Untargeted metabolite profiling and multivariate analysis
Untargeted metabolomic studies are exploratory in nature and usually result in extremely large and multi-dimensional datasets. Analyses of such datasets using chemometric tools can hugely aid data interpretation.
The representative averaged pre-processed spectra for each replicate belonging to the different plant species exhibit some visual distinction in mass spectra between the two plant genera (Fig. 2). This distinction between Centaurea and Geranium samples was further confirmed by unsupervised hierarchical clustering of the mass feature matrix (Fig. 3a). Within the two genera, the different plant species were mostly clearly separated based on their chemical features, with the exception of one of the C. stoebe replicates (Fig. 3a).
Visual comparison of the representative mass spectrum for each sample group can be used to broadly study the differing metabolic profiles. To further examine these differences and similarities between the root metabolic profiles of the four plant species, we employed PCA. The first two selected principal component axes explain over 75% of cumulative variance amongst the samples (Fig. 3b). Samples from different plant genera were strongly separated along the first PC-axis (~ 57%), whereas the separation along the second PC-axis (~ 18%) corresponded with within-genus variation (Fig. 3b). Together with hierarchical clustering (Fig. 3a), these results indicate a strong phylogenetic signal in root chemistry, as between-genus variation is considerably stronger than within-genus variation (Senior et al. 2016).
The number of mass features detected for each LAESI-MSI acquisition after performing data pre-processing and peak detection clearly shows that there are more mass features detected for the replicates of Centaurea as compared to those of Geranium (Table 1). The two Centaurea species shared 314 metabolites, whereas 53 metabolites were unique to either one of the species (Fig. 4a). Interestingly, 49 of these metabolites were unique for range-expanding C. stoebe, whereas only four were unique for native C. jacea. In contrast, for native G. molle more unique metabolites were detected than in range-expanding G. pyrenaicum (Fig. 4b). These results are in line with a previous study in which only root volatiles were examined (Wilschut et al. 2017) and indicate that range-expanding plants do not necessarily possess a more unique root chemistry than related natives.
To visualize the statistically significant metabolites for the two Centaurea species, a volcano plot was constructed (Fig. 5a). As seen in Fig. 5a, in total 367 metabolites were detected in genus Centaurea. Within this, ten mass features (shown in green) that are located in upper right quadrant of the plot, indicate that their concentration is significantly higher in native species C. jacea than in range-expanding species C. stoebe. The five mass features (shown in red) that are observed in the upper left quadrant indicate that their concentration is significantly lower in native species C. jacea than in range-expanding species C. stoebe. To examine the differences in metabolite concentrations for the C. jacea and C. stoebe pair, box-and-whisker plots were realized for four statistically significant metabolites chosen based on the volcano plot (Fig. 5b). The box-and-whisker plots and the ion intensity maps reveal that m/z 84.9607, m/z 159.0520 and m/z 557.290 are highly abundant in native species C. jacea, whereas m/z 272.9550 are highly abundant in range-expanding species C. stoebe. Additionally, the corresponding ion intensity maps for these metabolites were also generated to visualize the changes on the spatial level in the imaged roots. The ion intensity maps can be seen alongside the box-and-whisker plots in Fig. 5b. Each ion map is plotted on the same color scale (depicted below the ion maps) ranging from 0 (blue meaning least intense) to 1 (red meaning most intense), to allow comparison of relative ion intensity between images.
Similar analysis was performed for the two Geranium species (Fig. 5c). For this pair, in total 175 metabolites were detected. Within these, 15 mass features (shown in green) that are located in the upper right quadrant of the plot, which indicates that their concentrations are significantly higher in native species G. molle than in range-expanding species G. pyrenaicum. The four mass features (shown in red) that are observed in the upper left quadrant indicate that their concentration is significantly lower in native species G. molle than in range-expanding species G. pyrenaicum. The box-and-whisker plots for the four statistically significant metabolites selected from the volcano plot for the pair G. molle and G. pyrenaicum are shown in Fig. 5d. The ion intensity maps for these statistically significant metabolites are shown alongside box-and-whisker plots. As it can be seen, m/z 158.2647 and m/z 250.8271 show high abundance in native species G. molle, whereas m/z 172.3829 and m/z 196.5855 display high abundance in range-expanding species G. pyrenaicum. All significant metabolites detected for Centaurea and Geranium samples are listed in Online Resource 1 as Supplementary Table 1.
Taken together, we demonstrated the utility of the unique ambient ionization ability of LAESI coupled with MSI as a high-throughput method to explore the chemical differences in the root metabolome between two pairs of native and range-expanding plant species. This technology provided an in situ analysis method capable of revealing differentially produced metabolites linked to each group. We detected clear differences in root chemical profiles within both pairs of range-expanding plant species and congeneric natives using untargeted LAESI-MSI approach. Interestingly, the range-expanding plant species Centaurea stoebe showed a strongly unique root chemistry, which also may have enabled this species to become invasive in its introduced range in North America (Callaway and Ridenour 2004; Schaffner et al. 2011).
Furthermore, we demonstrated that LAESI-MSI can help to spatially elucidate the metabolite composition of the intact roots with minimal to no sample preparation. Our demonstration did not involve an exhaustive region-specific spatial analysis of the roots, but rather a ‘proof-of-concept’ by lateral profiling of the root samples. This allowed us to establish that LAESI-MSI of whole-root sections could reveal information on location-specific metabolite distribution without the need for any sample preparation. These results can help to reveal the role of single metabolites based on their location within the roots. The statistically significant biomolecules in the processed LAESI mass spectra can be putatively annotated by matching the observed masses with those of known metabolites using database searches. However, for confident compound identification it is best to couple LAESI mass spectrometry imaging with ion mobility separation (Stopka et al. 2017), which allows separation of isobaric species. Apart from this, LAESI-MSI with its feature of spatial mapping can hugely complement conventional extraction-based untargeted hyphenated-MS techniques like LC–MS or GC–MS. This can be further assisted by the application of tandem mass spectrometry for increased selectivity and structure assignments.
Overall, our results illustrate the feasibility of LAESI–MSI as a high-throughput technique for the detection and localization of metabolites from intact plant samples and gaining spatial information without the need for extensive sample preparation. The potential applications of this work could lead to rapid phenotyping of plant tissues as well as comparative untargeted metabolomics of different plant parts, a topic of considerable recent interest for plant research.
Author contribution statement
PG, KJFV and RAW devised the project. PG and RAW oversaw the sample collection and the data acquisition. PK planned and performed the bioinformatics analysis, interpretation of results and prepared the figures. PK, RAW and PG wrote the manuscript. WHvdP, PG, KJFV and RAW provided their comments and contributed to substantial revision of the manuscript.