Plantago lanceolata, L.

Gonda, Sándor, Kiss-Szikszai, Attila, Szucs, Zsolt, Máthé, Csaba & Vasas, Gábor, 2014, Effects of N source concentration and NH / NO ratio on phenylethanoid glycoside pattern in tissue cultures of Plantago lanceolata L.: A metabolomics driven full-factorial experiment with LC-ESI-MS, Phytochemistry 106, pp. 44-54 : 45-51

2.1. Effects of N source on growth of P. lanceolata calli

The calli of P. lanceolata were cultured on modified Murashige Skoog media with different NH 4 + /NO 3 – ratios and N source concentrations in a full factorial experiment. For experimental design, nomenclature of media and details, please see Section 4.4.

Growth indices on different media ranged from 4.52 ± 0.10 (40(0.33)) to 9.87 ± 3.31 (20(0)). Unsurprisingly, the concentration and composition of the N source influenced the growth of the P. lanceolata TC. A more detailed insight was obtained by subjecting all known medium parameters to ANOVA models, with the growth index being the response variable. It was shown, that N source concentration (ranging from 10 to 60 mM) did not significantly alter growth of the calli, while NH 4 + /NO 3 – ratio had a more pronounced effect. Addition of different amounts of NH 4 + and NO 3 – required counter-ions, which were Na + and Cl –, added in the range 0–41 and 0–16 mM, respectively, to the different media. The amount of these ions was shown to have only insignificant effects on growth of the TC. Higher concentrations of NH + 4 have led to decreased growth of the tissue cultures. This is in accordance with that described by Budzianowska et al. (2004) and is usually attributed to the direct toxicity of excess NH + 4 at higher concentrations ( George et al., 2007). Examination of the effects on growth was secondary, our main focus was to detect the changes in the metabolome, which will be detailed in the following sections.

2.2. Phenylethanoid glycosides in P. lanceolata calli

After a few preliminary tests, calli from media with 10 mM N were selected for the qualitative study. Ten microliters of the concentrated MeOH extract were injected into a semi-preparative HPLC system (Supelco C 18 column, 250 mm × 10 mm × 5 µm; gradient was adapted from the LC–MS method described in Section 4.5) for HPLC-UV–Vis measurement. Most of the NPs in the polarity range examined later were found to have the UV–Vis spectrum characteristic to phenylethanoid glycosides, i.e. maxima around 248, 290 and 330 nm were found ( Shi et al., 2013). These wavelengths are unaffected by methylation of the aglyca ( Shi et al., 2013). The same pattern was found by Budzianowska et al. (2004), who reported lack of other phenolic NP groups in P. lanceolata TC.

Negative mode LC–ESI–MS 3 was shown to be a very suitable method for characterization of the NPs from the MeOH extracts of P. lanceolat a calli. A high number of [M – H] – peaks were detected in the total ion chromatogram ( Fig. 2 ). Putative identification was carried out using detailed study of the fragmentation patterns of the two authentic standards plantamajoside and acteoside, application to the similar PGs, and consulting the scientific literature. Fragmentation patterns of most PGs also present in these samples are described in Li et al. (2009), Petreska et al. (2011), Qi et al. (2012) and Shi et al. (2013). For practical reasons and to show the ‘‘taxonomy’’ of the detected molecules and fragments, the abbreviations of the PG subunits (glucose (Glc), caffeic acid (CA), hydroxytyrosol (HT), etc.) will be used to indicate what parts make up a particular fragment: e.g. [HT-Glc(Glc)-CA – H] – is the m / z 639 ion shown in Fig. 3 . The NPs that could be putatively identified from the MS 3 fragmentation are shown in Table 1 .

