were the least types (P < 0.05) of VOCs in the bud stage of CK (Fig. 2A), but nano Fe fertilizer caused obvious increases in the stage. However, the varieties of VOCs were reduced with the increase of Fe concentrations in Fe treatment groups. Fe-1g groups were more than Fe-2g except for the blooming stage.
As shown in Fig. 2B, alkane, alcohol and aromatic compounds were dominant as the main types of VOCs in red clover flowers, the sum of which accounted for more than 85% of the total VOCs (85.07–95.92%). The sums in the Fe treatments were higher than CK. Among them, the most kinds of alkanes were found in CK-2, followed by Fe-2g-3 and Fe-1g-4, and CK-1 had the least kinds of alkanes. It was worth noting that 63 alkanes were detected in GC-MS, accounting for 63.9% of the content of VOCs. Among them, 8 identical alkanes were found in all treatments, including tetracontane, heptane, 2,6,10-trimethyltridecane, eicosane, dotriacontane, hexadecane, heneicosane and heptadecane (listed in order of content). It was interesting that tetracontane and heptane, the most two abundant alkanes, decreased in half-open stages, and increased in blooming and mature stages compared with CK, while 2,6,10-trimethyltridecane was increased in all stages, especially in the blooming stages. In addition, the types of alcohols and aromatics showed an increasing trend during flower growth, indicating that the main aromatic compounds gradually accumulated during the flowering process. A decreasing trend was observed in the senescent stage. As fragrant compounds were reduced, the aroma of the flower faded. Overall, there were the most types of VOCs detected in Fe-1g-1, while Fe-2g-2 had the lowest (P < 0.05). Compared with other samples, alcohols, esters and aromatics in Fe-1g-1 were obviously more, which indicated a rich aroma at this stage. Furthermore, the change of VOCs’ types was completely different from those in CK. Among five stages of Fe fertilization, the types of aroma and alcohol compounds in the bud stage were the highest, which were thought to be useful to attract pollinators and resist Fe fertilization (Zhou et al. 2018; Rachersberger et al. 2019).
The clustering relationship between the relative abundance values of different compounds was analyzed to evaluate the differences in VOC anabolism in the process of flowering of red clover under different treatments (Fig. 2C). Among them, alkanes made up the largest proportion in the relative abundance of VOCs, accounting for 45–81% of VOCs, and exhibited gradually increasing trend with the advancement of flowering stages.
Ketones might be an important precursor for other compound types. High proportions of ketones appeared in the first stage, of which CK-1 had the highest proportion. Moreover, the clustering analysis showed obvious differences in five stages of the flowering process for CK treatment. The clustering relationships of the first and second stages were close for Fe treatments, indicating that appropriate nano Fe fertilizer accelerated the synthesis of secondary metabolites.
3.2 Effects on the contents of VOCs in red clover flowers
A heatmap was used to analyze the contents of 77 VOCs in different flowering stages and different concentrations of nano Fe (Fig. 3), in which 15 treatments were clustered into two distinct groups.
Fifteen compounds, such as 1-octane-3-ol, phytone, eicosane and chlorophyll, were found in all fifteen treatment samples. Fe-2g-5 was clearly distinct from the other treatments due to the high level of esters and acid compounds (n-hexadecanoic acid, linoleic acid ethyl ester, octacosyl acetate, oleyl acetate, etc.). The remaining 14 treatments belonged to two groups (A and B). All flowering stages of CK were located in Group B. The classification of groups A and B were consistent with the application of nano Fe fertilizer. The cluster analysis results clearly illustrated that the influence of Fe on VOCs was very significant. Although it has been mentioned that an increase of Fe concentration reduced the variety of VOCs in the previous analysis (Fig. 2), it was interesting that the content of the same compound in different stages increased with the application of Fe fertilizer. Then groups A and B were divided into 4 subgroups (A1, A2, B1 and B2) according to the correlation. The flowering stages CK1, CK2 and CK3 were in B2, while the end of flowering stages CK4 and CK5 were in B1. The distinctions between subgroups B1 and B2 were caused by the contents of 1-hentetracontanol, fucosterol, 1-triacontanol, (Z)-3-hexen-1-ol, heptanal, diethyl phthalate, tetracontane, octadecane and antioxidant 168, which may also be used to distinguish the blooming condition.
