METABOLIC ANALYSIS METHOD

Abstract

The invention provides methods of characterising the mode of action of a test compound by exposing populations of a living system to control and treatment conditions, using an unlabeled and labeled isotope source materials. The measurement of the isotopomer distribution of metabolites reveals the effect of the test compounds on the metabolism in that living systems. Further embodiments include where the living system is a plant, fungus, invertebrate, bacteria or virus, and where the test compound is a screening lead in a pesticide product development program.

Claims

1. A method of characterising a mode of action of a test compound comprising a) a first biological sample and a second biological sample, wherein there is at least one isotopic difference between the samples, and wherein the first and second biological samples are isolated from a living system exposed to the test compound; b) a third biological sample and a fourth biological sample, wherein there is at least one isotopic difference between the samples, and wherein the third and fourth biological samples are isolated from a living system not exposed to the test compound; c) performing chromatograph and mass spectrometry analysis of metabolites extracted from the biological samples to provide mass spectra data; d) calculating a mass isotopomer distribution for the metabolites from the mass spectra; e) identifying metabolites that have a significant different isotopomer distribution when the living system was exposed to the test compound to the isotopomer distribution when the living system was not exposed to the test compound.

2. The method of claim 1 wherein the mass spectra data provided in step c) is deconvoluted and metabolites subsequently identified from the mass spectra by data comparison with a reference library.

3. The method of claim 1 wherein the mass isotopomer distribution generated by step d) is determined by: calculating the retention indices and deconvolution of the mass spectral data; pairing the deconvoluted mass spectra data across samples and identifying metabolites according to isotope incorporation; identifying those metabolites based on similarity of spectra, retention time and/or retention index by comparison to a reference library; calculating the mass isotopomer distribution of the metabolites.

4. The method of claim 1 wherein step e) comprises comparing the mass isotopomer distribution between metabolites having at least one isotopic difference and identifying those metabolites which have a statistically different deviation between the biological samples isolated from a living system exposed to the test compound and the biological samples isolated from a living system not exposed to the test compound.

5. The method of claim 1 wherein the mode of action of the test compound is identified according to the metabolites identified from step e).

6. The method of claim 1 wherein living system incorporates an isotopic label from an isotope-labeled substrate.

7. The method of claim 6 wherein the isotope-labeled substrate is selected from .sup.2H.sub.2O, H.sub.2 .sup.18O, .sup.2H-glucose, .sup.2H-labeled amino acids, .sup.2H-labeled organic molecules, .sup.13C-labeled organic molecules, .sup.13C-labeled glycerol, .sup.13CO.sub.2, .sup.15N-labeled organic molecules, .sup.3H.sub.2O, .sup.3H-labeled glucose, .sup.3H-labeled amino acids, and .sup.3H-labeled organic molecules.

8. The method of claim 1 wherein the living system is a plant, fungus, virus, bacteria, invertebrate, or vertebrate.

9. The method of claim 1 wherein the living system is a plant and the isotope-labeled substrate is .sup.13CO.sub.2.

10. The method of claim 1 wherein the living system is a fungus and the isotope-labeled substrate is a .sup.13C-labeled organic molecule.

11. The method of claim 1 wherein the living system is exposed to the isotope-labeled substrate for between 10 mins and 48 hours.

12. The method of claim 1 wherein the living system is exposed to the test substrate for between 10 mins and 48 hours.

13. The method of claim 1 wherein the test compound is a small molecule or a biological factor.

Description

FIGURE LEGENDS

[0120] FIG. 1: Outline for Calculating the mass isotopomer distribution for metabolites from the mass spectra data. (a) Samples to which the method can be applied; (b) Ion chromatographic deconvolution of mass spectral data; (c) deconvoluted mass spectral data of metabolites derived from different biological samples are then paired according to the retention index values and spectrum similarity; (d) paired mass spectra are applied to compute the difference spectrum d(x) for each detected compound. MID for unlabeled is put in the positive direction (above the axis), MID for labeled is put in the negative direction (below the axis). Labeled ion fragments form a shape similar to first derivative of peak function, while in contrast unlabeled ion fragments are eliminated (left); (e) This difference spectrum is integrated over the whole spectrum to obtain the integral of d(x); (f) MID is calculated using a set of linear equations to elucidate the composition of mixtures of isotopomers for each detected labeled fragment.