Major PGs were shown to be plantamajoside (18, [M – H] –: [HT-Glc(Glc)-CA – H] –) and acteoside (verbascoside, 27, [M – H] –: [HT-Glc(Rha)-CA – H] –), their retention times and mass spectra were identical with our authentic standards ( Fig. 2 ). Their isomers (43 for 27 and 37, 49 for 18) were also abundant, these can either be isoacteoside and isoplantamajoside, or cis-acteoside and cis-plantamajoside ( Li et al., 2009) – cis/trans pairs can appear as separate peaks in HPLC according to Budzianowska et al. (2004), even from solution of purified standards. The fragmentation routes of 18 and 27 were identical to that of acteoside ( Li et al., 2009). The proposed fragmentation pattern of plantamajoside is presented in Fig. 3 , that of acteoside is presented in Supplementary Fig. 1 .

More polar PGs were detected at lower retention times. 8 was putatively identified as a PG bearing an additional glucose (hexose) on plantamajoside, presenting the [HT-Glc(Glc,Glc)-CA – H] – deprotonated molecular ion (m / z 801). In MS 2, it yielded [HT-Glc(Glc)-CA – H] – ([M – H] – of 18), which could be fragmented further, and yielded the same fragments, as 18 (m / z 477 and m / z 315). 19 was putatively identified as lavandulifolioside, showing fragmentation described in Schmidt et al. (2013). The compound was described in P. lanceolata leaves, and micropropagated plants, but could not be detected from calli so far ( Budzianowska et al., 2004). Proposed fragmentation scheme is available in Supplementary Fig. 2 .

Simpler PGs were also found in the calli, as also presented in Budzianowska et al. (2004). 22 and 30 (m / z 477) were putatively identified as desrhamnosyl-acteoside (=calceorioside B) and its isomer. Distinguishment of the suggested deprotonated molecular ion [HT-Glc-CA – H] – from [HT-Glc-Glc – H] – (both m / z 477) was done on basis of retention time (polarity in RP-HPLC), and the presence of fragments, that can be linked to CA, and cannot be explained by fragmentation of glucose. The proposed fragmentation scheme of 22 is plotted in Supplementary Fig. 3 .

In the higher retention time range, several less polar PGs were detected. Some of them could also be putatively identified, some were found to be methylated PGs. Of these, many had methyl-caffeic acid (MeCA) as a phenylpropanoid aglycon, and this was found to be the case with 59 (leucosceptoside A, [M – H] –: [HT-Glc(Rha)- MeCA – H] –); 53 and 61 ([M – H] –: [HT-Glc(Glc)-MeCA – H] –). The decay routes of these NPs were identical to that described in Kırmızıbekmez et al. (2005). The proposed fragmentation schemes are shown in the Supplementary Fig. 4 . The structures of these three NPs were supported by the presence of fragments m / z 491[HT-Glc-MeCA – H] – and 315 [HT-Glc – H] –, indicating that it was not the HT aglycon, that carried the methyl group. The fragment ion m / z 477 can be interpreted as [HT-Glc-Glc – H] – in this case, unlike in most other PGs. Methylated plantamajoside derivates have only been described from Digitalis ( Jin et al., 2011) so far.

Even less polar, dimethyl- and/or acetylated PGs were also found. 64 was putatively identified as dimethyl-plantamajoside, both aglyca (HT and CA) were found to be methlyated, [M – H] –: [MeHT-Glc(Glc)-MeCA – H] –. The fragmentation pattern fully supports this structure, the most specific fragment was [MeHT-Glc- MeCA – H] – (m / z 505).

In the case of 71 and 65, the two, completely identical fragmentation patterns indicated the structure [HT-Glc(Ac)(Rha) Me 2 CA – H] –. The CH 3 CHO loss from the dimethyl CA aglycon was the main fragmentation route resulting in [MeHT-Glc(Ac) (Rha)-pCoumA – H] – (m / z 649), by conversion of the CA aglycon into p-coumaric acid (pCoumA). The ion m / z 649 yielded several fragments in MS 3, all supporting the proposed structure. Identification of fragments, and pathways of decay in MS/MS is summarized in Supplementary Fig. 5 .

Many unidentified peaks were also detected. These showed characteristic fragments common to the identified PGs. Several of them also showed CH 2 CO loss, indicating possible presence of an acetyl group, but most had insufficient fragments to deduce their structures, or the pattern led to conflicting evidence on substituent positions. Identification of these molecules was beyond the scope of our study. As most of the found PGs yielded MS 2 ions between 40 and 70% efficacy (ratio of abundance of main MS 2 ion and [M – H] –), the lack of MS 2 ions was considered to be indicative of a substantially different structure. These molecules could not be identified.

