Interactions of a medicinal climber Tinospora cordifolia with supportive interspecific plants trigger the modulation in its secondary metabolic profiles

Tinospora cordifolia (TC) is scientifically proven immunomodulatory drug being used for centuries. Ancient literature reported that inter-specific interactions change medicinal properties of TC. Thus, the current study is aimed to understand the influence of interspecific biotic interactions on chemo-profiles of TC. To explore it, TC samples collected from six co-occurring plants, i.e. Azarditchita indica, Acacia nilotica, Albezia lebbeck, Ficus benghalensis, Tamarandus indica and Acacia leucophloea were analyzed by HPLC-ESI-QTOF-MS. Mass data were subjected to multivariate analysis. Support vector machines (SVMs) was found to be best classifier (r2 < 0.93). Data analysis showed the specific compounds in all TC due to inter-specific interactions. Data were further analyzed with SNK post-hoc test followed by permutative (n = 50) Bonferroni FDR multiple testing correction. The compound without any missing values reduced the number of variables to 133 (p < 0.01). Statistical analysis revealed that TC having interactions with A.lebbeck and A. nilotica formed the most distant groups. However, TC co-occurred with A. indica showed the highest number of up-regulated metabolites, including jatrorrhizine, chrysin, peonidin, 6-methylcoumarin and some terpenoids. Some metabolites, including jatrorrhizine and magnoflorine were quantified to confirm the accuracy of qualitative analysis. Results demonstrated the influence of inter-specific biotic interactions on TC chemo-profiles, hence its medicinal properties.

Multivariate statistical analysis. Fifty nine major peaks were observed in the chromatogram when integrated. All the spectra were aligned using ion intensity, retention time (<0.2 min) and mass (<5 ppm) with the help of internal standards i.e. ions of m/z 296. 15, 373.13, 311.13, 230.24, and 436.44 present universally in all the samples. Final data were normalized using Z-transforms. Data sets were subjected to one way ANOVA (p < 0.05), fold change (>2.0) and coefficient variation (>15%) analysis. ALL and ANI groups showed the highest number of down-regulated metabolites. Box Whisker plots of the data revealed least variability in the ALL group as compared to other samples. All the groups showed more variability in the upper quartile portion of Box Whisker plot ( Fig. 2A). Supervised PCA was performed on all the datasets and visualized to check for outliers and classification trend among the samples (Table S1). Principal components have been extracted from the variables in the datasets. Statistical analysis involves principal component analysis projection to latent structures for identifying variation in spectral features of samples. PCA of 7 groups resulted in 1643 principal components. Each groups was observed to be distinct, forming their own cluster and lying far apart from each other. All the groups showed 23.42, 17.19 and 13.62% variations along the X, Y and Z axis respectively (Fig. 2B). Supervised PCA plot showed highest variations in AIN and ALC groups as compared to other groups. Further, data were subjected to multivariate analysis to identify and reveal differentially expressed metabolites in different groups. Initial analysis showed confidence in variability among the groups and the presence of distinct metabolites. Already established separation among the groups was sharpened by multivariate analysis. Data was further subjected to PLS-DA, SVM, NB, DT and NN classifiers for preparing respective classifier models 13 . Classifier models expressed confidence ranging from 1 to 0.896 (Table S2). All the models were trained for further prediction of unknown samples. Mass data files of different plant extracts and T. cordifolia (collected from different supporting trees) extracts were subjected to the trained models for classification and identification. A trained model has classified all the unknown samples and prediction measure were expressed as confidence measure. Trained model of PLS-DA, NB and DT failed to classify all known samples (Table S3), whereas SVM and NB classified all the extracts correctly and showed least r 2 for the samples other than T. cordifolia (Table S2). Furthermore, SVM was found to be a better discrimination model and powerful classification tool in real-world applications due to expression of better confidence levels in T. cordifolia samples and