[0121] FIG. 2: Mass isotope distribution (MID) of two unidentified metabolites, RI 3506 and RI 3309 for fungicide one and fungicide two. The 2 h and 6 h graphs differ in incubation time after treatment. Each individual graph shows the mass distribution (M0, M1, M2, . . . Mn for n=number of .sup.13C atoms in the molecule) of the metabolite for that particular combination of treatment and incubation time point. The decrease and increase triangles visualize how the MID has changed in comparison to the respective control. Grey circles represent unchanged MIDs. The dashed line indicates the fractional contribution of the metabolite (fraction of .sup.13C atoms in the metabolite). (A): RI 3506 MID values for control fungicide one (2 hours); (B): RI 3506 MID values for control fungicide one (6 hours); (C): RI 3506 MID values for fungicide one treatment (2 hours); (D): RI 3506 MID values for fungicide one treatment (6 hours); (E): RI 3506 MID values for control fungicide two (2 hours); (F): RI 3506 MID values for control fungicide two (6 hours); (G): RI 3506 MID values for fungicide two treatment (2 hours); (H): RI 3506 MID values for fungicide two treatment (6 hours); (I): RI 3309 MID values for control fungicide one (2 hours); (J): RI 3309 MID values for control fungicide one (6 hours); (K): RI 3309 MID values for fungicide one treatment (2 hours); (L): RI 3309 MID values for fungicide one treatment (6 hours); (M): RI 3309 MID values for control fungicide two (2 hours); (N): RI 3309 MID values for control fungicide two (6 hours); (0): RI 3309 MID values for fungicide two treatment (2 hours); (P): RI 3309 MID values for fungicide two treatment (6 hours).

[0122] FIG. 3: The metabolites displaying the largest treatment effects for fungicide one and two are listed. As none of these metabolites could be uniquely identified using the NIST data base, their retention indices (RI) are used as identifiers.

[0123] FIG. 4: Two hypotheses on the differences in mode-of-action between fungicide one and two. The boxed represent (unidentified) metabolites and contain the retention index ID. The connecting arrows indicate one or more enzymatic steps between the metabolites.

[0124] FIG. 5: Metabolomics results of fenpropidin treatment for samples taken 0.5, 2, 6 and 48 hours after treatment. The values shown represent fold-changes to the control. The grey background indicates significant increases in metabolite concentration.

[0125] FIG. 6: Mass isotope distribution (MID) of valine and glutamic acid for herbicide four. Each individual graph shows the mass distribution (M0, M1, M2, Mn for n=number of .sup.13C atoms in the molecule) of the metabolite for that particular combination of treatment and incubation time point. The decrease and increase triangles visualize how the MID has changed in comparison to the respective control. Grey circles represent unchanged MIDs. The dashed line indicates the fractional contribution of the metabolite (fraction of .sup.13C atoms in the metabolite). (A) Valine MID values for control herbicide four (2 hours); (B): Valine MID values for control herbicide four (5 hours); (C): Valine MID values for herbicide four treatment (2 hours); (D): Valine MID values for herbicide four treatment (5 hours); (E): glutamic acid values for control herbicide four (2 hours); (F): glutamic acid MID values for control herbicide four (5 hours); (G): glutamic acid MID values for herbicide four treatment (2 hours); (H): glutamic acid MID values for herbicide four treatment (5 hours).

[0126] FIG. 7: Mode-of-action hypothesis for herbicide four. The boxes represent metabolites and contain the annotation or the retention index ID of the metabolites. The connecting arrows indicate one or more enzymatic steps between the metabolites. Whenever an increase in label incorporation is detected, the metabolite box is equipped with an increase triangle. Decreases are represented by a triangle pointing downward. Unchanged label patterns are indicated by a grey circle, whereas the symbol for an empty set is used if no label could be detected. The crosses indicate the mode-of-action hypothesis. Measured metabolites are in black boxes, boxes with a dashed line and grey background represent groups of metabolites and if a metabolite was analyzed in a targeted way, this is represented by a light grey dashed box.

[0127] The invention will now be described further by way of the following non limiting examples.