2.3. Effects of N source composition on phenylethanoid glycoside pattern and concentration in P. lanceolata calli

2.3.1. Effects of N source composition on phenylethanoid glycoside pattern, and the metabolome as a whole

Of the many peaks found in the set of total ion chromatograms, 89 were selected for a more in-depth study (see Section 4.8. for filtering, and data processing). As the dataset had a high degree of multicollinearity, the scaled, centered abundance data of the 89 NPs were subjected to dimensionality reduction with principle component analysis (PCA). The first three PC scores were visualized in a pseudo-3D plot shown in Fig. 4a . It is visible, that high and low NH 4 + content media are separated along the axis PC2. The media without NH 4 + had very similar scores suggesting highly similar metabolome. It can also be seen, that the metabolome response seemed to be non-linear, as no other clear trends can be observed in the PCA plot. In the loading plots ( Fig. 4b ) it is obvious, that a small group of metabolites is responsible for most of the variance in PC2. We can state that the NP pattern as a whole was indicative of both N source concentration and NH + 4 /NO 3 – ratio.

2.3.2. Effects of N source composition on abundance of individual natural products

As we wanted to evaluate the effects of the treatments on the NPs directly, every metabolite abundance was subjected to ANOVA separately, the N source concentration and the NH + 4 /NO 3 – ratio being the independent variables. The highest abundances of individual NPs were compared to those found on the most frequently used TC medium, MSM (60(0.33)), and the 20(0) medium used for the culturing of P. lanceolata calli by Budzianowska et al. (2004). Overall, we can state, that the N source composition had a very significant effect on the abundance of individual NPs: the N source concentration significantly influenced the abundance of 42 metabolites (p <0.00011), while the NH 4 + /NO 3 – ratio significantly altered concentrations of 10 metabolites (of 89 NPs examined) (p <0.00011). These results are summarized in Supplementary Table 1 .

If we take a closer look at the abundance gains of individual NPs compared to MSM, 60(0.33) in Supplementary Table 1 , it can be observed that the conventional medium was one of the worst for production of NPs. For most NPs, the best media were those containing six times less (10 mM) N source only. Most metabolites were synthesized at highest rate, when the NH + 4 /NO 3 – ratio was kept at 0.33, including the chief compound, plantamajoside. The best medium was therefore 10(0.33) for overall production. But, in case the yield of a specific NP would be the aim to be optimized, the optimal medium can be a very different composition (Supplementary Table 1 ). In many of the studies regarding the optimization of medium N source for high NP yield and growth, media with lower NH + 4 content than that in MSM was found to be optimal for the production of secondary metabolites, for example in a study on Echinacea adventitious shoots ( Wu et al., 2006). In our case, the 10(0.33) or 10(0) media were the best for production of most of the different NPs (Supplementary Table 1 ). Despite several of the tested NPs were identified as phenylethanoid glycosides, very different responses to NH + 4 /NO 3 – ratio and N source were detected, as detailed later.

After obtaining the correlation matrix of the scaled and centered matrix of metabolite abundances, and clustering of the metabolites with respect to their relative abundances in calli grown on different media, a heatmap was generated in R (Supplementary Fig. 6). A relatively strong multi-correlation can be observed for most of the metabolites. The clustering of the metabolites resulted in six groups ( Fig. 4b , Supplementary Fig. 6), that reacted relatively similarly to different media compositions. Interestingly, the NPs putatively identified as PGs were sporadically spread across the clusters with different response to the treatments.

