Biogeography shaped the metabolome of the genus Espeletia: a phytochemical perspective on an Andean adaptive radiation

The páramo ecosystem has the highest rate of diversification across plant lineages on earth, of which the genus Espeletia (Asteraceae) is a prime example. The current distribution and molecular phylogeny of Espeletia suggest the influence of Andean geography and past climatic fluctuations on the diversification of this genus. However, molecular markers have failed to reveal subtle biogeographical trends in Espeletia diversification, and metabolomic evidence for allopatric segregation in plants has never been reported. Here, we present for the first time a metabolomics approach based on liquid chromatography-mass spectrometry for revealing subtle biogeographical trends in Espeletia diversification. We demonstrate that Espeletia lineages can be distinguished by means of different metabolic fingerprints correlated to the country of origin on a global scale and to the páramo massif on a regional scale. Distinctive patterns in the accumulation of secondary metabolites according to the main diversification centers of Espeletia are also identified and a comprehensive phytochemical characterization is reported. These findings demonstrate that a variation in the metabolic fingerprints of Espeletia lineages followed the biogeography of this genus, suggesting that our untargeted metabolomics approach can be potentially used as a model to understand the biogeographic history of additional plant groups in the páramo ecosystem.

suggesting a possible esterification of a caffeic acid unit with a quinic acid molecule. Its unambiguous identification was possible based on Rt and HRMS comparisons with a reference substance. Compounds 1, 2 and 4 (Table S1, Fig. S1), identified as quinic acid, protocatechuic acid and p-coumaric acid, respectively, were also identified based on Rt and spectral comparisons with reference substances. Four di-caffeoylquinic acid isomers (compounds 5, 13, 15 and 17, Table S1 (Table S1) was proposed as another di-caffeoylquinic acid isomer, considering that such compound showed a precursor ion at 515.11926 m/z in negative mode with fragments at 353.08780, 191.05530 and 179.03407 m/z, which is in accordance with the literature 3 . Additionally, its UV spectrum presented two maximum absorptions in ca. 300 and 325 nm, which characterize chlorogenic acids. Compound 12 (Table S1, Fig. S1) was proposed as methyl 3-Ocaffeoyl-4-O-feruloylquinate 4 based on its mass and UV spectra. This compound showed a precursor ion at 543.15009 m/z in the negative ionization mode and a UV spectrum characteristic of chlorogenic acids.
In the positive mode of ionization this molecule presented extensive fragmentation with a parent ion at 545.16455 m/z. The fragment observed at 527.15363 m/z indicates a possible loss of a neutral molecule of H2O, while the fragment at 351.10672 m/z suggests the loss of a water molecule and a feruloyl unit.
Additionally, we observed two peaks at 177.05429 m/z and 163.03853 m/z originated after the loss of a caffeoylquinate unit in the first case and the loss of a feruloylquinate unit in the second.
Compounds 9, 19 and 20 (Table S1, Fig. S1) were identified as caffeoylaltraric acids based on their UV and mass spectra. Compound 9 was identified as a di-caffeoylaltraric acid isomer. This compound presented a precursor ion at 533.09363 m/z in the negative mode and fragments peaks at 371.06198 and 209.02956 m/z, corresponding to successive losses of one and two caffeoyl units, respectively (Table S1). According to literature reports a product ion at 209 m/z is characteristic of galactaric acid 5 . On the other hand, compounds 19 and 20 (Table S1, (Table S1, Fig. S1) were identified as quercetin, 3-methoxy quercetin and kaempferol, respectively, based on their spectral and Rt comparison with reference standards. Compound 24 (Table S1, Fig. S1), also an aglycone, was proposed as pinobanksin 7 . This compound presented a UV spectrum with two maximum at ca. 230 and 290 nm, which is in accordance with the UV absorption values reported in the literature for the same compound 8 . Its MS 2 spectrum presented a deprotonated molecule at 271.06125 m/z and product ions at 165.01825, 151.00252, 119.04893 and 107.01253 m/z, which is in accordance with the MS 2 spectrum reported for pinobanksin 7 . Compound 26 (Table S1, Table S1).
Among the glycosylated flavonoids, compounds 8, 10 and 16 (Table S1,   This compound presented a precursor ion at 711.14093 m/z, a product ion at 463.08780 m/z, corresponding probably to the loss of a manoylglucosyl unit and a base peak at 301.03522 m/z corresponding to the aglycone. The same pattern of glycosyl losses and a subsequent peak representing the flavonoid aglycone was also observed in the MS 2 spectra of compounds 11 and 14 (Table S1, Fig.   S1), identified as quercetin-3-O-arabinoside 12 , isorhamnetin 3-glucoside 13 , respectively. Compounds 18 and 21 (Table S1, Fig. S1) correspond putatively to two flavonoids esterified with caffeoyl or cinnamoyl units based on their characteristic UV spectra, which presented the usual signals of chlorogenic acids (two maximum absorptions in ca. 300 and 325 nm), and fragmentation patterns. Compound 18 (Table S1, Fig. S1) Table S1, Fig. S1) correspond to ent-kaurane derivatives, which constitute the most commonly reported secondary metabolites in the subtribe Espeletiinae [15][16][17] .

Triterpenes
The chemical class of three putative triterpenes (Compounds 44, 45 and 46, Table S1, Fig. S1) was tentatively proposed based on their accurate mass measurements and data base screening. It is worth mentioning that the identity of each of these compounds was not reported here, as additional information regarding the MS 2 fragmentation patterns of the several possible hits (all of them triterpenes) identified in the DNP lacks in the literature. These three compounds showed the same fragmentation pattern in the MS 2 experiments with an initial loss of 150.0683 Da to form the base peak and a subsequent peak at 149.05972 m/z. Additionally, as previous phytochemical studies with species of the genus Espeletia have reported this class of metabolites it may be safe to assume its presence in some of the plant samples analyzed [18][19][20] .

Sesquiterpene lactones
We identified three sesquiterpene lactones, namely fluctuadin, longipilin acetate, and polymatin B (Compounds 31, 34 and 35, Table S1, Fig. S1, respectively) based on accurate mass comparisons with a reference standard (compound 34, Table S1, Fig. S1) and with compounds previously reported in Espeletiinae and in the genus Smallanthus (compounds 31 and 35, Table S1, Fig. S1). Compounds 34 and 35 were previously isolated from the leaves of E. killipii and E. tunjana and show chemotaxonomical significance 20 . On the other hand, compound 31 has not been previously reported in the subtribe Espeletiinae, but it is commonly found in the genus Smallanthus 21 , the sister group of Espeletiinae based on molecular markers 22 .