EXAMPLE 1: MODE OF ACTION IDENTIFICATION

Introduction

[0128] “Screening leads” in pesticide development are compounds with high efficacy and potency that additionally display an interesting mode-of-action during initial screening. Approximately half of these promising compounds have a mode-of-action that is not elucidated easily, and they proceed into the advanced mode-of-action analysis. These latter compounds are especially interesting as potentially new pesticide classes with entirely new modes-of-action are discovered. However, elucidating a new mode-of-action is particularly difficult with only a limited number of methods available in the advanced mode-of-action analysis pipeline.

[0129] Metabolite profiling for mode-of-action identification is one approach to determine the mode-of-action of screening leads. However, current metabolic profiling methodologies can only detect major metabolic changes, as interfering secondary effects disguise the primary mode-of-action.

[0130] Against this background the present inventors sought to use isotopic tracers (for example .sup.13C) to distinguish between primary and secondary effects. Surprisingly they determined that this new approach provides an improved methodology identify the mode-of-action for promising compounds. The improved methodology has been termed MIAMI (Mass Isotopolome Analysis for Mode-of-Action Identification).

[0131] Advantages of the improved method of the invention include: increase sensitivity and broader scope of application than existing methods, to create a generalized platform for identification of the primary mode-of-action.

[0132] Further benefits of MIAMI are: [0133] The method is untargeted and thus not limited to known metabolites or pathways. A broader analytical scope will allow for the identification of novel modes-of-action and new biomarkers. [0134] Using isotopic tracers will increase the capability to identify modes-of-action by metabolic profiling, as it has an increased sensitivity, allowing for the identification of primary effects in highly variable biological systems. [0135] A single combined analytical and computational platform can be used for multiple purposes, e.g. pesticide discover, toxicology assessment and white biotechnology and toxicology.

[0136] MIAMI improves existing methods for the identification of mode-of-action. This knowledge can allow more rational optimization of compounds for the on-target and to identify (toxicological) off-targets in pesticide development. Therefor this technology can further speed up research projects and identify possible risks early on.

Mass Isotopolome Analysis

[0137] The measurement of the isotopomer distribution (˜enrichment) of each metabolite adds an important dimension to metabolic data. The challenge lies in how to extract the knowledge from this big and complex data. Recent advances in systems biology (Weindl et al., (2016) Bioinformatics, vol 32(18) 2875-2876) deliver an intelligent solution by building similarity networks and identifying differential enzyme activities.

[0138] For this purpose, separate cultivations are done under control and treatment conditions using an unlabeled and labeled carbon source (for example labeled glucose or labeled CO.sub.2). Samples are taken, prepared and measured with gas chromatography/mass spectrometry in SCAN mode. After deconvolution of the mass spectra, NTFD (Hiller K, et al (2013) Bioinformatics, 29(9):1226-8) is used to detect all labeled metabolite fragments in a non-targeted manner. After quality control and ion selection, the mass isotopomer distribution (MID) is determined for all known and unknown labeled compounds. Metabolites are then sorted by their fractional contribution of tracer derived isotopes, which is correlated to their distance to the source of labeling. Using MID alignment and distance calculation, a network based on MID similarity can be constructed, which connects metabolites with similar MID patterns in the control dataset. A variability analysis is performed based on MIDs between control and treatment conditions, revealing flux changes between control and treatment conditions in a non-targeted manner. The identification of connected metabolites with variations in metabolic fluxes allows the identification of the MoA in known pathways and unknown pathways just by MID similarity.

Methods Used Herein

Tracer Experiments for Herbicides

Plant Material and Cultivation Conditions

[0139] Lemna paucicostata plants were grown under sterile conditions in nutrient solution [KNO3 (400 mg/L), CaCl2)*2 H2O (540 mg/L), MgSO4*7 H2O (614 mg/L), KH2PO4 (200 mg/L), Fetrilon 13% (2.81 mg/L), trace element solution: MnCl2*4 H2O (415 mg/L), H3BO3 (500 mg/L), Na2MoO4+2 H2O (120 mg/L), ZnSO4*7 H2O (50 mg/L), CuSO4*5 H2O (25 mg/L), CoCl2 (25 mg/L)] and constant light for one week prior to the experiments.

[0140] .sup.13C—CO.sub.2 Experiment

[0141] One-week-old Lemna plants were placed into nutrient solution in 30 mL Falcon tubes, treated with Imazapyr (a known ALS inhibitor, 10 μM), closed with air- and liquid-permeable gauze and incubated for one hour prior to .sup.13C—CO.sub.2 labeling. Incubation of the plants directly in the sample tubes ensured rapid sampling times and limited the contamination with .sup.12C—CO.sub.2.