While the abundance of NPs in clusters 1 and 2 was the highest on 10(0), maximal abundances for NPs in clusters 3, 4 or 5 could be achieved on 10(0.33). As 10 mM N media are substantially different from the reference media, the production of these metabolites could be increased to a high extent (Supplementary Table 1 ). In many cases, threshold-like responses were found, that were for example described for anthocyanin production response to N source concentration by Hirasuna et al. (1991) in grape cell-cultures. The NP with the highest abundance in the sample was plantamajoside (18) in cluster 4. For 18, the best yield was obtained on medium 10(0.33), where 3.54 ± 0.83% (dry wt.) content was found. The least content was 1.04 ± 0.63% (dry wt.), found on medium 40(0.11). Interestingly, the metabolite with the second highest abundance fell in a different cluster; both the optimal medium and the response to the treatments was found to be substantially different – as it can be seen by comparing the response surfaces ( Fig. 5d and f ). Acteoside content varied from 0.54 ± 0.29% (of dry wt.) on 20(0) to 1.30 ± 0.40% on 40(0.33).

2.3.3. Statistical interaction between N source and NH + 4 /NO 3 – ratio

Most of the NPs responded in a non-linear fashion to the combinations of these two factors tested. Responses are plotted as heatmaps for better evaluation possibility of this interaction ( Fig. 5 ). As stated before, there were many response types, despite many NPs were putatively identified as phenylethanoid glycosides. The presence of local maxima and/or minima renders the independent optimization of NH 4 + / NO 3 – ratio and N source concentration impossible.

This statistical interaction means that the response to the N concentration at different levels of the NH + 4 /NO 3 – ratio was different: for example in the case of 27 ( Fig. 5f ), at 0.33 NH 4 + /NO 3 – ratio, the response to the N concentration showed an optimum curve: the highest abundance was detected at 40 mM N, while at 0 NH + 4 /NO 3 – ratio, a threshold-like response was observed, with low abundance except at 10 mM N. This means that if we would have done the yield optimization for 27 using the one-factor at a time design, we might have ended up in a suboptimal medium composition, as shown later by in silico OFAT optimization studies.

The studies on TCs from Plantago spp. almost exclusively used the original MS, supplemented with various hormone concentrations ( Fons et al., 2008). A few studies used modified MS, these were either diluted variants (e.g. ½ strength) ( Fons et al., 2008), or reduction in some inorganic components were used, as in Mederos et al. (1997). The effects of the N source concentration and the NH 4 + /NO 3 – ratio on phenylethanoid glycoside production was investigated for the first time in our study. NH 4 + and NO 3 – are usually thought to modulate metabolism via enhancing growth and driving the TC towards more enhanced growth at the cost of production of less secondary metabolites. High carbon source to N source ratios in plant TCs usually lead to higher metabolite biosynthesis. As the growth speed decreases with the reduction of the relatively available N (or other limiting nutrient), the excess carbon pool is driven towards synthesis of secondary metabolites to a higher extent – thus, secondary metabolite accumulation often occurs after the period of maximum growth ( Collin, 2001). In our case, however, none of the metabolite abundances were shown to be significantly influenced by GI itself (as shown by a linear model). If the N source would have been limiting in any of the media, GI would have likely been higher at higher N concentrations, but GI was unaffected by N concentration (p = 0.54). This suggests that the availability of NO 3 – and NH + 4 in different ratios and net concentrations exerted more direct effects on biosynthesis of the examined NPs. These direct effects may include NH 4 + acidity stress ( George et al., 2007), or other, unknown pathways.

We can state, that while the widely used MSM is suboptimal with respect to NP abundance in the TCs, careful experimental design has to be implemented when reoptimizing the N source in the TC medium.

2.3.4. Simulated one-factor at a time optimization of natural product yields

As the abundance of several of the NPs were very good examples of the statistical interaction between the N source concentration and the NH + 4 /NO 3 – ratio, OFAT experiments were simulated in silico to assess the risk of ending up with a NP suboptimal yield. The inability to independently optimize media components was also suggested by Amdoun et al. (2009) (regarding NO 3 – /Ca 2+). Other studies also found, that NP accumulation depended on media N source, but the interactions between the NH 4 + /NO 3 – ratio and the N source concentrations could not be assessed for effects on the metabolome in most cases, because most studies used one-factor at a time strategy ( Guo et al., 2005; Wang and Tan, 2002), or did not use a factorial approach ( Jacob and Malpathak, 2005).