least confidence in other extracts along with its excellent learning performance as reported earlier 14 . SVM data shown to have 1083 ranked metabolites from all the seven samples. Being best model, SVM data were subjected to further analysis to reduce the number of metabolites and to keep only significant variables. Differential metabolites of T. cordifolia due biotic interactions with different trees. Venn diagrammatic representation of specific metabolites in respective groups showed the presence of 168, 167, 59, 126, 83, 80, and 150 differentially expressed metabolites in CON, AIN, ALL, ALC, ANI, TMI and FBG groups respectively. It showed the metabolites present in specific groups only, and/or may be due to missing values in data. Venn diagrammatic representation showed the presence of 6, 2, 7, 5, 5, 3, and 0 metabolites exclusively in CON, AIN, ALL, ALC, ANI, TMI and FBG groups, respectively (Fig. 3). It showed that 8-hydroxytinosporide and one unknown terpenoid were exclusively present in the AIN group while the control group contains 5-allyloxysalvigenin, trans-farnesol, reticuline, N-isovaleroylglycine and two unknown metabolites. The only ANI group contained palmatoside C, α-D-glucan, 5-aminovaleric acid and two unknown metabolites of higher molecular weight while, ALL group contained tinosporinone, baenzigeroside A, tinosinen, tinoridine and two unknown metabolites. The TMI group contains 11-hydroxymustakone along with two unknown metabolites ( Table 2). Data were subjected to One-way ANOVA with SNK post-hoc test and asymptotic p values were computed with permutative (n = 50) Bonferroni FDR multiple testing correction. After analysis, 229 metabolites Base peak chromatograms of all seven groups extracted from total ion current chromatograms (TIC), showing visual differences among the chromatographic profiles of various groups. A, B, C, D, E, F and G represents the ALC, ANI, ALL, AIN, CON, FBG, TMI groups respectively.

Intra-day (n = 9)
Inter-day (n = 9) values were removed due to the high number of variables that reduced the number of discriminated metabolite to 133 ( Table 3). The FBG group did not contain any specific metabolite. Observed mass differences and fragment ions are given in Table S3. Spearman correlation heat map of groups and metabolites without missing values is shown in Fig. 4. Spearman correlation heatmap analysis showed that TMI and FBG, control and ALC groups were close to each other, whereas ALL and ANI being more distant groups. Clusters of metabolites in red color and inter-spreading metabolites in blue color across the different groups are highly variable regions showing differential expression of metabolites in that particular group (Fig. 4) Other up-regulated compounds whose level was found to be >2 fold but <3 fold were tinosporaside, jatrorrhizine, glucoside 493, 1-{hydroxy[2-(trimethylammonio) ethyl]amino}-1-oxo-2-dodecanaminium, 6-methylcoumarin, and cinnamaldehyde (Table 3)  Up-regulated metabolites having fold change <2.0 as compared to control but >2.0 as compared to other groups were β,5α,14α-trihydroxyergosta-7,22-dien-6-one, cycloeucalenol, oblongine, 20-hydroxyecdysone, phenylmethanethiol, and cycloartane-24,25-diol-3-one ( Table 3) Palmarin, feruloyltyramine, betaine, 1,3 dimethylpteridine-2,4-dione, sodium thiosalicylate, haplopine, isoquinolone alkaloid, amino-tridecanoic acid, glucoside of 207, tetramethylazobenzene, scalar-17(25)-en-19-ol were identified metabolite having fold change <1.5 as compared to control but have significant differential levels when compared to other groups ( Table 3)   Differential metabolites in ALC group. In this group, compounds of 108, 389, 604, 871 Da and tinocordifolioside were identified exclusively. Precursor ion scan of the protonated molecule at m/z 413 afforded the products ion at m/z 251 die loss of 162 Da (glucose moiety), hence identified as tinocordifolioside. Other compounds uniquely present in this group could not be identified. ALC group also showed elevated levels (>2.0 fold) of isotanshinone IIA, tinosporaside, saponin, N-methylcoclaurine (described in AIN group), dehydrocorybulbine, and palmatine ( Table 3) ] + • due to the loss of methyl radical. It produced ions product ions at m/z 322 and 308 due to sequential losses of hydrogen radical and CO. Product ions including others at m/z 294, 278, 264 and 250 were equivalents to mass spectra obtained from synthetic palmatine (Fig. S4), hence compound was identified as palmatine.