[0142] .sup.13C—CO.sub.2 labeling was done in a gas tight valve-equipped plexiglass box connected to a gas absorber. Plant samples were placed into the box, the box was closed and .sup.12C—CO.sub.2 was absorbed for 30 sec. Immediately after absorption, 400 ppm of .sup.13C—CO.sub.2 were injected into the box with a syringe. In case of the unlabeled experiments, we used the same absorption conditions and then shortly opened and reclosed the box to allow .sup.12C—CO.sub.2 entry.

[0143] The experiments described herein used the herbicide imazapyr in one concentration, two labeling conditions and two different labeling timepoints (3 and 6 hours after CO.sub.2 injection). Five replicates were down per treatment. Due to the limited space in the box and the experimental design, the experiments had to be split into several rounds; in each round untreated controls were included for direct comparison.

[0144] After the labeling time, the box was opened, and the samples quickly harvested. Therefore, the herbicide solution was poured through the gauze, the sample was slightly dried on tissue paper, closed with a lid and directly frozen in liquid nitrogen.

Tracer Experiments for Fungicides

Media and Solutions

[0145] Media are prepared using commercially available malt extract (#70167, Sigma Aldrich). Malt-solution is prepared by dissolving 20 g malt extract in 1000 mL purified water and setting the pH to 6.8. Rice leaf agar is prepared by mixing 50 g of freshly frozen young rice leaves with 1000 mL purified water and pressing it through a household filter. Subsequently, 10 g of soluble starch (VWR 1.012.571.000), 2 g of baking yeast (Biolabor) and 20 g of agar-agar (granules, Becton Dickinson, 214510) were added. Both the rice leaf agar and the malt-solution media are sterilized for 15 min. at 121° C.

Strain

[0146] Pyricularia oryzae, Strain J1 was used for all experiments.

Cultivation Conditions

[0147] Initially, Pyricularia oryzae spores are harvested by adding 10 mL sterile malt-solution (2% w/v) to rice leaf agar plates with 14-day old fungus and loosening the spores with a drigalski spatula. The malt-solution containing the fungal spores is filtered through a double layer of sterile gauze. Subsequently, the spores are cultivated in 25 mL malt-solution in 100 mL Erlenmeyer flasks (20° C., 140 rpm, aluminium foil seal).

[0148] Four days after the start of the spore incubation two Erlenmeyer flasks, each containing 25 mL spore-solution, are mixed, diluted with 50 mL malt-solution, and homogenised using an Ultra-turrax homogenizer. One mL of the homogenised spore-solution is added to 25 mL sterile malt-solution and further cultivated in 100 mL Erlenmeyer flasks (20° C., 140 rpm, aluminium foil seal).

.SUP.13.C-Glucose Experiment

[0149] Seven days after the start of cultivation (spore-harvest), 12 Erlenmeyer flasks with homogenous fungal growth were selected per treatment. Two mL glucose-solution (2% w/v) were added to each flask, as well as 100 μL solution containing the active ingredient. The glucose-solution for half of the samples consisted of 100% D-glucose-12C.sub.6, whereas the other half contained 50% D-glucose-.sup.12C.sub.6 and 50% D-glucose-.sup.13C.sub.6. An overview on active ingredient solutions is given in Table 1. Control samples were treated with 100 μL DMSO (99.7%, Berndt Kraft). After subsequent incubation at 20° C. and 140 rpm for two or six hours, cells were harvested as follows. Cultivation broth was vacuum-filtrated using a Büchner-funnel with filter paper. The precipitate was scraped off the filter paper into 2 mL Eppendorf tubes and immediately frozen in liquid nitrogen. Samples were stored at −80° C. until analytical processing.

[0150] The 12 samples per treatment originate from .sup.13C-labeled and non-labelled glucose solutions, two time points and three replicates.

TABLE-US-00001 TABLE 1 Overview on the active ingredients used in the MIAMI experiments with the respective concentration of the active ingredient solution applied. Treatment Concentration [μM] Control Fung1 1490 Fung2  167

Analytics

Sample Preparation

[0151] Lemna paucicostata and Pyricularia oryzae samples were freeze dried overnight (Christ Epsilon 2-10D,) and ground (Bead Ruptor, Omni International Inc.) prior to extraction.