To show the advantage of the FF strategy, four virtual optimization studies were carried out in silico. The interaction between the tested parameters and the consequence of using OFAT is perhaps best presented this way. The composition of the media was optimized in different OFAT regimes for each of the 89 NPs individually. The virtual experiments either started from MSM (60(0.33)) or 20(0). The optimization of the NH + 4 /NO 3 – ratio was followed by the optimization of the N source concentration or vice versa. In strategy #1, the starting medium was MSM and the virtual optimization began with the NH + 4 /NO 3 – ratio, followed by the virtual optimization of the N source concentration. In this scenario, 53 of the 89 NP abundances were suboptimal, with the suboptimal NPs showing a mean 1.56-fold abundance advantage in FF compared to the OFAT optima. The list of suboptimal media obtained by the different OFAT simulations for each NP is shown in Supplementary Table 2. Starting with medium 20(0) in strategy #2 resulted in similar results: 44 NP abundances were sub-optimal with the FF optima being 1.28-fold better on average. When the scheme was reversed to the experimental setup less common in the literature (N source concentration was optimized first in strategies #3–4), the OFAT performance substantially increased. Despite this, there were still 12 and 19 NPs sub-optimal abundance NPs, when starting from MSM or 20(0), respectively. The average fold advantage of the FF for these NPs was 1.30 and 1.37. For some NPs, the difference between the best FF and OFAT medium was greater, than 2-fold (Supplementary Table 2). It is also noteworthy, that the abundances of 6 NPs did not reach the FF optimum, regardless of the starting medium. Thus, if NP abundance optima are reached with OFAT, it can also simply be by chance. This chance can be increased however, by optimizing the N source concentration first, followed by the NH 4 + /NO 3 – ratio.

3. Conclusions

In this study, we successfully applied LC–ESI– MS 3 for the characterization of tissue cultures of a medicinal plant, P. lanceolata L. Many phenylethanoid glycosides previously not found in these calli were detected, and their structures were putatively identified.

The main goal of the study was to examine the effects of the N composition of the medium on the production of phenylethanoid glycosides and other NPs in P. lanceolata calli . Four N concentrations and four levels of the NH + 4 / NO 3 – ratio were tested in a full-factorial experiment enabling the estimation of interactions between these two parameters. Many conclusions could be drawn with regard to secondary metabolite production in medicinal plant tissue cultures.

The metabolomic approach was shown to be a powerful one when evaluating the results. The data processing and visualization procedures provided high throughput and highlighted most of the phenomena. The presented approach is therefore strongly encouraged when optimizing secondary metabolite production in in vitro cultures.

The original Murashige Skoog medium was found unsuitable for high yield production of phenylethanoid glycosides. In fact, it was one of the media with least NP abundances. The medium proposed by Budzianowska et al. (2004) for P. lanceolata tissue cultures was optimal for growth, but not NP production.

However, with the manipulation of the NH 4 + / NO 3 – ratio and N source concentration, very significant increase in the natural product accumulation could be achieved. Compared to the reference media, 1.5–3-fold increases in abundance were achieved for most metabolites. The highest NP abundance media were those with low N source concentrations (10 mM N). Thus, we have shown that a very simple and inexpensive modification of the medium can dramatically increase phenylethanoid glycoside yields. What is more, the manipulation of these parameters can also be easily implemented in industrial scale cultures, and is compatible with elicitations as well. The results suggest that N source composition must be re-optimized for optimal production of these natural products and that the two-stage culture strategy is to be preferred when tissue cultures are used for production of phenylethanoid glycosides.

For major metabolites, the maximum yields could be observed on different media: optimal plantamajoside production was achieved on 10(0.33), 3.54 ± 0.83% (dry wt.), while, in the case of acteoside, the best yield was 1.30 ± 0.40% (dry wt.) on 40(0.33). For most metabolites, the medium 10(0.33) was found to be optimal, for most of the others, 10(0) or 40(0.33).