There were some specific compounds only found in the Fe treatments, such as 16-hentriacontanol, octacosan-14-ol, cholesterol, benzyl alcohol, butyl-2-methylvalerate, ethyl linoleate and cis-3-hexenyl acetate. Benzyl alcohol provided floral and rose aromas; ethyl linoleate exuded floral, fruity and chamomile aromas; cis-3-hexenyl acetate provide sweet, fruity, banana and apple aromas. These compounds with rich fruity and floral aromas likely attract pollinators, which may be beneficial to promote pollination, improve seed setting rates and increase economic potential (El-Sayed et al. 2018; Rachersberger et al. 2019; Bohman et al., 2020). These results indicated that Fe fertilizer was useful for promoting the metabolism of VOCs and improving the pollination efficiency of red clover.
Principal component analysis (PCA) was used to analyze the abundance of 77 compounds (covariance matrix) to assess the total variation in different treatment groups based on major VOC components (Fig. 4B). The first two principal components (PCs) explained 89.9% of the observed variation in VOCs, in which PC1 and PC2 explained 53.17% and 36.73%, respectively. The flowers of red clover were clearly classified into a nonfertilized group (CK group) and an Fe-fertilized group. The red clover flowers in the CK group were distributed throughout the flowering process, which demonstrated the excellent reproducibility of the analytical method. In addition, the Fe treatments were further divided into group A (Fe-1g-1, Fe-1g-2, Fe-1g-3, Fe-1g-4, Fe-1g-5, Fe-2g -1 and Fe-2g-3) and group B (Fe-2g-2, Fe-2g-4 and Fe-2g-5). The results of different types of compounds and different treatments were visualized via a chord diagram (Fig. 4A) to express the dynamic changes in each VOC content under different treatments. It was also observed that alkanes were the most abundant VOCs in each sample group and increased with the flowering stage, followed closely by alcohols and aromatics. The red clover flowers in the blooming stage contained the highest content of VOCs in CK. It is worth mentioning that Fe could delay VOCs release. Fe fertilizer was conducive to the enrichment of VOCs in the mature stage and senescent stage.
3.3 The effects on phytoestrogens in red clover flowers
To study the metabolism of phytoestrogens in red clover flowers, 30 known phytoestrogens were identified by LC–MS in the 15 treatments based on the accurate mass (Δm/z ≤ 5 ppm) (Fig. 5). Among treatments, the CK and Fe-2g groups had high phytoestrogen contents in the half-open stage and senescent stages, while significantly high contents were exhibited in mature stage for Fe-1g (P < 0.05). As reported in previous studies, Fe treatment could increase the content of phytoestrogens (polyphenols and isoflavones), especially total anthocyanins and flavonoids (Gryszczynska et al. 2018; Farhadi et al. 2020). A notable cause for concern is that some phytoestrogens were only present in early blooming stages (bud, half-open and blooming stages), such as (-)-epicatechin, cianidanol, astragalin, 6-methoxykaempferol 3-O-glycoside and isorhamnetin 3,7-O-di-beta-D-glucopyranoside. However, some phytoestrogens could be activated in advance in the half-open stage of Fe treatments, such as astragalin and 6-methoxykaempferol 3-O-glycoside. At the same time, several phytohormonal compounds were found to exist throughout the flowering process of Fe treatments, such as proanthocyanidin A2, 6-methoxykaempferol 3-O-glycoside and 3-hydroxycoumarin.