[0152] An aliquot of the samples was placed in an Eppendorf tube together with a 3 mm stainless steel ball (Iemna paucicostata 4.8-5.2 mg, Pyricularia oryzae 9.5-10.5 mg). 350 μL water (ultrapure) and 750 μL methanol p.a. were added. Extraction was performed in a Retsch MM300 mixer mill for 3 min., 30 Hz at ambient temperature. Subsequently sample tubes were centrifuged in a Sigma 4-16KS for 10 min at 5000 rpm and 24° C. and 800 μL the supernatant from each tube was transferred into individual 400 μL water containing Eppendorf tubes for extracts collection. The original tube, containing the pellet and remaining volume of the first extraction, was used for a second extraction with a mixture of 190 μL methanol and 660 μL dichloromethane p.a. Extraction and centrifugation was repeated as done in the first extraction step. 800 μL of the supernatant of this second extraction step was transferred into the water and first extract containing Eppendorf extracts collection tube, related to the respective tissue sample. It was shaken for 10 sec. (Vortex mixer) and for clear phase separation it was centrifuged as done before.

[0153] Aliquots of the upper polar layer (Iemna paucicostata 500 μL, Pyricularia oryzae 133 μL) and lower lipid layer (Iemna paucicostata 50 μL, Pyricularia oryzae 144 μL) were transferred into individual glass vials for derivatization and analysis.

[0154] Extracts were evaporated with a Genevac HT-12 evaporator centrifuge for the following derivatization steps. During the derivatization steps the vials, placed in a custom-made metal rack, have been tightly sealed by a silicon mat.

[0155] Only the lipid extract residues were trans-esterified, using a mixture of 140 μL dichloromethane, 38 μL hydrochloric acid 37% in water, 320 μL methanol and 20 μL toluene for 2 h at 100° C. Those samples were then evaporated, using a Hettlab IR Dancer infrared vortex-evaporator.

[0156] Dry lipid and polar extract residues were treated with 50 μL of O-methylhydroxylamine hydrochloride in pyridine (20 mg/mL) for 1.5 h at 60° C. and MSTFA (N-Methyl-N-(trimethylsilyl)-trifluoracetamid) for 30 min. at 60° C.

[0157] Additional samples contained only aliquots of an alkane standard solution C21-C40 (No. 04071, Sigma-Aldrich) for retention index calculation.

GC-MS Analysis

[0158] GC-MS analysis was performed using a CTC GC PAL autosampler, attached to an Agilent 6890 gas chromatograph which was coupled to an Agilent 5973 MSD mass spectrometer. 0.5 μL of the derivatized samples were injected in splitless mode at an injector temperature of 280° C. Separation was performed with helium in constant flow mode capillary columns with 30 m length, 0.25 mm i. d. and 0.25 μm film thickness. For lipid samples an Agilent HP-5MS column at 1.7 mL/min. (70° C., 50 K/min., 130° C., 10 K/min., 340° C., 9 min.) was used and for polar samples an Agilent DB-XLB column at 1.0 mL/min. (70° C., 50 K/min., 100° C., 8 K/min., 200° C., 14 K/min., 340° C., 4 min.). Mass spectra were acquired in scan mode at 3 spec./sec. from 70 to 600 m/z.

Computational Data Processing

[0159] The generated netCDF file for each sample is imported into the MetaboliteDetector software (MetaboliteDetector—Deconvolution and Analysis of GC/MS Data; Version 3.2; http://metabolitedetector.tu-bs.de). Using the RI-calibration wizard, an n-alkane mix, specific to the day and instrument of the sample measurement, was chosen as a reference chromatogram for retention index calculation.

[0160] Next, the MetaboliteDetector files (.bin) were imported into MIA (Mass Isotopolome Analyzer; Version 1.0; (https://mia.bioinfo.nat.tu-bs.de/) for the detection of relevant metabolites. For this purpose, the labeled and unlabeled files for a respective treatment were selected and the labeled peaks determined (RI tolerance 99). Compound identification was performed using Golm Metabolome Database (http://gmd.mpimp-golm.mpg.de/) with a cutoff score of 0.75.