Comparing the optimization strategies also led us to interesting conclusions. Interactions were detected between the two parameters for many metabolites – the response to the NH 4 + / NO 3 – ratio was not the same at different N source concentrations. The manipulation of the NH 4 + / NO 3 – ratio and N source concentrations has led to sub-optimal yields in case of in silico simulated one-factor at a time experimental designs, for many natural products. Therefore, factorial experiments with less repetitions must be preferred to one-factor at a time experiments with higher repetition counts. If OFAT is unavoidable due to some constraint, the N source concentration must be optimized first, as it was shown to lead to suboptimal yields much less often than OFAT protocols optimizing the NH + 4 / NO 3 – ratio first.

As the overall polyphenolic metabolite concentration was also heavily influenced by the tested parameters, our suggestions should be considered to be adopted when optimizing yields of other natural products in tissue cultures. These results are likely to apply for other metabolites, that are biosynthesized through the phenylpropanoid pathway.

Figures

Fig. 2. LC–ESI–MS chromatogram of the MeOH extract of a P.lanceolata callus grown on medium 10(0.33) (25 mg dry wt./ml, total ion chromatogram in negative ion mode). Note that in this injection, the area under curve of plantamajoside was beyond the linear range of determination, and thus, major NPs were quantified in the 100-fold dilution, and not from the chromatogram presented here.Main peaks are marked with the metabolite numbers used in text, see Table 1 for putative identification.NPs compared with standards are marked with an asterisk. For m/z values, and putative identification, see Table 1. Abbreviations: AC, acteoside; PM, plantamajoside.

Fig. 3. Proposed fragmentation pattern of plantamajoside, the main phenylethanoid glycoside of the P. lanceolata callus extract in negative mode LC–ESI–MS3. The main fragmentation route is highlighted with bold arrows and bold font.

Fig. 1. Selected phenylethanoid glycosides detected from the MeOH extracts of P. lanceolata calli by LC–ESI–MS3. Plantamajoside, R 1 = Glc, R 2 = H; Acteoside, R 1 = Rha, R2 = H; Desrhamnosyl-acteoside, R1 = R2 = H; Lavandulifolioside, R1 = Ara(1->2)Rha, R2 = H; Leucosceptoside A, R1 = Rha, R2 = Me. For details of identification, see text and Table 1. Abbreviations: Ara, arabinose; Glc, glucose; Rha, rhamnose.

Fig. 4. Principal component (PC) analysis score plot of the scaled, centered and per-composition averaged natural product abundance dataset. The plot shows the effect of media composition on the metabolome as a whole. The NP abundances were measured by LC–ESI–MS3 from MeOH extracts of P. lanceolata calli grown on media with different N source compositions. For more details, see text. Subplots: (a) Pseudo-3D score plot of PCs 1–3. PC2 and PC3 scores are plotted on axes x and y, respectively, while PC1 is made proportional to point size (see legend). Media codes are as follows: N concentration (mM), followed by NH + /NO – ratio (ratio of NH + in the total N source) in 4 3 4 parentheses. The NH + concentration in the media is coded by color (see legend). Abbreviations: NSC, N source concentration. (b), Loading plot for the first three PCs showing 4 contribution of the 89 individual metabolites to the PC scores. The order of NPs was obtained from hierarchical cluster analysis of the NP abundances.

Fig. 5. Abundance of selected natural products from extracts of P. lanceolata calli grown on media with different N source concentrations and NH + /NO – ratios (ratio of NH + in 4 3 4 the total N source).Color is proportional to the abundance (area under curve) for the given NPs, quantified in extracted ion chromatograms (a, b, c, e), or content of dry weight expressed as percent (d, f) Subplots: (a), 22, putatively identified as desrhamnosyl-acteoside, cluster 1; (b), 49, a plantamajoside isomer, cluster 2; (c), 8, putatively identified as a plantamajoside hexoside, cluster 3; (d), 18, plantamajoside (plantamajoside), cluster 4; (e), 65, putatively identified as a dimethyl-acetyl acteoside, cluster 4; (f), 27, acteoside, cluster 6.