Differential metabolites in ANI group.
In this group α-D-glucan, 5-aminovaleric acid, Palmatoside C and two unidentified compounds at m/z 775 and 795 were found to be uniquely present. Precursor ion [M + H] + at m/z 181 showed product ions as of in glucose when compared standard, hence named α-D-glucan (breakdown unit of glucans). Protonated molecule [M + H]+ at m/z 118 produce product ions at m/z 101 and 100 due to loss of NH 3 and H 2 O molecules, therefore identified as 5-aminovaleric acid also confirmed using the monoisotopic mass error <5 ppm. Biotic interactions of A. nilotica (ANI) with T. cordifolia displayed elevated levels of isotanshinone IIA and (9Z,14Z)-octadeca-9,14-dien-6-ynoic acid ( Table 3). Most of the other metabolites were found to be down-regulated as in case of ALL group that includes oblongine, palmatine, tinocordifolin, cinnamaldehyde, feruloyltyramine, magnoflorine, and jatrorrhizine well-known metabolites from T. cordifolia. A number of other metabolites were also found to be up-regulated in CON group as compared to other groups (Table 3) (Table 4). Calibration curve for the standard compounds exhibited good linearity in the measured range of 0.25-50 ng/mL. The coefficients of linear regression were determined from 5 independent experiments. The calibration curves by means of weighted (1/x2) least squares linear regression were y = 0.24x + 0.026 and r 2 = 0.9892 for berberine, y = 0.029x − 0.0025 and r 2 = 0.9987 for jatrorrhizine, y = 0.084x + 0.027 and r 2 = 0.9992 for palmatine, y = 0.122x + 0.028 and r 2 > 1.000 for choline and y = 0.0962x + 0.343 and r 2 = 0.9984 for magnoflorine. On the basis of calibration curves, the contents of berberine, choline, palmatine, magnoflorine, and jatrorrhizine in T. cordifolia extracts were quantified (Table 5). AIN group was found to have highest content of jatrorrhizine (2.88 ng/mL) as noticed in the qualitative analysis. FBG group was found to be endowed with (2019) 9:14327 | https://doi.org/10.1038/s41598-019-50801-0 www.nature.com/scientificreports www.nature.com/scientificreports/ significantly high concentrations of magnoflorine and choline confirmed the validity of qualitative analysis. Choline, a major component of cell wall was found to be highest in control (T. cordifolia climbed on the steel pole). ALC was found to have highest levels of palmatine (4.16 ng/mL) ( Table 5). Overall, quantitative analysis confirmed the authenticity of results obtained in qualitative analysis.

Discussion
To explore the chemoprofiles of T. cordifolia climbed on different trees, a simple but efficient HPLC-QTOF-MS based method was used. HPLC-QTOF-MS method is very precise having >98% interday and intra-day accuracy. It has also recovered > 98% of seck its efficiency (Table 1). Both factors are important to standardize in order to minimize the variability and error rate across the data, in particular for qualitative analysis. Retention time alignment and mass data intensity normalization also affect the quality of analysis. Hence, the minimum difference in retention time and mass error noticed in this study were <0.05 min and <5 ppm respectively that further enhanced the data quality. Multivariate analysis methods are strong tools to differentiate the variations across the data 21 . In this study SVM and NB classifiers have been found to be the most accurate with r 2 < 0.97. To decrease the false discovery rate, SNK post-hoc test with permutative (n = 50) Bonferroni FDR multiple testing correction was employed to increase the confidence in the analysis. The Venn diagram analysis has predicted the unique or the most abundant compounds across the different groups that can further be used as markers to identify the specific groups (Table 2).