[0161] Subsequently, labeled metabolites were identified and the mass isotopomer distribution (MID) was calculated using the NTFD algorithm. The MID based distance calculation used the Needleman-Wunsch algorithm for gap filling and the Canberra distance measure to determine the distance between all labeled metabolites.

[0162] Subsequently, all detected metabolites were gathered in an experiment specific MetaboliteDetector compound library, containing the unlabeled spectra of all labeled metabolites, ions, the identified name, retention time and retention index.

[0163] Using MetaboliteDetectors MID wizard, the labeled files were loaded and processed using the generated library. The MIDs for all conditions using a targeted search with the non-targeted generated library were calculated.

Variability Analysis

[0164] To identify the variability between the treatment and control the largest common ion with usable MIDs was detected. MIDs were excluded if the absolute sum of mass isotopes exceeded 1 (considering a tolerance of 0.02 for each carbon atom) (Eq 2). Ions were only used if they were present in all experimental datasets.

Equation 2

Σ.sub.i=0.sup.M|M.sub.i|>1+Σ.sub.j=1.sup.M0.02  (2)

[0165] The variability between the selected common ions was the difference between the unlabeled mass isotopomer M0 of treatment and control. A threshold of 5% for polar metabolites and 2% for non-polar metabolites was used.

Sorting

[0166] Under the assumption that the .sup.13C tracer distributes through the organism starting from the entry point with a decreasing fractional contribution towards more downstream metabolites, the detected metabolites can be sorted based on the total amount of labeling. The fractional contribution (FC) is the amount of labeling of the metabolite related to the number of carbon atoms M (Eq 3), where m.sub.i is the ith mass isotopomer.

[00002]$\begin{array}{cc}\mathrm{Equation}3& \\ \mathrm{FC}=\frac{{.Math.}_{i=0}^M{m}_i.Math.i}{M}& \left(3\right)\end{array}$

[0167] A higher FC indicates a closer proximity to the tracer than a metabolite with a lower FC.

Context Generation

[0168] Based on fractional contribution and MID distances, pathways can be constructed without a priori biochemical knowledge. Nonetheless, the generated data needs to be set into a biochemical context. Here the identification of the detected metabolites is crucial to extrapolate the enzymatic reactions taking place.

Results

[0169] MIAMI-results were generated for two fungicides and one herbicides in a blind study (without the data analyst knowing the mode-of-action) as a retrospective validation of the method.

[0170] In general, MIAMI-results consist of a set of metabolites that display the largest changes in MID between treatment and control. Contextualization of these metabolites is based on their enrichment and their MID similarity, without a priori biochemical knowledge on the metabolite. Data interpretation is supported through library-assisted metabolite identification and biochemical pathway knowledge. In the coming paragraphs an overview on the amount of detected and changed metabolites is given, as well as the contextualized MIAMI results and the data interpretation for each of the investigated pesticides.

Analysis of Known Fungicide Modes-of-Action

[0171] For the known fungicides a total of 127 and 137 metabolites showed label incorporation, of which approximately 60% were measured in the polar phase. In total, the amount of changed metabolites varied between treatments from 42 to 44%, with most changes in the non-polar phase (table 2).

TABLE-US-00002 TABLE 2 Number of detected and changed metabolite MIDs for known fungicides. Polar Polar Non-polar Non-polar metabolites metabolites metabolites metabolites Fungicide detected changed detected changed fung 1 77 19 50 34 fung 2 84 26 53 34

Ergosterol Biosynthesis Inhibitors: Fungicide 1—Fenpropidin and Fungicide 2—Epoxiconazole

[0172] The similarities in labeling patterns between fungicide one and two were apparent from the beginning (FIG. 2). For both treatments the metabolites showing the largest treatment effects are near-to the same (FIG. 3). Therefore, both fungicides are analyzed simultaneously.

[0173] FIG. 2 shows the Mass isotope distribution (MID) of two unidentified metabolites, RI 3506 and RI 3309 for fungicide one and fungicide two. The 2 h and 6 h graphs differ in incubation time after treatment. Each individual graph shows the mass distribution (M0, M1, M2, . . . Mn for n=number of .sup.13C atoms in the molecule) of the metabolite for that particular combination of treatment and incubation time point. The decrease and increase triangles visualize how the MID has changed in comparison to the respective control. Grey circles represent unchanged MIDs. The dashed line indicates the fractional contribution of the metabolite (fraction of .sup.13C atoms in the metabolite).

