In all three varieties studied and fruit tissues, most limonoid aglycones (C22, C24, C26, C27, C29, C30 and C31) along with glycoside C21 showed maximum increases at S2 and subsequently reduced their concentration to reach different levels. By contrast, the rest of glycosylated limonoids (C20, C23, C28 and C32) increased their concentration throughout all developmental stages, reaching their maximum concentration at full ripeness (S4). Nevertheless, some deviation from this accumulation pattern was detected regarding tissue and fruit type. For instance, limonoid aglycone C24 could be barely detected in Wase mandarin fruit tissues whereas downstream metabolites such as C22, C26, C27, C29 or C30 showed significantly higher levels in this genotype, suggesting enhanced hydrolase and deacetylase activities rendering C22 and, subsequently, C26 [
28]. In orange pulp, C24 showed an accumulation trend up to S3 stage and then dropped to minimal values, whereas in albedo a transient and isolated maximum was recorded at S2. Downstream metabolites were at significantly higher levels compared to grapefruit or mandarin, especially at S2 (
Figure 3) suggesting an overall activation of the pathway both supplying the precursor C24 and also transforming it. Interestingly, C24 showed low albeit constant levels in grapefruit albedo after S1 and higher levels in pulp, especially at S3. This trend was associated to reduced levels of C27 and C29 and the absence of C30 in this genotype (
Figure 2), consistent with the reported activity of limonin D-ring lactone hydrolase for Duncan grapefruit in comparison to sweet orange varieties [
28]. Finally, the bitter limonoid C31 showed a sharp accumulation at S2 in mandarin but it rapidly reduced reaching very low levels at S4, particularly in albedo tissues. Conversely, levels of C31 in grapefruit slightly increased after S1 and remained barely changed thereafter showing the highest values at the full ripe stage S4 (
Supplementary Figure S2). For this metabolite, the accumulation profile observed in orange fruit tissues resembled that of mandarin but with significantly lower levels throughout the developmental process. This reversion of C31 levels in albedo and pulp tissues of the two sweeter varieties could be a result of the enhanced expression of limonoid UDP-glucosyl transferase and the use of C31 as a substrate instead of C30. This proposal is consistent with the observed expression of limonoid UDP-glucosyl transferase gene showing a maximum at S3 in mandarin (
Supplementary Figure S5). The accumulation profile of the end-products of the pathway: C30, C31 and C32, known to have a role in citrus fruit bitterness, could be clearly associated with the expected taste trait for each citrus variety and tissue (
Supplementary Figure S3). Interestingly, the tasteless metabolic precursor of C31, C30, could not be detected in albedo or pulp tissues of Duncan grapefruit whereas its concentration was remarkably high in sweet varieties (mandarin > orange) especially at S4 (
Figure 3 and
Supplementary Figure S3).
The flavonoid pathway in citrus arises from naringenin chalcone rendering different chemical compounds differing in their polyphenolic core structures, namely flavanones, flavones, flavonols, etc and their derivatives. Most flavonoids studied in this work reduced their relative levels in pulp and albedo throughout the developmental process reaching different values depending on the citrus genotype (
Figure 4 and
Supplementary Figure S3), consistent with the reduced expression of chalcone isomerase gene after S1 (
Supplementary Figure S5). For most flavonoid compounds such as C1, C3, C7, C9, C10, C11, C12, C13, C14, C15 and C16 there was a clear partitioning towards the edible part showing significantly increased levels in pulp. For polymethoxylated flavones (C4–C6) the situation was the opposite (
Figure 4), showing higher values in albedo. The occurrence of some metabolites was restricted to specific genotypes such as C8, C10, C11 and C12, only present in orange and mandarin, C7, C14 and C15 only present in grapefruit and mandarin and C3 and C19, only present in grapefruit and orange. This could indicate that C13 could be the precursor of the rest of the kaempferol derivatives, particularly C12 and C14, with the addition of a deoxyhexose and a hexose and a caffeoyl moiety, respectively. The precursor of C15 is less clear, although it is possible that a hydroxymethyl glutaryl moiety is attached to a glycosylated kaempferol core structure, constituting C13 also a potential intermediate in the reaction. For isorhamnetin glycosyl derivatives, a clearer picture is devised, being C9 the plausible precursor and C10 and C11 the derivatives whose synthesis is abolished in grapefruit, either the addition of hexose or deoxyhexose. In this case, a different enzyme to that involved in glycosylation of kaempferol derivatives is likely involved. In the other branch, hesperidin (C2) attached to hesperidoside and neohesperidin (C3) attached to neohesperidoside showed nearly opposite trends, the former being highly accumulated in mandarin tissues, especially in albedo, and the latter showing a strong accumulation in pulp tissues in grapefruit, but not detected in mandarin. These results could be partially explained by the genetics of citrus: grapefruit is a hybrid between pummelo (
C. maxima) and sweet orange (
C. sinensis). In turn, sweet orange is likely a result of the hybridization of pummelo and the ancestral mandarin (
C. reticulata). Finally, Wase mandarin, a satsuma mandarin (
C. unshiu), is thought to be the result of backcrossing a pummelo and a mandarin hybrid [
34] as reflected in
Supplementary Table S1. The ability to synthesize C12, C14 and C15 is likely an ancestral mandarin trait somehow lost in grapefruit and sweet orange through several backcrosses and selection. Similarly, glycosylation of C9 to render C10 and C11 also seems to be an ancestral mandarin trait, absent in grapefruit. Regarding the synthesis of C3, this is possibly a pummelo trait, lost in satsuma mandarin. Although, the actual carbohydrate positioning of C7 and C8 could not be determined, mass spectrometry data allow us to conclude that both metabolites are two isomeric molecules sharing the eriodictyol core structure: C7 is absent in orange and C8 in grapefruit. In mandarin, despite the two molecules being detected, a tissue specialization in the accumulation of each isomer was observed. A plausible explanation to this observation is that the ability to synthesize both compounds is an ancestral mandarin trait partially inherited by sweet orange and subsequently grapefruit. Indeed, the origin of satsuma mandarin (late admixture mandarins type 3, according to [
34]) could explain the presence of both metabolites in this genotype. In grapefruit, metabolite profiles were predominantly composed by neohesperidosides (
Figure 4 and
Supplementary Figure S3). This is likely a result of the upregulation of genes involved in the transformation of core flavonoid structures rendering bitter 1,2-rhamnosyl derivatives (e.g., naringin) [
5,
20]. By contrast, sweet varieties such as mandarins and sweet oranges preferentially accumulated
O-rutinosides as described previously in [
15,
17] and also reported in this work (
Figure 3 and
Figure 4). Interestingly, no single compound was found to be absent in both oranges and mandarins, despite their ‘sweet’ trait, reinforcing the admixture origin of all three fruit types [
34].
Throughout the entire developmental process, flavonoid concentration, unlike limonoid glycosides, was progressively reduced showing very little values at S4 in both tissues (
Figure 4 and
Figure 5) except for C14 in albedo (
Figure 4,
Figure 5,
Figure 6 and
Figure 7, and
Table 3). These observations are consistent with the downregulation of chalcone isomerase gene (
Supplementary Figure S5) in pulp tissues. Despite this general trend, differences in the starting and final concentrations were observed among genotypes and tissues (
Figure 4). In this sense, typical ‘bitter’ compounds such as C19 showed the highest values in grapefruit followed by sweet orange and presented minimal levels or were not detected in mandarins (
Figure 4 and
Supplementary Figure S3).