T. cordifolia co-occurred with A. indica (AIN) contains 8-hydroxytinosporide and one unknown triterpenoid that can be used to differentiate this group from others. 5-Allyloxysalvigenin, trans-farnesol, reticuline and N-isovaleroylglycine are found in the control (T. cordifolia co-occuredwith steel pole), whereas, palmatoside C, D-glucan and 5-aminovaleric acid were found only in the ANI (T. cordifolia co-occurred with A. nilotica). Other group specific compounds include tinocordifolioside, tinoridine, tinosporinone, baenzigeroside A, tinosinen and 11-hydroxymustakone. All these compounds are found to be the group specific. However, medicinal importance of these compounds needs to be explored. Further fold change analysis have shown the highest variables in the T. cordifolia that co-occured with A. indica and steel pole or AIN and CON groups. Least number of variables are observed in TMI, ALL and FBG groups, i.e. T. cordifolia co-occured with T. indica, A. lebbeck and F. benghalensis respectively (Table 3). These qualitative results are further verified by quantitative analysis with some standard compounds. It is noticed that AIN contains the highest quantity of jatrorrhizine (2.88 ng/mL) and ALC group contains palmatine (4.16 ng/mL) and magnoflorine (7.65 ng/mL) ( Table 5). The results of quantitative analysis are in accordance with qualitative analysis and have increased the confidence in the data. AIN group is also found to be a rich source of specific and up-regulated terpenoids including 8-hydroxytinosporide, unidentified terpenoids of 457 Da, borapetoside D, N-methylcoclaurine, isotanshinone II, oblongine, tinosporaside and cycloeucalenol, alkaloid jatrorrhizine and flavanoids i.e. peonidin, 5-allyloxysalvigenin and chrysin (Table 4). More specifically, AIN have increased contents of terpenoids along with specific alkaloid that synthesized from different route and fllvanoids.
Plants are chemically divergent and differe in biochemicals with plant species [22][23][24][25][26] . The plant chemistry mediates ecological interactions and influence the selection traits including specific constitutes 27 . Inter and intra-specific ecological interactions or competition within plant communities has been reported to influence the selection of an allelopathic secondary plant compound 28 . Hence, various kinds of selection forces may lead to different chemotypes within a plant species. In the current study, AIN group which is reported with the highest immunomodulatory activity is found to have high levels of terpenoids, jatrorhizine, flavonoids, coumarins, phytosterols and other categories compounds. The 8-hydroxytinosporide, scalar-17(25)-en-19-ol, borapetosides D,  www.nature.com/scientificreports www.nature.com/scientificreports/ cycloartane-24,25-diol-3-one, palmarin, tinosporaside, and isotanshinone IIA are the terpenoids. One study has reported a significant decrease in concentrations of tinosporaside and berberine during winter season 29 . However, in the group AIN tinosporoside is found to be elevated as compared to other groups. High levels of borapetoside D, isotanshinone II, cycloeucalenol, chrysin, and methylcoumarin are found in AIN. These metabolites have been reported to have antioxidant properties and increase glucose uptake [30][31][32][33][34] . Tanshinone II, an analogue of isotanshinone II has been reported to have anti-fatigue properties by increasing the muscle glucose uptake and decreasing the lactic acid production 35,36 . Tanshinone II A is also found to be effective in treatment of cardiac problems despite of its less absorption through intestine 37 .