[0174] In total, seven metabolites show a noticeably large differential MID (FIG. 3). However, two of these metabolites can only be detected under fungicide two treatment. Meaning that although both fungicide treatments are very similar, still distinct and prominent changes are identified.

[0175] Usually, the differentially labeled metabolites are identified using a library in the next step. Unfortunately, in this case not a single metabolite of the list could be uniquely identified using the NIST library, although for all metabolites the identification could be narrowed down to ergosterol-related compounds. This is a known problem in the identification of ergosterol-related compounds due to their high degree of similarity. The possibility that different derivates of the same metabolite exist within the subset cannot be excluded.

[0176] One of the powerful features of MIAMI allows contextualization and biological interpretation of unknown metabolites. Based on MID similarity closely related metabolites can be assembled into groups that approximate pathways and ranked according to the assumed sequence of enzymatic reactions, without a priori biochemical knowledge. In the case of fungicides one and two, all differentially labeled metabolites belonged to one pathway (likely ergosterol-biosynthesis) and the sequence of reactions can be taken from FIG. 3.

[0177] The metabolites displaying the largest treatment effects for fungicide one and two are listed. As none of these metabolites could be uniquely identified using the NIST data base, their retention indices (RI) are used as identifiers. The ranking (#) of the seven metabolites is based on the MID similarity and their fractional contribution (FC) in the control. The ranking represents an approximation of the in vivo sequence of enzymatic reactions. In the two most-right columns the changes in isotope distribution between treatment and control are represented graphically. Increase triangles indicate an increase in label incorporation in the metabolite under treatment, whereas red triangles indicate a decrease. For metabolites that are not found in a particular treatment, the symbol for an empty is used.

[0178] These results indicate a clear breaking point in the metabolic sequence with an enzymatic inhibition taking place between metabolites RI 3507 and RI 3309 for fungicide one and between RI 3569 and RI 3309 for fungicide two. This can be concluded because all metabolites in the sequence up until RI 3569 show an increase in label incorporation, whereas the once further down in the sequence all show a clear decrease. If both fungicides were to be analyzed separately, the next step would be to identify the metabolites RI 3507, RI 3309 and RI 3569 to pinpoint the exact enzyme that is inhibited. Such identification can be performed using other available libraries or in silico identification based on the MS-spectrum. Although improving our libraries and enabling in silico identification would be a tremendous benefit for the interpretation of these results, this was out of scope for MIAMI and will likely be followed up in a different project.

[0179] Based on both metabolite sequences two valid hypotheses can be generated on the pathway (FIG. 4). In hypothesis A a linear pathway is assumed in which 3 or more enzymatic steps take place between RI 3507 and RI 3309, in which fungicide one inhibits the first enzyme and fungicide two inhibits the third. The sudden appearance of RI 3442 and RI 3569 under fungicide two treatment is then explained by an accumulation in metabolite concentration due to the proposed pathway blockage. As this would not occur in the control or the fungicide one treatment, this explains the absence of the metabolites RI 3442 and RI 3569 under those conditions.

[0180] In hypothesis B a non-linear pathway is assumed in which multiple enzymatic steps take place between RI 3507 and RI 3309, and RI 3442 and RI 3569 are produced in a side-branch of the pathway. Although according to this hypothesis both fungicides act on an enzyme between RI 3507 and RI 3309, it is sure that they do not act on the same enzyme. If they would act on the same enzyme, the appearance of RI 3442 and RI 3569 only under fungicide two treatment would not be explained. Also, because multiple time points were measured in the experiment, the time component of the treatment effect can be analyzed. For RI 3507 the fractional contribution levels in the control only vary between 0.09 and 0.12 across the time points, whereas both for fungicide one and two the levels are clearly increased between 0.16 and 0.37 (FIG. 4). A closer look at the time component reveals that 2 hours after treatment both fungicides have a similar fractional contribution (0.16 and 0.17), whereas after 6 hours the enrichment of metabolite RI 3507 has increased significantly more under fungicide two treatment (0.37) than under fungicide one treatment (0.18). This points to an enzyme inhibition earlier in the pathway for fungicide two than for fungicide one. Likely, this is also linked to the overflow of metabolites into the side-branch of the pathway producing RI 3442 and RI 3569.

[0181] In FIG. 4 the connecting arrows indicate one or more enzymatic steps between the metabolites. Whenever an increase in label incorporation is detected, the metabolite box is equipped with an upwards triangle. Decreases are represented by a triangle pointing downward. These triangles are placed in the upper left (fungicide one) and upper right (fungicide two) corner. Small crosses occur for metabolites that are not detectable, whereas big crosses indicate the mode-of-action hypothesis.

[0182] To assess the viability of both hypotheses, identification of the metabolites was attempted using an existing library. In this way, six of the seven metabolites could be associated to a metabolite. RI 3507 was annotated as 24-methylene-cycloartenol, RI 3309, RI3069, RI3278 and RI3219 as ergosterol derivatives. Biochemical pathway information (https://www.genome.jp/kegg/pathway.html) shows that 24-methylene-cycloartenol is indeed upstream of both fungicide targets in the phytosterol pathway, which is normally not active in fungi. The ergosterol derivatives are downstream of the targets. RI3342 matched a lipophilic metabolite of unknown identity, a so-called “known Unknown”. Its unknown identity, in presence of known spectra for lanosterol, ergosterol and cycloartenol, make it more likely to represent a side product as an intermediate. Therefore, hypothesis B is more likely than hypothesis A. This is in agreement with the target of fungicide 1 being later in the pathway than that of fungicide 2.

[0183] It is clear that MIAMI has successfully pinpointed the modes-of-action of fenpropidin and epoxiconazole to a hand full of potential targets in steroid biosynthesis. As both of these fungicides have a known mode-of-action within this subset of potential targets, this result provides a retrospective validation of MIAMI.

[0184] In the present use case (mode-of-action identification of fungicides) MIAMI showed a greater potential to elucidate the mode-of-action as compared to metabolomics. Metabolomics results on four time points for fenpropidin show reductions in ergosterol, however not significantly after 6 hours of treatment (FIG. 5). With MIAMI we can show significant differences already after 2 hours of treatment. This indicates that MIAMI is the method of choice for fast-acting pesticides as it captures effects earlier after treatment. Metabolomics can pinpoint the mode-of-action at the 48-hour time point to ergosterol biosynthesis, which means that still approximately 40 enzymes are potentially inhibited, whereas MIAMI could narrow this down to a hand full of enzymes due to its unique contextualization feature. This also enables MIAMI to analyze and interpret metabolites that were previously unknown. As the analytical method is untargeted, it is also clear that potentially more metabolites are captured, increasing the coverage of metabolic pathways and therefore increasing the likelihood of discovering new modes-of-action.

Analysis of Known Herbicides Modes-of-Action

[0185] For the known herbicide a total of 110 metabolites showed label incorporation, of which approximately 90% were measured in the polar phase. The low number of detected metabolites in the non-polar phase is very likely due to a low .sup.13C-incorporation into these metabolites. Unfortunately, this decreases the pathway coverage significantly, and with that, also the potential to discover a mode-of-action drops noticeably. In total, the amount of changed metabolites was 46%.

Acetolactate Synthase Inhibitor: Herbicide 4—Imazapyr

[0186] Overall the label incorporation was very small, although present, for many metabolites after herbicide four treatment. All differentially labeled metabolites displayed an increase in fractional contribution, except for valine (FIG. 6). This indicates a mode-of-action in the valine biosynthesis pathway (FIG. 7), however to be more accurate on which enzyme is inhibited specifically, a higher coverage is needed. The limited resolution for herbicide four is definitely due to the lack of label incorporation for this treatment, because of which less metabolites could be detected, decreasing the coverage significantly. Nonetheless, MIAMI identified a potential mode-of-action in the branched-chain amino acid synthesis pathway. This is correct, however doesn't yield a satisfactory result as a specific enzyme could not be identified.

[0187] In FIG. 6, each individual graph shows the mass distribution (M0, M1, M2, . . . Mn for n=number of .sup.13C atoms in the molecule) of the metabolite for that particular combination of treatment and incubation time point. The decrease and increase triangles visualize how the MID has changed in comparison to the respective control. Grey circles represent unchanged MIDs. The dashed line indicates the fractional contribution of the metabolite (fraction of .sup.13C atoms in the metabolite).