3.4 The characteristic compounds of red clover flowers
Multivariate statistical analysis via the PLS-DA model was performed to evaluate the characteristic compounds from 77 VOCs and 30 phytoestrogens in red clover flowers (Belmonte-Sánchez et al. 2019; Song et al. 2020). The relationship among the map of loadings and the scores was simultaneously displayed on the biplot via k-fold cross-validation (Fig. 6A). Through the PLS-DA biplot of red clover flower samples, it was observed that red clover flowers of Fe-1g-1, Fe-2g-1, Fe-2g-2 and Fe-2g-5 were significantly affected by Fe fertilizer. At the same time, the CK groups were clustered in the same quadrant, which also showed that this model could well distinguish CK and Fe fertilizer groups.
The PLS-DA model was also used to analyze the relationship between Fe fertilizer and secondary metabolites of red clover flowers. There were differences in the VOCs and phytoestrogens among red clover flowers of different treatments, especially Fe fertilizer, which completely changed the correlations of VOCs and phytoestrogens in different stages (Fig. 6A). The compounds with a strong correlation in CK-1 stage were (3R)-4'-methoxy-2',3,7-trihydroxyisoflavanone, diethyl phthalate and triacontyl acetate (Supplementary Table S4). Two alcohol compounds, 1-hexacosanol and (E)-2-hexen-1-ol, were closely related to CK-3, CK-4 and CK-5. The highly correlated compounds of CK-3 and CK-5 were the same. Although the related compound anthricin in CK-4 was different from 2,2',5,5'-tetramethyl-1,1'-biphenyl and genistein in CK-3 and CK-5, the correlations between their structures were high. The same phenomena were also observed in the Fe treatment groups. For example, the same related compounds were found in Fe-1g-3 and Fe-1g-4 or Fe-2g-1 and Fe-2g-2. It was interesting that the related compounds 6-methoxykaempferol 3-O-glycoside in Fe-1g-1, isorhamnetin 3,7-O-di-beta-D-glucopyranoside in Fe-1g-2, and isoquercitrin in Fe-1g-3 and Fe-1g-4 contained to the same flavone carbon skeleton, 3,4,7-trihydroxyflavone. These results indicated that the main secondary metabolites have obvious continuity during the flowering process, although nano Fe fertilizer significantly changed the highly related VOCs and phytoestrogens in different flowering stages. Moreover, the same related compounds were also in Fe-1g-2 and Fe-2g-3, which may be because the accumulation mechanisms of secondary metabolites in the plant flowering process were consistent for different Fe treatments, and a high concentration of Fe evidently delayed the accumulation of secondary metabolites.
The characteristic components of red clover flowers from different treatments were screened by the VIP method (Fig. 6B). Except 10 phytoestrogens (3-hydroxycoumarin, aloe-emodin, 3'-methoxydaidzein, biochanin A, 5-methoxy-6,7-methylenedioxyflavone, coumafuryl, anthricin, 6-methoxykaempferol 3-O-glycoside, (-)-olivil-4''-O-beta-D-glucopyranoside, and daturametelin E), 25 volatile compounds were screened out, including 8 alcohols (1-hexacosanol, 1-hexanol, campesterol, 2-methyl-1-butanol, 2-ethyl-1-hexanol, (E)-2-hexen-1-ol, and (E)-3-hexen-1-ol, octacosanal), 2 aldehydes (2-hexenal and heptanal), 1 ester (palmitic acid ethyl ester), 10 alkanes (tetrapentacontane, heptane, 2,2,3,3-Tetramethyl-butane, pentacosane, dotriacontane, heptadecane, 2-methylhexacosane, 2,4-dimethyl-undecane, 7-acetoxyeicosane, and 4,6-dimethyl-dodecane), and 3 aromatics ((3,3-dimethyldecyl) benzene, 2,3,6-trimethylnaphthalene, and 2,3-dimethylnaphthalene). Among them, (E)-3-hexen-1-ol has the highest importance in VIP values, and the VIP values of 3-hydroxycoumarin and (-)-olivil-4''-O-beta-D-glucopyranoside in phytoestrogens are also significantly high (P < 0.05). It is worth noting that palmitic acid ethyl ester and (E)-3-hexen-1-ol with high VIP values were reported to be indicative of an adaptation to butterfly pollination (Andreas 2004).