Concentrations of isoquinoline alkaloids i.e. N-methylcoclaurine, isoquinolone alkaloid, jatrorrhizine, oblongine, and haplopine are also found to be the highest in AIN group. N-methylcoclaurine inhibits the solute carrier organic anion transporter family member 1B1 and 1B3 38 . It has also been established that jatrorrhizine, oblongine, N-methylcoclaurine and magnoflorine are agonist of dopamine receptors 39,40 . Compounds that are monoamine depletors, could be useful to treat Parkinson or Huntington's disease. Interaction of these alkaloids with dopamine receptors also modulate c-AMP metabolic process and cytosolic calcium ion concentration that are important in regulation of energy balance and other metabolic processes [41][42][43] . These alkaloids being dopamine receptors agonists activate trigeminal motoneurons that increase the muscle tone 44 . These compounds along-with above reported terpenoids, flavonoids and coumarin are also capable to modulate MAPK, ERK1, ERK2 and protein kinase A that are involved in signal transduction pathways to control the innate and adaptive immunity and cell death [45][46][47] . Cinnamaldehyde and sodium thiosalicylate along-with 20-hydroxyecdysone elevated in AIN group are known anti-inflammatory compounds and boost up immune system [48][49][50] . Hence, synergistic effect of these compounds along with jatrorrhizine might be a possible reason of the highest immunodulatory activities of the crude extract of T. cordifolia co-occured with A. indica. However, the role of other compounds present in minute quantities cannot be ignored and need to studied carefully.
The present study has clearly established that different chemotypes of T. cordifolia are due to biotic interactions with other plants. Levels of choline contents is affected as observed in the present study by interspecific interactions and therefore, the cell membrane composition for the ease of working of interactive compounds from both species. Hence, T. cordifolia co-occured with A. indica have the best medicinal efficacy due to the presence of high concentrations of specific terpenoids, alkaloids, and flavonoids which affect ERK1, ERK2 and MAPK cascade, c-AMP metabolic process, cytosolic calcium ion concentration and angiogenesis. However, more studies are required to explore the individual or syngesitic mechanism of these compounds. The study also emphasizes the necessity to study secondary metabolites accumulation in the plants under normal and biotic interaction conditions. Identification and understanding of the biotic and abiotic factors which influence the secondary metabolites production in T. cordifolia may also help to increase its medicinal efficacy. Better understanding of the ecological interactions due to co-occurrence of T. cordifolia with other plants will help to increase its medicinal efficacy and also to understand physiology of plant. Preparation of extracts. The stem samples were washed with tap water followed by deionized water to remove soil and other traces. These were dried in the air for 4 weeks. The air-dried stems were chopped and further converted into fine small pieces by mixer grinder (Philips, India). Dried stems (10 g) was extracted overnight with deionized water (Direct-Q, Millipore) (1:1 w/v) in orbital shaker at 37 °C and 180 rpm to yield the thick juice. Extracts were then centrifuged at 15,000 g for 10 min at 4 °C. The extraction was repeated three times for each sample and the supernatant was collected. The percolate from three repeats of each sample was then concentrated in a rotary vacuum evaporator at 50-60 °C. The supernatant juice was quick-freezed at −80 °C (Thermo-Fisher, Germany) and lyophilized (Freezone 4.5 Labconco, CA, USA) to yield a dry homogenous powder (0.3 g) and stored at −80 °C. The lyophilized powder from various samples was reconstituted in LC-MS grade water to make a solution of 5.0 mg/ml. Solutions were vortexed and centrifuged at 15,000 g for 20 minutes at 4 °C temperature. The supernatants were carefully removed and filtered through 0.22 µm syringe filters and transferred to 96 well plates. The complete workflow of the study design is depected in Fig. S1.

Methods
HPLC-ESI-QTOF-MS. T. cordifolia stem extracts were resolved over ZORBAX Eclipse Plus reversed phase column (C18, 2.1 × 250 mm) having particle size 5 µm. Auto-sampler and column temperatures were maintained at 4 °C and 40 °C. Injection volume was kept constant, i.e. 20 µl for all samples. Chromatographic separation was carried out with Agilent 1200 Series HPLC interfaced to an Agilent 6520 Accurate-Mass QTOF-MS, with mobile phase A (water containing 0.1% formic acid) and B (acetonitrile containing 0.1% formic acid). The gradient program was carried as follows: